summaryrefslogtreecommitdiffstats
path: root/lib/zstd/compress
diff options
context:
space:
mode:
Diffstat (limited to 'lib/zstd/compress')
-rw-r--r--lib/zstd/compress/clevels.h132
-rw-r--r--lib/zstd/compress/fse_compress.c668
-rw-r--r--lib/zstd/compress/hist.c165
-rw-r--r--lib/zstd/compress/hist.h75
-rw-r--r--lib/zstd/compress/huf_compress.c1335
-rw-r--r--lib/zstd/compress/zstd_compress.c6127
-rw-r--r--lib/zstd/compress/zstd_compress_internal.h1399
-rw-r--r--lib/zstd/compress/zstd_compress_literals.c159
-rw-r--r--lib/zstd/compress/zstd_compress_literals.h31
-rw-r--r--lib/zstd/compress/zstd_compress_sequences.c442
-rw-r--r--lib/zstd/compress/zstd_compress_sequences.h54
-rw-r--r--lib/zstd/compress/zstd_compress_superblock.c573
-rw-r--r--lib/zstd/compress/zstd_compress_superblock.h32
-rw-r--r--lib/zstd/compress/zstd_cwksp.h595
-rw-r--r--lib/zstd/compress/zstd_double_fast.c696
-rw-r--r--lib/zstd/compress/zstd_double_fast.h32
-rw-r--r--lib/zstd/compress/zstd_fast.c675
-rw-r--r--lib/zstd/compress/zstd_fast.h31
-rw-r--r--lib/zstd/compress/zstd_lazy.c2102
-rw-r--r--lib/zstd/compress/zstd_lazy.h119
-rw-r--r--lib/zstd/compress/zstd_ldm.c724
-rw-r--r--lib/zstd/compress/zstd_ldm.h111
-rw-r--r--lib/zstd/compress/zstd_ldm_geartab.h106
-rw-r--r--lib/zstd/compress/zstd_opt.c1446
-rw-r--r--lib/zstd/compress/zstd_opt.h50
25 files changed, 17879 insertions, 0 deletions
diff --git a/lib/zstd/compress/clevels.h b/lib/zstd/compress/clevels.h
new file mode 100644
index 0000000000..d9a76112ec
--- /dev/null
+++ b/lib/zstd/compress/clevels.h
@@ -0,0 +1,132 @@
+/*
+ * Copyright (c) Yann Collet, Facebook, Inc.
+ * All rights reserved.
+ *
+ * This source code is licensed under both the BSD-style license (found in the
+ * LICENSE file in the root directory of this source tree) and the GPLv2 (found
+ * in the COPYING file in the root directory of this source tree).
+ * You may select, at your option, one of the above-listed licenses.
+ */
+
+#ifndef ZSTD_CLEVELS_H
+#define ZSTD_CLEVELS_H
+
+#define ZSTD_STATIC_LINKING_ONLY /* ZSTD_compressionParameters */
+#include <linux/zstd.h>
+
+/*-===== Pre-defined compression levels =====-*/
+
+#define ZSTD_MAX_CLEVEL 22
+
+__attribute__((__unused__))
+
+static const ZSTD_compressionParameters ZSTD_defaultCParameters[4][ZSTD_MAX_CLEVEL+1] = {
+{ /* "default" - for any srcSize > 256 KB */
+ /* W, C, H, S, L, TL, strat */
+ { 19, 12, 13, 1, 6, 1, ZSTD_fast }, /* base for negative levels */
+ { 19, 13, 14, 1, 7, 0, ZSTD_fast }, /* level 1 */
+ { 20, 15, 16, 1, 6, 0, ZSTD_fast }, /* level 2 */
+ { 21, 16, 17, 1, 5, 0, ZSTD_dfast }, /* level 3 */
+ { 21, 18, 18, 1, 5, 0, ZSTD_dfast }, /* level 4 */
+ { 21, 18, 19, 3, 5, 2, ZSTD_greedy }, /* level 5 */
+ { 21, 18, 19, 3, 5, 4, ZSTD_lazy }, /* level 6 */
+ { 21, 19, 20, 4, 5, 8, ZSTD_lazy }, /* level 7 */
+ { 21, 19, 20, 4, 5, 16, ZSTD_lazy2 }, /* level 8 */
+ { 22, 20, 21, 4, 5, 16, ZSTD_lazy2 }, /* level 9 */
+ { 22, 21, 22, 5, 5, 16, ZSTD_lazy2 }, /* level 10 */
+ { 22, 21, 22, 6, 5, 16, ZSTD_lazy2 }, /* level 11 */
+ { 22, 22, 23, 6, 5, 32, ZSTD_lazy2 }, /* level 12 */
+ { 22, 22, 22, 4, 5, 32, ZSTD_btlazy2 }, /* level 13 */
+ { 22, 22, 23, 5, 5, 32, ZSTD_btlazy2 }, /* level 14 */
+ { 22, 23, 23, 6, 5, 32, ZSTD_btlazy2 }, /* level 15 */
+ { 22, 22, 22, 5, 5, 48, ZSTD_btopt }, /* level 16 */
+ { 23, 23, 22, 5, 4, 64, ZSTD_btopt }, /* level 17 */
+ { 23, 23, 22, 6, 3, 64, ZSTD_btultra }, /* level 18 */
+ { 23, 24, 22, 7, 3,256, ZSTD_btultra2}, /* level 19 */
+ { 25, 25, 23, 7, 3,256, ZSTD_btultra2}, /* level 20 */
+ { 26, 26, 24, 7, 3,512, ZSTD_btultra2}, /* level 21 */
+ { 27, 27, 25, 9, 3,999, ZSTD_btultra2}, /* level 22 */
+},
+{ /* for srcSize <= 256 KB */
+ /* W, C, H, S, L, T, strat */
+ { 18, 12, 13, 1, 5, 1, ZSTD_fast }, /* base for negative levels */
+ { 18, 13, 14, 1, 6, 0, ZSTD_fast }, /* level 1 */
+ { 18, 14, 14, 1, 5, 0, ZSTD_dfast }, /* level 2 */
+ { 18, 16, 16, 1, 4, 0, ZSTD_dfast }, /* level 3 */
+ { 18, 16, 17, 3, 5, 2, ZSTD_greedy }, /* level 4.*/
+ { 18, 17, 18, 5, 5, 2, ZSTD_greedy }, /* level 5.*/
+ { 18, 18, 19, 3, 5, 4, ZSTD_lazy }, /* level 6.*/
+ { 18, 18, 19, 4, 4, 4, ZSTD_lazy }, /* level 7 */
+ { 18, 18, 19, 4, 4, 8, ZSTD_lazy2 }, /* level 8 */
+ { 18, 18, 19, 5, 4, 8, ZSTD_lazy2 }, /* level 9 */
+ { 18, 18, 19, 6, 4, 8, ZSTD_lazy2 }, /* level 10 */
+ { 18, 18, 19, 5, 4, 12, ZSTD_btlazy2 }, /* level 11.*/
+ { 18, 19, 19, 7, 4, 12, ZSTD_btlazy2 }, /* level 12.*/
+ { 18, 18, 19, 4, 4, 16, ZSTD_btopt }, /* level 13 */
+ { 18, 18, 19, 4, 3, 32, ZSTD_btopt }, /* level 14.*/
+ { 18, 18, 19, 6, 3,128, ZSTD_btopt }, /* level 15.*/
+ { 18, 19, 19, 6, 3,128, ZSTD_btultra }, /* level 16.*/
+ { 18, 19, 19, 8, 3,256, ZSTD_btultra }, /* level 17.*/
+ { 18, 19, 19, 6, 3,128, ZSTD_btultra2}, /* level 18.*/
+ { 18, 19, 19, 8, 3,256, ZSTD_btultra2}, /* level 19.*/
+ { 18, 19, 19, 10, 3,512, ZSTD_btultra2}, /* level 20.*/
+ { 18, 19, 19, 12, 3,512, ZSTD_btultra2}, /* level 21.*/
+ { 18, 19, 19, 13, 3,999, ZSTD_btultra2}, /* level 22.*/
+},
+{ /* for srcSize <= 128 KB */
+ /* W, C, H, S, L, T, strat */
+ { 17, 12, 12, 1, 5, 1, ZSTD_fast }, /* base for negative levels */
+ { 17, 12, 13, 1, 6, 0, ZSTD_fast }, /* level 1 */
+ { 17, 13, 15, 1, 5, 0, ZSTD_fast }, /* level 2 */
+ { 17, 15, 16, 2, 5, 0, ZSTD_dfast }, /* level 3 */
+ { 17, 17, 17, 2, 4, 0, ZSTD_dfast }, /* level 4 */
+ { 17, 16, 17, 3, 4, 2, ZSTD_greedy }, /* level 5 */
+ { 17, 16, 17, 3, 4, 4, ZSTD_lazy }, /* level 6 */
+ { 17, 16, 17, 3, 4, 8, ZSTD_lazy2 }, /* level 7 */
+ { 17, 16, 17, 4, 4, 8, ZSTD_lazy2 }, /* level 8 */
+ { 17, 16, 17, 5, 4, 8, ZSTD_lazy2 }, /* level 9 */
+ { 17, 16, 17, 6, 4, 8, ZSTD_lazy2 }, /* level 10 */
+ { 17, 17, 17, 5, 4, 8, ZSTD_btlazy2 }, /* level 11 */
+ { 17, 18, 17, 7, 4, 12, ZSTD_btlazy2 }, /* level 12 */
+ { 17, 18, 17, 3, 4, 12, ZSTD_btopt }, /* level 13.*/
+ { 17, 18, 17, 4, 3, 32, ZSTD_btopt }, /* level 14.*/
+ { 17, 18, 17, 6, 3,256, ZSTD_btopt }, /* level 15.*/
+ { 17, 18, 17, 6, 3,128, ZSTD_btultra }, /* level 16.*/
+ { 17, 18, 17, 8, 3,256, ZSTD_btultra }, /* level 17.*/
+ { 17, 18, 17, 10, 3,512, ZSTD_btultra }, /* level 18.*/
+ { 17, 18, 17, 5, 3,256, ZSTD_btultra2}, /* level 19.*/
+ { 17, 18, 17, 7, 3,512, ZSTD_btultra2}, /* level 20.*/
+ { 17, 18, 17, 9, 3,512, ZSTD_btultra2}, /* level 21.*/
+ { 17, 18, 17, 11, 3,999, ZSTD_btultra2}, /* level 22.*/
+},
+{ /* for srcSize <= 16 KB */
+ /* W, C, H, S, L, T, strat */
+ { 14, 12, 13, 1, 5, 1, ZSTD_fast }, /* base for negative levels */
+ { 14, 14, 15, 1, 5, 0, ZSTD_fast }, /* level 1 */
+ { 14, 14, 15, 1, 4, 0, ZSTD_fast }, /* level 2 */
+ { 14, 14, 15, 2, 4, 0, ZSTD_dfast }, /* level 3 */
+ { 14, 14, 14, 4, 4, 2, ZSTD_greedy }, /* level 4 */
+ { 14, 14, 14, 3, 4, 4, ZSTD_lazy }, /* level 5.*/
+ { 14, 14, 14, 4, 4, 8, ZSTD_lazy2 }, /* level 6 */
+ { 14, 14, 14, 6, 4, 8, ZSTD_lazy2 }, /* level 7 */
+ { 14, 14, 14, 8, 4, 8, ZSTD_lazy2 }, /* level 8.*/
+ { 14, 15, 14, 5, 4, 8, ZSTD_btlazy2 }, /* level 9.*/
+ { 14, 15, 14, 9, 4, 8, ZSTD_btlazy2 }, /* level 10.*/
+ { 14, 15, 14, 3, 4, 12, ZSTD_btopt }, /* level 11.*/
+ { 14, 15, 14, 4, 3, 24, ZSTD_btopt }, /* level 12.*/
+ { 14, 15, 14, 5, 3, 32, ZSTD_btultra }, /* level 13.*/
+ { 14, 15, 15, 6, 3, 64, ZSTD_btultra }, /* level 14.*/
+ { 14, 15, 15, 7, 3,256, ZSTD_btultra }, /* level 15.*/
+ { 14, 15, 15, 5, 3, 48, ZSTD_btultra2}, /* level 16.*/
+ { 14, 15, 15, 6, 3,128, ZSTD_btultra2}, /* level 17.*/
+ { 14, 15, 15, 7, 3,256, ZSTD_btultra2}, /* level 18.*/
+ { 14, 15, 15, 8, 3,256, ZSTD_btultra2}, /* level 19.*/
+ { 14, 15, 15, 8, 3,512, ZSTD_btultra2}, /* level 20.*/
+ { 14, 15, 15, 9, 3,512, ZSTD_btultra2}, /* level 21.*/
+ { 14, 15, 15, 10, 3,999, ZSTD_btultra2}, /* level 22.*/
+},
+};
+
+
+
+#endif /* ZSTD_CLEVELS_H */
diff --git a/lib/zstd/compress/fse_compress.c b/lib/zstd/compress/fse_compress.c
new file mode 100644
index 0000000000..ec5b1ca6d7
--- /dev/null
+++ b/lib/zstd/compress/fse_compress.c
@@ -0,0 +1,668 @@
+/* ******************************************************************
+ * FSE : Finite State Entropy encoder
+ * Copyright (c) Yann Collet, Facebook, Inc.
+ *
+ * You can contact the author at :
+ * - FSE source repository : https://github.com/Cyan4973/FiniteStateEntropy
+ * - Public forum : https://groups.google.com/forum/#!forum/lz4c
+ *
+ * This source code is licensed under both the BSD-style license (found in the
+ * LICENSE file in the root directory of this source tree) and the GPLv2 (found
+ * in the COPYING file in the root directory of this source tree).
+ * You may select, at your option, one of the above-listed licenses.
+****************************************************************** */
+
+/* **************************************************************
+* Includes
+****************************************************************/
+#include "../common/compiler.h"
+#include "../common/mem.h" /* U32, U16, etc. */
+#include "../common/debug.h" /* assert, DEBUGLOG */
+#include "hist.h" /* HIST_count_wksp */
+#include "../common/bitstream.h"
+#define FSE_STATIC_LINKING_ONLY
+#include "../common/fse.h"
+#include "../common/error_private.h"
+#define ZSTD_DEPS_NEED_MALLOC
+#define ZSTD_DEPS_NEED_MATH64
+#include "../common/zstd_deps.h" /* ZSTD_malloc, ZSTD_free, ZSTD_memcpy, ZSTD_memset */
+
+
+/* **************************************************************
+* Error Management
+****************************************************************/
+#define FSE_isError ERR_isError
+
+
+/* **************************************************************
+* Templates
+****************************************************************/
+/*
+ designed to be included
+ for type-specific functions (template emulation in C)
+ Objective is to write these functions only once, for improved maintenance
+*/
+
+/* safety checks */
+#ifndef FSE_FUNCTION_EXTENSION
+# error "FSE_FUNCTION_EXTENSION must be defined"
+#endif
+#ifndef FSE_FUNCTION_TYPE
+# error "FSE_FUNCTION_TYPE must be defined"
+#endif
+
+/* Function names */
+#define FSE_CAT(X,Y) X##Y
+#define FSE_FUNCTION_NAME(X,Y) FSE_CAT(X,Y)
+#define FSE_TYPE_NAME(X,Y) FSE_CAT(X,Y)
+
+
+/* Function templates */
+
+/* FSE_buildCTable_wksp() :
+ * Same as FSE_buildCTable(), but using an externally allocated scratch buffer (`workSpace`).
+ * wkspSize should be sized to handle worst case situation, which is `1<<max_tableLog * sizeof(FSE_FUNCTION_TYPE)`
+ * workSpace must also be properly aligned with FSE_FUNCTION_TYPE requirements
+ */
+size_t FSE_buildCTable_wksp(FSE_CTable* ct,
+ const short* normalizedCounter, unsigned maxSymbolValue, unsigned tableLog,
+ void* workSpace, size_t wkspSize)
+{
+ U32 const tableSize = 1 << tableLog;
+ U32 const tableMask = tableSize - 1;
+ void* const ptr = ct;
+ U16* const tableU16 = ( (U16*) ptr) + 2;
+ void* const FSCT = ((U32*)ptr) + 1 /* header */ + (tableLog ? tableSize>>1 : 1) ;
+ FSE_symbolCompressionTransform* const symbolTT = (FSE_symbolCompressionTransform*) (FSCT);
+ U32 const step = FSE_TABLESTEP(tableSize);
+ U32 const maxSV1 = maxSymbolValue+1;
+
+ U16* cumul = (U16*)workSpace; /* size = maxSV1 */
+ FSE_FUNCTION_TYPE* const tableSymbol = (FSE_FUNCTION_TYPE*)(cumul + (maxSV1+1)); /* size = tableSize */
+
+ U32 highThreshold = tableSize-1;
+
+ assert(((size_t)workSpace & 1) == 0); /* Must be 2 bytes-aligned */
+ if (FSE_BUILD_CTABLE_WORKSPACE_SIZE(maxSymbolValue, tableLog) > wkspSize) return ERROR(tableLog_tooLarge);
+ /* CTable header */
+ tableU16[-2] = (U16) tableLog;
+ tableU16[-1] = (U16) maxSymbolValue;
+ assert(tableLog < 16); /* required for threshold strategy to work */
+
+ /* For explanations on how to distribute symbol values over the table :
+ * http://fastcompression.blogspot.fr/2014/02/fse-distributing-symbol-values.html */
+
+ #ifdef __clang_analyzer__
+ ZSTD_memset(tableSymbol, 0, sizeof(*tableSymbol) * tableSize); /* useless initialization, just to keep scan-build happy */
+ #endif
+
+ /* symbol start positions */
+ { U32 u;
+ cumul[0] = 0;
+ for (u=1; u <= maxSV1; u++) {
+ if (normalizedCounter[u-1]==-1) { /* Low proba symbol */
+ cumul[u] = cumul[u-1] + 1;
+ tableSymbol[highThreshold--] = (FSE_FUNCTION_TYPE)(u-1);
+ } else {
+ assert(normalizedCounter[u-1] >= 0);
+ cumul[u] = cumul[u-1] + (U16)normalizedCounter[u-1];
+ assert(cumul[u] >= cumul[u-1]); /* no overflow */
+ } }
+ cumul[maxSV1] = (U16)(tableSize+1);
+ }
+
+ /* Spread symbols */
+ if (highThreshold == tableSize - 1) {
+ /* Case for no low prob count symbols. Lay down 8 bytes at a time
+ * to reduce branch misses since we are operating on a small block
+ */
+ BYTE* const spread = tableSymbol + tableSize; /* size = tableSize + 8 (may write beyond tableSize) */
+ { U64 const add = 0x0101010101010101ull;
+ size_t pos = 0;
+ U64 sv = 0;
+ U32 s;
+ for (s=0; s<maxSV1; ++s, sv += add) {
+ int i;
+ int const n = normalizedCounter[s];
+ MEM_write64(spread + pos, sv);
+ for (i = 8; i < n; i += 8) {
+ MEM_write64(spread + pos + i, sv);
+ }
+ assert(n>=0);
+ pos += (size_t)n;
+ }
+ }
+ /* Spread symbols across the table. Lack of lowprob symbols means that
+ * we don't need variable sized inner loop, so we can unroll the loop and
+ * reduce branch misses.
+ */
+ { size_t position = 0;
+ size_t s;
+ size_t const unroll = 2; /* Experimentally determined optimal unroll */
+ assert(tableSize % unroll == 0); /* FSE_MIN_TABLELOG is 5 */
+ for (s = 0; s < (size_t)tableSize; s += unroll) {
+ size_t u;
+ for (u = 0; u < unroll; ++u) {
+ size_t const uPosition = (position + (u * step)) & tableMask;
+ tableSymbol[uPosition] = spread[s + u];
+ }
+ position = (position + (unroll * step)) & tableMask;
+ }
+ assert(position == 0); /* Must have initialized all positions */
+ }
+ } else {
+ U32 position = 0;
+ U32 symbol;
+ for (symbol=0; symbol<maxSV1; symbol++) {
+ int nbOccurrences;
+ int const freq = normalizedCounter[symbol];
+ for (nbOccurrences=0; nbOccurrences<freq; nbOccurrences++) {
+ tableSymbol[position] = (FSE_FUNCTION_TYPE)symbol;
+ position = (position + step) & tableMask;
+ while (position > highThreshold)
+ position = (position + step) & tableMask; /* Low proba area */
+ } }
+ assert(position==0); /* Must have initialized all positions */
+ }
+
+ /* Build table */
+ { U32 u; for (u=0; u<tableSize; u++) {
+ FSE_FUNCTION_TYPE s = tableSymbol[u]; /* note : static analyzer may not understand tableSymbol is properly initialized */
+ tableU16[cumul[s]++] = (U16) (tableSize+u); /* TableU16 : sorted by symbol order; gives next state value */
+ } }
+
+ /* Build Symbol Transformation Table */
+ { unsigned total = 0;
+ unsigned s;
+ for (s=0; s<=maxSymbolValue; s++) {
+ switch (normalizedCounter[s])
+ {
+ case 0:
+ /* filling nonetheless, for compatibility with FSE_getMaxNbBits() */
+ symbolTT[s].deltaNbBits = ((tableLog+1) << 16) - (1<<tableLog);
+ break;
+
+ case -1:
+ case 1:
+ symbolTT[s].deltaNbBits = (tableLog << 16) - (1<<tableLog);
+ assert(total <= INT_MAX);
+ symbolTT[s].deltaFindState = (int)(total - 1);
+ total ++;
+ break;
+ default :
+ assert(normalizedCounter[s] > 1);
+ { U32 const maxBitsOut = tableLog - BIT_highbit32 ((U32)normalizedCounter[s]-1);
+ U32 const minStatePlus = (U32)normalizedCounter[s] << maxBitsOut;
+ symbolTT[s].deltaNbBits = (maxBitsOut << 16) - minStatePlus;
+ symbolTT[s].deltaFindState = (int)(total - (unsigned)normalizedCounter[s]);
+ total += (unsigned)normalizedCounter[s];
+ } } } }
+
+#if 0 /* debug : symbol costs */
+ DEBUGLOG(5, "\n --- table statistics : ");
+ { U32 symbol;
+ for (symbol=0; symbol<=maxSymbolValue; symbol++) {
+ DEBUGLOG(5, "%3u: w=%3i, maxBits=%u, fracBits=%.2f",
+ symbol, normalizedCounter[symbol],
+ FSE_getMaxNbBits(symbolTT, symbol),
+ (double)FSE_bitCost(symbolTT, tableLog, symbol, 8) / 256);
+ } }
+#endif
+
+ return 0;
+}
+
+
+
+#ifndef FSE_COMMONDEFS_ONLY
+
+/*-**************************************************************
+* FSE NCount encoding
+****************************************************************/
+size_t FSE_NCountWriteBound(unsigned maxSymbolValue, unsigned tableLog)
+{
+ size_t const maxHeaderSize = (((maxSymbolValue+1) * tableLog
+ + 4 /* bitCount initialized at 4 */
+ + 2 /* first two symbols may use one additional bit each */) / 8)
+ + 1 /* round up to whole nb bytes */
+ + 2 /* additional two bytes for bitstream flush */;
+ return maxSymbolValue ? maxHeaderSize : FSE_NCOUNTBOUND; /* maxSymbolValue==0 ? use default */
+}
+
+static size_t
+FSE_writeNCount_generic (void* header, size_t headerBufferSize,
+ const short* normalizedCounter, unsigned maxSymbolValue, unsigned tableLog,
+ unsigned writeIsSafe)
+{
+ BYTE* const ostart = (BYTE*) header;
+ BYTE* out = ostart;
+ BYTE* const oend = ostart + headerBufferSize;
+ int nbBits;
+ const int tableSize = 1 << tableLog;
+ int remaining;
+ int threshold;
+ U32 bitStream = 0;
+ int bitCount = 0;
+ unsigned symbol = 0;
+ unsigned const alphabetSize = maxSymbolValue + 1;
+ int previousIs0 = 0;
+
+ /* Table Size */
+ bitStream += (tableLog-FSE_MIN_TABLELOG) << bitCount;
+ bitCount += 4;
+
+ /* Init */
+ remaining = tableSize+1; /* +1 for extra accuracy */
+ threshold = tableSize;
+ nbBits = tableLog+1;
+
+ while ((symbol < alphabetSize) && (remaining>1)) { /* stops at 1 */
+ if (previousIs0) {
+ unsigned start = symbol;
+ while ((symbol < alphabetSize) && !normalizedCounter[symbol]) symbol++;
+ if (symbol == alphabetSize) break; /* incorrect distribution */
+ while (symbol >= start+24) {
+ start+=24;
+ bitStream += 0xFFFFU << bitCount;
+ if ((!writeIsSafe) && (out > oend-2))
+ return ERROR(dstSize_tooSmall); /* Buffer overflow */
+ out[0] = (BYTE) bitStream;
+ out[1] = (BYTE)(bitStream>>8);
+ out+=2;
+ bitStream>>=16;
+ }
+ while (symbol >= start+3) {
+ start+=3;
+ bitStream += 3 << bitCount;
+ bitCount += 2;
+ }
+ bitStream += (symbol-start) << bitCount;
+ bitCount += 2;
+ if (bitCount>16) {
+ if ((!writeIsSafe) && (out > oend - 2))
+ return ERROR(dstSize_tooSmall); /* Buffer overflow */
+ out[0] = (BYTE)bitStream;
+ out[1] = (BYTE)(bitStream>>8);
+ out += 2;
+ bitStream >>= 16;
+ bitCount -= 16;
+ } }
+ { int count = normalizedCounter[symbol++];
+ int const max = (2*threshold-1) - remaining;
+ remaining -= count < 0 ? -count : count;
+ count++; /* +1 for extra accuracy */
+ if (count>=threshold)
+ count += max; /* [0..max[ [max..threshold[ (...) [threshold+max 2*threshold[ */
+ bitStream += count << bitCount;
+ bitCount += nbBits;
+ bitCount -= (count<max);
+ previousIs0 = (count==1);
+ if (remaining<1) return ERROR(GENERIC);
+ while (remaining<threshold) { nbBits--; threshold>>=1; }
+ }
+ if (bitCount>16) {
+ if ((!writeIsSafe) && (out > oend - 2))
+ return ERROR(dstSize_tooSmall); /* Buffer overflow */
+ out[0] = (BYTE)bitStream;
+ out[1] = (BYTE)(bitStream>>8);
+ out += 2;
+ bitStream >>= 16;
+ bitCount -= 16;
+ } }
+
+ if (remaining != 1)
+ return ERROR(GENERIC); /* incorrect normalized distribution */
+ assert(symbol <= alphabetSize);
+
+ /* flush remaining bitStream */
+ if ((!writeIsSafe) && (out > oend - 2))
+ return ERROR(dstSize_tooSmall); /* Buffer overflow */
+ out[0] = (BYTE)bitStream;
+ out[1] = (BYTE)(bitStream>>8);
+ out+= (bitCount+7) /8;
+
+ return (out-ostart);
+}
+
+
+size_t FSE_writeNCount (void* buffer, size_t bufferSize,
+ const short* normalizedCounter, unsigned maxSymbolValue, unsigned tableLog)
+{
+ if (tableLog > FSE_MAX_TABLELOG) return ERROR(tableLog_tooLarge); /* Unsupported */
+ if (tableLog < FSE_MIN_TABLELOG) return ERROR(GENERIC); /* Unsupported */
+
+ if (bufferSize < FSE_NCountWriteBound(maxSymbolValue, tableLog))
+ return FSE_writeNCount_generic(buffer, bufferSize, normalizedCounter, maxSymbolValue, tableLog, 0);
+
+ return FSE_writeNCount_generic(buffer, bufferSize, normalizedCounter, maxSymbolValue, tableLog, 1 /* write in buffer is safe */);
+}
+
+
+/*-**************************************************************
+* FSE Compression Code
+****************************************************************/
+
+FSE_CTable* FSE_createCTable (unsigned maxSymbolValue, unsigned tableLog)
+{
+ size_t size;
+ if (tableLog > FSE_TABLELOG_ABSOLUTE_MAX) tableLog = FSE_TABLELOG_ABSOLUTE_MAX;
+ size = FSE_CTABLE_SIZE_U32 (tableLog, maxSymbolValue) * sizeof(U32);
+ return (FSE_CTable*)ZSTD_malloc(size);
+}
+
+void FSE_freeCTable (FSE_CTable* ct) { ZSTD_free(ct); }
+
+/* provides the minimum logSize to safely represent a distribution */
+static unsigned FSE_minTableLog(size_t srcSize, unsigned maxSymbolValue)
+{
+ U32 minBitsSrc = BIT_highbit32((U32)(srcSize)) + 1;
+ U32 minBitsSymbols = BIT_highbit32(maxSymbolValue) + 2;
+ U32 minBits = minBitsSrc < minBitsSymbols ? minBitsSrc : minBitsSymbols;
+ assert(srcSize > 1); /* Not supported, RLE should be used instead */
+ return minBits;
+}
+
+unsigned FSE_optimalTableLog_internal(unsigned maxTableLog, size_t srcSize, unsigned maxSymbolValue, unsigned minus)
+{
+ U32 maxBitsSrc = BIT_highbit32((U32)(srcSize - 1)) - minus;
+ U32 tableLog = maxTableLog;
+ U32 minBits = FSE_minTableLog(srcSize, maxSymbolValue);
+ assert(srcSize > 1); /* Not supported, RLE should be used instead */
+ if (tableLog==0) tableLog = FSE_DEFAULT_TABLELOG;
+ if (maxBitsSrc < tableLog) tableLog = maxBitsSrc; /* Accuracy can be reduced */
+ if (minBits > tableLog) tableLog = minBits; /* Need a minimum to safely represent all symbol values */
+ if (tableLog < FSE_MIN_TABLELOG) tableLog = FSE_MIN_TABLELOG;
+ if (tableLog > FSE_MAX_TABLELOG) tableLog = FSE_MAX_TABLELOG;
+ return tableLog;
+}
+
+unsigned FSE_optimalTableLog(unsigned maxTableLog, size_t srcSize, unsigned maxSymbolValue)
+{
+ return FSE_optimalTableLog_internal(maxTableLog, srcSize, maxSymbolValue, 2);
+}
+
+/* Secondary normalization method.
+ To be used when primary method fails. */
+
+static size_t FSE_normalizeM2(short* norm, U32 tableLog, const unsigned* count, size_t total, U32 maxSymbolValue, short lowProbCount)
+{
+ short const NOT_YET_ASSIGNED = -2;
+ U32 s;
+ U32 distributed = 0;
+ U32 ToDistribute;
+
+ /* Init */
+ U32 const lowThreshold = (U32)(total >> tableLog);
+ U32 lowOne = (U32)((total * 3) >> (tableLog + 1));
+
+ for (s=0; s<=maxSymbolValue; s++) {
+ if (count[s] == 0) {
+ norm[s]=0;
+ continue;
+ }
+ if (count[s] <= lowThreshold) {
+ norm[s] = lowProbCount;
+ distributed++;
+ total -= count[s];
+ continue;
+ }
+ if (count[s] <= lowOne) {
+ norm[s] = 1;
+ distributed++;
+ total -= count[s];
+ continue;
+ }
+
+ norm[s]=NOT_YET_ASSIGNED;
+ }
+ ToDistribute = (1 << tableLog) - distributed;
+
+ if (ToDistribute == 0)
+ return 0;
+
+ if ((total / ToDistribute) > lowOne) {
+ /* risk of rounding to zero */
+ lowOne = (U32)((total * 3) / (ToDistribute * 2));
+ for (s=0; s<=maxSymbolValue; s++) {
+ if ((norm[s] == NOT_YET_ASSIGNED) && (count[s] <= lowOne)) {
+ norm[s] = 1;
+ distributed++;
+ total -= count[s];
+ continue;
+ } }
+ ToDistribute = (1 << tableLog) - distributed;
+ }
+
+ if (distributed == maxSymbolValue+1) {
+ /* all values are pretty poor;
+ probably incompressible data (should have already been detected);
+ find max, then give all remaining points to max */
+ U32 maxV = 0, maxC = 0;
+ for (s=0; s<=maxSymbolValue; s++)
+ if (count[s] > maxC) { maxV=s; maxC=count[s]; }
+ norm[maxV] += (short)ToDistribute;
+ return 0;
+ }
+
+ if (total == 0) {
+ /* all of the symbols were low enough for the lowOne or lowThreshold */
+ for (s=0; ToDistribute > 0; s = (s+1)%(maxSymbolValue+1))
+ if (norm[s] > 0) { ToDistribute--; norm[s]++; }
+ return 0;
+ }
+
+ { U64 const vStepLog = 62 - tableLog;
+ U64 const mid = (1ULL << (vStepLog-1)) - 1;
+ U64 const rStep = ZSTD_div64((((U64)1<<vStepLog) * ToDistribute) + mid, (U32)total); /* scale on remaining */
+ U64 tmpTotal = mid;
+ for (s=0; s<=maxSymbolValue; s++) {
+ if (norm[s]==NOT_YET_ASSIGNED) {
+ U64 const end = tmpTotal + (count[s] * rStep);
+ U32 const sStart = (U32)(tmpTotal >> vStepLog);
+ U32 const sEnd = (U32)(end >> vStepLog);
+ U32 const weight = sEnd - sStart;
+ if (weight < 1)
+ return ERROR(GENERIC);
+ norm[s] = (short)weight;
+ tmpTotal = end;
+ } } }
+
+ return 0;
+}
+
+size_t FSE_normalizeCount (short* normalizedCounter, unsigned tableLog,
+ const unsigned* count, size_t total,
+ unsigned maxSymbolValue, unsigned useLowProbCount)
+{
+ /* Sanity checks */
+ if (tableLog==0) tableLog = FSE_DEFAULT_TABLELOG;
+ if (tableLog < FSE_MIN_TABLELOG) return ERROR(GENERIC); /* Unsupported size */
+ if (tableLog > FSE_MAX_TABLELOG) return ERROR(tableLog_tooLarge); /* Unsupported size */
+ if (tableLog < FSE_minTableLog(total, maxSymbolValue)) return ERROR(GENERIC); /* Too small tableLog, compression potentially impossible */
+
+ { static U32 const rtbTable[] = { 0, 473195, 504333, 520860, 550000, 700000, 750000, 830000 };
+ short const lowProbCount = useLowProbCount ? -1 : 1;
+ U64 const scale = 62 - tableLog;
+ U64 const step = ZSTD_div64((U64)1<<62, (U32)total); /* <== here, one division ! */
+ U64 const vStep = 1ULL<<(scale-20);
+ int stillToDistribute = 1<<tableLog;
+ unsigned s;
+ unsigned largest=0;
+ short largestP=0;
+ U32 lowThreshold = (U32)(total >> tableLog);
+
+ for (s=0; s<=maxSymbolValue; s++) {
+ if (count[s] == total) return 0; /* rle special case */
+ if (count[s] == 0) { normalizedCounter[s]=0; continue; }
+ if (count[s] <= lowThreshold) {
+ normalizedCounter[s] = lowProbCount;
+ stillToDistribute--;
+ } else {
+ short proba = (short)((count[s]*step) >> scale);
+ if (proba<8) {
+ U64 restToBeat = vStep * rtbTable[proba];
+ proba += (count[s]*step) - ((U64)proba<<scale) > restToBeat;
+ }
+ if (proba > largestP) { largestP=proba; largest=s; }
+ normalizedCounter[s] = proba;
+ stillToDistribute -= proba;
+ } }
+ if (-stillToDistribute >= (normalizedCounter[largest] >> 1)) {
+ /* corner case, need another normalization method */
+ size_t const errorCode = FSE_normalizeM2(normalizedCounter, tableLog, count, total, maxSymbolValue, lowProbCount);
+ if (FSE_isError(errorCode)) return errorCode;
+ }
+ else normalizedCounter[largest] += (short)stillToDistribute;
+ }
+
+#if 0
+ { /* Print Table (debug) */
+ U32 s;
+ U32 nTotal = 0;
+ for (s=0; s<=maxSymbolValue; s++)
+ RAWLOG(2, "%3i: %4i \n", s, normalizedCounter[s]);
+ for (s=0; s<=maxSymbolValue; s++)
+ nTotal += abs(normalizedCounter[s]);
+ if (nTotal != (1U<<tableLog))
+ RAWLOG(2, "Warning !!! Total == %u != %u !!!", nTotal, 1U<<tableLog);
+ getchar();
+ }
+#endif
+
+ return tableLog;
+}
+
+
+/* fake FSE_CTable, for raw (uncompressed) input */
+size_t FSE_buildCTable_raw (FSE_CTable* ct, unsigned nbBits)
+{
+ const unsigned tableSize = 1 << nbBits;
+ const unsigned tableMask = tableSize - 1;
+ const unsigned maxSymbolValue = tableMask;
+ void* const ptr = ct;
+ U16* const tableU16 = ( (U16*) ptr) + 2;
+ void* const FSCT = ((U32*)ptr) + 1 /* header */ + (tableSize>>1); /* assumption : tableLog >= 1 */
+ FSE_symbolCompressionTransform* const symbolTT = (FSE_symbolCompressionTransform*) (FSCT);
+ unsigned s;
+
+ /* Sanity checks */
+ if (nbBits < 1) return ERROR(GENERIC); /* min size */
+
+ /* header */
+ tableU16[-2] = (U16) nbBits;
+ tableU16[-1] = (U16) maxSymbolValue;
+
+ /* Build table */
+ for (s=0; s<tableSize; s++)
+ tableU16[s] = (U16)(tableSize + s);
+
+ /* Build Symbol Transformation Table */
+ { const U32 deltaNbBits = (nbBits << 16) - (1 << nbBits);
+ for (s=0; s<=maxSymbolValue; s++) {
+ symbolTT[s].deltaNbBits = deltaNbBits;
+ symbolTT[s].deltaFindState = s-1;
+ } }
+
+ return 0;
+}
+
+/* fake FSE_CTable, for rle input (always same symbol) */
+size_t FSE_buildCTable_rle (FSE_CTable* ct, BYTE symbolValue)
+{
+ void* ptr = ct;
+ U16* tableU16 = ( (U16*) ptr) + 2;
+ void* FSCTptr = (U32*)ptr + 2;
+ FSE_symbolCompressionTransform* symbolTT = (FSE_symbolCompressionTransform*) FSCTptr;
+
+ /* header */
+ tableU16[-2] = (U16) 0;
+ tableU16[-1] = (U16) symbolValue;
+
+ /* Build table */
+ tableU16[0] = 0;
+ tableU16[1] = 0; /* just in case */
+
+ /* Build Symbol Transformation Table */
+ symbolTT[symbolValue].deltaNbBits = 0;
+ symbolTT[symbolValue].deltaFindState = 0;
+
+ return 0;
+}
+
+
+static size_t FSE_compress_usingCTable_generic (void* dst, size_t dstSize,
+ const void* src, size_t srcSize,
+ const FSE_CTable* ct, const unsigned fast)
+{
+ const BYTE* const istart = (const BYTE*) src;
+ const BYTE* const iend = istart + srcSize;
+ const BYTE* ip=iend;
+
+ BIT_CStream_t bitC;
+ FSE_CState_t CState1, CState2;
+
+ /* init */
+ if (srcSize <= 2) return 0;
+ { size_t const initError = BIT_initCStream(&bitC, dst, dstSize);
+ if (FSE_isError(initError)) return 0; /* not enough space available to write a bitstream */ }
+
+#define FSE_FLUSHBITS(s) (fast ? BIT_flushBitsFast(s) : BIT_flushBits(s))
+
+ if (srcSize & 1) {
+ FSE_initCState2(&CState1, ct, *--ip);
+ FSE_initCState2(&CState2, ct, *--ip);
+ FSE_encodeSymbol(&bitC, &CState1, *--ip);
+ FSE_FLUSHBITS(&bitC);
+ } else {
+ FSE_initCState2(&CState2, ct, *--ip);
+ FSE_initCState2(&CState1, ct, *--ip);
+ }
+
+ /* join to mod 4 */
+ srcSize -= 2;
+ if ((sizeof(bitC.bitContainer)*8 > FSE_MAX_TABLELOG*4+7 ) && (srcSize & 2)) { /* test bit 2 */
+ FSE_encodeSymbol(&bitC, &CState2, *--ip);
+ FSE_encodeSymbol(&bitC, &CState1, *--ip);
+ FSE_FLUSHBITS(&bitC);
+ }
+
+ /* 2 or 4 encoding per loop */
+ while ( ip>istart ) {
+
+ FSE_encodeSymbol(&bitC, &CState2, *--ip);
+
+ if (sizeof(bitC.bitContainer)*8 < FSE_MAX_TABLELOG*2+7 ) /* this test must be static */
+ FSE_FLUSHBITS(&bitC);
+
+ FSE_encodeSymbol(&bitC, &CState1, *--ip);
+
+ if (sizeof(bitC.bitContainer)*8 > FSE_MAX_TABLELOG*4+7 ) { /* this test must be static */
+ FSE_encodeSymbol(&bitC, &CState2, *--ip);
+ FSE_encodeSymbol(&bitC, &CState1, *--ip);
+ }
+
+ FSE_FLUSHBITS(&bitC);
+ }
+
+ FSE_flushCState(&bitC, &CState2);
+ FSE_flushCState(&bitC, &CState1);
+ return BIT_closeCStream(&bitC);
+}
+
+size_t FSE_compress_usingCTable (void* dst, size_t dstSize,
+ const void* src, size_t srcSize,
+ const FSE_CTable* ct)
+{
+ unsigned const fast = (dstSize >= FSE_BLOCKBOUND(srcSize));
+
+ if (fast)
+ return FSE_compress_usingCTable_generic(dst, dstSize, src, srcSize, ct, 1);
+ else
+ return FSE_compress_usingCTable_generic(dst, dstSize, src, srcSize, ct, 0);
+}
+
+
+size_t FSE_compressBound(size_t size) { return FSE_COMPRESSBOUND(size); }
+
+
+#endif /* FSE_COMMONDEFS_ONLY */
diff --git a/lib/zstd/compress/hist.c b/lib/zstd/compress/hist.c
new file mode 100644
index 0000000000..3ddc6dfb68
--- /dev/null
+++ b/lib/zstd/compress/hist.c
@@ -0,0 +1,165 @@
+/* ******************************************************************
+ * hist : Histogram functions
+ * part of Finite State Entropy project
+ * Copyright (c) Yann Collet, Facebook, Inc.
+ *
+ * You can contact the author at :
+ * - FSE source repository : https://github.com/Cyan4973/FiniteStateEntropy
+ * - Public forum : https://groups.google.com/forum/#!forum/lz4c
+ *
+ * This source code is licensed under both the BSD-style license (found in the
+ * LICENSE file in the root directory of this source tree) and the GPLv2 (found
+ * in the COPYING file in the root directory of this source tree).
+ * You may select, at your option, one of the above-listed licenses.
+****************************************************************** */
+
+/* --- dependencies --- */
+#include "../common/mem.h" /* U32, BYTE, etc. */
+#include "../common/debug.h" /* assert, DEBUGLOG */
+#include "../common/error_private.h" /* ERROR */
+#include "hist.h"
+
+
+/* --- Error management --- */
+unsigned HIST_isError(size_t code) { return ERR_isError(code); }
+
+/*-**************************************************************
+ * Histogram functions
+ ****************************************************************/
+unsigned HIST_count_simple(unsigned* count, unsigned* maxSymbolValuePtr,
+ const void* src, size_t srcSize)
+{
+ const BYTE* ip = (const BYTE*)src;
+ const BYTE* const end = ip + srcSize;
+ unsigned maxSymbolValue = *maxSymbolValuePtr;
+ unsigned largestCount=0;
+
+ ZSTD_memset(count, 0, (maxSymbolValue+1) * sizeof(*count));
+ if (srcSize==0) { *maxSymbolValuePtr = 0; return 0; }
+
+ while (ip<end) {
+ assert(*ip <= maxSymbolValue);
+ count[*ip++]++;
+ }
+
+ while (!count[maxSymbolValue]) maxSymbolValue--;
+ *maxSymbolValuePtr = maxSymbolValue;
+
+ { U32 s;
+ for (s=0; s<=maxSymbolValue; s++)
+ if (count[s] > largestCount) largestCount = count[s];
+ }
+
+ return largestCount;
+}
+
+typedef enum { trustInput, checkMaxSymbolValue } HIST_checkInput_e;
+
+/* HIST_count_parallel_wksp() :
+ * store histogram into 4 intermediate tables, recombined at the end.
+ * this design makes better use of OoO cpus,
+ * and is noticeably faster when some values are heavily repeated.
+ * But it needs some additional workspace for intermediate tables.
+ * `workSpace` must be a U32 table of size >= HIST_WKSP_SIZE_U32.
+ * @return : largest histogram frequency,
+ * or an error code (notably when histogram's alphabet is larger than *maxSymbolValuePtr) */
+static size_t HIST_count_parallel_wksp(
+ unsigned* count, unsigned* maxSymbolValuePtr,
+ const void* source, size_t sourceSize,
+ HIST_checkInput_e check,
+ U32* const workSpace)
+{
+ const BYTE* ip = (const BYTE*)source;
+ const BYTE* const iend = ip+sourceSize;
+ size_t const countSize = (*maxSymbolValuePtr + 1) * sizeof(*count);
+ unsigned max=0;
+ U32* const Counting1 = workSpace;
+ U32* const Counting2 = Counting1 + 256;
+ U32* const Counting3 = Counting2 + 256;
+ U32* const Counting4 = Counting3 + 256;
+
+ /* safety checks */
+ assert(*maxSymbolValuePtr <= 255);
+ if (!sourceSize) {
+ ZSTD_memset(count, 0, countSize);
+ *maxSymbolValuePtr = 0;
+ return 0;
+ }
+ ZSTD_memset(workSpace, 0, 4*256*sizeof(unsigned));
+
+ /* by stripes of 16 bytes */
+ { U32 cached = MEM_read32(ip); ip += 4;
+ while (ip < iend-15) {
+ U32 c = cached; cached = MEM_read32(ip); ip += 4;
+ Counting1[(BYTE) c ]++;
+ Counting2[(BYTE)(c>>8) ]++;
+ Counting3[(BYTE)(c>>16)]++;
+ Counting4[ c>>24 ]++;
+ c = cached; cached = MEM_read32(ip); ip += 4;
+ Counting1[(BYTE) c ]++;
+ Counting2[(BYTE)(c>>8) ]++;
+ Counting3[(BYTE)(c>>16)]++;
+ Counting4[ c>>24 ]++;
+ c = cached; cached = MEM_read32(ip); ip += 4;
+ Counting1[(BYTE) c ]++;
+ Counting2[(BYTE)(c>>8) ]++;
+ Counting3[(BYTE)(c>>16)]++;
+ Counting4[ c>>24 ]++;
+ c = cached; cached = MEM_read32(ip); ip += 4;
+ Counting1[(BYTE) c ]++;
+ Counting2[(BYTE)(c>>8) ]++;
+ Counting3[(BYTE)(c>>16)]++;
+ Counting4[ c>>24 ]++;
+ }
+ ip-=4;
+ }
+
+ /* finish last symbols */
+ while (ip<iend) Counting1[*ip++]++;
+
+ { U32 s;
+ for (s=0; s<256; s++) {
+ Counting1[s] += Counting2[s] + Counting3[s] + Counting4[s];
+ if (Counting1[s] > max) max = Counting1[s];
+ } }
+
+ { unsigned maxSymbolValue = 255;
+ while (!Counting1[maxSymbolValue]) maxSymbolValue--;
+ if (check && maxSymbolValue > *maxSymbolValuePtr) return ERROR(maxSymbolValue_tooSmall);
+ *maxSymbolValuePtr = maxSymbolValue;
+ ZSTD_memmove(count, Counting1, countSize); /* in case count & Counting1 are overlapping */
+ }
+ return (size_t)max;
+}
+
+/* HIST_countFast_wksp() :
+ * Same as HIST_countFast(), but using an externally provided scratch buffer.
+ * `workSpace` is a writable buffer which must be 4-bytes aligned,
+ * `workSpaceSize` must be >= HIST_WKSP_SIZE
+ */
+size_t HIST_countFast_wksp(unsigned* count, unsigned* maxSymbolValuePtr,
+ const void* source, size_t sourceSize,
+ void* workSpace, size_t workSpaceSize)
+{
+ if (sourceSize < 1500) /* heuristic threshold */
+ return HIST_count_simple(count, maxSymbolValuePtr, source, sourceSize);
+ if ((size_t)workSpace & 3) return ERROR(GENERIC); /* must be aligned on 4-bytes boundaries */
+ if (workSpaceSize < HIST_WKSP_SIZE) return ERROR(workSpace_tooSmall);
+ return HIST_count_parallel_wksp(count, maxSymbolValuePtr, source, sourceSize, trustInput, (U32*)workSpace);
+}
+
+/* HIST_count_wksp() :
+ * Same as HIST_count(), but using an externally provided scratch buffer.
+ * `workSpace` size must be table of >= HIST_WKSP_SIZE_U32 unsigned */
+size_t HIST_count_wksp(unsigned* count, unsigned* maxSymbolValuePtr,
+ const void* source, size_t sourceSize,
+ void* workSpace, size_t workSpaceSize)
+{
+ if ((size_t)workSpace & 3) return ERROR(GENERIC); /* must be aligned on 4-bytes boundaries */
+ if (workSpaceSize < HIST_WKSP_SIZE) return ERROR(workSpace_tooSmall);
+ if (*maxSymbolValuePtr < 255)
+ return HIST_count_parallel_wksp(count, maxSymbolValuePtr, source, sourceSize, checkMaxSymbolValue, (U32*)workSpace);
+ *maxSymbolValuePtr = 255;
+ return HIST_countFast_wksp(count, maxSymbolValuePtr, source, sourceSize, workSpace, workSpaceSize);
+}
+
diff --git a/lib/zstd/compress/hist.h b/lib/zstd/compress/hist.h
new file mode 100644
index 0000000000..fc1830abc9
--- /dev/null
+++ b/lib/zstd/compress/hist.h
@@ -0,0 +1,75 @@
+/* ******************************************************************
+ * hist : Histogram functions
+ * part of Finite State Entropy project
+ * Copyright (c) Yann Collet, Facebook, Inc.
+ *
+ * You can contact the author at :
+ * - FSE source repository : https://github.com/Cyan4973/FiniteStateEntropy
+ * - Public forum : https://groups.google.com/forum/#!forum/lz4c
+ *
+ * This source code is licensed under both the BSD-style license (found in the
+ * LICENSE file in the root directory of this source tree) and the GPLv2 (found
+ * in the COPYING file in the root directory of this source tree).
+ * You may select, at your option, one of the above-listed licenses.
+****************************************************************** */
+
+/* --- dependencies --- */
+#include "../common/zstd_deps.h" /* size_t */
+
+
+/* --- simple histogram functions --- */
+
+/*! HIST_count():
+ * Provides the precise count of each byte within a table 'count'.
+ * 'count' is a table of unsigned int, of minimum size (*maxSymbolValuePtr+1).
+ * Updates *maxSymbolValuePtr with actual largest symbol value detected.
+ * @return : count of the most frequent symbol (which isn't identified).
+ * or an error code, which can be tested using HIST_isError().
+ * note : if return == srcSize, there is only one symbol.
+ */
+size_t HIST_count(unsigned* count, unsigned* maxSymbolValuePtr,
+ const void* src, size_t srcSize);
+
+unsigned HIST_isError(size_t code); /*< tells if a return value is an error code */
+
+
+/* --- advanced histogram functions --- */
+
+#define HIST_WKSP_SIZE_U32 1024
+#define HIST_WKSP_SIZE (HIST_WKSP_SIZE_U32 * sizeof(unsigned))
+/* HIST_count_wksp() :
+ * Same as HIST_count(), but using an externally provided scratch buffer.
+ * Benefit is this function will use very little stack space.
+ * `workSpace` is a writable buffer which must be 4-bytes aligned,
+ * `workSpaceSize` must be >= HIST_WKSP_SIZE
+ */
+size_t HIST_count_wksp(unsigned* count, unsigned* maxSymbolValuePtr,
+ const void* src, size_t srcSize,
+ void* workSpace, size_t workSpaceSize);
+
+/* HIST_countFast() :
+ * same as HIST_count(), but blindly trusts that all byte values within src are <= *maxSymbolValuePtr.
+ * This function is unsafe, and will segfault if any value within `src` is `> *maxSymbolValuePtr`
+ */
+size_t HIST_countFast(unsigned* count, unsigned* maxSymbolValuePtr,
+ const void* src, size_t srcSize);
+
+/* HIST_countFast_wksp() :
+ * Same as HIST_countFast(), but using an externally provided scratch buffer.
+ * `workSpace` is a writable buffer which must be 4-bytes aligned,
+ * `workSpaceSize` must be >= HIST_WKSP_SIZE
+ */
+size_t HIST_countFast_wksp(unsigned* count, unsigned* maxSymbolValuePtr,
+ const void* src, size_t srcSize,
+ void* workSpace, size_t workSpaceSize);
+
+/*! HIST_count_simple() :
+ * Same as HIST_countFast(), this function is unsafe,
+ * and will segfault if any value within `src` is `> *maxSymbolValuePtr`.
+ * It is also a bit slower for large inputs.
+ * However, it does not need any additional memory (not even on stack).
+ * @return : count of the most frequent symbol.
+ * Note this function doesn't produce any error (i.e. it must succeed).
+ */
+unsigned HIST_count_simple(unsigned* count, unsigned* maxSymbolValuePtr,
+ const void* src, size_t srcSize);
diff --git a/lib/zstd/compress/huf_compress.c b/lib/zstd/compress/huf_compress.c
new file mode 100644
index 0000000000..74ef0db476
--- /dev/null
+++ b/lib/zstd/compress/huf_compress.c
@@ -0,0 +1,1335 @@
+/* ******************************************************************
+ * Huffman encoder, part of New Generation Entropy library
+ * Copyright (c) Yann Collet, Facebook, Inc.
+ *
+ * You can contact the author at :
+ * - FSE+HUF source repository : https://github.com/Cyan4973/FiniteStateEntropy
+ * - Public forum : https://groups.google.com/forum/#!forum/lz4c
+ *
+ * This source code is licensed under both the BSD-style license (found in the
+ * LICENSE file in the root directory of this source tree) and the GPLv2 (found
+ * in the COPYING file in the root directory of this source tree).
+ * You may select, at your option, one of the above-listed licenses.
+****************************************************************** */
+
+/* **************************************************************
+* Compiler specifics
+****************************************************************/
+
+
+/* **************************************************************
+* Includes
+****************************************************************/
+#include "../common/zstd_deps.h" /* ZSTD_memcpy, ZSTD_memset */
+#include "../common/compiler.h"
+#include "../common/bitstream.h"
+#include "hist.h"
+#define FSE_STATIC_LINKING_ONLY /* FSE_optimalTableLog_internal */
+#include "../common/fse.h" /* header compression */
+#define HUF_STATIC_LINKING_ONLY
+#include "../common/huf.h"
+#include "../common/error_private.h"
+
+
+/* **************************************************************
+* Error Management
+****************************************************************/
+#define HUF_isError ERR_isError
+#define HUF_STATIC_ASSERT(c) DEBUG_STATIC_ASSERT(c) /* use only *after* variable declarations */
+
+
+/* **************************************************************
+* Utils
+****************************************************************/
+unsigned HUF_optimalTableLog(unsigned maxTableLog, size_t srcSize, unsigned maxSymbolValue)
+{
+ return FSE_optimalTableLog_internal(maxTableLog, srcSize, maxSymbolValue, 1);
+}
+
+
+/* *******************************************************
+* HUF : Huffman block compression
+*********************************************************/
+#define HUF_WORKSPACE_MAX_ALIGNMENT 8
+
+static void* HUF_alignUpWorkspace(void* workspace, size_t* workspaceSizePtr, size_t align)
+{
+ size_t const mask = align - 1;
+ size_t const rem = (size_t)workspace & mask;
+ size_t const add = (align - rem) & mask;
+ BYTE* const aligned = (BYTE*)workspace + add;
+ assert((align & (align - 1)) == 0); /* pow 2 */
+ assert(align <= HUF_WORKSPACE_MAX_ALIGNMENT);
+ if (*workspaceSizePtr >= add) {
+ assert(add < align);
+ assert(((size_t)aligned & mask) == 0);
+ *workspaceSizePtr -= add;
+ return aligned;
+ } else {
+ *workspaceSizePtr = 0;
+ return NULL;
+ }
+}
+
+
+/* HUF_compressWeights() :
+ * Same as FSE_compress(), but dedicated to huff0's weights compression.
+ * The use case needs much less stack memory.
+ * Note : all elements within weightTable are supposed to be <= HUF_TABLELOG_MAX.
+ */
+#define MAX_FSE_TABLELOG_FOR_HUFF_HEADER 6
+
+typedef struct {
+ FSE_CTable CTable[FSE_CTABLE_SIZE_U32(MAX_FSE_TABLELOG_FOR_HUFF_HEADER, HUF_TABLELOG_MAX)];
+ U32 scratchBuffer[FSE_BUILD_CTABLE_WORKSPACE_SIZE_U32(HUF_TABLELOG_MAX, MAX_FSE_TABLELOG_FOR_HUFF_HEADER)];
+ unsigned count[HUF_TABLELOG_MAX+1];
+ S16 norm[HUF_TABLELOG_MAX+1];
+} HUF_CompressWeightsWksp;
+
+static size_t HUF_compressWeights(void* dst, size_t dstSize, const void* weightTable, size_t wtSize, void* workspace, size_t workspaceSize)
+{
+ BYTE* const ostart = (BYTE*) dst;
+ BYTE* op = ostart;
+ BYTE* const oend = ostart + dstSize;
+
+ unsigned maxSymbolValue = HUF_TABLELOG_MAX;
+ U32 tableLog = MAX_FSE_TABLELOG_FOR_HUFF_HEADER;
+ HUF_CompressWeightsWksp* wksp = (HUF_CompressWeightsWksp*)HUF_alignUpWorkspace(workspace, &workspaceSize, ZSTD_ALIGNOF(U32));
+
+ if (workspaceSize < sizeof(HUF_CompressWeightsWksp)) return ERROR(GENERIC);
+
+ /* init conditions */
+ if (wtSize <= 1) return 0; /* Not compressible */
+
+ /* Scan input and build symbol stats */
+ { unsigned const maxCount = HIST_count_simple(wksp->count, &maxSymbolValue, weightTable, wtSize); /* never fails */
+ if (maxCount == wtSize) return 1; /* only a single symbol in src : rle */
+ if (maxCount == 1) return 0; /* each symbol present maximum once => not compressible */
+ }
+
+ tableLog = FSE_optimalTableLog(tableLog, wtSize, maxSymbolValue);
+ CHECK_F( FSE_normalizeCount(wksp->norm, tableLog, wksp->count, wtSize, maxSymbolValue, /* useLowProbCount */ 0) );
+
+ /* Write table description header */
+ { CHECK_V_F(hSize, FSE_writeNCount(op, (size_t)(oend-op), wksp->norm, maxSymbolValue, tableLog) );
+ op += hSize;
+ }
+
+ /* Compress */
+ CHECK_F( FSE_buildCTable_wksp(wksp->CTable, wksp->norm, maxSymbolValue, tableLog, wksp->scratchBuffer, sizeof(wksp->scratchBuffer)) );
+ { CHECK_V_F(cSize, FSE_compress_usingCTable(op, (size_t)(oend - op), weightTable, wtSize, wksp->CTable) );
+ if (cSize == 0) return 0; /* not enough space for compressed data */
+ op += cSize;
+ }
+
+ return (size_t)(op-ostart);
+}
+
+static size_t HUF_getNbBits(HUF_CElt elt)
+{
+ return elt & 0xFF;
+}
+
+static size_t HUF_getNbBitsFast(HUF_CElt elt)
+{
+ return elt;
+}
+
+static size_t HUF_getValue(HUF_CElt elt)
+{
+ return elt & ~0xFF;
+}
+
+static size_t HUF_getValueFast(HUF_CElt elt)
+{
+ return elt;
+}
+
+static void HUF_setNbBits(HUF_CElt* elt, size_t nbBits)
+{
+ assert(nbBits <= HUF_TABLELOG_ABSOLUTEMAX);
+ *elt = nbBits;
+}
+
+static void HUF_setValue(HUF_CElt* elt, size_t value)
+{
+ size_t const nbBits = HUF_getNbBits(*elt);
+ if (nbBits > 0) {
+ assert((value >> nbBits) == 0);
+ *elt |= value << (sizeof(HUF_CElt) * 8 - nbBits);
+ }
+}
+
+typedef struct {
+ HUF_CompressWeightsWksp wksp;
+ BYTE bitsToWeight[HUF_TABLELOG_MAX + 1]; /* precomputed conversion table */
+ BYTE huffWeight[HUF_SYMBOLVALUE_MAX];
+} HUF_WriteCTableWksp;
+
+size_t HUF_writeCTable_wksp(void* dst, size_t maxDstSize,
+ const HUF_CElt* CTable, unsigned maxSymbolValue, unsigned huffLog,
+ void* workspace, size_t workspaceSize)
+{
+ HUF_CElt const* const ct = CTable + 1;
+ BYTE* op = (BYTE*)dst;
+ U32 n;
+ HUF_WriteCTableWksp* wksp = (HUF_WriteCTableWksp*)HUF_alignUpWorkspace(workspace, &workspaceSize, ZSTD_ALIGNOF(U32));
+
+ /* check conditions */
+ if (workspaceSize < sizeof(HUF_WriteCTableWksp)) return ERROR(GENERIC);
+ if (maxSymbolValue > HUF_SYMBOLVALUE_MAX) return ERROR(maxSymbolValue_tooLarge);
+
+ /* convert to weight */
+ wksp->bitsToWeight[0] = 0;
+ for (n=1; n<huffLog+1; n++)
+ wksp->bitsToWeight[n] = (BYTE)(huffLog + 1 - n);
+ for (n=0; n<maxSymbolValue; n++)
+ wksp->huffWeight[n] = wksp->bitsToWeight[HUF_getNbBits(ct[n])];
+
+ /* attempt weights compression by FSE */
+ if (maxDstSize < 1) return ERROR(dstSize_tooSmall);
+ { CHECK_V_F(hSize, HUF_compressWeights(op+1, maxDstSize-1, wksp->huffWeight, maxSymbolValue, &wksp->wksp, sizeof(wksp->wksp)) );
+ if ((hSize>1) & (hSize < maxSymbolValue/2)) { /* FSE compressed */
+ op[0] = (BYTE)hSize;
+ return hSize+1;
+ } }
+
+ /* write raw values as 4-bits (max : 15) */
+ if (maxSymbolValue > (256-128)) return ERROR(GENERIC); /* should not happen : likely means source cannot be compressed */
+ if (((maxSymbolValue+1)/2) + 1 > maxDstSize) return ERROR(dstSize_tooSmall); /* not enough space within dst buffer */
+ op[0] = (BYTE)(128 /*special case*/ + (maxSymbolValue-1));
+ wksp->huffWeight[maxSymbolValue] = 0; /* to be sure it doesn't cause msan issue in final combination */
+ for (n=0; n<maxSymbolValue; n+=2)
+ op[(n/2)+1] = (BYTE)((wksp->huffWeight[n] << 4) + wksp->huffWeight[n+1]);
+ return ((maxSymbolValue+1)/2) + 1;
+}
+
+/*! HUF_writeCTable() :
+ `CTable` : Huffman tree to save, using huf representation.
+ @return : size of saved CTable */
+size_t HUF_writeCTable (void* dst, size_t maxDstSize,
+ const HUF_CElt* CTable, unsigned maxSymbolValue, unsigned huffLog)
+{
+ HUF_WriteCTableWksp wksp;
+ return HUF_writeCTable_wksp(dst, maxDstSize, CTable, maxSymbolValue, huffLog, &wksp, sizeof(wksp));
+}
+
+
+size_t HUF_readCTable (HUF_CElt* CTable, unsigned* maxSymbolValuePtr, const void* src, size_t srcSize, unsigned* hasZeroWeights)
+{
+ BYTE huffWeight[HUF_SYMBOLVALUE_MAX + 1]; /* init not required, even though some static analyzer may complain */
+ U32 rankVal[HUF_TABLELOG_ABSOLUTEMAX + 1]; /* large enough for values from 0 to 16 */
+ U32 tableLog = 0;
+ U32 nbSymbols = 0;
+ HUF_CElt* const ct = CTable + 1;
+
+ /* get symbol weights */
+ CHECK_V_F(readSize, HUF_readStats(huffWeight, HUF_SYMBOLVALUE_MAX+1, rankVal, &nbSymbols, &tableLog, src, srcSize));
+ *hasZeroWeights = (rankVal[0] > 0);
+
+ /* check result */
+ if (tableLog > HUF_TABLELOG_MAX) return ERROR(tableLog_tooLarge);
+ if (nbSymbols > *maxSymbolValuePtr+1) return ERROR(maxSymbolValue_tooSmall);
+
+ CTable[0] = tableLog;
+
+ /* Prepare base value per rank */
+ { U32 n, nextRankStart = 0;
+ for (n=1; n<=tableLog; n++) {
+ U32 curr = nextRankStart;
+ nextRankStart += (rankVal[n] << (n-1));
+ rankVal[n] = curr;
+ } }
+
+ /* fill nbBits */
+ { U32 n; for (n=0; n<nbSymbols; n++) {
+ const U32 w = huffWeight[n];
+ HUF_setNbBits(ct + n, (BYTE)(tableLog + 1 - w) & -(w != 0));
+ } }
+
+ /* fill val */
+ { U16 nbPerRank[HUF_TABLELOG_MAX+2] = {0}; /* support w=0=>n=tableLog+1 */
+ U16 valPerRank[HUF_TABLELOG_MAX+2] = {0};
+ { U32 n; for (n=0; n<nbSymbols; n++) nbPerRank[HUF_getNbBits(ct[n])]++; }
+ /* determine stating value per rank */
+ valPerRank[tableLog+1] = 0; /* for w==0 */
+ { U16 min = 0;
+ U32 n; for (n=tableLog; n>0; n--) { /* start at n=tablelog <-> w=1 */
+ valPerRank[n] = min; /* get starting value within each rank */
+ min += nbPerRank[n];
+ min >>= 1;
+ } }
+ /* assign value within rank, symbol order */
+ { U32 n; for (n=0; n<nbSymbols; n++) HUF_setValue(ct + n, valPerRank[HUF_getNbBits(ct[n])]++); }
+ }
+
+ *maxSymbolValuePtr = nbSymbols - 1;
+ return readSize;
+}
+
+U32 HUF_getNbBitsFromCTable(HUF_CElt const* CTable, U32 symbolValue)
+{
+ const HUF_CElt* ct = CTable + 1;
+ assert(symbolValue <= HUF_SYMBOLVALUE_MAX);
+ return (U32)HUF_getNbBits(ct[symbolValue]);
+}
+
+
+typedef struct nodeElt_s {
+ U32 count;
+ U16 parent;
+ BYTE byte;
+ BYTE nbBits;
+} nodeElt;
+
+/*
+ * HUF_setMaxHeight():
+ * Enforces maxNbBits on the Huffman tree described in huffNode.
+ *
+ * It sets all nodes with nbBits > maxNbBits to be maxNbBits. Then it adjusts
+ * the tree to so that it is a valid canonical Huffman tree.
+ *
+ * @pre The sum of the ranks of each symbol == 2^largestBits,
+ * where largestBits == huffNode[lastNonNull].nbBits.
+ * @post The sum of the ranks of each symbol == 2^largestBits,
+ * where largestBits is the return value <= maxNbBits.
+ *
+ * @param huffNode The Huffman tree modified in place to enforce maxNbBits.
+ * @param lastNonNull The symbol with the lowest count in the Huffman tree.
+ * @param maxNbBits The maximum allowed number of bits, which the Huffman tree
+ * may not respect. After this function the Huffman tree will
+ * respect maxNbBits.
+ * @return The maximum number of bits of the Huffman tree after adjustment,
+ * necessarily no more than maxNbBits.
+ */
+static U32 HUF_setMaxHeight(nodeElt* huffNode, U32 lastNonNull, U32 maxNbBits)
+{
+ const U32 largestBits = huffNode[lastNonNull].nbBits;
+ /* early exit : no elt > maxNbBits, so the tree is already valid. */
+ if (largestBits <= maxNbBits) return largestBits;
+
+ /* there are several too large elements (at least >= 2) */
+ { int totalCost = 0;
+ const U32 baseCost = 1 << (largestBits - maxNbBits);
+ int n = (int)lastNonNull;
+
+ /* Adjust any ranks > maxNbBits to maxNbBits.
+ * Compute totalCost, which is how far the sum of the ranks is
+ * we are over 2^largestBits after adjust the offending ranks.
+ */
+ while (huffNode[n].nbBits > maxNbBits) {
+ totalCost += baseCost - (1 << (largestBits - huffNode[n].nbBits));
+ huffNode[n].nbBits = (BYTE)maxNbBits;
+ n--;
+ }
+ /* n stops at huffNode[n].nbBits <= maxNbBits */
+ assert(huffNode[n].nbBits <= maxNbBits);
+ /* n end at index of smallest symbol using < maxNbBits */
+ while (huffNode[n].nbBits == maxNbBits) --n;
+
+ /* renorm totalCost from 2^largestBits to 2^maxNbBits
+ * note : totalCost is necessarily a multiple of baseCost */
+ assert((totalCost & (baseCost - 1)) == 0);
+ totalCost >>= (largestBits - maxNbBits);
+ assert(totalCost > 0);
+
+ /* repay normalized cost */
+ { U32 const noSymbol = 0xF0F0F0F0;
+ U32 rankLast[HUF_TABLELOG_MAX+2];
+
+ /* Get pos of last (smallest = lowest cum. count) symbol per rank */
+ ZSTD_memset(rankLast, 0xF0, sizeof(rankLast));
+ { U32 currentNbBits = maxNbBits;
+ int pos;
+ for (pos=n ; pos >= 0; pos--) {
+ if (huffNode[pos].nbBits >= currentNbBits) continue;
+ currentNbBits = huffNode[pos].nbBits; /* < maxNbBits */
+ rankLast[maxNbBits-currentNbBits] = (U32)pos;
+ } }
+
+ while (totalCost > 0) {
+ /* Try to reduce the next power of 2 above totalCost because we
+ * gain back half the rank.
+ */
+ U32 nBitsToDecrease = BIT_highbit32((U32)totalCost) + 1;
+ for ( ; nBitsToDecrease > 1; nBitsToDecrease--) {
+ U32 const highPos = rankLast[nBitsToDecrease];
+ U32 const lowPos = rankLast[nBitsToDecrease-1];
+ if (highPos == noSymbol) continue;
+ /* Decrease highPos if no symbols of lowPos or if it is
+ * not cheaper to remove 2 lowPos than highPos.
+ */
+ if (lowPos == noSymbol) break;
+ { U32 const highTotal = huffNode[highPos].count;
+ U32 const lowTotal = 2 * huffNode[lowPos].count;
+ if (highTotal <= lowTotal) break;
+ } }
+ /* only triggered when no more rank 1 symbol left => find closest one (note : there is necessarily at least one !) */
+ assert(rankLast[nBitsToDecrease] != noSymbol || nBitsToDecrease == 1);
+ /* HUF_MAX_TABLELOG test just to please gcc 5+; but it should not be necessary */
+ while ((nBitsToDecrease<=HUF_TABLELOG_MAX) && (rankLast[nBitsToDecrease] == noSymbol))
+ nBitsToDecrease++;
+ assert(rankLast[nBitsToDecrease] != noSymbol);
+ /* Increase the number of bits to gain back half the rank cost. */
+ totalCost -= 1 << (nBitsToDecrease-1);
+ huffNode[rankLast[nBitsToDecrease]].nbBits++;
+
+ /* Fix up the new rank.
+ * If the new rank was empty, this symbol is now its smallest.
+ * Otherwise, this symbol will be the largest in the new rank so no adjustment.
+ */
+ if (rankLast[nBitsToDecrease-1] == noSymbol)
+ rankLast[nBitsToDecrease-1] = rankLast[nBitsToDecrease];
+ /* Fix up the old rank.
+ * If the symbol was at position 0, meaning it was the highest weight symbol in the tree,
+ * it must be the only symbol in its rank, so the old rank now has no symbols.
+ * Otherwise, since the Huffman nodes are sorted by count, the previous position is now
+ * the smallest node in the rank. If the previous position belongs to a different rank,
+ * then the rank is now empty.
+ */
+ if (rankLast[nBitsToDecrease] == 0) /* special case, reached largest symbol */
+ rankLast[nBitsToDecrease] = noSymbol;
+ else {
+ rankLast[nBitsToDecrease]--;
+ if (huffNode[rankLast[nBitsToDecrease]].nbBits != maxNbBits-nBitsToDecrease)
+ rankLast[nBitsToDecrease] = noSymbol; /* this rank is now empty */
+ }
+ } /* while (totalCost > 0) */
+
+ /* If we've removed too much weight, then we have to add it back.
+ * To avoid overshooting again, we only adjust the smallest rank.
+ * We take the largest nodes from the lowest rank 0 and move them
+ * to rank 1. There's guaranteed to be enough rank 0 symbols because
+ * TODO.
+ */
+ while (totalCost < 0) { /* Sometimes, cost correction overshoot */
+ /* special case : no rank 1 symbol (using maxNbBits-1);
+ * let's create one from largest rank 0 (using maxNbBits).
+ */
+ if (rankLast[1] == noSymbol) {
+ while (huffNode[n].nbBits == maxNbBits) n--;
+ huffNode[n+1].nbBits--;
+ assert(n >= 0);
+ rankLast[1] = (U32)(n+1);
+ totalCost++;
+ continue;
+ }
+ huffNode[ rankLast[1] + 1 ].nbBits--;
+ rankLast[1]++;
+ totalCost ++;
+ }
+ } /* repay normalized cost */
+ } /* there are several too large elements (at least >= 2) */
+
+ return maxNbBits;
+}
+
+typedef struct {
+ U16 base;
+ U16 curr;
+} rankPos;
+
+typedef nodeElt huffNodeTable[HUF_CTABLE_WORKSPACE_SIZE_U32];
+
+/* Number of buckets available for HUF_sort() */
+#define RANK_POSITION_TABLE_SIZE 192
+
+typedef struct {
+ huffNodeTable huffNodeTbl;
+ rankPos rankPosition[RANK_POSITION_TABLE_SIZE];
+} HUF_buildCTable_wksp_tables;
+
+/* RANK_POSITION_DISTINCT_COUNT_CUTOFF == Cutoff point in HUF_sort() buckets for which we use log2 bucketing.
+ * Strategy is to use as many buckets as possible for representing distinct
+ * counts while using the remainder to represent all "large" counts.
+ *
+ * To satisfy this requirement for 192 buckets, we can do the following:
+ * Let buckets 0-166 represent distinct counts of [0, 166]
+ * Let buckets 166 to 192 represent all remaining counts up to RANK_POSITION_MAX_COUNT_LOG using log2 bucketing.
+ */
+#define RANK_POSITION_MAX_COUNT_LOG 32
+#define RANK_POSITION_LOG_BUCKETS_BEGIN (RANK_POSITION_TABLE_SIZE - 1) - RANK_POSITION_MAX_COUNT_LOG - 1 /* == 158 */
+#define RANK_POSITION_DISTINCT_COUNT_CUTOFF RANK_POSITION_LOG_BUCKETS_BEGIN + BIT_highbit32(RANK_POSITION_LOG_BUCKETS_BEGIN) /* == 166 */
+
+/* Return the appropriate bucket index for a given count. See definition of
+ * RANK_POSITION_DISTINCT_COUNT_CUTOFF for explanation of bucketing strategy.
+ */
+static U32 HUF_getIndex(U32 const count) {
+ return (count < RANK_POSITION_DISTINCT_COUNT_CUTOFF)
+ ? count
+ : BIT_highbit32(count) + RANK_POSITION_LOG_BUCKETS_BEGIN;
+}
+
+/* Helper swap function for HUF_quickSortPartition() */
+static void HUF_swapNodes(nodeElt* a, nodeElt* b) {
+ nodeElt tmp = *a;
+ *a = *b;
+ *b = tmp;
+}
+
+/* Returns 0 if the huffNode array is not sorted by descending count */
+MEM_STATIC int HUF_isSorted(nodeElt huffNode[], U32 const maxSymbolValue1) {
+ U32 i;
+ for (i = 1; i < maxSymbolValue1; ++i) {
+ if (huffNode[i].count > huffNode[i-1].count) {
+ return 0;
+ }
+ }
+ return 1;
+}
+
+/* Insertion sort by descending order */
+HINT_INLINE void HUF_insertionSort(nodeElt huffNode[], int const low, int const high) {
+ int i;
+ int const size = high-low+1;
+ huffNode += low;
+ for (i = 1; i < size; ++i) {
+ nodeElt const key = huffNode[i];
+ int j = i - 1;
+ while (j >= 0 && huffNode[j].count < key.count) {
+ huffNode[j + 1] = huffNode[j];
+ j--;
+ }
+ huffNode[j + 1] = key;
+ }
+}
+
+/* Pivot helper function for quicksort. */
+static int HUF_quickSortPartition(nodeElt arr[], int const low, int const high) {
+ /* Simply select rightmost element as pivot. "Better" selectors like
+ * median-of-three don't experimentally appear to have any benefit.
+ */
+ U32 const pivot = arr[high].count;
+ int i = low - 1;
+ int j = low;
+ for ( ; j < high; j++) {
+ if (arr[j].count > pivot) {
+ i++;
+ HUF_swapNodes(&arr[i], &arr[j]);
+ }
+ }
+ HUF_swapNodes(&arr[i + 1], &arr[high]);
+ return i + 1;
+}
+
+/* Classic quicksort by descending with partially iterative calls
+ * to reduce worst case callstack size.
+ */
+static void HUF_simpleQuickSort(nodeElt arr[], int low, int high) {
+ int const kInsertionSortThreshold = 8;
+ if (high - low < kInsertionSortThreshold) {
+ HUF_insertionSort(arr, low, high);
+ return;
+ }
+ while (low < high) {
+ int const idx = HUF_quickSortPartition(arr, low, high);
+ if (idx - low < high - idx) {
+ HUF_simpleQuickSort(arr, low, idx - 1);
+ low = idx + 1;
+ } else {
+ HUF_simpleQuickSort(arr, idx + 1, high);
+ high = idx - 1;
+ }
+ }
+}
+
+/*
+ * HUF_sort():
+ * Sorts the symbols [0, maxSymbolValue] by count[symbol] in decreasing order.
+ * This is a typical bucket sorting strategy that uses either quicksort or insertion sort to sort each bucket.
+ *
+ * @param[out] huffNode Sorted symbols by decreasing count. Only members `.count` and `.byte` are filled.
+ * Must have (maxSymbolValue + 1) entries.
+ * @param[in] count Histogram of the symbols.
+ * @param[in] maxSymbolValue Maximum symbol value.
+ * @param rankPosition This is a scratch workspace. Must have RANK_POSITION_TABLE_SIZE entries.
+ */
+static void HUF_sort(nodeElt huffNode[], const unsigned count[], U32 const maxSymbolValue, rankPos rankPosition[]) {
+ U32 n;
+ U32 const maxSymbolValue1 = maxSymbolValue+1;
+
+ /* Compute base and set curr to base.
+ * For symbol s let lowerRank = HUF_getIndex(count[n]) and rank = lowerRank + 1.
+ * See HUF_getIndex to see bucketing strategy.
+ * We attribute each symbol to lowerRank's base value, because we want to know where
+ * each rank begins in the output, so for rank R we want to count ranks R+1 and above.
+ */
+ ZSTD_memset(rankPosition, 0, sizeof(*rankPosition) * RANK_POSITION_TABLE_SIZE);
+ for (n = 0; n < maxSymbolValue1; ++n) {
+ U32 lowerRank = HUF_getIndex(count[n]);
+ assert(lowerRank < RANK_POSITION_TABLE_SIZE - 1);
+ rankPosition[lowerRank].base++;
+ }
+
+ assert(rankPosition[RANK_POSITION_TABLE_SIZE - 1].base == 0);
+ /* Set up the rankPosition table */
+ for (n = RANK_POSITION_TABLE_SIZE - 1; n > 0; --n) {
+ rankPosition[n-1].base += rankPosition[n].base;
+ rankPosition[n-1].curr = rankPosition[n-1].base;
+ }
+
+ /* Insert each symbol into their appropriate bucket, setting up rankPosition table. */
+ for (n = 0; n < maxSymbolValue1; ++n) {
+ U32 const c = count[n];
+ U32 const r = HUF_getIndex(c) + 1;
+ U32 const pos = rankPosition[r].curr++;
+ assert(pos < maxSymbolValue1);
+ huffNode[pos].count = c;
+ huffNode[pos].byte = (BYTE)n;
+ }
+
+ /* Sort each bucket. */
+ for (n = RANK_POSITION_DISTINCT_COUNT_CUTOFF; n < RANK_POSITION_TABLE_SIZE - 1; ++n) {
+ U32 const bucketSize = rankPosition[n].curr-rankPosition[n].base;
+ U32 const bucketStartIdx = rankPosition[n].base;
+ if (bucketSize > 1) {
+ assert(bucketStartIdx < maxSymbolValue1);
+ HUF_simpleQuickSort(huffNode + bucketStartIdx, 0, bucketSize-1);
+ }
+ }
+
+ assert(HUF_isSorted(huffNode, maxSymbolValue1));
+}
+
+/* HUF_buildCTable_wksp() :
+ * Same as HUF_buildCTable(), but using externally allocated scratch buffer.
+ * `workSpace` must be aligned on 4-bytes boundaries, and be at least as large as sizeof(HUF_buildCTable_wksp_tables).
+ */
+#define STARTNODE (HUF_SYMBOLVALUE_MAX+1)
+
+/* HUF_buildTree():
+ * Takes the huffNode array sorted by HUF_sort() and builds an unlimited-depth Huffman tree.
+ *
+ * @param huffNode The array sorted by HUF_sort(). Builds the Huffman tree in this array.
+ * @param maxSymbolValue The maximum symbol value.
+ * @return The smallest node in the Huffman tree (by count).
+ */
+static int HUF_buildTree(nodeElt* huffNode, U32 maxSymbolValue)
+{
+ nodeElt* const huffNode0 = huffNode - 1;
+ int nonNullRank;
+ int lowS, lowN;
+ int nodeNb = STARTNODE;
+ int n, nodeRoot;
+ /* init for parents */
+ nonNullRank = (int)maxSymbolValue;
+ while(huffNode[nonNullRank].count == 0) nonNullRank--;
+ lowS = nonNullRank; nodeRoot = nodeNb + lowS - 1; lowN = nodeNb;
+ huffNode[nodeNb].count = huffNode[lowS].count + huffNode[lowS-1].count;
+ huffNode[lowS].parent = huffNode[lowS-1].parent = (U16)nodeNb;
+ nodeNb++; lowS-=2;
+ for (n=nodeNb; n<=nodeRoot; n++) huffNode[n].count = (U32)(1U<<30);
+ huffNode0[0].count = (U32)(1U<<31); /* fake entry, strong barrier */
+
+ /* create parents */
+ while (nodeNb <= nodeRoot) {
+ int const n1 = (huffNode[lowS].count < huffNode[lowN].count) ? lowS-- : lowN++;
+ int const n2 = (huffNode[lowS].count < huffNode[lowN].count) ? lowS-- : lowN++;
+ huffNode[nodeNb].count = huffNode[n1].count + huffNode[n2].count;
+ huffNode[n1].parent = huffNode[n2].parent = (U16)nodeNb;
+ nodeNb++;
+ }
+
+ /* distribute weights (unlimited tree height) */
+ huffNode[nodeRoot].nbBits = 0;
+ for (n=nodeRoot-1; n>=STARTNODE; n--)
+ huffNode[n].nbBits = huffNode[ huffNode[n].parent ].nbBits + 1;
+ for (n=0; n<=nonNullRank; n++)
+ huffNode[n].nbBits = huffNode[ huffNode[n].parent ].nbBits + 1;
+
+ return nonNullRank;
+}
+
+/*
+ * HUF_buildCTableFromTree():
+ * Build the CTable given the Huffman tree in huffNode.
+ *
+ * @param[out] CTable The output Huffman CTable.
+ * @param huffNode The Huffman tree.
+ * @param nonNullRank The last and smallest node in the Huffman tree.
+ * @param maxSymbolValue The maximum symbol value.
+ * @param maxNbBits The exact maximum number of bits used in the Huffman tree.
+ */
+static void HUF_buildCTableFromTree(HUF_CElt* CTable, nodeElt const* huffNode, int nonNullRank, U32 maxSymbolValue, U32 maxNbBits)
+{
+ HUF_CElt* const ct = CTable + 1;
+ /* fill result into ctable (val, nbBits) */
+ int n;
+ U16 nbPerRank[HUF_TABLELOG_MAX+1] = {0};
+ U16 valPerRank[HUF_TABLELOG_MAX+1] = {0};
+ int const alphabetSize = (int)(maxSymbolValue + 1);
+ for (n=0; n<=nonNullRank; n++)
+ nbPerRank[huffNode[n].nbBits]++;
+ /* determine starting value per rank */
+ { U16 min = 0;
+ for (n=(int)maxNbBits; n>0; n--) {
+ valPerRank[n] = min; /* get starting value within each rank */
+ min += nbPerRank[n];
+ min >>= 1;
+ } }
+ for (n=0; n<alphabetSize; n++)
+ HUF_setNbBits(ct + huffNode[n].byte, huffNode[n].nbBits); /* push nbBits per symbol, symbol order */
+ for (n=0; n<alphabetSize; n++)
+ HUF_setValue(ct + n, valPerRank[HUF_getNbBits(ct[n])]++); /* assign value within rank, symbol order */
+ CTable[0] = maxNbBits;
+}
+
+size_t HUF_buildCTable_wksp (HUF_CElt* CTable, const unsigned* count, U32 maxSymbolValue, U32 maxNbBits, void* workSpace, size_t wkspSize)
+{
+ HUF_buildCTable_wksp_tables* const wksp_tables = (HUF_buildCTable_wksp_tables*)HUF_alignUpWorkspace(workSpace, &wkspSize, ZSTD_ALIGNOF(U32));
+ nodeElt* const huffNode0 = wksp_tables->huffNodeTbl;
+ nodeElt* const huffNode = huffNode0+1;
+ int nonNullRank;
+
+ /* safety checks */
+ if (wkspSize < sizeof(HUF_buildCTable_wksp_tables))
+ return ERROR(workSpace_tooSmall);
+ if (maxNbBits == 0) maxNbBits = HUF_TABLELOG_DEFAULT;
+ if (maxSymbolValue > HUF_SYMBOLVALUE_MAX)
+ return ERROR(maxSymbolValue_tooLarge);
+ ZSTD_memset(huffNode0, 0, sizeof(huffNodeTable));
+
+ /* sort, decreasing order */
+ HUF_sort(huffNode, count, maxSymbolValue, wksp_tables->rankPosition);
+
+ /* build tree */
+ nonNullRank = HUF_buildTree(huffNode, maxSymbolValue);
+
+ /* enforce maxTableLog */
+ maxNbBits = HUF_setMaxHeight(huffNode, (U32)nonNullRank, maxNbBits);
+ if (maxNbBits > HUF_TABLELOG_MAX) return ERROR(GENERIC); /* check fit into table */
+
+ HUF_buildCTableFromTree(CTable, huffNode, nonNullRank, maxSymbolValue, maxNbBits);
+
+ return maxNbBits;
+}
+
+size_t HUF_estimateCompressedSize(const HUF_CElt* CTable, const unsigned* count, unsigned maxSymbolValue)
+{
+ HUF_CElt const* ct = CTable + 1;
+ size_t nbBits = 0;
+ int s;
+ for (s = 0; s <= (int)maxSymbolValue; ++s) {
+ nbBits += HUF_getNbBits(ct[s]) * count[s];
+ }
+ return nbBits >> 3;
+}
+
+int HUF_validateCTable(const HUF_CElt* CTable, const unsigned* count, unsigned maxSymbolValue) {
+ HUF_CElt const* ct = CTable + 1;
+ int bad = 0;
+ int s;
+ for (s = 0; s <= (int)maxSymbolValue; ++s) {
+ bad |= (count[s] != 0) & (HUF_getNbBits(ct[s]) == 0);
+ }
+ return !bad;
+}
+
+size_t HUF_compressBound(size_t size) { return HUF_COMPRESSBOUND(size); }
+
+/* HUF_CStream_t:
+ * Huffman uses its own BIT_CStream_t implementation.
+ * There are three major differences from BIT_CStream_t:
+ * 1. HUF_addBits() takes a HUF_CElt (size_t) which is
+ * the pair (nbBits, value) in the format:
+ * format:
+ * - Bits [0, 4) = nbBits
+ * - Bits [4, 64 - nbBits) = 0
+ * - Bits [64 - nbBits, 64) = value
+ * 2. The bitContainer is built from the upper bits and
+ * right shifted. E.g. to add a new value of N bits
+ * you right shift the bitContainer by N, then or in
+ * the new value into the N upper bits.
+ * 3. The bitstream has two bit containers. You can add
+ * bits to the second container and merge them into
+ * the first container.
+ */
+
+#define HUF_BITS_IN_CONTAINER (sizeof(size_t) * 8)
+
+typedef struct {
+ size_t bitContainer[2];
+ size_t bitPos[2];
+
+ BYTE* startPtr;
+ BYTE* ptr;
+ BYTE* endPtr;
+} HUF_CStream_t;
+
+/*! HUF_initCStream():
+ * Initializes the bitstream.
+ * @returns 0 or an error code.
+ */
+static size_t HUF_initCStream(HUF_CStream_t* bitC,
+ void* startPtr, size_t dstCapacity)
+{
+ ZSTD_memset(bitC, 0, sizeof(*bitC));
+ bitC->startPtr = (BYTE*)startPtr;
+ bitC->ptr = bitC->startPtr;
+ bitC->endPtr = bitC->startPtr + dstCapacity - sizeof(bitC->bitContainer[0]);
+ if (dstCapacity <= sizeof(bitC->bitContainer[0])) return ERROR(dstSize_tooSmall);
+ return 0;
+}
+
+/*! HUF_addBits():
+ * Adds the symbol stored in HUF_CElt elt to the bitstream.
+ *
+ * @param elt The element we're adding. This is a (nbBits, value) pair.
+ * See the HUF_CStream_t docs for the format.
+ * @param idx Insert into the bitstream at this idx.
+ * @param kFast This is a template parameter. If the bitstream is guaranteed
+ * to have at least 4 unused bits after this call it may be 1,
+ * otherwise it must be 0. HUF_addBits() is faster when fast is set.
+ */
+FORCE_INLINE_TEMPLATE void HUF_addBits(HUF_CStream_t* bitC, HUF_CElt elt, int idx, int kFast)
+{
+ assert(idx <= 1);
+ assert(HUF_getNbBits(elt) <= HUF_TABLELOG_ABSOLUTEMAX);
+ /* This is efficient on x86-64 with BMI2 because shrx
+ * only reads the low 6 bits of the register. The compiler
+ * knows this and elides the mask. When fast is set,
+ * every operation can use the same value loaded from elt.
+ */
+ bitC->bitContainer[idx] >>= HUF_getNbBits(elt);
+ bitC->bitContainer[idx] |= kFast ? HUF_getValueFast(elt) : HUF_getValue(elt);
+ /* We only read the low 8 bits of bitC->bitPos[idx] so it
+ * doesn't matter that the high bits have noise from the value.
+ */
+ bitC->bitPos[idx] += HUF_getNbBitsFast(elt);
+ assert((bitC->bitPos[idx] & 0xFF) <= HUF_BITS_IN_CONTAINER);
+ /* The last 4-bits of elt are dirty if fast is set,
+ * so we must not be overwriting bits that have already been
+ * inserted into the bit container.
+ */
+#if DEBUGLEVEL >= 1
+ {
+ size_t const nbBits = HUF_getNbBits(elt);
+ size_t const dirtyBits = nbBits == 0 ? 0 : BIT_highbit32((U32)nbBits) + 1;
+ (void)dirtyBits;
+ /* Middle bits are 0. */
+ assert(((elt >> dirtyBits) << (dirtyBits + nbBits)) == 0);
+ /* We didn't overwrite any bits in the bit container. */
+ assert(!kFast || (bitC->bitPos[idx] & 0xFF) <= HUF_BITS_IN_CONTAINER);
+ (void)dirtyBits;
+ }
+#endif
+}
+
+FORCE_INLINE_TEMPLATE void HUF_zeroIndex1(HUF_CStream_t* bitC)
+{
+ bitC->bitContainer[1] = 0;
+ bitC->bitPos[1] = 0;
+}
+
+/*! HUF_mergeIndex1() :
+ * Merges the bit container @ index 1 into the bit container @ index 0
+ * and zeros the bit container @ index 1.
+ */
+FORCE_INLINE_TEMPLATE void HUF_mergeIndex1(HUF_CStream_t* bitC)
+{
+ assert((bitC->bitPos[1] & 0xFF) < HUF_BITS_IN_CONTAINER);
+ bitC->bitContainer[0] >>= (bitC->bitPos[1] & 0xFF);
+ bitC->bitContainer[0] |= bitC->bitContainer[1];
+ bitC->bitPos[0] += bitC->bitPos[1];
+ assert((bitC->bitPos[0] & 0xFF) <= HUF_BITS_IN_CONTAINER);
+}
+
+/*! HUF_flushBits() :
+* Flushes the bits in the bit container @ index 0.
+*
+* @post bitPos will be < 8.
+* @param kFast If kFast is set then we must know a-priori that
+* the bit container will not overflow.
+*/
+FORCE_INLINE_TEMPLATE void HUF_flushBits(HUF_CStream_t* bitC, int kFast)
+{
+ /* The upper bits of bitPos are noisy, so we must mask by 0xFF. */
+ size_t const nbBits = bitC->bitPos[0] & 0xFF;
+ size_t const nbBytes = nbBits >> 3;
+ /* The top nbBits bits of bitContainer are the ones we need. */
+ size_t const bitContainer = bitC->bitContainer[0] >> (HUF_BITS_IN_CONTAINER - nbBits);
+ /* Mask bitPos to account for the bytes we consumed. */
+ bitC->bitPos[0] &= 7;
+ assert(nbBits > 0);
+ assert(nbBits <= sizeof(bitC->bitContainer[0]) * 8);
+ assert(bitC->ptr <= bitC->endPtr);
+ MEM_writeLEST(bitC->ptr, bitContainer);
+ bitC->ptr += nbBytes;
+ assert(!kFast || bitC->ptr <= bitC->endPtr);
+ if (!kFast && bitC->ptr > bitC->endPtr) bitC->ptr = bitC->endPtr;
+ /* bitContainer doesn't need to be modified because the leftover
+ * bits are already the top bitPos bits. And we don't care about
+ * noise in the lower values.
+ */
+}
+
+/*! HUF_endMark()
+ * @returns The Huffman stream end mark: A 1-bit value = 1.
+ */
+static HUF_CElt HUF_endMark(void)
+{
+ HUF_CElt endMark;
+ HUF_setNbBits(&endMark, 1);
+ HUF_setValue(&endMark, 1);
+ return endMark;
+}
+
+/*! HUF_closeCStream() :
+ * @return Size of CStream, in bytes,
+ * or 0 if it could not fit into dstBuffer */
+static size_t HUF_closeCStream(HUF_CStream_t* bitC)
+{
+ HUF_addBits(bitC, HUF_endMark(), /* idx */ 0, /* kFast */ 0);
+ HUF_flushBits(bitC, /* kFast */ 0);
+ {
+ size_t const nbBits = bitC->bitPos[0] & 0xFF;
+ if (bitC->ptr >= bitC->endPtr) return 0; /* overflow detected */
+ return (bitC->ptr - bitC->startPtr) + (nbBits > 0);
+ }
+}
+
+FORCE_INLINE_TEMPLATE void
+HUF_encodeSymbol(HUF_CStream_t* bitCPtr, U32 symbol, const HUF_CElt* CTable, int idx, int fast)
+{
+ HUF_addBits(bitCPtr, CTable[symbol], idx, fast);
+}
+
+FORCE_INLINE_TEMPLATE void
+HUF_compress1X_usingCTable_internal_body_loop(HUF_CStream_t* bitC,
+ const BYTE* ip, size_t srcSize,
+ const HUF_CElt* ct,
+ int kUnroll, int kFastFlush, int kLastFast)
+{
+ /* Join to kUnroll */
+ int n = (int)srcSize;
+ int rem = n % kUnroll;
+ if (rem > 0) {
+ for (; rem > 0; --rem) {
+ HUF_encodeSymbol(bitC, ip[--n], ct, 0, /* fast */ 0);
+ }
+ HUF_flushBits(bitC, kFastFlush);
+ }
+ assert(n % kUnroll == 0);
+
+ /* Join to 2 * kUnroll */
+ if (n % (2 * kUnroll)) {
+ int u;
+ for (u = 1; u < kUnroll; ++u) {
+ HUF_encodeSymbol(bitC, ip[n - u], ct, 0, 1);
+ }
+ HUF_encodeSymbol(bitC, ip[n - kUnroll], ct, 0, kLastFast);
+ HUF_flushBits(bitC, kFastFlush);
+ n -= kUnroll;
+ }
+ assert(n % (2 * kUnroll) == 0);
+
+ for (; n>0; n-= 2 * kUnroll) {
+ /* Encode kUnroll symbols into the bitstream @ index 0. */
+ int u;
+ for (u = 1; u < kUnroll; ++u) {
+ HUF_encodeSymbol(bitC, ip[n - u], ct, /* idx */ 0, /* fast */ 1);
+ }
+ HUF_encodeSymbol(bitC, ip[n - kUnroll], ct, /* idx */ 0, /* fast */ kLastFast);
+ HUF_flushBits(bitC, kFastFlush);
+ /* Encode kUnroll symbols into the bitstream @ index 1.
+ * This allows us to start filling the bit container
+ * without any data dependencies.
+ */
+ HUF_zeroIndex1(bitC);
+ for (u = 1; u < kUnroll; ++u) {
+ HUF_encodeSymbol(bitC, ip[n - kUnroll - u], ct, /* idx */ 1, /* fast */ 1);
+ }
+ HUF_encodeSymbol(bitC, ip[n - kUnroll - kUnroll], ct, /* idx */ 1, /* fast */ kLastFast);
+ /* Merge bitstream @ index 1 into the bitstream @ index 0 */
+ HUF_mergeIndex1(bitC);
+ HUF_flushBits(bitC, kFastFlush);
+ }
+ assert(n == 0);
+
+}
+
+/*
+ * Returns a tight upper bound on the output space needed by Huffman
+ * with 8 bytes buffer to handle over-writes. If the output is at least
+ * this large we don't need to do bounds checks during Huffman encoding.
+ */
+static size_t HUF_tightCompressBound(size_t srcSize, size_t tableLog)
+{
+ return ((srcSize * tableLog) >> 3) + 8;
+}
+
+
+FORCE_INLINE_TEMPLATE size_t
+HUF_compress1X_usingCTable_internal_body(void* dst, size_t dstSize,
+ const void* src, size_t srcSize,
+ const HUF_CElt* CTable)
+{
+ U32 const tableLog = (U32)CTable[0];
+ HUF_CElt const* ct = CTable + 1;
+ const BYTE* ip = (const BYTE*) src;
+ BYTE* const ostart = (BYTE*)dst;
+ BYTE* const oend = ostart + dstSize;
+ BYTE* op = ostart;
+ HUF_CStream_t bitC;
+
+ /* init */
+ if (dstSize < 8) return 0; /* not enough space to compress */
+ { size_t const initErr = HUF_initCStream(&bitC, op, (size_t)(oend-op));
+ if (HUF_isError(initErr)) return 0; }
+
+ if (dstSize < HUF_tightCompressBound(srcSize, (size_t)tableLog) || tableLog > 11)
+ HUF_compress1X_usingCTable_internal_body_loop(&bitC, ip, srcSize, ct, /* kUnroll */ MEM_32bits() ? 2 : 4, /* kFast */ 0, /* kLastFast */ 0);
+ else {
+ if (MEM_32bits()) {
+ switch (tableLog) {
+ case 11:
+ HUF_compress1X_usingCTable_internal_body_loop(&bitC, ip, srcSize, ct, /* kUnroll */ 2, /* kFastFlush */ 1, /* kLastFast */ 0);
+ break;
+ case 10: ZSTD_FALLTHROUGH;
+ case 9: ZSTD_FALLTHROUGH;
+ case 8:
+ HUF_compress1X_usingCTable_internal_body_loop(&bitC, ip, srcSize, ct, /* kUnroll */ 2, /* kFastFlush */ 1, /* kLastFast */ 1);
+ break;
+ case 7: ZSTD_FALLTHROUGH;
+ default:
+ HUF_compress1X_usingCTable_internal_body_loop(&bitC, ip, srcSize, ct, /* kUnroll */ 3, /* kFastFlush */ 1, /* kLastFast */ 1);
+ break;
+ }
+ } else {
+ switch (tableLog) {
+ case 11:
+ HUF_compress1X_usingCTable_internal_body_loop(&bitC, ip, srcSize, ct, /* kUnroll */ 5, /* kFastFlush */ 1, /* kLastFast */ 0);
+ break;
+ case 10:
+ HUF_compress1X_usingCTable_internal_body_loop(&bitC, ip, srcSize, ct, /* kUnroll */ 5, /* kFastFlush */ 1, /* kLastFast */ 1);
+ break;
+ case 9:
+ HUF_compress1X_usingCTable_internal_body_loop(&bitC, ip, srcSize, ct, /* kUnroll */ 6, /* kFastFlush */ 1, /* kLastFast */ 0);
+ break;
+ case 8:
+ HUF_compress1X_usingCTable_internal_body_loop(&bitC, ip, srcSize, ct, /* kUnroll */ 7, /* kFastFlush */ 1, /* kLastFast */ 0);
+ break;
+ case 7:
+ HUF_compress1X_usingCTable_internal_body_loop(&bitC, ip, srcSize, ct, /* kUnroll */ 8, /* kFastFlush */ 1, /* kLastFast */ 0);
+ break;
+ case 6: ZSTD_FALLTHROUGH;
+ default:
+ HUF_compress1X_usingCTable_internal_body_loop(&bitC, ip, srcSize, ct, /* kUnroll */ 9, /* kFastFlush */ 1, /* kLastFast */ 1);
+ break;
+ }
+ }
+ }
+ assert(bitC.ptr <= bitC.endPtr);
+
+ return HUF_closeCStream(&bitC);
+}
+
+#if DYNAMIC_BMI2
+
+static BMI2_TARGET_ATTRIBUTE size_t
+HUF_compress1X_usingCTable_internal_bmi2(void* dst, size_t dstSize,
+ const void* src, size_t srcSize,
+ const HUF_CElt* CTable)
+{
+ return HUF_compress1X_usingCTable_internal_body(dst, dstSize, src, srcSize, CTable);
+}
+
+static size_t
+HUF_compress1X_usingCTable_internal_default(void* dst, size_t dstSize,
+ const void* src, size_t srcSize,
+ const HUF_CElt* CTable)
+{
+ return HUF_compress1X_usingCTable_internal_body(dst, dstSize, src, srcSize, CTable);
+}
+
+static size_t
+HUF_compress1X_usingCTable_internal(void* dst, size_t dstSize,
+ const void* src, size_t srcSize,
+ const HUF_CElt* CTable, const int bmi2)
+{
+ if (bmi2) {
+ return HUF_compress1X_usingCTable_internal_bmi2(dst, dstSize, src, srcSize, CTable);
+ }
+ return HUF_compress1X_usingCTable_internal_default(dst, dstSize, src, srcSize, CTable);
+}
+
+#else
+
+static size_t
+HUF_compress1X_usingCTable_internal(void* dst, size_t dstSize,
+ const void* src, size_t srcSize,
+ const HUF_CElt* CTable, const int bmi2)
+{
+ (void)bmi2;
+ return HUF_compress1X_usingCTable_internal_body(dst, dstSize, src, srcSize, CTable);
+}
+
+#endif
+
+size_t HUF_compress1X_usingCTable(void* dst, size_t dstSize, const void* src, size_t srcSize, const HUF_CElt* CTable)
+{
+ return HUF_compress1X_usingCTable_bmi2(dst, dstSize, src, srcSize, CTable, /* bmi2 */ 0);
+}
+
+size_t HUF_compress1X_usingCTable_bmi2(void* dst, size_t dstSize, const void* src, size_t srcSize, const HUF_CElt* CTable, int bmi2)
+{
+ return HUF_compress1X_usingCTable_internal(dst, dstSize, src, srcSize, CTable, bmi2);
+}
+
+static size_t
+HUF_compress4X_usingCTable_internal(void* dst, size_t dstSize,
+ const void* src, size_t srcSize,
+ const HUF_CElt* CTable, int bmi2)
+{
+ size_t const segmentSize = (srcSize+3)/4; /* first 3 segments */
+ const BYTE* ip = (const BYTE*) src;
+ const BYTE* const iend = ip + srcSize;
+ BYTE* const ostart = (BYTE*) dst;
+ BYTE* const oend = ostart + dstSize;
+ BYTE* op = ostart;
+
+ if (dstSize < 6 + 1 + 1 + 1 + 8) return 0; /* minimum space to compress successfully */
+ if (srcSize < 12) return 0; /* no saving possible : too small input */
+ op += 6; /* jumpTable */
+
+ assert(op <= oend);
+ { CHECK_V_F(cSize, HUF_compress1X_usingCTable_internal(op, (size_t)(oend-op), ip, segmentSize, CTable, bmi2) );
+ if (cSize == 0 || cSize > 65535) return 0;
+ MEM_writeLE16(ostart, (U16)cSize);
+ op += cSize;
+ }
+
+ ip += segmentSize;
+ assert(op <= oend);
+ { CHECK_V_F(cSize, HUF_compress1X_usingCTable_internal(op, (size_t)(oend-op), ip, segmentSize, CTable, bmi2) );
+ if (cSize == 0 || cSize > 65535) return 0;
+ MEM_writeLE16(ostart+2, (U16)cSize);
+ op += cSize;
+ }
+
+ ip += segmentSize;
+ assert(op <= oend);
+ { CHECK_V_F(cSize, HUF_compress1X_usingCTable_internal(op, (size_t)(oend-op), ip, segmentSize, CTable, bmi2) );
+ if (cSize == 0 || cSize > 65535) return 0;
+ MEM_writeLE16(ostart+4, (U16)cSize);
+ op += cSize;
+ }
+
+ ip += segmentSize;
+ assert(op <= oend);
+ assert(ip <= iend);
+ { CHECK_V_F(cSize, HUF_compress1X_usingCTable_internal(op, (size_t)(oend-op), ip, (size_t)(iend-ip), CTable, bmi2) );
+ if (cSize == 0 || cSize > 65535) return 0;
+ op += cSize;
+ }
+
+ return (size_t)(op-ostart);
+}
+
+size_t HUF_compress4X_usingCTable(void* dst, size_t dstSize, const void* src, size_t srcSize, const HUF_CElt* CTable)
+{
+ return HUF_compress4X_usingCTable_bmi2(dst, dstSize, src, srcSize, CTable, /* bmi2 */ 0);
+}
+
+size_t HUF_compress4X_usingCTable_bmi2(void* dst, size_t dstSize, const void* src, size_t srcSize, const HUF_CElt* CTable, int bmi2)
+{
+ return HUF_compress4X_usingCTable_internal(dst, dstSize, src, srcSize, CTable, bmi2);
+}
+
+typedef enum { HUF_singleStream, HUF_fourStreams } HUF_nbStreams_e;
+
+static size_t HUF_compressCTable_internal(
+ BYTE* const ostart, BYTE* op, BYTE* const oend,
+ const void* src, size_t srcSize,
+ HUF_nbStreams_e nbStreams, const HUF_CElt* CTable, const int bmi2)
+{
+ size_t const cSize = (nbStreams==HUF_singleStream) ?
+ HUF_compress1X_usingCTable_internal(op, (size_t)(oend - op), src, srcSize, CTable, bmi2) :
+ HUF_compress4X_usingCTable_internal(op, (size_t)(oend - op), src, srcSize, CTable, bmi2);
+ if (HUF_isError(cSize)) { return cSize; }
+ if (cSize==0) { return 0; } /* uncompressible */
+ op += cSize;
+ /* check compressibility */
+ assert(op >= ostart);
+ if ((size_t)(op-ostart) >= srcSize-1) { return 0; }
+ return (size_t)(op-ostart);
+}
+
+typedef struct {
+ unsigned count[HUF_SYMBOLVALUE_MAX + 1];
+ HUF_CElt CTable[HUF_CTABLE_SIZE_ST(HUF_SYMBOLVALUE_MAX)];
+ union {
+ HUF_buildCTable_wksp_tables buildCTable_wksp;
+ HUF_WriteCTableWksp writeCTable_wksp;
+ U32 hist_wksp[HIST_WKSP_SIZE_U32];
+ } wksps;
+} HUF_compress_tables_t;
+
+#define SUSPECT_INCOMPRESSIBLE_SAMPLE_SIZE 4096
+#define SUSPECT_INCOMPRESSIBLE_SAMPLE_RATIO 10 /* Must be >= 2 */
+
+/* HUF_compress_internal() :
+ * `workSpace_align4` must be aligned on 4-bytes boundaries,
+ * and occupies the same space as a table of HUF_WORKSPACE_SIZE_U64 unsigned */
+static size_t
+HUF_compress_internal (void* dst, size_t dstSize,
+ const void* src, size_t srcSize,
+ unsigned maxSymbolValue, unsigned huffLog,
+ HUF_nbStreams_e nbStreams,
+ void* workSpace, size_t wkspSize,
+ HUF_CElt* oldHufTable, HUF_repeat* repeat, int preferRepeat,
+ const int bmi2, unsigned suspectUncompressible)
+{
+ HUF_compress_tables_t* const table = (HUF_compress_tables_t*)HUF_alignUpWorkspace(workSpace, &wkspSize, ZSTD_ALIGNOF(size_t));
+ BYTE* const ostart = (BYTE*)dst;
+ BYTE* const oend = ostart + dstSize;
+ BYTE* op = ostart;
+
+ HUF_STATIC_ASSERT(sizeof(*table) + HUF_WORKSPACE_MAX_ALIGNMENT <= HUF_WORKSPACE_SIZE);
+
+ /* checks & inits */
+ if (wkspSize < sizeof(*table)) return ERROR(workSpace_tooSmall);
+ if (!srcSize) return 0; /* Uncompressed */
+ if (!dstSize) return 0; /* cannot fit anything within dst budget */
+ if (srcSize > HUF_BLOCKSIZE_MAX) return ERROR(srcSize_wrong); /* current block size limit */
+ if (huffLog > HUF_TABLELOG_MAX) return ERROR(tableLog_tooLarge);
+ if (maxSymbolValue > HUF_SYMBOLVALUE_MAX) return ERROR(maxSymbolValue_tooLarge);
+ if (!maxSymbolValue) maxSymbolValue = HUF_SYMBOLVALUE_MAX;
+ if (!huffLog) huffLog = HUF_TABLELOG_DEFAULT;
+
+ /* Heuristic : If old table is valid, use it for small inputs */
+ if (preferRepeat && repeat && *repeat == HUF_repeat_valid) {
+ return HUF_compressCTable_internal(ostart, op, oend,
+ src, srcSize,
+ nbStreams, oldHufTable, bmi2);
+ }
+
+ /* If uncompressible data is suspected, do a smaller sampling first */
+ DEBUG_STATIC_ASSERT(SUSPECT_INCOMPRESSIBLE_SAMPLE_RATIO >= 2);
+ if (suspectUncompressible && srcSize >= (SUSPECT_INCOMPRESSIBLE_SAMPLE_SIZE * SUSPECT_INCOMPRESSIBLE_SAMPLE_RATIO)) {
+ size_t largestTotal = 0;
+ { unsigned maxSymbolValueBegin = maxSymbolValue;
+ CHECK_V_F(largestBegin, HIST_count_simple (table->count, &maxSymbolValueBegin, (const BYTE*)src, SUSPECT_INCOMPRESSIBLE_SAMPLE_SIZE) );
+ largestTotal += largestBegin;
+ }
+ { unsigned maxSymbolValueEnd = maxSymbolValue;
+ CHECK_V_F(largestEnd, HIST_count_simple (table->count, &maxSymbolValueEnd, (const BYTE*)src + srcSize - SUSPECT_INCOMPRESSIBLE_SAMPLE_SIZE, SUSPECT_INCOMPRESSIBLE_SAMPLE_SIZE) );
+ largestTotal += largestEnd;
+ }
+ if (largestTotal <= ((2 * SUSPECT_INCOMPRESSIBLE_SAMPLE_SIZE) >> 7)+4) return 0; /* heuristic : probably not compressible enough */
+ }
+
+ /* Scan input and build symbol stats */
+ { CHECK_V_F(largest, HIST_count_wksp (table->count, &maxSymbolValue, (const BYTE*)src, srcSize, table->wksps.hist_wksp, sizeof(table->wksps.hist_wksp)) );
+ if (largest == srcSize) { *ostart = ((const BYTE*)src)[0]; return 1; } /* single symbol, rle */
+ if (largest <= (srcSize >> 7)+4) return 0; /* heuristic : probably not compressible enough */
+ }
+
+ /* Check validity of previous table */
+ if ( repeat
+ && *repeat == HUF_repeat_check
+ && !HUF_validateCTable(oldHufTable, table->count, maxSymbolValue)) {
+ *repeat = HUF_repeat_none;
+ }
+ /* Heuristic : use existing table for small inputs */
+ if (preferRepeat && repeat && *repeat != HUF_repeat_none) {
+ return HUF_compressCTable_internal(ostart, op, oend,
+ src, srcSize,
+ nbStreams, oldHufTable, bmi2);
+ }
+
+ /* Build Huffman Tree */
+ huffLog = HUF_optimalTableLog(huffLog, srcSize, maxSymbolValue);
+ { size_t const maxBits = HUF_buildCTable_wksp(table->CTable, table->count,
+ maxSymbolValue, huffLog,
+ &table->wksps.buildCTable_wksp, sizeof(table->wksps.buildCTable_wksp));
+ CHECK_F(maxBits);
+ huffLog = (U32)maxBits;
+ }
+ /* Zero unused symbols in CTable, so we can check it for validity */
+ {
+ size_t const ctableSize = HUF_CTABLE_SIZE_ST(maxSymbolValue);
+ size_t const unusedSize = sizeof(table->CTable) - ctableSize * sizeof(HUF_CElt);
+ ZSTD_memset(table->CTable + ctableSize, 0, unusedSize);
+ }
+
+ /* Write table description header */
+ { CHECK_V_F(hSize, HUF_writeCTable_wksp(op, dstSize, table->CTable, maxSymbolValue, huffLog,
+ &table->wksps.writeCTable_wksp, sizeof(table->wksps.writeCTable_wksp)) );
+ /* Check if using previous huffman table is beneficial */
+ if (repeat && *repeat != HUF_repeat_none) {
+ size_t const oldSize = HUF_estimateCompressedSize(oldHufTable, table->count, maxSymbolValue);
+ size_t const newSize = HUF_estimateCompressedSize(table->CTable, table->count, maxSymbolValue);
+ if (oldSize <= hSize + newSize || hSize + 12 >= srcSize) {
+ return HUF_compressCTable_internal(ostart, op, oend,
+ src, srcSize,
+ nbStreams, oldHufTable, bmi2);
+ } }
+
+ /* Use the new huffman table */
+ if (hSize + 12ul >= srcSize) { return 0; }
+ op += hSize;
+ if (repeat) { *repeat = HUF_repeat_none; }
+ if (oldHufTable)
+ ZSTD_memcpy(oldHufTable, table->CTable, sizeof(table->CTable)); /* Save new table */
+ }
+ return HUF_compressCTable_internal(ostart, op, oend,
+ src, srcSize,
+ nbStreams, table->CTable, bmi2);
+}
+
+
+size_t HUF_compress1X_wksp (void* dst, size_t dstSize,
+ const void* src, size_t srcSize,
+ unsigned maxSymbolValue, unsigned huffLog,
+ void* workSpace, size_t wkspSize)
+{
+ return HUF_compress_internal(dst, dstSize, src, srcSize,
+ maxSymbolValue, huffLog, HUF_singleStream,
+ workSpace, wkspSize,
+ NULL, NULL, 0, 0 /*bmi2*/, 0);
+}
+
+size_t HUF_compress1X_repeat (void* dst, size_t dstSize,
+ const void* src, size_t srcSize,
+ unsigned maxSymbolValue, unsigned huffLog,
+ void* workSpace, size_t wkspSize,
+ HUF_CElt* hufTable, HUF_repeat* repeat, int preferRepeat,
+ int bmi2, unsigned suspectUncompressible)
+{
+ return HUF_compress_internal(dst, dstSize, src, srcSize,
+ maxSymbolValue, huffLog, HUF_singleStream,
+ workSpace, wkspSize, hufTable,
+ repeat, preferRepeat, bmi2, suspectUncompressible);
+}
+
+/* HUF_compress4X_repeat():
+ * compress input using 4 streams.
+ * provide workspace to generate compression tables */
+size_t HUF_compress4X_wksp (void* dst, size_t dstSize,
+ const void* src, size_t srcSize,
+ unsigned maxSymbolValue, unsigned huffLog,
+ void* workSpace, size_t wkspSize)
+{
+ return HUF_compress_internal(dst, dstSize, src, srcSize,
+ maxSymbolValue, huffLog, HUF_fourStreams,
+ workSpace, wkspSize,
+ NULL, NULL, 0, 0 /*bmi2*/, 0);
+}
+
+/* HUF_compress4X_repeat():
+ * compress input using 4 streams.
+ * consider skipping quickly
+ * re-use an existing huffman compression table */
+size_t HUF_compress4X_repeat (void* dst, size_t dstSize,
+ const void* src, size_t srcSize,
+ unsigned maxSymbolValue, unsigned huffLog,
+ void* workSpace, size_t wkspSize,
+ HUF_CElt* hufTable, HUF_repeat* repeat, int preferRepeat, int bmi2, unsigned suspectUncompressible)
+{
+ return HUF_compress_internal(dst, dstSize, src, srcSize,
+ maxSymbolValue, huffLog, HUF_fourStreams,
+ workSpace, wkspSize,
+ hufTable, repeat, preferRepeat, bmi2, suspectUncompressible);
+}
+
diff --git a/lib/zstd/compress/zstd_compress.c b/lib/zstd/compress/zstd_compress.c
new file mode 100644
index 0000000000..f620cafca6
--- /dev/null
+++ b/lib/zstd/compress/zstd_compress.c
@@ -0,0 +1,6127 @@
+/*
+ * Copyright (c) Yann Collet, Facebook, Inc.
+ * All rights reserved.
+ *
+ * This source code is licensed under both the BSD-style license (found in the
+ * LICENSE file in the root directory of this source tree) and the GPLv2 (found
+ * in the COPYING file in the root directory of this source tree).
+ * You may select, at your option, one of the above-listed licenses.
+ */
+
+/*-*************************************
+* Dependencies
+***************************************/
+#include "../common/zstd_deps.h" /* INT_MAX, ZSTD_memset, ZSTD_memcpy */
+#include "../common/mem.h"
+#include "hist.h" /* HIST_countFast_wksp */
+#define FSE_STATIC_LINKING_ONLY /* FSE_encodeSymbol */
+#include "../common/fse.h"
+#define HUF_STATIC_LINKING_ONLY
+#include "../common/huf.h"
+#include "zstd_compress_internal.h"
+#include "zstd_compress_sequences.h"
+#include "zstd_compress_literals.h"
+#include "zstd_fast.h"
+#include "zstd_double_fast.h"
+#include "zstd_lazy.h"
+#include "zstd_opt.h"
+#include "zstd_ldm.h"
+#include "zstd_compress_superblock.h"
+
+/* ***************************************************************
+* Tuning parameters
+*****************************************************************/
+/*!
+ * COMPRESS_HEAPMODE :
+ * Select how default decompression function ZSTD_compress() allocates its context,
+ * on stack (0, default), or into heap (1).
+ * Note that functions with explicit context such as ZSTD_compressCCtx() are unaffected.
+ */
+
+/*!
+ * ZSTD_HASHLOG3_MAX :
+ * Maximum size of the hash table dedicated to find 3-bytes matches,
+ * in log format, aka 17 => 1 << 17 == 128Ki positions.
+ * This structure is only used in zstd_opt.
+ * Since allocation is centralized for all strategies, it has to be known here.
+ * The actual (selected) size of the hash table is then stored in ZSTD_matchState_t.hashLog3,
+ * so that zstd_opt.c doesn't need to know about this constant.
+ */
+#ifndef ZSTD_HASHLOG3_MAX
+# define ZSTD_HASHLOG3_MAX 17
+#endif
+
+/*-*************************************
+* Helper functions
+***************************************/
+/* ZSTD_compressBound()
+ * Note that the result from this function is only compatible with the "normal"
+ * full-block strategy.
+ * When there are a lot of small blocks due to frequent flush in streaming mode
+ * the overhead of headers can make the compressed data to be larger than the
+ * return value of ZSTD_compressBound().
+ */
+size_t ZSTD_compressBound(size_t srcSize) {
+ return ZSTD_COMPRESSBOUND(srcSize);
+}
+
+
+/*-*************************************
+* Context memory management
+***************************************/
+struct ZSTD_CDict_s {
+ const void* dictContent;
+ size_t dictContentSize;
+ ZSTD_dictContentType_e dictContentType; /* The dictContentType the CDict was created with */
+ U32* entropyWorkspace; /* entropy workspace of HUF_WORKSPACE_SIZE bytes */
+ ZSTD_cwksp workspace;
+ ZSTD_matchState_t matchState;
+ ZSTD_compressedBlockState_t cBlockState;
+ ZSTD_customMem customMem;
+ U32 dictID;
+ int compressionLevel; /* 0 indicates that advanced API was used to select CDict params */
+ ZSTD_paramSwitch_e useRowMatchFinder; /* Indicates whether the CDict was created with params that would use
+ * row-based matchfinder. Unless the cdict is reloaded, we will use
+ * the same greedy/lazy matchfinder at compression time.
+ */
+}; /* typedef'd to ZSTD_CDict within "zstd.h" */
+
+ZSTD_CCtx* ZSTD_createCCtx(void)
+{
+ return ZSTD_createCCtx_advanced(ZSTD_defaultCMem);
+}
+
+static void ZSTD_initCCtx(ZSTD_CCtx* cctx, ZSTD_customMem memManager)
+{
+ assert(cctx != NULL);
+ ZSTD_memset(cctx, 0, sizeof(*cctx));
+ cctx->customMem = memManager;
+ cctx->bmi2 = ZSTD_cpuSupportsBmi2();
+ { size_t const err = ZSTD_CCtx_reset(cctx, ZSTD_reset_parameters);
+ assert(!ZSTD_isError(err));
+ (void)err;
+ }
+}
+
+ZSTD_CCtx* ZSTD_createCCtx_advanced(ZSTD_customMem customMem)
+{
+ ZSTD_STATIC_ASSERT(zcss_init==0);
+ ZSTD_STATIC_ASSERT(ZSTD_CONTENTSIZE_UNKNOWN==(0ULL - 1));
+ if ((!customMem.customAlloc) ^ (!customMem.customFree)) return NULL;
+ { ZSTD_CCtx* const cctx = (ZSTD_CCtx*)ZSTD_customMalloc(sizeof(ZSTD_CCtx), customMem);
+ if (!cctx) return NULL;
+ ZSTD_initCCtx(cctx, customMem);
+ return cctx;
+ }
+}
+
+ZSTD_CCtx* ZSTD_initStaticCCtx(void* workspace, size_t workspaceSize)
+{
+ ZSTD_cwksp ws;
+ ZSTD_CCtx* cctx;
+ if (workspaceSize <= sizeof(ZSTD_CCtx)) return NULL; /* minimum size */
+ if ((size_t)workspace & 7) return NULL; /* must be 8-aligned */
+ ZSTD_cwksp_init(&ws, workspace, workspaceSize, ZSTD_cwksp_static_alloc);
+
+ cctx = (ZSTD_CCtx*)ZSTD_cwksp_reserve_object(&ws, sizeof(ZSTD_CCtx));
+ if (cctx == NULL) return NULL;
+
+ ZSTD_memset(cctx, 0, sizeof(ZSTD_CCtx));
+ ZSTD_cwksp_move(&cctx->workspace, &ws);
+ cctx->staticSize = workspaceSize;
+
+ /* statically sized space. entropyWorkspace never moves (but prev/next block swap places) */
+ if (!ZSTD_cwksp_check_available(&cctx->workspace, ENTROPY_WORKSPACE_SIZE + 2 * sizeof(ZSTD_compressedBlockState_t))) return NULL;
+ cctx->blockState.prevCBlock = (ZSTD_compressedBlockState_t*)ZSTD_cwksp_reserve_object(&cctx->workspace, sizeof(ZSTD_compressedBlockState_t));
+ cctx->blockState.nextCBlock = (ZSTD_compressedBlockState_t*)ZSTD_cwksp_reserve_object(&cctx->workspace, sizeof(ZSTD_compressedBlockState_t));
+ cctx->entropyWorkspace = (U32*)ZSTD_cwksp_reserve_object(&cctx->workspace, ENTROPY_WORKSPACE_SIZE);
+ cctx->bmi2 = ZSTD_cpuid_bmi2(ZSTD_cpuid());
+ return cctx;
+}
+
+/*
+ * Clears and frees all of the dictionaries in the CCtx.
+ */
+static void ZSTD_clearAllDicts(ZSTD_CCtx* cctx)
+{
+ ZSTD_customFree(cctx->localDict.dictBuffer, cctx->customMem);
+ ZSTD_freeCDict(cctx->localDict.cdict);
+ ZSTD_memset(&cctx->localDict, 0, sizeof(cctx->localDict));
+ ZSTD_memset(&cctx->prefixDict, 0, sizeof(cctx->prefixDict));
+ cctx->cdict = NULL;
+}
+
+static size_t ZSTD_sizeof_localDict(ZSTD_localDict dict)
+{
+ size_t const bufferSize = dict.dictBuffer != NULL ? dict.dictSize : 0;
+ size_t const cdictSize = ZSTD_sizeof_CDict(dict.cdict);
+ return bufferSize + cdictSize;
+}
+
+static void ZSTD_freeCCtxContent(ZSTD_CCtx* cctx)
+{
+ assert(cctx != NULL);
+ assert(cctx->staticSize == 0);
+ ZSTD_clearAllDicts(cctx);
+ ZSTD_cwksp_free(&cctx->workspace, cctx->customMem);
+}
+
+size_t ZSTD_freeCCtx(ZSTD_CCtx* cctx)
+{
+ if (cctx==NULL) return 0; /* support free on NULL */
+ RETURN_ERROR_IF(cctx->staticSize, memory_allocation,
+ "not compatible with static CCtx");
+ {
+ int cctxInWorkspace = ZSTD_cwksp_owns_buffer(&cctx->workspace, cctx);
+ ZSTD_freeCCtxContent(cctx);
+ if (!cctxInWorkspace) {
+ ZSTD_customFree(cctx, cctx->customMem);
+ }
+ }
+ return 0;
+}
+
+
+static size_t ZSTD_sizeof_mtctx(const ZSTD_CCtx* cctx)
+{
+ (void)cctx;
+ return 0;
+}
+
+
+size_t ZSTD_sizeof_CCtx(const ZSTD_CCtx* cctx)
+{
+ if (cctx==NULL) return 0; /* support sizeof on NULL */
+ /* cctx may be in the workspace */
+ return (cctx->workspace.workspace == cctx ? 0 : sizeof(*cctx))
+ + ZSTD_cwksp_sizeof(&cctx->workspace)
+ + ZSTD_sizeof_localDict(cctx->localDict)
+ + ZSTD_sizeof_mtctx(cctx);
+}
+
+size_t ZSTD_sizeof_CStream(const ZSTD_CStream* zcs)
+{
+ return ZSTD_sizeof_CCtx(zcs); /* same object */
+}
+
+/* private API call, for dictBuilder only */
+const seqStore_t* ZSTD_getSeqStore(const ZSTD_CCtx* ctx) { return &(ctx->seqStore); }
+
+/* Returns true if the strategy supports using a row based matchfinder */
+static int ZSTD_rowMatchFinderSupported(const ZSTD_strategy strategy) {
+ return (strategy >= ZSTD_greedy && strategy <= ZSTD_lazy2);
+}
+
+/* Returns true if the strategy and useRowMatchFinder mode indicate that we will use the row based matchfinder
+ * for this compression.
+ */
+static int ZSTD_rowMatchFinderUsed(const ZSTD_strategy strategy, const ZSTD_paramSwitch_e mode) {
+ assert(mode != ZSTD_ps_auto);
+ return ZSTD_rowMatchFinderSupported(strategy) && (mode == ZSTD_ps_enable);
+}
+
+/* Returns row matchfinder usage given an initial mode and cParams */
+static ZSTD_paramSwitch_e ZSTD_resolveRowMatchFinderMode(ZSTD_paramSwitch_e mode,
+ const ZSTD_compressionParameters* const cParams) {
+#if defined(ZSTD_ARCH_X86_SSE2) || defined(ZSTD_ARCH_ARM_NEON)
+ int const kHasSIMD128 = 1;
+#else
+ int const kHasSIMD128 = 0;
+#endif
+ if (mode != ZSTD_ps_auto) return mode; /* if requested enabled, but no SIMD, we still will use row matchfinder */
+ mode = ZSTD_ps_disable;
+ if (!ZSTD_rowMatchFinderSupported(cParams->strategy)) return mode;
+ if (kHasSIMD128) {
+ if (cParams->windowLog > 14) mode = ZSTD_ps_enable;
+ } else {
+ if (cParams->windowLog > 17) mode = ZSTD_ps_enable;
+ }
+ return mode;
+}
+
+/* Returns block splitter usage (generally speaking, when using slower/stronger compression modes) */
+static ZSTD_paramSwitch_e ZSTD_resolveBlockSplitterMode(ZSTD_paramSwitch_e mode,
+ const ZSTD_compressionParameters* const cParams) {
+ if (mode != ZSTD_ps_auto) return mode;
+ return (cParams->strategy >= ZSTD_btopt && cParams->windowLog >= 17) ? ZSTD_ps_enable : ZSTD_ps_disable;
+}
+
+/* Returns 1 if the arguments indicate that we should allocate a chainTable, 0 otherwise */
+static int ZSTD_allocateChainTable(const ZSTD_strategy strategy,
+ const ZSTD_paramSwitch_e useRowMatchFinder,
+ const U32 forDDSDict) {
+ assert(useRowMatchFinder != ZSTD_ps_auto);
+ /* We always should allocate a chaintable if we are allocating a matchstate for a DDS dictionary matchstate.
+ * We do not allocate a chaintable if we are using ZSTD_fast, or are using the row-based matchfinder.
+ */
+ return forDDSDict || ((strategy != ZSTD_fast) && !ZSTD_rowMatchFinderUsed(strategy, useRowMatchFinder));
+}
+
+/* Returns 1 if compression parameters are such that we should
+ * enable long distance matching (wlog >= 27, strategy >= btopt).
+ * Returns 0 otherwise.
+ */
+static ZSTD_paramSwitch_e ZSTD_resolveEnableLdm(ZSTD_paramSwitch_e mode,
+ const ZSTD_compressionParameters* const cParams) {
+ if (mode != ZSTD_ps_auto) return mode;
+ return (cParams->strategy >= ZSTD_btopt && cParams->windowLog >= 27) ? ZSTD_ps_enable : ZSTD_ps_disable;
+}
+
+static ZSTD_CCtx_params ZSTD_makeCCtxParamsFromCParams(
+ ZSTD_compressionParameters cParams)
+{
+ ZSTD_CCtx_params cctxParams;
+ /* should not matter, as all cParams are presumed properly defined */
+ ZSTD_CCtxParams_init(&cctxParams, ZSTD_CLEVEL_DEFAULT);
+ cctxParams.cParams = cParams;
+
+ /* Adjust advanced params according to cParams */
+ cctxParams.ldmParams.enableLdm = ZSTD_resolveEnableLdm(cctxParams.ldmParams.enableLdm, &cParams);
+ if (cctxParams.ldmParams.enableLdm == ZSTD_ps_enable) {
+ ZSTD_ldm_adjustParameters(&cctxParams.ldmParams, &cParams);
+ assert(cctxParams.ldmParams.hashLog >= cctxParams.ldmParams.bucketSizeLog);
+ assert(cctxParams.ldmParams.hashRateLog < 32);
+ }
+ cctxParams.useBlockSplitter = ZSTD_resolveBlockSplitterMode(cctxParams.useBlockSplitter, &cParams);
+ cctxParams.useRowMatchFinder = ZSTD_resolveRowMatchFinderMode(cctxParams.useRowMatchFinder, &cParams);
+ assert(!ZSTD_checkCParams(cParams));
+ return cctxParams;
+}
+
+static ZSTD_CCtx_params* ZSTD_createCCtxParams_advanced(
+ ZSTD_customMem customMem)
+{
+ ZSTD_CCtx_params* params;
+ if ((!customMem.customAlloc) ^ (!customMem.customFree)) return NULL;
+ params = (ZSTD_CCtx_params*)ZSTD_customCalloc(
+ sizeof(ZSTD_CCtx_params), customMem);
+ if (!params) { return NULL; }
+ ZSTD_CCtxParams_init(params, ZSTD_CLEVEL_DEFAULT);
+ params->customMem = customMem;
+ return params;
+}
+
+ZSTD_CCtx_params* ZSTD_createCCtxParams(void)
+{
+ return ZSTD_createCCtxParams_advanced(ZSTD_defaultCMem);
+}
+
+size_t ZSTD_freeCCtxParams(ZSTD_CCtx_params* params)
+{
+ if (params == NULL) { return 0; }
+ ZSTD_customFree(params, params->customMem);
+ return 0;
+}
+
+size_t ZSTD_CCtxParams_reset(ZSTD_CCtx_params* params)
+{
+ return ZSTD_CCtxParams_init(params, ZSTD_CLEVEL_DEFAULT);
+}
+
+size_t ZSTD_CCtxParams_init(ZSTD_CCtx_params* cctxParams, int compressionLevel) {
+ RETURN_ERROR_IF(!cctxParams, GENERIC, "NULL pointer!");
+ ZSTD_memset(cctxParams, 0, sizeof(*cctxParams));
+ cctxParams->compressionLevel = compressionLevel;
+ cctxParams->fParams.contentSizeFlag = 1;
+ return 0;
+}
+
+#define ZSTD_NO_CLEVEL 0
+
+/*
+ * Initializes the cctxParams from params and compressionLevel.
+ * @param compressionLevel If params are derived from a compression level then that compression level, otherwise ZSTD_NO_CLEVEL.
+ */
+static void ZSTD_CCtxParams_init_internal(ZSTD_CCtx_params* cctxParams, ZSTD_parameters const* params, int compressionLevel)
+{
+ assert(!ZSTD_checkCParams(params->cParams));
+ ZSTD_memset(cctxParams, 0, sizeof(*cctxParams));
+ cctxParams->cParams = params->cParams;
+ cctxParams->fParams = params->fParams;
+ /* Should not matter, as all cParams are presumed properly defined.
+ * But, set it for tracing anyway.
+ */
+ cctxParams->compressionLevel = compressionLevel;
+ cctxParams->useRowMatchFinder = ZSTD_resolveRowMatchFinderMode(cctxParams->useRowMatchFinder, &params->cParams);
+ cctxParams->useBlockSplitter = ZSTD_resolveBlockSplitterMode(cctxParams->useBlockSplitter, &params->cParams);
+ cctxParams->ldmParams.enableLdm = ZSTD_resolveEnableLdm(cctxParams->ldmParams.enableLdm, &params->cParams);
+ DEBUGLOG(4, "ZSTD_CCtxParams_init_internal: useRowMatchFinder=%d, useBlockSplitter=%d ldm=%d",
+ cctxParams->useRowMatchFinder, cctxParams->useBlockSplitter, cctxParams->ldmParams.enableLdm);
+}
+
+size_t ZSTD_CCtxParams_init_advanced(ZSTD_CCtx_params* cctxParams, ZSTD_parameters params)
+{
+ RETURN_ERROR_IF(!cctxParams, GENERIC, "NULL pointer!");
+ FORWARD_IF_ERROR( ZSTD_checkCParams(params.cParams) , "");
+ ZSTD_CCtxParams_init_internal(cctxParams, &params, ZSTD_NO_CLEVEL);
+ return 0;
+}
+
+/*
+ * Sets cctxParams' cParams and fParams from params, but otherwise leaves them alone.
+ * @param param Validated zstd parameters.
+ */
+static void ZSTD_CCtxParams_setZstdParams(
+ ZSTD_CCtx_params* cctxParams, const ZSTD_parameters* params)
+{
+ assert(!ZSTD_checkCParams(params->cParams));
+ cctxParams->cParams = params->cParams;
+ cctxParams->fParams = params->fParams;
+ /* Should not matter, as all cParams are presumed properly defined.
+ * But, set it for tracing anyway.
+ */
+ cctxParams->compressionLevel = ZSTD_NO_CLEVEL;
+}
+
+ZSTD_bounds ZSTD_cParam_getBounds(ZSTD_cParameter param)
+{
+ ZSTD_bounds bounds = { 0, 0, 0 };
+
+ switch(param)
+ {
+ case ZSTD_c_compressionLevel:
+ bounds.lowerBound = ZSTD_minCLevel();
+ bounds.upperBound = ZSTD_maxCLevel();
+ return bounds;
+
+ case ZSTD_c_windowLog:
+ bounds.lowerBound = ZSTD_WINDOWLOG_MIN;
+ bounds.upperBound = ZSTD_WINDOWLOG_MAX;
+ return bounds;
+
+ case ZSTD_c_hashLog:
+ bounds.lowerBound = ZSTD_HASHLOG_MIN;
+ bounds.upperBound = ZSTD_HASHLOG_MAX;
+ return bounds;
+
+ case ZSTD_c_chainLog:
+ bounds.lowerBound = ZSTD_CHAINLOG_MIN;
+ bounds.upperBound = ZSTD_CHAINLOG_MAX;
+ return bounds;
+
+ case ZSTD_c_searchLog:
+ bounds.lowerBound = ZSTD_SEARCHLOG_MIN;
+ bounds.upperBound = ZSTD_SEARCHLOG_MAX;
+ return bounds;
+
+ case ZSTD_c_minMatch:
+ bounds.lowerBound = ZSTD_MINMATCH_MIN;
+ bounds.upperBound = ZSTD_MINMATCH_MAX;
+ return bounds;
+
+ case ZSTD_c_targetLength:
+ bounds.lowerBound = ZSTD_TARGETLENGTH_MIN;
+ bounds.upperBound = ZSTD_TARGETLENGTH_MAX;
+ return bounds;
+
+ case ZSTD_c_strategy:
+ bounds.lowerBound = ZSTD_STRATEGY_MIN;
+ bounds.upperBound = ZSTD_STRATEGY_MAX;
+ return bounds;
+
+ case ZSTD_c_contentSizeFlag:
+ bounds.lowerBound = 0;
+ bounds.upperBound = 1;
+ return bounds;
+
+ case ZSTD_c_checksumFlag:
+ bounds.lowerBound = 0;
+ bounds.upperBound = 1;
+ return bounds;
+
+ case ZSTD_c_dictIDFlag:
+ bounds.lowerBound = 0;
+ bounds.upperBound = 1;
+ return bounds;
+
+ case ZSTD_c_nbWorkers:
+ bounds.lowerBound = 0;
+ bounds.upperBound = 0;
+ return bounds;
+
+ case ZSTD_c_jobSize:
+ bounds.lowerBound = 0;
+ bounds.upperBound = 0;
+ return bounds;
+
+ case ZSTD_c_overlapLog:
+ bounds.lowerBound = 0;
+ bounds.upperBound = 0;
+ return bounds;
+
+ case ZSTD_c_enableDedicatedDictSearch:
+ bounds.lowerBound = 0;
+ bounds.upperBound = 1;
+ return bounds;
+
+ case ZSTD_c_enableLongDistanceMatching:
+ bounds.lowerBound = 0;
+ bounds.upperBound = 1;
+ return bounds;
+
+ case ZSTD_c_ldmHashLog:
+ bounds.lowerBound = ZSTD_LDM_HASHLOG_MIN;
+ bounds.upperBound = ZSTD_LDM_HASHLOG_MAX;
+ return bounds;
+
+ case ZSTD_c_ldmMinMatch:
+ bounds.lowerBound = ZSTD_LDM_MINMATCH_MIN;
+ bounds.upperBound = ZSTD_LDM_MINMATCH_MAX;
+ return bounds;
+
+ case ZSTD_c_ldmBucketSizeLog:
+ bounds.lowerBound = ZSTD_LDM_BUCKETSIZELOG_MIN;
+ bounds.upperBound = ZSTD_LDM_BUCKETSIZELOG_MAX;
+ return bounds;
+
+ case ZSTD_c_ldmHashRateLog:
+ bounds.lowerBound = ZSTD_LDM_HASHRATELOG_MIN;
+ bounds.upperBound = ZSTD_LDM_HASHRATELOG_MAX;
+ return bounds;
+
+ /* experimental parameters */
+ case ZSTD_c_rsyncable:
+ bounds.lowerBound = 0;
+ bounds.upperBound = 1;
+ return bounds;
+
+ case ZSTD_c_forceMaxWindow :
+ bounds.lowerBound = 0;
+ bounds.upperBound = 1;
+ return bounds;
+
+ case ZSTD_c_format:
+ ZSTD_STATIC_ASSERT(ZSTD_f_zstd1 < ZSTD_f_zstd1_magicless);
+ bounds.lowerBound = ZSTD_f_zstd1;
+ bounds.upperBound = ZSTD_f_zstd1_magicless; /* note : how to ensure at compile time that this is the highest value enum ? */
+ return bounds;
+
+ case ZSTD_c_forceAttachDict:
+ ZSTD_STATIC_ASSERT(ZSTD_dictDefaultAttach < ZSTD_dictForceLoad);
+ bounds.lowerBound = ZSTD_dictDefaultAttach;
+ bounds.upperBound = ZSTD_dictForceLoad; /* note : how to ensure at compile time that this is the highest value enum ? */
+ return bounds;
+
+ case ZSTD_c_literalCompressionMode:
+ ZSTD_STATIC_ASSERT(ZSTD_ps_auto < ZSTD_ps_enable && ZSTD_ps_enable < ZSTD_ps_disable);
+ bounds.lowerBound = (int)ZSTD_ps_auto;
+ bounds.upperBound = (int)ZSTD_ps_disable;
+ return bounds;
+
+ case ZSTD_c_targetCBlockSize:
+ bounds.lowerBound = ZSTD_TARGETCBLOCKSIZE_MIN;
+ bounds.upperBound = ZSTD_TARGETCBLOCKSIZE_MAX;
+ return bounds;
+
+ case ZSTD_c_srcSizeHint:
+ bounds.lowerBound = ZSTD_SRCSIZEHINT_MIN;
+ bounds.upperBound = ZSTD_SRCSIZEHINT_MAX;
+ return bounds;
+
+ case ZSTD_c_stableInBuffer:
+ case ZSTD_c_stableOutBuffer:
+ bounds.lowerBound = (int)ZSTD_bm_buffered;
+ bounds.upperBound = (int)ZSTD_bm_stable;
+ return bounds;
+
+ case ZSTD_c_blockDelimiters:
+ bounds.lowerBound = (int)ZSTD_sf_noBlockDelimiters;
+ bounds.upperBound = (int)ZSTD_sf_explicitBlockDelimiters;
+ return bounds;
+
+ case ZSTD_c_validateSequences:
+ bounds.lowerBound = 0;
+ bounds.upperBound = 1;
+ return bounds;
+
+ case ZSTD_c_useBlockSplitter:
+ bounds.lowerBound = (int)ZSTD_ps_auto;
+ bounds.upperBound = (int)ZSTD_ps_disable;
+ return bounds;
+
+ case ZSTD_c_useRowMatchFinder:
+ bounds.lowerBound = (int)ZSTD_ps_auto;
+ bounds.upperBound = (int)ZSTD_ps_disable;
+ return bounds;
+
+ case ZSTD_c_deterministicRefPrefix:
+ bounds.lowerBound = 0;
+ bounds.upperBound = 1;
+ return bounds;
+
+ default:
+ bounds.error = ERROR(parameter_unsupported);
+ return bounds;
+ }
+}
+
+/* ZSTD_cParam_clampBounds:
+ * Clamps the value into the bounded range.
+ */
+static size_t ZSTD_cParam_clampBounds(ZSTD_cParameter cParam, int* value)
+{
+ ZSTD_bounds const bounds = ZSTD_cParam_getBounds(cParam);
+ if (ZSTD_isError(bounds.error)) return bounds.error;
+ if (*value < bounds.lowerBound) *value = bounds.lowerBound;
+ if (*value > bounds.upperBound) *value = bounds.upperBound;
+ return 0;
+}
+
+#define BOUNDCHECK(cParam, val) { \
+ RETURN_ERROR_IF(!ZSTD_cParam_withinBounds(cParam,val), \
+ parameter_outOfBound, "Param out of bounds"); \
+}
+
+
+static int ZSTD_isUpdateAuthorized(ZSTD_cParameter param)
+{
+ switch(param)
+ {
+ case ZSTD_c_compressionLevel:
+ case ZSTD_c_hashLog:
+ case ZSTD_c_chainLog:
+ case ZSTD_c_searchLog:
+ case ZSTD_c_minMatch:
+ case ZSTD_c_targetLength:
+ case ZSTD_c_strategy:
+ return 1;
+
+ case ZSTD_c_format:
+ case ZSTD_c_windowLog:
+ case ZSTD_c_contentSizeFlag:
+ case ZSTD_c_checksumFlag:
+ case ZSTD_c_dictIDFlag:
+ case ZSTD_c_forceMaxWindow :
+ case ZSTD_c_nbWorkers:
+ case ZSTD_c_jobSize:
+ case ZSTD_c_overlapLog:
+ case ZSTD_c_rsyncable:
+ case ZSTD_c_enableDedicatedDictSearch:
+ case ZSTD_c_enableLongDistanceMatching:
+ case ZSTD_c_ldmHashLog:
+ case ZSTD_c_ldmMinMatch:
+ case ZSTD_c_ldmBucketSizeLog:
+ case ZSTD_c_ldmHashRateLog:
+ case ZSTD_c_forceAttachDict:
+ case ZSTD_c_literalCompressionMode:
+ case ZSTD_c_targetCBlockSize:
+ case ZSTD_c_srcSizeHint:
+ case ZSTD_c_stableInBuffer:
+ case ZSTD_c_stableOutBuffer:
+ case ZSTD_c_blockDelimiters:
+ case ZSTD_c_validateSequences:
+ case ZSTD_c_useBlockSplitter:
+ case ZSTD_c_useRowMatchFinder:
+ case ZSTD_c_deterministicRefPrefix:
+ default:
+ return 0;
+ }
+}
+
+size_t ZSTD_CCtx_setParameter(ZSTD_CCtx* cctx, ZSTD_cParameter param, int value)
+{
+ DEBUGLOG(4, "ZSTD_CCtx_setParameter (%i, %i)", (int)param, value);
+ if (cctx->streamStage != zcss_init) {
+ if (ZSTD_isUpdateAuthorized(param)) {
+ cctx->cParamsChanged = 1;
+ } else {
+ RETURN_ERROR(stage_wrong, "can only set params in ctx init stage");
+ } }
+
+ switch(param)
+ {
+ case ZSTD_c_nbWorkers:
+ RETURN_ERROR_IF((value!=0) && cctx->staticSize, parameter_unsupported,
+ "MT not compatible with static alloc");
+ break;
+
+ case ZSTD_c_compressionLevel:
+ case ZSTD_c_windowLog:
+ case ZSTD_c_hashLog:
+ case ZSTD_c_chainLog:
+ case ZSTD_c_searchLog:
+ case ZSTD_c_minMatch:
+ case ZSTD_c_targetLength:
+ case ZSTD_c_strategy:
+ case ZSTD_c_ldmHashRateLog:
+ case ZSTD_c_format:
+ case ZSTD_c_contentSizeFlag:
+ case ZSTD_c_checksumFlag:
+ case ZSTD_c_dictIDFlag:
+ case ZSTD_c_forceMaxWindow:
+ case ZSTD_c_forceAttachDict:
+ case ZSTD_c_literalCompressionMode:
+ case ZSTD_c_jobSize:
+ case ZSTD_c_overlapLog:
+ case ZSTD_c_rsyncable:
+ case ZSTD_c_enableDedicatedDictSearch:
+ case ZSTD_c_enableLongDistanceMatching:
+ case ZSTD_c_ldmHashLog:
+ case ZSTD_c_ldmMinMatch:
+ case ZSTD_c_ldmBucketSizeLog:
+ case ZSTD_c_targetCBlockSize:
+ case ZSTD_c_srcSizeHint:
+ case ZSTD_c_stableInBuffer:
+ case ZSTD_c_stableOutBuffer:
+ case ZSTD_c_blockDelimiters:
+ case ZSTD_c_validateSequences:
+ case ZSTD_c_useBlockSplitter:
+ case ZSTD_c_useRowMatchFinder:
+ case ZSTD_c_deterministicRefPrefix:
+ break;
+
+ default: RETURN_ERROR(parameter_unsupported, "unknown parameter");
+ }
+ return ZSTD_CCtxParams_setParameter(&cctx->requestedParams, param, value);
+}
+
+size_t ZSTD_CCtxParams_setParameter(ZSTD_CCtx_params* CCtxParams,
+ ZSTD_cParameter param, int value)
+{
+ DEBUGLOG(4, "ZSTD_CCtxParams_setParameter (%i, %i)", (int)param, value);
+ switch(param)
+ {
+ case ZSTD_c_format :
+ BOUNDCHECK(ZSTD_c_format, value);
+ CCtxParams->format = (ZSTD_format_e)value;
+ return (size_t)CCtxParams->format;
+
+ case ZSTD_c_compressionLevel : {
+ FORWARD_IF_ERROR(ZSTD_cParam_clampBounds(param, &value), "");
+ if (value == 0)
+ CCtxParams->compressionLevel = ZSTD_CLEVEL_DEFAULT; /* 0 == default */
+ else
+ CCtxParams->compressionLevel = value;
+ if (CCtxParams->compressionLevel >= 0) return (size_t)CCtxParams->compressionLevel;
+ return 0; /* return type (size_t) cannot represent negative values */
+ }
+
+ case ZSTD_c_windowLog :
+ if (value!=0) /* 0 => use default */
+ BOUNDCHECK(ZSTD_c_windowLog, value);
+ CCtxParams->cParams.windowLog = (U32)value;
+ return CCtxParams->cParams.windowLog;
+
+ case ZSTD_c_hashLog :
+ if (value!=0) /* 0 => use default */
+ BOUNDCHECK(ZSTD_c_hashLog, value);
+ CCtxParams->cParams.hashLog = (U32)value;
+ return CCtxParams->cParams.hashLog;
+
+ case ZSTD_c_chainLog :
+ if (value!=0) /* 0 => use default */
+ BOUNDCHECK(ZSTD_c_chainLog, value);
+ CCtxParams->cParams.chainLog = (U32)value;
+ return CCtxParams->cParams.chainLog;
+
+ case ZSTD_c_searchLog :
+ if (value!=0) /* 0 => use default */
+ BOUNDCHECK(ZSTD_c_searchLog, value);
+ CCtxParams->cParams.searchLog = (U32)value;
+ return (size_t)value;
+
+ case ZSTD_c_minMatch :
+ if (value!=0) /* 0 => use default */
+ BOUNDCHECK(ZSTD_c_minMatch, value);
+ CCtxParams->cParams.minMatch = value;
+ return CCtxParams->cParams.minMatch;
+
+ case ZSTD_c_targetLength :
+ BOUNDCHECK(ZSTD_c_targetLength, value);
+ CCtxParams->cParams.targetLength = value;
+ return CCtxParams->cParams.targetLength;
+
+ case ZSTD_c_strategy :
+ if (value!=0) /* 0 => use default */
+ BOUNDCHECK(ZSTD_c_strategy, value);
+ CCtxParams->cParams.strategy = (ZSTD_strategy)value;
+ return (size_t)CCtxParams->cParams.strategy;
+
+ case ZSTD_c_contentSizeFlag :
+ /* Content size written in frame header _when known_ (default:1) */
+ DEBUGLOG(4, "set content size flag = %u", (value!=0));
+ CCtxParams->fParams.contentSizeFlag = value != 0;
+ return CCtxParams->fParams.contentSizeFlag;
+
+ case ZSTD_c_checksumFlag :
+ /* A 32-bits content checksum will be calculated and written at end of frame (default:0) */
+ CCtxParams->fParams.checksumFlag = value != 0;
+ return CCtxParams->fParams.checksumFlag;
+
+ case ZSTD_c_dictIDFlag : /* When applicable, dictionary's dictID is provided in frame header (default:1) */
+ DEBUGLOG(4, "set dictIDFlag = %u", (value!=0));
+ CCtxParams->fParams.noDictIDFlag = !value;
+ return !CCtxParams->fParams.noDictIDFlag;
+
+ case ZSTD_c_forceMaxWindow :
+ CCtxParams->forceWindow = (value != 0);
+ return CCtxParams->forceWindow;
+
+ case ZSTD_c_forceAttachDict : {
+ const ZSTD_dictAttachPref_e pref = (ZSTD_dictAttachPref_e)value;
+ BOUNDCHECK(ZSTD_c_forceAttachDict, pref);
+ CCtxParams->attachDictPref = pref;
+ return CCtxParams->attachDictPref;
+ }
+
+ case ZSTD_c_literalCompressionMode : {
+ const ZSTD_paramSwitch_e lcm = (ZSTD_paramSwitch_e)value;
+ BOUNDCHECK(ZSTD_c_literalCompressionMode, lcm);
+ CCtxParams->literalCompressionMode = lcm;
+ return CCtxParams->literalCompressionMode;
+ }
+
+ case ZSTD_c_nbWorkers :
+ RETURN_ERROR_IF(value!=0, parameter_unsupported, "not compiled with multithreading");
+ return 0;
+
+ case ZSTD_c_jobSize :
+ RETURN_ERROR_IF(value!=0, parameter_unsupported, "not compiled with multithreading");
+ return 0;
+
+ case ZSTD_c_overlapLog :
+ RETURN_ERROR_IF(value!=0, parameter_unsupported, "not compiled with multithreading");
+ return 0;
+
+ case ZSTD_c_rsyncable :
+ RETURN_ERROR_IF(value!=0, parameter_unsupported, "not compiled with multithreading");
+ return 0;
+
+ case ZSTD_c_enableDedicatedDictSearch :
+ CCtxParams->enableDedicatedDictSearch = (value!=0);
+ return CCtxParams->enableDedicatedDictSearch;
+
+ case ZSTD_c_enableLongDistanceMatching :
+ CCtxParams->ldmParams.enableLdm = (ZSTD_paramSwitch_e)value;
+ return CCtxParams->ldmParams.enableLdm;
+
+ case ZSTD_c_ldmHashLog :
+ if (value!=0) /* 0 ==> auto */
+ BOUNDCHECK(ZSTD_c_ldmHashLog, value);
+ CCtxParams->ldmParams.hashLog = value;
+ return CCtxParams->ldmParams.hashLog;
+
+ case ZSTD_c_ldmMinMatch :
+ if (value!=0) /* 0 ==> default */
+ BOUNDCHECK(ZSTD_c_ldmMinMatch, value);
+ CCtxParams->ldmParams.minMatchLength = value;
+ return CCtxParams->ldmParams.minMatchLength;
+
+ case ZSTD_c_ldmBucketSizeLog :
+ if (value!=0) /* 0 ==> default */
+ BOUNDCHECK(ZSTD_c_ldmBucketSizeLog, value);
+ CCtxParams->ldmParams.bucketSizeLog = value;
+ return CCtxParams->ldmParams.bucketSizeLog;
+
+ case ZSTD_c_ldmHashRateLog :
+ if (value!=0) /* 0 ==> default */
+ BOUNDCHECK(ZSTD_c_ldmHashRateLog, value);
+ CCtxParams->ldmParams.hashRateLog = value;
+ return CCtxParams->ldmParams.hashRateLog;
+
+ case ZSTD_c_targetCBlockSize :
+ if (value!=0) /* 0 ==> default */
+ BOUNDCHECK(ZSTD_c_targetCBlockSize, value);
+ CCtxParams->targetCBlockSize = value;
+ return CCtxParams->targetCBlockSize;
+
+ case ZSTD_c_srcSizeHint :
+ if (value!=0) /* 0 ==> default */
+ BOUNDCHECK(ZSTD_c_srcSizeHint, value);
+ CCtxParams->srcSizeHint = value;
+ return CCtxParams->srcSizeHint;
+
+ case ZSTD_c_stableInBuffer:
+ BOUNDCHECK(ZSTD_c_stableInBuffer, value);
+ CCtxParams->inBufferMode = (ZSTD_bufferMode_e)value;
+ return CCtxParams->inBufferMode;
+
+ case ZSTD_c_stableOutBuffer:
+ BOUNDCHECK(ZSTD_c_stableOutBuffer, value);
+ CCtxParams->outBufferMode = (ZSTD_bufferMode_e)value;
+ return CCtxParams->outBufferMode;
+
+ case ZSTD_c_blockDelimiters:
+ BOUNDCHECK(ZSTD_c_blockDelimiters, value);
+ CCtxParams->blockDelimiters = (ZSTD_sequenceFormat_e)value;
+ return CCtxParams->blockDelimiters;
+
+ case ZSTD_c_validateSequences:
+ BOUNDCHECK(ZSTD_c_validateSequences, value);
+ CCtxParams->validateSequences = value;
+ return CCtxParams->validateSequences;
+
+ case ZSTD_c_useBlockSplitter:
+ BOUNDCHECK(ZSTD_c_useBlockSplitter, value);
+ CCtxParams->useBlockSplitter = (ZSTD_paramSwitch_e)value;
+ return CCtxParams->useBlockSplitter;
+
+ case ZSTD_c_useRowMatchFinder:
+ BOUNDCHECK(ZSTD_c_useRowMatchFinder, value);
+ CCtxParams->useRowMatchFinder = (ZSTD_paramSwitch_e)value;
+ return CCtxParams->useRowMatchFinder;
+
+ case ZSTD_c_deterministicRefPrefix:
+ BOUNDCHECK(ZSTD_c_deterministicRefPrefix, value);
+ CCtxParams->deterministicRefPrefix = !!value;
+ return CCtxParams->deterministicRefPrefix;
+
+ default: RETURN_ERROR(parameter_unsupported, "unknown parameter");
+ }
+}
+
+size_t ZSTD_CCtx_getParameter(ZSTD_CCtx const* cctx, ZSTD_cParameter param, int* value)
+{
+ return ZSTD_CCtxParams_getParameter(&cctx->requestedParams, param, value);
+}
+
+size_t ZSTD_CCtxParams_getParameter(
+ ZSTD_CCtx_params const* CCtxParams, ZSTD_cParameter param, int* value)
+{
+ switch(param)
+ {
+ case ZSTD_c_format :
+ *value = CCtxParams->format;
+ break;
+ case ZSTD_c_compressionLevel :
+ *value = CCtxParams->compressionLevel;
+ break;
+ case ZSTD_c_windowLog :
+ *value = (int)CCtxParams->cParams.windowLog;
+ break;
+ case ZSTD_c_hashLog :
+ *value = (int)CCtxParams->cParams.hashLog;
+ break;
+ case ZSTD_c_chainLog :
+ *value = (int)CCtxParams->cParams.chainLog;
+ break;
+ case ZSTD_c_searchLog :
+ *value = CCtxParams->cParams.searchLog;
+ break;
+ case ZSTD_c_minMatch :
+ *value = CCtxParams->cParams.minMatch;
+ break;
+ case ZSTD_c_targetLength :
+ *value = CCtxParams->cParams.targetLength;
+ break;
+ case ZSTD_c_strategy :
+ *value = (unsigned)CCtxParams->cParams.strategy;
+ break;
+ case ZSTD_c_contentSizeFlag :
+ *value = CCtxParams->fParams.contentSizeFlag;
+ break;
+ case ZSTD_c_checksumFlag :
+ *value = CCtxParams->fParams.checksumFlag;
+ break;
+ case ZSTD_c_dictIDFlag :
+ *value = !CCtxParams->fParams.noDictIDFlag;
+ break;
+ case ZSTD_c_forceMaxWindow :
+ *value = CCtxParams->forceWindow;
+ break;
+ case ZSTD_c_forceAttachDict :
+ *value = CCtxParams->attachDictPref;
+ break;
+ case ZSTD_c_literalCompressionMode :
+ *value = CCtxParams->literalCompressionMode;
+ break;
+ case ZSTD_c_nbWorkers :
+ assert(CCtxParams->nbWorkers == 0);
+ *value = CCtxParams->nbWorkers;
+ break;
+ case ZSTD_c_jobSize :
+ RETURN_ERROR(parameter_unsupported, "not compiled with multithreading");
+ case ZSTD_c_overlapLog :
+ RETURN_ERROR(parameter_unsupported, "not compiled with multithreading");
+ case ZSTD_c_rsyncable :
+ RETURN_ERROR(parameter_unsupported, "not compiled with multithreading");
+ case ZSTD_c_enableDedicatedDictSearch :
+ *value = CCtxParams->enableDedicatedDictSearch;
+ break;
+ case ZSTD_c_enableLongDistanceMatching :
+ *value = CCtxParams->ldmParams.enableLdm;
+ break;
+ case ZSTD_c_ldmHashLog :
+ *value = CCtxParams->ldmParams.hashLog;
+ break;
+ case ZSTD_c_ldmMinMatch :
+ *value = CCtxParams->ldmParams.minMatchLength;
+ break;
+ case ZSTD_c_ldmBucketSizeLog :
+ *value = CCtxParams->ldmParams.bucketSizeLog;
+ break;
+ case ZSTD_c_ldmHashRateLog :
+ *value = CCtxParams->ldmParams.hashRateLog;
+ break;
+ case ZSTD_c_targetCBlockSize :
+ *value = (int)CCtxParams->targetCBlockSize;
+ break;
+ case ZSTD_c_srcSizeHint :
+ *value = (int)CCtxParams->srcSizeHint;
+ break;
+ case ZSTD_c_stableInBuffer :
+ *value = (int)CCtxParams->inBufferMode;
+ break;
+ case ZSTD_c_stableOutBuffer :
+ *value = (int)CCtxParams->outBufferMode;
+ break;
+ case ZSTD_c_blockDelimiters :
+ *value = (int)CCtxParams->blockDelimiters;
+ break;
+ case ZSTD_c_validateSequences :
+ *value = (int)CCtxParams->validateSequences;
+ break;
+ case ZSTD_c_useBlockSplitter :
+ *value = (int)CCtxParams->useBlockSplitter;
+ break;
+ case ZSTD_c_useRowMatchFinder :
+ *value = (int)CCtxParams->useRowMatchFinder;
+ break;
+ case ZSTD_c_deterministicRefPrefix:
+ *value = (int)CCtxParams->deterministicRefPrefix;
+ break;
+ default: RETURN_ERROR(parameter_unsupported, "unknown parameter");
+ }
+ return 0;
+}
+
+/* ZSTD_CCtx_setParametersUsingCCtxParams() :
+ * just applies `params` into `cctx`
+ * no action is performed, parameters are merely stored.
+ * If ZSTDMT is enabled, parameters are pushed to cctx->mtctx.
+ * This is possible even if a compression is ongoing.
+ * In which case, new parameters will be applied on the fly, starting with next compression job.
+ */
+size_t ZSTD_CCtx_setParametersUsingCCtxParams(
+ ZSTD_CCtx* cctx, const ZSTD_CCtx_params* params)
+{
+ DEBUGLOG(4, "ZSTD_CCtx_setParametersUsingCCtxParams");
+ RETURN_ERROR_IF(cctx->streamStage != zcss_init, stage_wrong,
+ "The context is in the wrong stage!");
+ RETURN_ERROR_IF(cctx->cdict, stage_wrong,
+ "Can't override parameters with cdict attached (some must "
+ "be inherited from the cdict).");
+
+ cctx->requestedParams = *params;
+ return 0;
+}
+
+size_t ZSTD_CCtx_setPledgedSrcSize(ZSTD_CCtx* cctx, unsigned long long pledgedSrcSize)
+{
+ DEBUGLOG(4, "ZSTD_CCtx_setPledgedSrcSize to %u bytes", (U32)pledgedSrcSize);
+ RETURN_ERROR_IF(cctx->streamStage != zcss_init, stage_wrong,
+ "Can't set pledgedSrcSize when not in init stage.");
+ cctx->pledgedSrcSizePlusOne = pledgedSrcSize+1;
+ return 0;
+}
+
+static ZSTD_compressionParameters ZSTD_dedicatedDictSearch_getCParams(
+ int const compressionLevel,
+ size_t const dictSize);
+static int ZSTD_dedicatedDictSearch_isSupported(
+ const ZSTD_compressionParameters* cParams);
+static void ZSTD_dedicatedDictSearch_revertCParams(
+ ZSTD_compressionParameters* cParams);
+
+/*
+ * Initializes the local dict using the requested parameters.
+ * NOTE: This does not use the pledged src size, because it may be used for more
+ * than one compression.
+ */
+static size_t ZSTD_initLocalDict(ZSTD_CCtx* cctx)
+{
+ ZSTD_localDict* const dl = &cctx->localDict;
+ if (dl->dict == NULL) {
+ /* No local dictionary. */
+ assert(dl->dictBuffer == NULL);
+ assert(dl->cdict == NULL);
+ assert(dl->dictSize == 0);
+ return 0;
+ }
+ if (dl->cdict != NULL) {
+ assert(cctx->cdict == dl->cdict);
+ /* Local dictionary already initialized. */
+ return 0;
+ }
+ assert(dl->dictSize > 0);
+ assert(cctx->cdict == NULL);
+ assert(cctx->prefixDict.dict == NULL);
+
+ dl->cdict = ZSTD_createCDict_advanced2(
+ dl->dict,
+ dl->dictSize,
+ ZSTD_dlm_byRef,
+ dl->dictContentType,
+ &cctx->requestedParams,
+ cctx->customMem);
+ RETURN_ERROR_IF(!dl->cdict, memory_allocation, "ZSTD_createCDict_advanced failed");
+ cctx->cdict = dl->cdict;
+ return 0;
+}
+
+size_t ZSTD_CCtx_loadDictionary_advanced(
+ ZSTD_CCtx* cctx, const void* dict, size_t dictSize,
+ ZSTD_dictLoadMethod_e dictLoadMethod, ZSTD_dictContentType_e dictContentType)
+{
+ RETURN_ERROR_IF(cctx->streamStage != zcss_init, stage_wrong,
+ "Can't load a dictionary when ctx is not in init stage.");
+ DEBUGLOG(4, "ZSTD_CCtx_loadDictionary_advanced (size: %u)", (U32)dictSize);
+ ZSTD_clearAllDicts(cctx); /* in case one already exists */
+ if (dict == NULL || dictSize == 0) /* no dictionary mode */
+ return 0;
+ if (dictLoadMethod == ZSTD_dlm_byRef) {
+ cctx->localDict.dict = dict;
+ } else {
+ void* dictBuffer;
+ RETURN_ERROR_IF(cctx->staticSize, memory_allocation,
+ "no malloc for static CCtx");
+ dictBuffer = ZSTD_customMalloc(dictSize, cctx->customMem);
+ RETURN_ERROR_IF(!dictBuffer, memory_allocation, "NULL pointer!");
+ ZSTD_memcpy(dictBuffer, dict, dictSize);
+ cctx->localDict.dictBuffer = dictBuffer;
+ cctx->localDict.dict = dictBuffer;
+ }
+ cctx->localDict.dictSize = dictSize;
+ cctx->localDict.dictContentType = dictContentType;
+ return 0;
+}
+
+size_t ZSTD_CCtx_loadDictionary_byReference(
+ ZSTD_CCtx* cctx, const void* dict, size_t dictSize)
+{
+ return ZSTD_CCtx_loadDictionary_advanced(
+ cctx, dict, dictSize, ZSTD_dlm_byRef, ZSTD_dct_auto);
+}
+
+size_t ZSTD_CCtx_loadDictionary(ZSTD_CCtx* cctx, const void* dict, size_t dictSize)
+{
+ return ZSTD_CCtx_loadDictionary_advanced(
+ cctx, dict, dictSize, ZSTD_dlm_byCopy, ZSTD_dct_auto);
+}
+
+
+size_t ZSTD_CCtx_refCDict(ZSTD_CCtx* cctx, const ZSTD_CDict* cdict)
+{
+ RETURN_ERROR_IF(cctx->streamStage != zcss_init, stage_wrong,
+ "Can't ref a dict when ctx not in init stage.");
+ /* Free the existing local cdict (if any) to save memory. */
+ ZSTD_clearAllDicts(cctx);
+ cctx->cdict = cdict;
+ return 0;
+}
+
+size_t ZSTD_CCtx_refThreadPool(ZSTD_CCtx* cctx, ZSTD_threadPool* pool)
+{
+ RETURN_ERROR_IF(cctx->streamStage != zcss_init, stage_wrong,
+ "Can't ref a pool when ctx not in init stage.");
+ cctx->pool = pool;
+ return 0;
+}
+
+size_t ZSTD_CCtx_refPrefix(ZSTD_CCtx* cctx, const void* prefix, size_t prefixSize)
+{
+ return ZSTD_CCtx_refPrefix_advanced(cctx, prefix, prefixSize, ZSTD_dct_rawContent);
+}
+
+size_t ZSTD_CCtx_refPrefix_advanced(
+ ZSTD_CCtx* cctx, const void* prefix, size_t prefixSize, ZSTD_dictContentType_e dictContentType)
+{
+ RETURN_ERROR_IF(cctx->streamStage != zcss_init, stage_wrong,
+ "Can't ref a prefix when ctx not in init stage.");
+ ZSTD_clearAllDicts(cctx);
+ if (prefix != NULL && prefixSize > 0) {
+ cctx->prefixDict.dict = prefix;
+ cctx->prefixDict.dictSize = prefixSize;
+ cctx->prefixDict.dictContentType = dictContentType;
+ }
+ return 0;
+}
+
+/*! ZSTD_CCtx_reset() :
+ * Also dumps dictionary */
+size_t ZSTD_CCtx_reset(ZSTD_CCtx* cctx, ZSTD_ResetDirective reset)
+{
+ if ( (reset == ZSTD_reset_session_only)
+ || (reset == ZSTD_reset_session_and_parameters) ) {
+ cctx->streamStage = zcss_init;
+ cctx->pledgedSrcSizePlusOne = 0;
+ }
+ if ( (reset == ZSTD_reset_parameters)
+ || (reset == ZSTD_reset_session_and_parameters) ) {
+ RETURN_ERROR_IF(cctx->streamStage != zcss_init, stage_wrong,
+ "Can't reset parameters only when not in init stage.");
+ ZSTD_clearAllDicts(cctx);
+ return ZSTD_CCtxParams_reset(&cctx->requestedParams);
+ }
+ return 0;
+}
+
+
+/* ZSTD_checkCParams() :
+ control CParam values remain within authorized range.
+ @return : 0, or an error code if one value is beyond authorized range */
+size_t ZSTD_checkCParams(ZSTD_compressionParameters cParams)
+{
+ BOUNDCHECK(ZSTD_c_windowLog, (int)cParams.windowLog);
+ BOUNDCHECK(ZSTD_c_chainLog, (int)cParams.chainLog);
+ BOUNDCHECK(ZSTD_c_hashLog, (int)cParams.hashLog);
+ BOUNDCHECK(ZSTD_c_searchLog, (int)cParams.searchLog);
+ BOUNDCHECK(ZSTD_c_minMatch, (int)cParams.minMatch);
+ BOUNDCHECK(ZSTD_c_targetLength,(int)cParams.targetLength);
+ BOUNDCHECK(ZSTD_c_strategy, cParams.strategy);
+ return 0;
+}
+
+/* ZSTD_clampCParams() :
+ * make CParam values within valid range.
+ * @return : valid CParams */
+static ZSTD_compressionParameters
+ZSTD_clampCParams(ZSTD_compressionParameters cParams)
+{
+# define CLAMP_TYPE(cParam, val, type) { \
+ ZSTD_bounds const bounds = ZSTD_cParam_getBounds(cParam); \
+ if ((int)val<bounds.lowerBound) val=(type)bounds.lowerBound; \
+ else if ((int)val>bounds.upperBound) val=(type)bounds.upperBound; \
+ }
+# define CLAMP(cParam, val) CLAMP_TYPE(cParam, val, unsigned)
+ CLAMP(ZSTD_c_windowLog, cParams.windowLog);
+ CLAMP(ZSTD_c_chainLog, cParams.chainLog);
+ CLAMP(ZSTD_c_hashLog, cParams.hashLog);
+ CLAMP(ZSTD_c_searchLog, cParams.searchLog);
+ CLAMP(ZSTD_c_minMatch, cParams.minMatch);
+ CLAMP(ZSTD_c_targetLength,cParams.targetLength);
+ CLAMP_TYPE(ZSTD_c_strategy,cParams.strategy, ZSTD_strategy);
+ return cParams;
+}
+
+/* ZSTD_cycleLog() :
+ * condition for correct operation : hashLog > 1 */
+U32 ZSTD_cycleLog(U32 hashLog, ZSTD_strategy strat)
+{
+ U32 const btScale = ((U32)strat >= (U32)ZSTD_btlazy2);
+ return hashLog - btScale;
+}
+
+/* ZSTD_dictAndWindowLog() :
+ * Returns an adjusted window log that is large enough to fit the source and the dictionary.
+ * The zstd format says that the entire dictionary is valid if one byte of the dictionary
+ * is within the window. So the hashLog and chainLog should be large enough to reference both
+ * the dictionary and the window. So we must use this adjusted dictAndWindowLog when downsizing
+ * the hashLog and windowLog.
+ * NOTE: srcSize must not be ZSTD_CONTENTSIZE_UNKNOWN.
+ */
+static U32 ZSTD_dictAndWindowLog(U32 windowLog, U64 srcSize, U64 dictSize)
+{
+ const U64 maxWindowSize = 1ULL << ZSTD_WINDOWLOG_MAX;
+ /* No dictionary ==> No change */
+ if (dictSize == 0) {
+ return windowLog;
+ }
+ assert(windowLog <= ZSTD_WINDOWLOG_MAX);
+ assert(srcSize != ZSTD_CONTENTSIZE_UNKNOWN); /* Handled in ZSTD_adjustCParams_internal() */
+ {
+ U64 const windowSize = 1ULL << windowLog;
+ U64 const dictAndWindowSize = dictSize + windowSize;
+ /* If the window size is already large enough to fit both the source and the dictionary
+ * then just use the window size. Otherwise adjust so that it fits the dictionary and
+ * the window.
+ */
+ if (windowSize >= dictSize + srcSize) {
+ return windowLog; /* Window size large enough already */
+ } else if (dictAndWindowSize >= maxWindowSize) {
+ return ZSTD_WINDOWLOG_MAX; /* Larger than max window log */
+ } else {
+ return ZSTD_highbit32((U32)dictAndWindowSize - 1) + 1;
+ }
+ }
+}
+
+/* ZSTD_adjustCParams_internal() :
+ * optimize `cPar` for a specified input (`srcSize` and `dictSize`).
+ * mostly downsize to reduce memory consumption and initialization latency.
+ * `srcSize` can be ZSTD_CONTENTSIZE_UNKNOWN when not known.
+ * `mode` is the mode for parameter adjustment. See docs for `ZSTD_cParamMode_e`.
+ * note : `srcSize==0` means 0!
+ * condition : cPar is presumed validated (can be checked using ZSTD_checkCParams()). */
+static ZSTD_compressionParameters
+ZSTD_adjustCParams_internal(ZSTD_compressionParameters cPar,
+ unsigned long long srcSize,
+ size_t dictSize,
+ ZSTD_cParamMode_e mode)
+{
+ const U64 minSrcSize = 513; /* (1<<9) + 1 */
+ const U64 maxWindowResize = 1ULL << (ZSTD_WINDOWLOG_MAX-1);
+ assert(ZSTD_checkCParams(cPar)==0);
+
+ switch (mode) {
+ case ZSTD_cpm_unknown:
+ case ZSTD_cpm_noAttachDict:
+ /* If we don't know the source size, don't make any
+ * assumptions about it. We will already have selected
+ * smaller parameters if a dictionary is in use.
+ */
+ break;
+ case ZSTD_cpm_createCDict:
+ /* Assume a small source size when creating a dictionary
+ * with an unknown source size.
+ */
+ if (dictSize && srcSize == ZSTD_CONTENTSIZE_UNKNOWN)
+ srcSize = minSrcSize;
+ break;
+ case ZSTD_cpm_attachDict:
+ /* Dictionary has its own dedicated parameters which have
+ * already been selected. We are selecting parameters
+ * for only the source.
+ */
+ dictSize = 0;
+ break;
+ default:
+ assert(0);
+ break;
+ }
+
+ /* resize windowLog if input is small enough, to use less memory */
+ if ( (srcSize < maxWindowResize)
+ && (dictSize < maxWindowResize) ) {
+ U32 const tSize = (U32)(srcSize + dictSize);
+ static U32 const hashSizeMin = 1 << ZSTD_HASHLOG_MIN;
+ U32 const srcLog = (tSize < hashSizeMin) ? ZSTD_HASHLOG_MIN :
+ ZSTD_highbit32(tSize-1) + 1;
+ if (cPar.windowLog > srcLog) cPar.windowLog = srcLog;
+ }
+ if (srcSize != ZSTD_CONTENTSIZE_UNKNOWN) {
+ U32 const dictAndWindowLog = ZSTD_dictAndWindowLog(cPar.windowLog, (U64)srcSize, (U64)dictSize);
+ U32 const cycleLog = ZSTD_cycleLog(cPar.chainLog, cPar.strategy);
+ if (cPar.hashLog > dictAndWindowLog+1) cPar.hashLog = dictAndWindowLog+1;
+ if (cycleLog > dictAndWindowLog)
+ cPar.chainLog -= (cycleLog - dictAndWindowLog);
+ }
+
+ if (cPar.windowLog < ZSTD_WINDOWLOG_ABSOLUTEMIN)
+ cPar.windowLog = ZSTD_WINDOWLOG_ABSOLUTEMIN; /* minimum wlog required for valid frame header */
+
+ return cPar;
+}
+
+ZSTD_compressionParameters
+ZSTD_adjustCParams(ZSTD_compressionParameters cPar,
+ unsigned long long srcSize,
+ size_t dictSize)
+{
+ cPar = ZSTD_clampCParams(cPar); /* resulting cPar is necessarily valid (all parameters within range) */
+ if (srcSize == 0) srcSize = ZSTD_CONTENTSIZE_UNKNOWN;
+ return ZSTD_adjustCParams_internal(cPar, srcSize, dictSize, ZSTD_cpm_unknown);
+}
+
+static ZSTD_compressionParameters ZSTD_getCParams_internal(int compressionLevel, unsigned long long srcSizeHint, size_t dictSize, ZSTD_cParamMode_e mode);
+static ZSTD_parameters ZSTD_getParams_internal(int compressionLevel, unsigned long long srcSizeHint, size_t dictSize, ZSTD_cParamMode_e mode);
+
+static void ZSTD_overrideCParams(
+ ZSTD_compressionParameters* cParams,
+ const ZSTD_compressionParameters* overrides)
+{
+ if (overrides->windowLog) cParams->windowLog = overrides->windowLog;
+ if (overrides->hashLog) cParams->hashLog = overrides->hashLog;
+ if (overrides->chainLog) cParams->chainLog = overrides->chainLog;
+ if (overrides->searchLog) cParams->searchLog = overrides->searchLog;
+ if (overrides->minMatch) cParams->minMatch = overrides->minMatch;
+ if (overrides->targetLength) cParams->targetLength = overrides->targetLength;
+ if (overrides->strategy) cParams->strategy = overrides->strategy;
+}
+
+ZSTD_compressionParameters ZSTD_getCParamsFromCCtxParams(
+ const ZSTD_CCtx_params* CCtxParams, U64 srcSizeHint, size_t dictSize, ZSTD_cParamMode_e mode)
+{
+ ZSTD_compressionParameters cParams;
+ if (srcSizeHint == ZSTD_CONTENTSIZE_UNKNOWN && CCtxParams->srcSizeHint > 0) {
+ srcSizeHint = CCtxParams->srcSizeHint;
+ }
+ cParams = ZSTD_getCParams_internal(CCtxParams->compressionLevel, srcSizeHint, dictSize, mode);
+ if (CCtxParams->ldmParams.enableLdm == ZSTD_ps_enable) cParams.windowLog = ZSTD_LDM_DEFAULT_WINDOW_LOG;
+ ZSTD_overrideCParams(&cParams, &CCtxParams->cParams);
+ assert(!ZSTD_checkCParams(cParams));
+ /* srcSizeHint == 0 means 0 */
+ return ZSTD_adjustCParams_internal(cParams, srcSizeHint, dictSize, mode);
+}
+
+static size_t
+ZSTD_sizeof_matchState(const ZSTD_compressionParameters* const cParams,
+ const ZSTD_paramSwitch_e useRowMatchFinder,
+ const U32 enableDedicatedDictSearch,
+ const U32 forCCtx)
+{
+ /* chain table size should be 0 for fast or row-hash strategies */
+ size_t const chainSize = ZSTD_allocateChainTable(cParams->strategy, useRowMatchFinder, enableDedicatedDictSearch && !forCCtx)
+ ? ((size_t)1 << cParams->chainLog)
+ : 0;
+ size_t const hSize = ((size_t)1) << cParams->hashLog;
+ U32 const hashLog3 = (forCCtx && cParams->minMatch==3) ? MIN(ZSTD_HASHLOG3_MAX, cParams->windowLog) : 0;
+ size_t const h3Size = hashLog3 ? ((size_t)1) << hashLog3 : 0;
+ /* We don't use ZSTD_cwksp_alloc_size() here because the tables aren't
+ * surrounded by redzones in ASAN. */
+ size_t const tableSpace = chainSize * sizeof(U32)
+ + hSize * sizeof(U32)
+ + h3Size * sizeof(U32);
+ size_t const optPotentialSpace =
+ ZSTD_cwksp_aligned_alloc_size((MaxML+1) * sizeof(U32))
+ + ZSTD_cwksp_aligned_alloc_size((MaxLL+1) * sizeof(U32))
+ + ZSTD_cwksp_aligned_alloc_size((MaxOff+1) * sizeof(U32))
+ + ZSTD_cwksp_aligned_alloc_size((1<<Litbits) * sizeof(U32))
+ + ZSTD_cwksp_aligned_alloc_size((ZSTD_OPT_NUM+1) * sizeof(ZSTD_match_t))
+ + ZSTD_cwksp_aligned_alloc_size((ZSTD_OPT_NUM+1) * sizeof(ZSTD_optimal_t));
+ size_t const lazyAdditionalSpace = ZSTD_rowMatchFinderUsed(cParams->strategy, useRowMatchFinder)
+ ? ZSTD_cwksp_aligned_alloc_size(hSize*sizeof(U16))
+ : 0;
+ size_t const optSpace = (forCCtx && (cParams->strategy >= ZSTD_btopt))
+ ? optPotentialSpace
+ : 0;
+ size_t const slackSpace = ZSTD_cwksp_slack_space_required();
+
+ /* tables are guaranteed to be sized in multiples of 64 bytes (or 16 uint32_t) */
+ ZSTD_STATIC_ASSERT(ZSTD_HASHLOG_MIN >= 4 && ZSTD_WINDOWLOG_MIN >= 4 && ZSTD_CHAINLOG_MIN >= 4);
+ assert(useRowMatchFinder != ZSTD_ps_auto);
+
+ DEBUGLOG(4, "chainSize: %u - hSize: %u - h3Size: %u",
+ (U32)chainSize, (U32)hSize, (U32)h3Size);
+ return tableSpace + optSpace + slackSpace + lazyAdditionalSpace;
+}
+
+static size_t ZSTD_estimateCCtxSize_usingCCtxParams_internal(
+ const ZSTD_compressionParameters* cParams,
+ const ldmParams_t* ldmParams,
+ const int isStatic,
+ const ZSTD_paramSwitch_e useRowMatchFinder,
+ const size_t buffInSize,
+ const size_t buffOutSize,
+ const U64 pledgedSrcSize)
+{
+ size_t const windowSize = (size_t) BOUNDED(1ULL, 1ULL << cParams->windowLog, pledgedSrcSize);
+ size_t const blockSize = MIN(ZSTD_BLOCKSIZE_MAX, windowSize);
+ U32 const divider = (cParams->minMatch==3) ? 3 : 4;
+ size_t const maxNbSeq = blockSize / divider;
+ size_t const tokenSpace = ZSTD_cwksp_alloc_size(WILDCOPY_OVERLENGTH + blockSize)
+ + ZSTD_cwksp_aligned_alloc_size(maxNbSeq * sizeof(seqDef))
+ + 3 * ZSTD_cwksp_alloc_size(maxNbSeq * sizeof(BYTE));
+ size_t const entropySpace = ZSTD_cwksp_alloc_size(ENTROPY_WORKSPACE_SIZE);
+ size_t const blockStateSpace = 2 * ZSTD_cwksp_alloc_size(sizeof(ZSTD_compressedBlockState_t));
+ size_t const matchStateSize = ZSTD_sizeof_matchState(cParams, useRowMatchFinder, /* enableDedicatedDictSearch */ 0, /* forCCtx */ 1);
+
+ size_t const ldmSpace = ZSTD_ldm_getTableSize(*ldmParams);
+ size_t const maxNbLdmSeq = ZSTD_ldm_getMaxNbSeq(*ldmParams, blockSize);
+ size_t const ldmSeqSpace = ldmParams->enableLdm == ZSTD_ps_enable ?
+ ZSTD_cwksp_aligned_alloc_size(maxNbLdmSeq * sizeof(rawSeq)) : 0;
+
+
+ size_t const bufferSpace = ZSTD_cwksp_alloc_size(buffInSize)
+ + ZSTD_cwksp_alloc_size(buffOutSize);
+
+ size_t const cctxSpace = isStatic ? ZSTD_cwksp_alloc_size(sizeof(ZSTD_CCtx)) : 0;
+
+ size_t const neededSpace =
+ cctxSpace +
+ entropySpace +
+ blockStateSpace +
+ ldmSpace +
+ ldmSeqSpace +
+ matchStateSize +
+ tokenSpace +
+ bufferSpace;
+
+ DEBUGLOG(5, "estimate workspace : %u", (U32)neededSpace);
+ return neededSpace;
+}
+
+size_t ZSTD_estimateCCtxSize_usingCCtxParams(const ZSTD_CCtx_params* params)
+{
+ ZSTD_compressionParameters const cParams =
+ ZSTD_getCParamsFromCCtxParams(params, ZSTD_CONTENTSIZE_UNKNOWN, 0, ZSTD_cpm_noAttachDict);
+ ZSTD_paramSwitch_e const useRowMatchFinder = ZSTD_resolveRowMatchFinderMode(params->useRowMatchFinder,
+ &cParams);
+
+ RETURN_ERROR_IF(params->nbWorkers > 0, GENERIC, "Estimate CCtx size is supported for single-threaded compression only.");
+ /* estimateCCtxSize is for one-shot compression. So no buffers should
+ * be needed. However, we still allocate two 0-sized buffers, which can
+ * take space under ASAN. */
+ return ZSTD_estimateCCtxSize_usingCCtxParams_internal(
+ &cParams, &params->ldmParams, 1, useRowMatchFinder, 0, 0, ZSTD_CONTENTSIZE_UNKNOWN);
+}
+
+size_t ZSTD_estimateCCtxSize_usingCParams(ZSTD_compressionParameters cParams)
+{
+ ZSTD_CCtx_params initialParams = ZSTD_makeCCtxParamsFromCParams(cParams);
+ if (ZSTD_rowMatchFinderSupported(cParams.strategy)) {
+ /* Pick bigger of not using and using row-based matchfinder for greedy and lazy strategies */
+ size_t noRowCCtxSize;
+ size_t rowCCtxSize;
+ initialParams.useRowMatchFinder = ZSTD_ps_disable;
+ noRowCCtxSize = ZSTD_estimateCCtxSize_usingCCtxParams(&initialParams);
+ initialParams.useRowMatchFinder = ZSTD_ps_enable;
+ rowCCtxSize = ZSTD_estimateCCtxSize_usingCCtxParams(&initialParams);
+ return MAX(noRowCCtxSize, rowCCtxSize);
+ } else {
+ return ZSTD_estimateCCtxSize_usingCCtxParams(&initialParams);
+ }
+}
+
+static size_t ZSTD_estimateCCtxSize_internal(int compressionLevel)
+{
+ int tier = 0;
+ size_t largestSize = 0;
+ static const unsigned long long srcSizeTiers[4] = {16 KB, 128 KB, 256 KB, ZSTD_CONTENTSIZE_UNKNOWN};
+ for (; tier < 4; ++tier) {
+ /* Choose the set of cParams for a given level across all srcSizes that give the largest cctxSize */
+ ZSTD_compressionParameters const cParams = ZSTD_getCParams_internal(compressionLevel, srcSizeTiers[tier], 0, ZSTD_cpm_noAttachDict);
+ largestSize = MAX(ZSTD_estimateCCtxSize_usingCParams(cParams), largestSize);
+ }
+ return largestSize;
+}
+
+size_t ZSTD_estimateCCtxSize(int compressionLevel)
+{
+ int level;
+ size_t memBudget = 0;
+ for (level=MIN(compressionLevel, 1); level<=compressionLevel; level++) {
+ /* Ensure monotonically increasing memory usage as compression level increases */
+ size_t const newMB = ZSTD_estimateCCtxSize_internal(level);
+ if (newMB > memBudget) memBudget = newMB;
+ }
+ return memBudget;
+}
+
+size_t ZSTD_estimateCStreamSize_usingCCtxParams(const ZSTD_CCtx_params* params)
+{
+ RETURN_ERROR_IF(params->nbWorkers > 0, GENERIC, "Estimate CCtx size is supported for single-threaded compression only.");
+ { ZSTD_compressionParameters const cParams =
+ ZSTD_getCParamsFromCCtxParams(params, ZSTD_CONTENTSIZE_UNKNOWN, 0, ZSTD_cpm_noAttachDict);
+ size_t const blockSize = MIN(ZSTD_BLOCKSIZE_MAX, (size_t)1 << cParams.windowLog);
+ size_t const inBuffSize = (params->inBufferMode == ZSTD_bm_buffered)
+ ? ((size_t)1 << cParams.windowLog) + blockSize
+ : 0;
+ size_t const outBuffSize = (params->outBufferMode == ZSTD_bm_buffered)
+ ? ZSTD_compressBound(blockSize) + 1
+ : 0;
+ ZSTD_paramSwitch_e const useRowMatchFinder = ZSTD_resolveRowMatchFinderMode(params->useRowMatchFinder, &params->cParams);
+
+ return ZSTD_estimateCCtxSize_usingCCtxParams_internal(
+ &cParams, &params->ldmParams, 1, useRowMatchFinder, inBuffSize, outBuffSize,
+ ZSTD_CONTENTSIZE_UNKNOWN);
+ }
+}
+
+size_t ZSTD_estimateCStreamSize_usingCParams(ZSTD_compressionParameters cParams)
+{
+ ZSTD_CCtx_params initialParams = ZSTD_makeCCtxParamsFromCParams(cParams);
+ if (ZSTD_rowMatchFinderSupported(cParams.strategy)) {
+ /* Pick bigger of not using and using row-based matchfinder for greedy and lazy strategies */
+ size_t noRowCCtxSize;
+ size_t rowCCtxSize;
+ initialParams.useRowMatchFinder = ZSTD_ps_disable;
+ noRowCCtxSize = ZSTD_estimateCStreamSize_usingCCtxParams(&initialParams);
+ initialParams.useRowMatchFinder = ZSTD_ps_enable;
+ rowCCtxSize = ZSTD_estimateCStreamSize_usingCCtxParams(&initialParams);
+ return MAX(noRowCCtxSize, rowCCtxSize);
+ } else {
+ return ZSTD_estimateCStreamSize_usingCCtxParams(&initialParams);
+ }
+}
+
+static size_t ZSTD_estimateCStreamSize_internal(int compressionLevel)
+{
+ ZSTD_compressionParameters const cParams = ZSTD_getCParams_internal(compressionLevel, ZSTD_CONTENTSIZE_UNKNOWN, 0, ZSTD_cpm_noAttachDict);
+ return ZSTD_estimateCStreamSize_usingCParams(cParams);
+}
+
+size_t ZSTD_estimateCStreamSize(int compressionLevel)
+{
+ int level;
+ size_t memBudget = 0;
+ for (level=MIN(compressionLevel, 1); level<=compressionLevel; level++) {
+ size_t const newMB = ZSTD_estimateCStreamSize_internal(level);
+ if (newMB > memBudget) memBudget = newMB;
+ }
+ return memBudget;
+}
+
+/* ZSTD_getFrameProgression():
+ * tells how much data has been consumed (input) and produced (output) for current frame.
+ * able to count progression inside worker threads (non-blocking mode).
+ */
+ZSTD_frameProgression ZSTD_getFrameProgression(const ZSTD_CCtx* cctx)
+{
+ { ZSTD_frameProgression fp;
+ size_t const buffered = (cctx->inBuff == NULL) ? 0 :
+ cctx->inBuffPos - cctx->inToCompress;
+ if (buffered) assert(cctx->inBuffPos >= cctx->inToCompress);
+ assert(buffered <= ZSTD_BLOCKSIZE_MAX);
+ fp.ingested = cctx->consumedSrcSize + buffered;
+ fp.consumed = cctx->consumedSrcSize;
+ fp.produced = cctx->producedCSize;
+ fp.flushed = cctx->producedCSize; /* simplified; some data might still be left within streaming output buffer */
+ fp.currentJobID = 0;
+ fp.nbActiveWorkers = 0;
+ return fp;
+} }
+
+/*! ZSTD_toFlushNow()
+ * Only useful for multithreading scenarios currently (nbWorkers >= 1).
+ */
+size_t ZSTD_toFlushNow(ZSTD_CCtx* cctx)
+{
+ (void)cctx;
+ return 0; /* over-simplification; could also check if context is currently running in streaming mode, and in which case, report how many bytes are left to be flushed within output buffer */
+}
+
+static void ZSTD_assertEqualCParams(ZSTD_compressionParameters cParams1,
+ ZSTD_compressionParameters cParams2)
+{
+ (void)cParams1;
+ (void)cParams2;
+ assert(cParams1.windowLog == cParams2.windowLog);
+ assert(cParams1.chainLog == cParams2.chainLog);
+ assert(cParams1.hashLog == cParams2.hashLog);
+ assert(cParams1.searchLog == cParams2.searchLog);
+ assert(cParams1.minMatch == cParams2.minMatch);
+ assert(cParams1.targetLength == cParams2.targetLength);
+ assert(cParams1.strategy == cParams2.strategy);
+}
+
+void ZSTD_reset_compressedBlockState(ZSTD_compressedBlockState_t* bs)
+{
+ int i;
+ for (i = 0; i < ZSTD_REP_NUM; ++i)
+ bs->rep[i] = repStartValue[i];
+ bs->entropy.huf.repeatMode = HUF_repeat_none;
+ bs->entropy.fse.offcode_repeatMode = FSE_repeat_none;
+ bs->entropy.fse.matchlength_repeatMode = FSE_repeat_none;
+ bs->entropy.fse.litlength_repeatMode = FSE_repeat_none;
+}
+
+/*! ZSTD_invalidateMatchState()
+ * Invalidate all the matches in the match finder tables.
+ * Requires nextSrc and base to be set (can be NULL).
+ */
+static void ZSTD_invalidateMatchState(ZSTD_matchState_t* ms)
+{
+ ZSTD_window_clear(&ms->window);
+
+ ms->nextToUpdate = ms->window.dictLimit;
+ ms->loadedDictEnd = 0;
+ ms->opt.litLengthSum = 0; /* force reset of btopt stats */
+ ms->dictMatchState = NULL;
+}
+
+/*
+ * Controls, for this matchState reset, whether the tables need to be cleared /
+ * prepared for the coming compression (ZSTDcrp_makeClean), or whether the
+ * tables can be left unclean (ZSTDcrp_leaveDirty), because we know that a
+ * subsequent operation will overwrite the table space anyways (e.g., copying
+ * the matchState contents in from a CDict).
+ */
+typedef enum {
+ ZSTDcrp_makeClean,
+ ZSTDcrp_leaveDirty
+} ZSTD_compResetPolicy_e;
+
+/*
+ * Controls, for this matchState reset, whether indexing can continue where it
+ * left off (ZSTDirp_continue), or whether it needs to be restarted from zero
+ * (ZSTDirp_reset).
+ */
+typedef enum {
+ ZSTDirp_continue,
+ ZSTDirp_reset
+} ZSTD_indexResetPolicy_e;
+
+typedef enum {
+ ZSTD_resetTarget_CDict,
+ ZSTD_resetTarget_CCtx
+} ZSTD_resetTarget_e;
+
+
+static size_t
+ZSTD_reset_matchState(ZSTD_matchState_t* ms,
+ ZSTD_cwksp* ws,
+ const ZSTD_compressionParameters* cParams,
+ const ZSTD_paramSwitch_e useRowMatchFinder,
+ const ZSTD_compResetPolicy_e crp,
+ const ZSTD_indexResetPolicy_e forceResetIndex,
+ const ZSTD_resetTarget_e forWho)
+{
+ /* disable chain table allocation for fast or row-based strategies */
+ size_t const chainSize = ZSTD_allocateChainTable(cParams->strategy, useRowMatchFinder,
+ ms->dedicatedDictSearch && (forWho == ZSTD_resetTarget_CDict))
+ ? ((size_t)1 << cParams->chainLog)
+ : 0;
+ size_t const hSize = ((size_t)1) << cParams->hashLog;
+ U32 const hashLog3 = ((forWho == ZSTD_resetTarget_CCtx) && cParams->minMatch==3) ? MIN(ZSTD_HASHLOG3_MAX, cParams->windowLog) : 0;
+ size_t const h3Size = hashLog3 ? ((size_t)1) << hashLog3 : 0;
+
+ DEBUGLOG(4, "reset indices : %u", forceResetIndex == ZSTDirp_reset);
+ assert(useRowMatchFinder != ZSTD_ps_auto);
+ if (forceResetIndex == ZSTDirp_reset) {
+ ZSTD_window_init(&ms->window);
+ ZSTD_cwksp_mark_tables_dirty(ws);
+ }
+
+ ms->hashLog3 = hashLog3;
+
+ ZSTD_invalidateMatchState(ms);
+
+ assert(!ZSTD_cwksp_reserve_failed(ws)); /* check that allocation hasn't already failed */
+
+ ZSTD_cwksp_clear_tables(ws);
+
+ DEBUGLOG(5, "reserving table space");
+ /* table Space */
+ ms->hashTable = (U32*)ZSTD_cwksp_reserve_table(ws, hSize * sizeof(U32));
+ ms->chainTable = (U32*)ZSTD_cwksp_reserve_table(ws, chainSize * sizeof(U32));
+ ms->hashTable3 = (U32*)ZSTD_cwksp_reserve_table(ws, h3Size * sizeof(U32));
+ RETURN_ERROR_IF(ZSTD_cwksp_reserve_failed(ws), memory_allocation,
+ "failed a workspace allocation in ZSTD_reset_matchState");
+
+ DEBUGLOG(4, "reset table : %u", crp!=ZSTDcrp_leaveDirty);
+ if (crp!=ZSTDcrp_leaveDirty) {
+ /* reset tables only */
+ ZSTD_cwksp_clean_tables(ws);
+ }
+
+ /* opt parser space */
+ if ((forWho == ZSTD_resetTarget_CCtx) && (cParams->strategy >= ZSTD_btopt)) {
+ DEBUGLOG(4, "reserving optimal parser space");
+ ms->opt.litFreq = (unsigned*)ZSTD_cwksp_reserve_aligned(ws, (1<<Litbits) * sizeof(unsigned));
+ ms->opt.litLengthFreq = (unsigned*)ZSTD_cwksp_reserve_aligned(ws, (MaxLL+1) * sizeof(unsigned));
+ ms->opt.matchLengthFreq = (unsigned*)ZSTD_cwksp_reserve_aligned(ws, (MaxML+1) * sizeof(unsigned));
+ ms->opt.offCodeFreq = (unsigned*)ZSTD_cwksp_reserve_aligned(ws, (MaxOff+1) * sizeof(unsigned));
+ ms->opt.matchTable = (ZSTD_match_t*)ZSTD_cwksp_reserve_aligned(ws, (ZSTD_OPT_NUM+1) * sizeof(ZSTD_match_t));
+ ms->opt.priceTable = (ZSTD_optimal_t*)ZSTD_cwksp_reserve_aligned(ws, (ZSTD_OPT_NUM+1) * sizeof(ZSTD_optimal_t));
+ }
+
+ if (ZSTD_rowMatchFinderUsed(cParams->strategy, useRowMatchFinder)) {
+ { /* Row match finder needs an additional table of hashes ("tags") */
+ size_t const tagTableSize = hSize*sizeof(U16);
+ ms->tagTable = (U16*)ZSTD_cwksp_reserve_aligned(ws, tagTableSize);
+ if (ms->tagTable) ZSTD_memset(ms->tagTable, 0, tagTableSize);
+ }
+ { /* Switch to 32-entry rows if searchLog is 5 (or more) */
+ U32 const rowLog = BOUNDED(4, cParams->searchLog, 6);
+ assert(cParams->hashLog >= rowLog);
+ ms->rowHashLog = cParams->hashLog - rowLog;
+ }
+ }
+
+ ms->cParams = *cParams;
+
+ RETURN_ERROR_IF(ZSTD_cwksp_reserve_failed(ws), memory_allocation,
+ "failed a workspace allocation in ZSTD_reset_matchState");
+ return 0;
+}
+
+/* ZSTD_indexTooCloseToMax() :
+ * minor optimization : prefer memset() rather than reduceIndex()
+ * which is measurably slow in some circumstances (reported for Visual Studio).
+ * Works when re-using a context for a lot of smallish inputs :
+ * if all inputs are smaller than ZSTD_INDEXOVERFLOW_MARGIN,
+ * memset() will be triggered before reduceIndex().
+ */
+#define ZSTD_INDEXOVERFLOW_MARGIN (16 MB)
+static int ZSTD_indexTooCloseToMax(ZSTD_window_t w)
+{
+ return (size_t)(w.nextSrc - w.base) > (ZSTD_CURRENT_MAX - ZSTD_INDEXOVERFLOW_MARGIN);
+}
+
+/* ZSTD_dictTooBig():
+ * When dictionaries are larger than ZSTD_CHUNKSIZE_MAX they can't be loaded in
+ * one go generically. So we ensure that in that case we reset the tables to zero,
+ * so that we can load as much of the dictionary as possible.
+ */
+static int ZSTD_dictTooBig(size_t const loadedDictSize)
+{
+ return loadedDictSize > ZSTD_CHUNKSIZE_MAX;
+}
+
+/*! ZSTD_resetCCtx_internal() :
+ * @param loadedDictSize The size of the dictionary to be loaded
+ * into the context, if any. If no dictionary is used, or the
+ * dictionary is being attached / copied, then pass 0.
+ * note : `params` are assumed fully validated at this stage.
+ */
+static size_t ZSTD_resetCCtx_internal(ZSTD_CCtx* zc,
+ ZSTD_CCtx_params const* params,
+ U64 const pledgedSrcSize,
+ size_t const loadedDictSize,
+ ZSTD_compResetPolicy_e const crp,
+ ZSTD_buffered_policy_e const zbuff)
+{
+ ZSTD_cwksp* const ws = &zc->workspace;
+ DEBUGLOG(4, "ZSTD_resetCCtx_internal: pledgedSrcSize=%u, wlog=%u, useRowMatchFinder=%d useBlockSplitter=%d",
+ (U32)pledgedSrcSize, params->cParams.windowLog, (int)params->useRowMatchFinder, (int)params->useBlockSplitter);
+ assert(!ZSTD_isError(ZSTD_checkCParams(params->cParams)));
+
+ zc->isFirstBlock = 1;
+
+ /* Set applied params early so we can modify them for LDM,
+ * and point params at the applied params.
+ */
+ zc->appliedParams = *params;
+ params = &zc->appliedParams;
+
+ assert(params->useRowMatchFinder != ZSTD_ps_auto);
+ assert(params->useBlockSplitter != ZSTD_ps_auto);
+ assert(params->ldmParams.enableLdm != ZSTD_ps_auto);
+ if (params->ldmParams.enableLdm == ZSTD_ps_enable) {
+ /* Adjust long distance matching parameters */
+ ZSTD_ldm_adjustParameters(&zc->appliedParams.ldmParams, &params->cParams);
+ assert(params->ldmParams.hashLog >= params->ldmParams.bucketSizeLog);
+ assert(params->ldmParams.hashRateLog < 32);
+ }
+
+ { size_t const windowSize = MAX(1, (size_t)MIN(((U64)1 << params->cParams.windowLog), pledgedSrcSize));
+ size_t const blockSize = MIN(ZSTD_BLOCKSIZE_MAX, windowSize);
+ U32 const divider = (params->cParams.minMatch==3) ? 3 : 4;
+ size_t const maxNbSeq = blockSize / divider;
+ size_t const buffOutSize = (zbuff == ZSTDb_buffered && params->outBufferMode == ZSTD_bm_buffered)
+ ? ZSTD_compressBound(blockSize) + 1
+ : 0;
+ size_t const buffInSize = (zbuff == ZSTDb_buffered && params->inBufferMode == ZSTD_bm_buffered)
+ ? windowSize + blockSize
+ : 0;
+ size_t const maxNbLdmSeq = ZSTD_ldm_getMaxNbSeq(params->ldmParams, blockSize);
+
+ int const indexTooClose = ZSTD_indexTooCloseToMax(zc->blockState.matchState.window);
+ int const dictTooBig = ZSTD_dictTooBig(loadedDictSize);
+ ZSTD_indexResetPolicy_e needsIndexReset =
+ (indexTooClose || dictTooBig || !zc->initialized) ? ZSTDirp_reset : ZSTDirp_continue;
+
+ size_t const neededSpace =
+ ZSTD_estimateCCtxSize_usingCCtxParams_internal(
+ &params->cParams, &params->ldmParams, zc->staticSize != 0, params->useRowMatchFinder,
+ buffInSize, buffOutSize, pledgedSrcSize);
+ int resizeWorkspace;
+
+ FORWARD_IF_ERROR(neededSpace, "cctx size estimate failed!");
+
+ if (!zc->staticSize) ZSTD_cwksp_bump_oversized_duration(ws, 0);
+
+ { /* Check if workspace is large enough, alloc a new one if needed */
+ int const workspaceTooSmall = ZSTD_cwksp_sizeof(ws) < neededSpace;
+ int const workspaceWasteful = ZSTD_cwksp_check_wasteful(ws, neededSpace);
+ resizeWorkspace = workspaceTooSmall || workspaceWasteful;
+ DEBUGLOG(4, "Need %zu B workspace", neededSpace);
+ DEBUGLOG(4, "windowSize: %zu - blockSize: %zu", windowSize, blockSize);
+
+ if (resizeWorkspace) {
+ DEBUGLOG(4, "Resize workspaceSize from %zuKB to %zuKB",
+ ZSTD_cwksp_sizeof(ws) >> 10,
+ neededSpace >> 10);
+
+ RETURN_ERROR_IF(zc->staticSize, memory_allocation, "static cctx : no resize");
+
+ needsIndexReset = ZSTDirp_reset;
+
+ ZSTD_cwksp_free(ws, zc->customMem);
+ FORWARD_IF_ERROR(ZSTD_cwksp_create(ws, neededSpace, zc->customMem), "");
+
+ DEBUGLOG(5, "reserving object space");
+ /* Statically sized space.
+ * entropyWorkspace never moves,
+ * though prev/next block swap places */
+ assert(ZSTD_cwksp_check_available(ws, 2 * sizeof(ZSTD_compressedBlockState_t)));
+ zc->blockState.prevCBlock = (ZSTD_compressedBlockState_t*) ZSTD_cwksp_reserve_object(ws, sizeof(ZSTD_compressedBlockState_t));
+ RETURN_ERROR_IF(zc->blockState.prevCBlock == NULL, memory_allocation, "couldn't allocate prevCBlock");
+ zc->blockState.nextCBlock = (ZSTD_compressedBlockState_t*) ZSTD_cwksp_reserve_object(ws, sizeof(ZSTD_compressedBlockState_t));
+ RETURN_ERROR_IF(zc->blockState.nextCBlock == NULL, memory_allocation, "couldn't allocate nextCBlock");
+ zc->entropyWorkspace = (U32*) ZSTD_cwksp_reserve_object(ws, ENTROPY_WORKSPACE_SIZE);
+ RETURN_ERROR_IF(zc->entropyWorkspace == NULL, memory_allocation, "couldn't allocate entropyWorkspace");
+ } }
+
+ ZSTD_cwksp_clear(ws);
+
+ /* init params */
+ zc->blockState.matchState.cParams = params->cParams;
+ zc->pledgedSrcSizePlusOne = pledgedSrcSize+1;
+ zc->consumedSrcSize = 0;
+ zc->producedCSize = 0;
+ if (pledgedSrcSize == ZSTD_CONTENTSIZE_UNKNOWN)
+ zc->appliedParams.fParams.contentSizeFlag = 0;
+ DEBUGLOG(4, "pledged content size : %u ; flag : %u",
+ (unsigned)pledgedSrcSize, zc->appliedParams.fParams.contentSizeFlag);
+ zc->blockSize = blockSize;
+
+ xxh64_reset(&zc->xxhState, 0);
+ zc->stage = ZSTDcs_init;
+ zc->dictID = 0;
+ zc->dictContentSize = 0;
+
+ ZSTD_reset_compressedBlockState(zc->blockState.prevCBlock);
+
+ /* ZSTD_wildcopy() is used to copy into the literals buffer,
+ * so we have to oversize the buffer by WILDCOPY_OVERLENGTH bytes.
+ */
+ zc->seqStore.litStart = ZSTD_cwksp_reserve_buffer(ws, blockSize + WILDCOPY_OVERLENGTH);
+ zc->seqStore.maxNbLit = blockSize;
+
+ /* buffers */
+ zc->bufferedPolicy = zbuff;
+ zc->inBuffSize = buffInSize;
+ zc->inBuff = (char*)ZSTD_cwksp_reserve_buffer(ws, buffInSize);
+ zc->outBuffSize = buffOutSize;
+ zc->outBuff = (char*)ZSTD_cwksp_reserve_buffer(ws, buffOutSize);
+
+ /* ldm bucketOffsets table */
+ if (params->ldmParams.enableLdm == ZSTD_ps_enable) {
+ /* TODO: avoid memset? */
+ size_t const numBuckets =
+ ((size_t)1) << (params->ldmParams.hashLog -
+ params->ldmParams.bucketSizeLog);
+ zc->ldmState.bucketOffsets = ZSTD_cwksp_reserve_buffer(ws, numBuckets);
+ ZSTD_memset(zc->ldmState.bucketOffsets, 0, numBuckets);
+ }
+
+ /* sequences storage */
+ ZSTD_referenceExternalSequences(zc, NULL, 0);
+ zc->seqStore.maxNbSeq = maxNbSeq;
+ zc->seqStore.llCode = ZSTD_cwksp_reserve_buffer(ws, maxNbSeq * sizeof(BYTE));
+ zc->seqStore.mlCode = ZSTD_cwksp_reserve_buffer(ws, maxNbSeq * sizeof(BYTE));
+ zc->seqStore.ofCode = ZSTD_cwksp_reserve_buffer(ws, maxNbSeq * sizeof(BYTE));
+ zc->seqStore.sequencesStart = (seqDef*)ZSTD_cwksp_reserve_aligned(ws, maxNbSeq * sizeof(seqDef));
+
+ FORWARD_IF_ERROR(ZSTD_reset_matchState(
+ &zc->blockState.matchState,
+ ws,
+ &params->cParams,
+ params->useRowMatchFinder,
+ crp,
+ needsIndexReset,
+ ZSTD_resetTarget_CCtx), "");
+
+ /* ldm hash table */
+ if (params->ldmParams.enableLdm == ZSTD_ps_enable) {
+ /* TODO: avoid memset? */
+ size_t const ldmHSize = ((size_t)1) << params->ldmParams.hashLog;
+ zc->ldmState.hashTable = (ldmEntry_t*)ZSTD_cwksp_reserve_aligned(ws, ldmHSize * sizeof(ldmEntry_t));
+ ZSTD_memset(zc->ldmState.hashTable, 0, ldmHSize * sizeof(ldmEntry_t));
+ zc->ldmSequences = (rawSeq*)ZSTD_cwksp_reserve_aligned(ws, maxNbLdmSeq * sizeof(rawSeq));
+ zc->maxNbLdmSequences = maxNbLdmSeq;
+
+ ZSTD_window_init(&zc->ldmState.window);
+ zc->ldmState.loadedDictEnd = 0;
+ }
+
+ DEBUGLOG(3, "wksp: finished allocating, %zd bytes remain available", ZSTD_cwksp_available_space(ws));
+ assert(ZSTD_cwksp_estimated_space_within_bounds(ws, neededSpace, resizeWorkspace));
+
+ zc->initialized = 1;
+
+ return 0;
+ }
+}
+
+/* ZSTD_invalidateRepCodes() :
+ * ensures next compression will not use repcodes from previous block.
+ * Note : only works with regular variant;
+ * do not use with extDict variant ! */
+void ZSTD_invalidateRepCodes(ZSTD_CCtx* cctx) {
+ int i;
+ for (i=0; i<ZSTD_REP_NUM; i++) cctx->blockState.prevCBlock->rep[i] = 0;
+ assert(!ZSTD_window_hasExtDict(cctx->blockState.matchState.window));
+}
+
+/* These are the approximate sizes for each strategy past which copying the
+ * dictionary tables into the working context is faster than using them
+ * in-place.
+ */
+static const size_t attachDictSizeCutoffs[ZSTD_STRATEGY_MAX+1] = {
+ 8 KB, /* unused */
+ 8 KB, /* ZSTD_fast */
+ 16 KB, /* ZSTD_dfast */
+ 32 KB, /* ZSTD_greedy */
+ 32 KB, /* ZSTD_lazy */
+ 32 KB, /* ZSTD_lazy2 */
+ 32 KB, /* ZSTD_btlazy2 */
+ 32 KB, /* ZSTD_btopt */
+ 8 KB, /* ZSTD_btultra */
+ 8 KB /* ZSTD_btultra2 */
+};
+
+static int ZSTD_shouldAttachDict(const ZSTD_CDict* cdict,
+ const ZSTD_CCtx_params* params,
+ U64 pledgedSrcSize)
+{
+ size_t cutoff = attachDictSizeCutoffs[cdict->matchState.cParams.strategy];
+ int const dedicatedDictSearch = cdict->matchState.dedicatedDictSearch;
+ return dedicatedDictSearch
+ || ( ( pledgedSrcSize <= cutoff
+ || pledgedSrcSize == ZSTD_CONTENTSIZE_UNKNOWN
+ || params->attachDictPref == ZSTD_dictForceAttach )
+ && params->attachDictPref != ZSTD_dictForceCopy
+ && !params->forceWindow ); /* dictMatchState isn't correctly
+ * handled in _enforceMaxDist */
+}
+
+static size_t
+ZSTD_resetCCtx_byAttachingCDict(ZSTD_CCtx* cctx,
+ const ZSTD_CDict* cdict,
+ ZSTD_CCtx_params params,
+ U64 pledgedSrcSize,
+ ZSTD_buffered_policy_e zbuff)
+{
+ DEBUGLOG(4, "ZSTD_resetCCtx_byAttachingCDict() pledgedSrcSize=%llu",
+ (unsigned long long)pledgedSrcSize);
+ {
+ ZSTD_compressionParameters adjusted_cdict_cParams = cdict->matchState.cParams;
+ unsigned const windowLog = params.cParams.windowLog;
+ assert(windowLog != 0);
+ /* Resize working context table params for input only, since the dict
+ * has its own tables. */
+ /* pledgedSrcSize == 0 means 0! */
+
+ if (cdict->matchState.dedicatedDictSearch) {
+ ZSTD_dedicatedDictSearch_revertCParams(&adjusted_cdict_cParams);
+ }
+
+ params.cParams = ZSTD_adjustCParams_internal(adjusted_cdict_cParams, pledgedSrcSize,
+ cdict->dictContentSize, ZSTD_cpm_attachDict);
+ params.cParams.windowLog = windowLog;
+ params.useRowMatchFinder = cdict->useRowMatchFinder; /* cdict overrides */
+ FORWARD_IF_ERROR(ZSTD_resetCCtx_internal(cctx, &params, pledgedSrcSize,
+ /* loadedDictSize */ 0,
+ ZSTDcrp_makeClean, zbuff), "");
+ assert(cctx->appliedParams.cParams.strategy == adjusted_cdict_cParams.strategy);
+ }
+
+ { const U32 cdictEnd = (U32)( cdict->matchState.window.nextSrc
+ - cdict->matchState.window.base);
+ const U32 cdictLen = cdictEnd - cdict->matchState.window.dictLimit;
+ if (cdictLen == 0) {
+ /* don't even attach dictionaries with no contents */
+ DEBUGLOG(4, "skipping attaching empty dictionary");
+ } else {
+ DEBUGLOG(4, "attaching dictionary into context");
+ cctx->blockState.matchState.dictMatchState = &cdict->matchState;
+
+ /* prep working match state so dict matches never have negative indices
+ * when they are translated to the working context's index space. */
+ if (cctx->blockState.matchState.window.dictLimit < cdictEnd) {
+ cctx->blockState.matchState.window.nextSrc =
+ cctx->blockState.matchState.window.base + cdictEnd;
+ ZSTD_window_clear(&cctx->blockState.matchState.window);
+ }
+ /* loadedDictEnd is expressed within the referential of the active context */
+ cctx->blockState.matchState.loadedDictEnd = cctx->blockState.matchState.window.dictLimit;
+ } }
+
+ cctx->dictID = cdict->dictID;
+ cctx->dictContentSize = cdict->dictContentSize;
+
+ /* copy block state */
+ ZSTD_memcpy(cctx->blockState.prevCBlock, &cdict->cBlockState, sizeof(cdict->cBlockState));
+
+ return 0;
+}
+
+static size_t ZSTD_resetCCtx_byCopyingCDict(ZSTD_CCtx* cctx,
+ const ZSTD_CDict* cdict,
+ ZSTD_CCtx_params params,
+ U64 pledgedSrcSize,
+ ZSTD_buffered_policy_e zbuff)
+{
+ const ZSTD_compressionParameters *cdict_cParams = &cdict->matchState.cParams;
+
+ assert(!cdict->matchState.dedicatedDictSearch);
+ DEBUGLOG(4, "ZSTD_resetCCtx_byCopyingCDict() pledgedSrcSize=%llu",
+ (unsigned long long)pledgedSrcSize);
+
+ { unsigned const windowLog = params.cParams.windowLog;
+ assert(windowLog != 0);
+ /* Copy only compression parameters related to tables. */
+ params.cParams = *cdict_cParams;
+ params.cParams.windowLog = windowLog;
+ params.useRowMatchFinder = cdict->useRowMatchFinder;
+ FORWARD_IF_ERROR(ZSTD_resetCCtx_internal(cctx, &params, pledgedSrcSize,
+ /* loadedDictSize */ 0,
+ ZSTDcrp_leaveDirty, zbuff), "");
+ assert(cctx->appliedParams.cParams.strategy == cdict_cParams->strategy);
+ assert(cctx->appliedParams.cParams.hashLog == cdict_cParams->hashLog);
+ assert(cctx->appliedParams.cParams.chainLog == cdict_cParams->chainLog);
+ }
+
+ ZSTD_cwksp_mark_tables_dirty(&cctx->workspace);
+ assert(params.useRowMatchFinder != ZSTD_ps_auto);
+
+ /* copy tables */
+ { size_t const chainSize = ZSTD_allocateChainTable(cdict_cParams->strategy, cdict->useRowMatchFinder, 0 /* DDS guaranteed disabled */)
+ ? ((size_t)1 << cdict_cParams->chainLog)
+ : 0;
+ size_t const hSize = (size_t)1 << cdict_cParams->hashLog;
+
+ ZSTD_memcpy(cctx->blockState.matchState.hashTable,
+ cdict->matchState.hashTable,
+ hSize * sizeof(U32));
+ /* Do not copy cdict's chainTable if cctx has parameters such that it would not use chainTable */
+ if (ZSTD_allocateChainTable(cctx->appliedParams.cParams.strategy, cctx->appliedParams.useRowMatchFinder, 0 /* forDDSDict */)) {
+ ZSTD_memcpy(cctx->blockState.matchState.chainTable,
+ cdict->matchState.chainTable,
+ chainSize * sizeof(U32));
+ }
+ /* copy tag table */
+ if (ZSTD_rowMatchFinderUsed(cdict_cParams->strategy, cdict->useRowMatchFinder)) {
+ size_t const tagTableSize = hSize*sizeof(U16);
+ ZSTD_memcpy(cctx->blockState.matchState.tagTable,
+ cdict->matchState.tagTable,
+ tagTableSize);
+ }
+ }
+
+ /* Zero the hashTable3, since the cdict never fills it */
+ { int const h3log = cctx->blockState.matchState.hashLog3;
+ size_t const h3Size = h3log ? ((size_t)1 << h3log) : 0;
+ assert(cdict->matchState.hashLog3 == 0);
+ ZSTD_memset(cctx->blockState.matchState.hashTable3, 0, h3Size * sizeof(U32));
+ }
+
+ ZSTD_cwksp_mark_tables_clean(&cctx->workspace);
+
+ /* copy dictionary offsets */
+ { ZSTD_matchState_t const* srcMatchState = &cdict->matchState;
+ ZSTD_matchState_t* dstMatchState = &cctx->blockState.matchState;
+ dstMatchState->window = srcMatchState->window;
+ dstMatchState->nextToUpdate = srcMatchState->nextToUpdate;
+ dstMatchState->loadedDictEnd= srcMatchState->loadedDictEnd;
+ }
+
+ cctx->dictID = cdict->dictID;
+ cctx->dictContentSize = cdict->dictContentSize;
+
+ /* copy block state */
+ ZSTD_memcpy(cctx->blockState.prevCBlock, &cdict->cBlockState, sizeof(cdict->cBlockState));
+
+ return 0;
+}
+
+/* We have a choice between copying the dictionary context into the working
+ * context, or referencing the dictionary context from the working context
+ * in-place. We decide here which strategy to use. */
+static size_t ZSTD_resetCCtx_usingCDict(ZSTD_CCtx* cctx,
+ const ZSTD_CDict* cdict,
+ const ZSTD_CCtx_params* params,
+ U64 pledgedSrcSize,
+ ZSTD_buffered_policy_e zbuff)
+{
+
+ DEBUGLOG(4, "ZSTD_resetCCtx_usingCDict (pledgedSrcSize=%u)",
+ (unsigned)pledgedSrcSize);
+
+ if (ZSTD_shouldAttachDict(cdict, params, pledgedSrcSize)) {
+ return ZSTD_resetCCtx_byAttachingCDict(
+ cctx, cdict, *params, pledgedSrcSize, zbuff);
+ } else {
+ return ZSTD_resetCCtx_byCopyingCDict(
+ cctx, cdict, *params, pledgedSrcSize, zbuff);
+ }
+}
+
+/*! ZSTD_copyCCtx_internal() :
+ * Duplicate an existing context `srcCCtx` into another one `dstCCtx`.
+ * Only works during stage ZSTDcs_init (i.e. after creation, but before first call to ZSTD_compressContinue()).
+ * The "context", in this case, refers to the hash and chain tables,
+ * entropy tables, and dictionary references.
+ * `windowLog` value is enforced if != 0, otherwise value is copied from srcCCtx.
+ * @return : 0, or an error code */
+static size_t ZSTD_copyCCtx_internal(ZSTD_CCtx* dstCCtx,
+ const ZSTD_CCtx* srcCCtx,
+ ZSTD_frameParameters fParams,
+ U64 pledgedSrcSize,
+ ZSTD_buffered_policy_e zbuff)
+{
+ RETURN_ERROR_IF(srcCCtx->stage!=ZSTDcs_init, stage_wrong,
+ "Can't copy a ctx that's not in init stage.");
+ DEBUGLOG(5, "ZSTD_copyCCtx_internal");
+ ZSTD_memcpy(&dstCCtx->customMem, &srcCCtx->customMem, sizeof(ZSTD_customMem));
+ { ZSTD_CCtx_params params = dstCCtx->requestedParams;
+ /* Copy only compression parameters related to tables. */
+ params.cParams = srcCCtx->appliedParams.cParams;
+ assert(srcCCtx->appliedParams.useRowMatchFinder != ZSTD_ps_auto);
+ assert(srcCCtx->appliedParams.useBlockSplitter != ZSTD_ps_auto);
+ assert(srcCCtx->appliedParams.ldmParams.enableLdm != ZSTD_ps_auto);
+ params.useRowMatchFinder = srcCCtx->appliedParams.useRowMatchFinder;
+ params.useBlockSplitter = srcCCtx->appliedParams.useBlockSplitter;
+ params.ldmParams = srcCCtx->appliedParams.ldmParams;
+ params.fParams = fParams;
+ ZSTD_resetCCtx_internal(dstCCtx, &params, pledgedSrcSize,
+ /* loadedDictSize */ 0,
+ ZSTDcrp_leaveDirty, zbuff);
+ assert(dstCCtx->appliedParams.cParams.windowLog == srcCCtx->appliedParams.cParams.windowLog);
+ assert(dstCCtx->appliedParams.cParams.strategy == srcCCtx->appliedParams.cParams.strategy);
+ assert(dstCCtx->appliedParams.cParams.hashLog == srcCCtx->appliedParams.cParams.hashLog);
+ assert(dstCCtx->appliedParams.cParams.chainLog == srcCCtx->appliedParams.cParams.chainLog);
+ assert(dstCCtx->blockState.matchState.hashLog3 == srcCCtx->blockState.matchState.hashLog3);
+ }
+
+ ZSTD_cwksp_mark_tables_dirty(&dstCCtx->workspace);
+
+ /* copy tables */
+ { size_t const chainSize = ZSTD_allocateChainTable(srcCCtx->appliedParams.cParams.strategy,
+ srcCCtx->appliedParams.useRowMatchFinder,
+ 0 /* forDDSDict */)
+ ? ((size_t)1 << srcCCtx->appliedParams.cParams.chainLog)
+ : 0;
+ size_t const hSize = (size_t)1 << srcCCtx->appliedParams.cParams.hashLog;
+ int const h3log = srcCCtx->blockState.matchState.hashLog3;
+ size_t const h3Size = h3log ? ((size_t)1 << h3log) : 0;
+
+ ZSTD_memcpy(dstCCtx->blockState.matchState.hashTable,
+ srcCCtx->blockState.matchState.hashTable,
+ hSize * sizeof(U32));
+ ZSTD_memcpy(dstCCtx->blockState.matchState.chainTable,
+ srcCCtx->blockState.matchState.chainTable,
+ chainSize * sizeof(U32));
+ ZSTD_memcpy(dstCCtx->blockState.matchState.hashTable3,
+ srcCCtx->blockState.matchState.hashTable3,
+ h3Size * sizeof(U32));
+ }
+
+ ZSTD_cwksp_mark_tables_clean(&dstCCtx->workspace);
+
+ /* copy dictionary offsets */
+ {
+ const ZSTD_matchState_t* srcMatchState = &srcCCtx->blockState.matchState;
+ ZSTD_matchState_t* dstMatchState = &dstCCtx->blockState.matchState;
+ dstMatchState->window = srcMatchState->window;
+ dstMatchState->nextToUpdate = srcMatchState->nextToUpdate;
+ dstMatchState->loadedDictEnd= srcMatchState->loadedDictEnd;
+ }
+ dstCCtx->dictID = srcCCtx->dictID;
+ dstCCtx->dictContentSize = srcCCtx->dictContentSize;
+
+ /* copy block state */
+ ZSTD_memcpy(dstCCtx->blockState.prevCBlock, srcCCtx->blockState.prevCBlock, sizeof(*srcCCtx->blockState.prevCBlock));
+
+ return 0;
+}
+
+/*! ZSTD_copyCCtx() :
+ * Duplicate an existing context `srcCCtx` into another one `dstCCtx`.
+ * Only works during stage ZSTDcs_init (i.e. after creation, but before first call to ZSTD_compressContinue()).
+ * pledgedSrcSize==0 means "unknown".
+* @return : 0, or an error code */
+size_t ZSTD_copyCCtx(ZSTD_CCtx* dstCCtx, const ZSTD_CCtx* srcCCtx, unsigned long long pledgedSrcSize)
+{
+ ZSTD_frameParameters fParams = { 1 /*content*/, 0 /*checksum*/, 0 /*noDictID*/ };
+ ZSTD_buffered_policy_e const zbuff = srcCCtx->bufferedPolicy;
+ ZSTD_STATIC_ASSERT((U32)ZSTDb_buffered==1);
+ if (pledgedSrcSize==0) pledgedSrcSize = ZSTD_CONTENTSIZE_UNKNOWN;
+ fParams.contentSizeFlag = (pledgedSrcSize != ZSTD_CONTENTSIZE_UNKNOWN);
+
+ return ZSTD_copyCCtx_internal(dstCCtx, srcCCtx,
+ fParams, pledgedSrcSize,
+ zbuff);
+}
+
+
+#define ZSTD_ROWSIZE 16
+/*! ZSTD_reduceTable() :
+ * reduce table indexes by `reducerValue`, or squash to zero.
+ * PreserveMark preserves "unsorted mark" for btlazy2 strategy.
+ * It must be set to a clear 0/1 value, to remove branch during inlining.
+ * Presume table size is a multiple of ZSTD_ROWSIZE
+ * to help auto-vectorization */
+FORCE_INLINE_TEMPLATE void
+ZSTD_reduceTable_internal (U32* const table, U32 const size, U32 const reducerValue, int const preserveMark)
+{
+ int const nbRows = (int)size / ZSTD_ROWSIZE;
+ int cellNb = 0;
+ int rowNb;
+ /* Protect special index values < ZSTD_WINDOW_START_INDEX. */
+ U32 const reducerThreshold = reducerValue + ZSTD_WINDOW_START_INDEX;
+ assert((size & (ZSTD_ROWSIZE-1)) == 0); /* multiple of ZSTD_ROWSIZE */
+ assert(size < (1U<<31)); /* can be casted to int */
+
+
+ for (rowNb=0 ; rowNb < nbRows ; rowNb++) {
+ int column;
+ for (column=0; column<ZSTD_ROWSIZE; column++) {
+ U32 newVal;
+ if (preserveMark && table[cellNb] == ZSTD_DUBT_UNSORTED_MARK) {
+ /* This write is pointless, but is required(?) for the compiler
+ * to auto-vectorize the loop. */
+ newVal = ZSTD_DUBT_UNSORTED_MARK;
+ } else if (table[cellNb] < reducerThreshold) {
+ newVal = 0;
+ } else {
+ newVal = table[cellNb] - reducerValue;
+ }
+ table[cellNb] = newVal;
+ cellNb++;
+ } }
+}
+
+static void ZSTD_reduceTable(U32* const table, U32 const size, U32 const reducerValue)
+{
+ ZSTD_reduceTable_internal(table, size, reducerValue, 0);
+}
+
+static void ZSTD_reduceTable_btlazy2(U32* const table, U32 const size, U32 const reducerValue)
+{
+ ZSTD_reduceTable_internal(table, size, reducerValue, 1);
+}
+
+/*! ZSTD_reduceIndex() :
+* rescale all indexes to avoid future overflow (indexes are U32) */
+static void ZSTD_reduceIndex (ZSTD_matchState_t* ms, ZSTD_CCtx_params const* params, const U32 reducerValue)
+{
+ { U32 const hSize = (U32)1 << params->cParams.hashLog;
+ ZSTD_reduceTable(ms->hashTable, hSize, reducerValue);
+ }
+
+ if (ZSTD_allocateChainTable(params->cParams.strategy, params->useRowMatchFinder, (U32)ms->dedicatedDictSearch)) {
+ U32 const chainSize = (U32)1 << params->cParams.chainLog;
+ if (params->cParams.strategy == ZSTD_btlazy2)
+ ZSTD_reduceTable_btlazy2(ms->chainTable, chainSize, reducerValue);
+ else
+ ZSTD_reduceTable(ms->chainTable, chainSize, reducerValue);
+ }
+
+ if (ms->hashLog3) {
+ U32 const h3Size = (U32)1 << ms->hashLog3;
+ ZSTD_reduceTable(ms->hashTable3, h3Size, reducerValue);
+ }
+}
+
+
+/*-*******************************************************
+* Block entropic compression
+*********************************************************/
+
+/* See doc/zstd_compression_format.md for detailed format description */
+
+void ZSTD_seqToCodes(const seqStore_t* seqStorePtr)
+{
+ const seqDef* const sequences = seqStorePtr->sequencesStart;
+ BYTE* const llCodeTable = seqStorePtr->llCode;
+ BYTE* const ofCodeTable = seqStorePtr->ofCode;
+ BYTE* const mlCodeTable = seqStorePtr->mlCode;
+ U32 const nbSeq = (U32)(seqStorePtr->sequences - seqStorePtr->sequencesStart);
+ U32 u;
+ assert(nbSeq <= seqStorePtr->maxNbSeq);
+ for (u=0; u<nbSeq; u++) {
+ U32 const llv = sequences[u].litLength;
+ U32 const mlv = sequences[u].mlBase;
+ llCodeTable[u] = (BYTE)ZSTD_LLcode(llv);
+ ofCodeTable[u] = (BYTE)ZSTD_highbit32(sequences[u].offBase);
+ mlCodeTable[u] = (BYTE)ZSTD_MLcode(mlv);
+ }
+ if (seqStorePtr->longLengthType==ZSTD_llt_literalLength)
+ llCodeTable[seqStorePtr->longLengthPos] = MaxLL;
+ if (seqStorePtr->longLengthType==ZSTD_llt_matchLength)
+ mlCodeTable[seqStorePtr->longLengthPos] = MaxML;
+}
+
+/* ZSTD_useTargetCBlockSize():
+ * Returns if target compressed block size param is being used.
+ * If used, compression will do best effort to make a compressed block size to be around targetCBlockSize.
+ * Returns 1 if true, 0 otherwise. */
+static int ZSTD_useTargetCBlockSize(const ZSTD_CCtx_params* cctxParams)
+{
+ DEBUGLOG(5, "ZSTD_useTargetCBlockSize (targetCBlockSize=%zu)", cctxParams->targetCBlockSize);
+ return (cctxParams->targetCBlockSize != 0);
+}
+
+/* ZSTD_blockSplitterEnabled():
+ * Returns if block splitting param is being used
+ * If used, compression will do best effort to split a block in order to improve compression ratio.
+ * At the time this function is called, the parameter must be finalized.
+ * Returns 1 if true, 0 otherwise. */
+static int ZSTD_blockSplitterEnabled(ZSTD_CCtx_params* cctxParams)
+{
+ DEBUGLOG(5, "ZSTD_blockSplitterEnabled (useBlockSplitter=%d)", cctxParams->useBlockSplitter);
+ assert(cctxParams->useBlockSplitter != ZSTD_ps_auto);
+ return (cctxParams->useBlockSplitter == ZSTD_ps_enable);
+}
+
+/* Type returned by ZSTD_buildSequencesStatistics containing finalized symbol encoding types
+ * and size of the sequences statistics
+ */
+typedef struct {
+ U32 LLtype;
+ U32 Offtype;
+ U32 MLtype;
+ size_t size;
+ size_t lastCountSize; /* Accounts for bug in 1.3.4. More detail in ZSTD_entropyCompressSeqStore_internal() */
+} ZSTD_symbolEncodingTypeStats_t;
+
+/* ZSTD_buildSequencesStatistics():
+ * Returns a ZSTD_symbolEncodingTypeStats_t, or a zstd error code in the `size` field.
+ * Modifies `nextEntropy` to have the appropriate values as a side effect.
+ * nbSeq must be greater than 0.
+ *
+ * entropyWkspSize must be of size at least ENTROPY_WORKSPACE_SIZE - (MaxSeq + 1)*sizeof(U32)
+ */
+static ZSTD_symbolEncodingTypeStats_t
+ZSTD_buildSequencesStatistics(seqStore_t* seqStorePtr, size_t nbSeq,
+ const ZSTD_fseCTables_t* prevEntropy, ZSTD_fseCTables_t* nextEntropy,
+ BYTE* dst, const BYTE* const dstEnd,
+ ZSTD_strategy strategy, unsigned* countWorkspace,
+ void* entropyWorkspace, size_t entropyWkspSize) {
+ BYTE* const ostart = dst;
+ const BYTE* const oend = dstEnd;
+ BYTE* op = ostart;
+ FSE_CTable* CTable_LitLength = nextEntropy->litlengthCTable;
+ FSE_CTable* CTable_OffsetBits = nextEntropy->offcodeCTable;
+ FSE_CTable* CTable_MatchLength = nextEntropy->matchlengthCTable;
+ const BYTE* const ofCodeTable = seqStorePtr->ofCode;
+ const BYTE* const llCodeTable = seqStorePtr->llCode;
+ const BYTE* const mlCodeTable = seqStorePtr->mlCode;
+ ZSTD_symbolEncodingTypeStats_t stats;
+
+ stats.lastCountSize = 0;
+ /* convert length/distances into codes */
+ ZSTD_seqToCodes(seqStorePtr);
+ assert(op <= oend);
+ assert(nbSeq != 0); /* ZSTD_selectEncodingType() divides by nbSeq */
+ /* build CTable for Literal Lengths */
+ { unsigned max = MaxLL;
+ size_t const mostFrequent = HIST_countFast_wksp(countWorkspace, &max, llCodeTable, nbSeq, entropyWorkspace, entropyWkspSize); /* can't fail */
+ DEBUGLOG(5, "Building LL table");
+ nextEntropy->litlength_repeatMode = prevEntropy->litlength_repeatMode;
+ stats.LLtype = ZSTD_selectEncodingType(&nextEntropy->litlength_repeatMode,
+ countWorkspace, max, mostFrequent, nbSeq,
+ LLFSELog, prevEntropy->litlengthCTable,
+ LL_defaultNorm, LL_defaultNormLog,
+ ZSTD_defaultAllowed, strategy);
+ assert(set_basic < set_compressed && set_rle < set_compressed);
+ assert(!(stats.LLtype < set_compressed && nextEntropy->litlength_repeatMode != FSE_repeat_none)); /* We don't copy tables */
+ { size_t const countSize = ZSTD_buildCTable(
+ op, (size_t)(oend - op),
+ CTable_LitLength, LLFSELog, (symbolEncodingType_e)stats.LLtype,
+ countWorkspace, max, llCodeTable, nbSeq,
+ LL_defaultNorm, LL_defaultNormLog, MaxLL,
+ prevEntropy->litlengthCTable,
+ sizeof(prevEntropy->litlengthCTable),
+ entropyWorkspace, entropyWkspSize);
+ if (ZSTD_isError(countSize)) {
+ DEBUGLOG(3, "ZSTD_buildCTable for LitLens failed");
+ stats.size = countSize;
+ return stats;
+ }
+ if (stats.LLtype == set_compressed)
+ stats.lastCountSize = countSize;
+ op += countSize;
+ assert(op <= oend);
+ } }
+ /* build CTable for Offsets */
+ { unsigned max = MaxOff;
+ size_t const mostFrequent = HIST_countFast_wksp(
+ countWorkspace, &max, ofCodeTable, nbSeq, entropyWorkspace, entropyWkspSize); /* can't fail */
+ /* We can only use the basic table if max <= DefaultMaxOff, otherwise the offsets are too large */
+ ZSTD_defaultPolicy_e const defaultPolicy = (max <= DefaultMaxOff) ? ZSTD_defaultAllowed : ZSTD_defaultDisallowed;
+ DEBUGLOG(5, "Building OF table");
+ nextEntropy->offcode_repeatMode = prevEntropy->offcode_repeatMode;
+ stats.Offtype = ZSTD_selectEncodingType(&nextEntropy->offcode_repeatMode,
+ countWorkspace, max, mostFrequent, nbSeq,
+ OffFSELog, prevEntropy->offcodeCTable,
+ OF_defaultNorm, OF_defaultNormLog,
+ defaultPolicy, strategy);
+ assert(!(stats.Offtype < set_compressed && nextEntropy->offcode_repeatMode != FSE_repeat_none)); /* We don't copy tables */
+ { size_t const countSize = ZSTD_buildCTable(
+ op, (size_t)(oend - op),
+ CTable_OffsetBits, OffFSELog, (symbolEncodingType_e)stats.Offtype,
+ countWorkspace, max, ofCodeTable, nbSeq,
+ OF_defaultNorm, OF_defaultNormLog, DefaultMaxOff,
+ prevEntropy->offcodeCTable,
+ sizeof(prevEntropy->offcodeCTable),
+ entropyWorkspace, entropyWkspSize);
+ if (ZSTD_isError(countSize)) {
+ DEBUGLOG(3, "ZSTD_buildCTable for Offsets failed");
+ stats.size = countSize;
+ return stats;
+ }
+ if (stats.Offtype == set_compressed)
+ stats.lastCountSize = countSize;
+ op += countSize;
+ assert(op <= oend);
+ } }
+ /* build CTable for MatchLengths */
+ { unsigned max = MaxML;
+ size_t const mostFrequent = HIST_countFast_wksp(
+ countWorkspace, &max, mlCodeTable, nbSeq, entropyWorkspace, entropyWkspSize); /* can't fail */
+ DEBUGLOG(5, "Building ML table (remaining space : %i)", (int)(oend-op));
+ nextEntropy->matchlength_repeatMode = prevEntropy->matchlength_repeatMode;
+ stats.MLtype = ZSTD_selectEncodingType(&nextEntropy->matchlength_repeatMode,
+ countWorkspace, max, mostFrequent, nbSeq,
+ MLFSELog, prevEntropy->matchlengthCTable,
+ ML_defaultNorm, ML_defaultNormLog,
+ ZSTD_defaultAllowed, strategy);
+ assert(!(stats.MLtype < set_compressed && nextEntropy->matchlength_repeatMode != FSE_repeat_none)); /* We don't copy tables */
+ { size_t const countSize = ZSTD_buildCTable(
+ op, (size_t)(oend - op),
+ CTable_MatchLength, MLFSELog, (symbolEncodingType_e)stats.MLtype,
+ countWorkspace, max, mlCodeTable, nbSeq,
+ ML_defaultNorm, ML_defaultNormLog, MaxML,
+ prevEntropy->matchlengthCTable,
+ sizeof(prevEntropy->matchlengthCTable),
+ entropyWorkspace, entropyWkspSize);
+ if (ZSTD_isError(countSize)) {
+ DEBUGLOG(3, "ZSTD_buildCTable for MatchLengths failed");
+ stats.size = countSize;
+ return stats;
+ }
+ if (stats.MLtype == set_compressed)
+ stats.lastCountSize = countSize;
+ op += countSize;
+ assert(op <= oend);
+ } }
+ stats.size = (size_t)(op-ostart);
+ return stats;
+}
+
+/* ZSTD_entropyCompressSeqStore_internal():
+ * compresses both literals and sequences
+ * Returns compressed size of block, or a zstd error.
+ */
+#define SUSPECT_UNCOMPRESSIBLE_LITERAL_RATIO 20
+MEM_STATIC size_t
+ZSTD_entropyCompressSeqStore_internal(seqStore_t* seqStorePtr,
+ const ZSTD_entropyCTables_t* prevEntropy,
+ ZSTD_entropyCTables_t* nextEntropy,
+ const ZSTD_CCtx_params* cctxParams,
+ void* dst, size_t dstCapacity,
+ void* entropyWorkspace, size_t entropyWkspSize,
+ const int bmi2)
+{
+ const int longOffsets = cctxParams->cParams.windowLog > STREAM_ACCUMULATOR_MIN;
+ ZSTD_strategy const strategy = cctxParams->cParams.strategy;
+ unsigned* count = (unsigned*)entropyWorkspace;
+ FSE_CTable* CTable_LitLength = nextEntropy->fse.litlengthCTable;
+ FSE_CTable* CTable_OffsetBits = nextEntropy->fse.offcodeCTable;
+ FSE_CTable* CTable_MatchLength = nextEntropy->fse.matchlengthCTable;
+ const seqDef* const sequences = seqStorePtr->sequencesStart;
+ const size_t nbSeq = seqStorePtr->sequences - seqStorePtr->sequencesStart;
+ const BYTE* const ofCodeTable = seqStorePtr->ofCode;
+ const BYTE* const llCodeTable = seqStorePtr->llCode;
+ const BYTE* const mlCodeTable = seqStorePtr->mlCode;
+ BYTE* const ostart = (BYTE*)dst;
+ BYTE* const oend = ostart + dstCapacity;
+ BYTE* op = ostart;
+ size_t lastCountSize;
+
+ entropyWorkspace = count + (MaxSeq + 1);
+ entropyWkspSize -= (MaxSeq + 1) * sizeof(*count);
+
+ DEBUGLOG(4, "ZSTD_entropyCompressSeqStore_internal (nbSeq=%zu)", nbSeq);
+ ZSTD_STATIC_ASSERT(HUF_WORKSPACE_SIZE >= (1<<MAX(MLFSELog,LLFSELog)));
+ assert(entropyWkspSize >= HUF_WORKSPACE_SIZE);
+
+ /* Compress literals */
+ { const BYTE* const literals = seqStorePtr->litStart;
+ size_t const numSequences = seqStorePtr->sequences - seqStorePtr->sequencesStart;
+ size_t const numLiterals = seqStorePtr->lit - seqStorePtr->litStart;
+ /* Base suspicion of uncompressibility on ratio of literals to sequences */
+ unsigned const suspectUncompressible = (numSequences == 0) || (numLiterals / numSequences >= SUSPECT_UNCOMPRESSIBLE_LITERAL_RATIO);
+ size_t const litSize = (size_t)(seqStorePtr->lit - literals);
+ size_t const cSize = ZSTD_compressLiterals(
+ &prevEntropy->huf, &nextEntropy->huf,
+ cctxParams->cParams.strategy,
+ ZSTD_literalsCompressionIsDisabled(cctxParams),
+ op, dstCapacity,
+ literals, litSize,
+ entropyWorkspace, entropyWkspSize,
+ bmi2, suspectUncompressible);
+ FORWARD_IF_ERROR(cSize, "ZSTD_compressLiterals failed");
+ assert(cSize <= dstCapacity);
+ op += cSize;
+ }
+
+ /* Sequences Header */
+ RETURN_ERROR_IF((oend-op) < 3 /*max nbSeq Size*/ + 1 /*seqHead*/,
+ dstSize_tooSmall, "Can't fit seq hdr in output buf!");
+ if (nbSeq < 128) {
+ *op++ = (BYTE)nbSeq;
+ } else if (nbSeq < LONGNBSEQ) {
+ op[0] = (BYTE)((nbSeq>>8) + 0x80);
+ op[1] = (BYTE)nbSeq;
+ op+=2;
+ } else {
+ op[0]=0xFF;
+ MEM_writeLE16(op+1, (U16)(nbSeq - LONGNBSEQ));
+ op+=3;
+ }
+ assert(op <= oend);
+ if (nbSeq==0) {
+ /* Copy the old tables over as if we repeated them */
+ ZSTD_memcpy(&nextEntropy->fse, &prevEntropy->fse, sizeof(prevEntropy->fse));
+ return (size_t)(op - ostart);
+ }
+ {
+ ZSTD_symbolEncodingTypeStats_t stats;
+ BYTE* seqHead = op++;
+ /* build stats for sequences */
+ stats = ZSTD_buildSequencesStatistics(seqStorePtr, nbSeq,
+ &prevEntropy->fse, &nextEntropy->fse,
+ op, oend,
+ strategy, count,
+ entropyWorkspace, entropyWkspSize);
+ FORWARD_IF_ERROR(stats.size, "ZSTD_buildSequencesStatistics failed!");
+ *seqHead = (BYTE)((stats.LLtype<<6) + (stats.Offtype<<4) + (stats.MLtype<<2));
+ lastCountSize = stats.lastCountSize;
+ op += stats.size;
+ }
+
+ { size_t const bitstreamSize = ZSTD_encodeSequences(
+ op, (size_t)(oend - op),
+ CTable_MatchLength, mlCodeTable,
+ CTable_OffsetBits, ofCodeTable,
+ CTable_LitLength, llCodeTable,
+ sequences, nbSeq,
+ longOffsets, bmi2);
+ FORWARD_IF_ERROR(bitstreamSize, "ZSTD_encodeSequences failed");
+ op += bitstreamSize;
+ assert(op <= oend);
+ /* zstd versions <= 1.3.4 mistakenly report corruption when
+ * FSE_readNCount() receives a buffer < 4 bytes.
+ * Fixed by https://github.com/facebook/zstd/pull/1146.
+ * This can happen when the last set_compressed table present is 2
+ * bytes and the bitstream is only one byte.
+ * In this exceedingly rare case, we will simply emit an uncompressed
+ * block, since it isn't worth optimizing.
+ */
+ if (lastCountSize && (lastCountSize + bitstreamSize) < 4) {
+ /* lastCountSize >= 2 && bitstreamSize > 0 ==> lastCountSize == 3 */
+ assert(lastCountSize + bitstreamSize == 3);
+ DEBUGLOG(5, "Avoiding bug in zstd decoder in versions <= 1.3.4 by "
+ "emitting an uncompressed block.");
+ return 0;
+ }
+ }
+
+ DEBUGLOG(5, "compressed block size : %u", (unsigned)(op - ostart));
+ return (size_t)(op - ostart);
+}
+
+MEM_STATIC size_t
+ZSTD_entropyCompressSeqStore(seqStore_t* seqStorePtr,
+ const ZSTD_entropyCTables_t* prevEntropy,
+ ZSTD_entropyCTables_t* nextEntropy,
+ const ZSTD_CCtx_params* cctxParams,
+ void* dst, size_t dstCapacity,
+ size_t srcSize,
+ void* entropyWorkspace, size_t entropyWkspSize,
+ int bmi2)
+{
+ size_t const cSize = ZSTD_entropyCompressSeqStore_internal(
+ seqStorePtr, prevEntropy, nextEntropy, cctxParams,
+ dst, dstCapacity,
+ entropyWorkspace, entropyWkspSize, bmi2);
+ if (cSize == 0) return 0;
+ /* When srcSize <= dstCapacity, there is enough space to write a raw uncompressed block.
+ * Since we ran out of space, block must be not compressible, so fall back to raw uncompressed block.
+ */
+ if ((cSize == ERROR(dstSize_tooSmall)) & (srcSize <= dstCapacity))
+ return 0; /* block not compressed */
+ FORWARD_IF_ERROR(cSize, "ZSTD_entropyCompressSeqStore_internal failed");
+
+ /* Check compressibility */
+ { size_t const maxCSize = srcSize - ZSTD_minGain(srcSize, cctxParams->cParams.strategy);
+ if (cSize >= maxCSize) return 0; /* block not compressed */
+ }
+ DEBUGLOG(4, "ZSTD_entropyCompressSeqStore() cSize: %zu", cSize);
+ return cSize;
+}
+
+/* ZSTD_selectBlockCompressor() :
+ * Not static, but internal use only (used by long distance matcher)
+ * assumption : strat is a valid strategy */
+ZSTD_blockCompressor ZSTD_selectBlockCompressor(ZSTD_strategy strat, ZSTD_paramSwitch_e useRowMatchFinder, ZSTD_dictMode_e dictMode)
+{
+ static const ZSTD_blockCompressor blockCompressor[4][ZSTD_STRATEGY_MAX+1] = {
+ { ZSTD_compressBlock_fast /* default for 0 */,
+ ZSTD_compressBlock_fast,
+ ZSTD_compressBlock_doubleFast,
+ ZSTD_compressBlock_greedy,
+ ZSTD_compressBlock_lazy,
+ ZSTD_compressBlock_lazy2,
+ ZSTD_compressBlock_btlazy2,
+ ZSTD_compressBlock_btopt,
+ ZSTD_compressBlock_btultra,
+ ZSTD_compressBlock_btultra2 },
+ { ZSTD_compressBlock_fast_extDict /* default for 0 */,
+ ZSTD_compressBlock_fast_extDict,
+ ZSTD_compressBlock_doubleFast_extDict,
+ ZSTD_compressBlock_greedy_extDict,
+ ZSTD_compressBlock_lazy_extDict,
+ ZSTD_compressBlock_lazy2_extDict,
+ ZSTD_compressBlock_btlazy2_extDict,
+ ZSTD_compressBlock_btopt_extDict,
+ ZSTD_compressBlock_btultra_extDict,
+ ZSTD_compressBlock_btultra_extDict },
+ { ZSTD_compressBlock_fast_dictMatchState /* default for 0 */,
+ ZSTD_compressBlock_fast_dictMatchState,
+ ZSTD_compressBlock_doubleFast_dictMatchState,
+ ZSTD_compressBlock_greedy_dictMatchState,
+ ZSTD_compressBlock_lazy_dictMatchState,
+ ZSTD_compressBlock_lazy2_dictMatchState,
+ ZSTD_compressBlock_btlazy2_dictMatchState,
+ ZSTD_compressBlock_btopt_dictMatchState,
+ ZSTD_compressBlock_btultra_dictMatchState,
+ ZSTD_compressBlock_btultra_dictMatchState },
+ { NULL /* default for 0 */,
+ NULL,
+ NULL,
+ ZSTD_compressBlock_greedy_dedicatedDictSearch,
+ ZSTD_compressBlock_lazy_dedicatedDictSearch,
+ ZSTD_compressBlock_lazy2_dedicatedDictSearch,
+ NULL,
+ NULL,
+ NULL,
+ NULL }
+ };
+ ZSTD_blockCompressor selectedCompressor;
+ ZSTD_STATIC_ASSERT((unsigned)ZSTD_fast == 1);
+
+ assert(ZSTD_cParam_withinBounds(ZSTD_c_strategy, strat));
+ DEBUGLOG(4, "Selected block compressor: dictMode=%d strat=%d rowMatchfinder=%d", (int)dictMode, (int)strat, (int)useRowMatchFinder);
+ if (ZSTD_rowMatchFinderUsed(strat, useRowMatchFinder)) {
+ static const ZSTD_blockCompressor rowBasedBlockCompressors[4][3] = {
+ { ZSTD_compressBlock_greedy_row,
+ ZSTD_compressBlock_lazy_row,
+ ZSTD_compressBlock_lazy2_row },
+ { ZSTD_compressBlock_greedy_extDict_row,
+ ZSTD_compressBlock_lazy_extDict_row,
+ ZSTD_compressBlock_lazy2_extDict_row },
+ { ZSTD_compressBlock_greedy_dictMatchState_row,
+ ZSTD_compressBlock_lazy_dictMatchState_row,
+ ZSTD_compressBlock_lazy2_dictMatchState_row },
+ { ZSTD_compressBlock_greedy_dedicatedDictSearch_row,
+ ZSTD_compressBlock_lazy_dedicatedDictSearch_row,
+ ZSTD_compressBlock_lazy2_dedicatedDictSearch_row }
+ };
+ DEBUGLOG(4, "Selecting a row-based matchfinder");
+ assert(useRowMatchFinder != ZSTD_ps_auto);
+ selectedCompressor = rowBasedBlockCompressors[(int)dictMode][(int)strat - (int)ZSTD_greedy];
+ } else {
+ selectedCompressor = blockCompressor[(int)dictMode][(int)strat];
+ }
+ assert(selectedCompressor != NULL);
+ return selectedCompressor;
+}
+
+static void ZSTD_storeLastLiterals(seqStore_t* seqStorePtr,
+ const BYTE* anchor, size_t lastLLSize)
+{
+ ZSTD_memcpy(seqStorePtr->lit, anchor, lastLLSize);
+ seqStorePtr->lit += lastLLSize;
+}
+
+void ZSTD_resetSeqStore(seqStore_t* ssPtr)
+{
+ ssPtr->lit = ssPtr->litStart;
+ ssPtr->sequences = ssPtr->sequencesStart;
+ ssPtr->longLengthType = ZSTD_llt_none;
+}
+
+typedef enum { ZSTDbss_compress, ZSTDbss_noCompress } ZSTD_buildSeqStore_e;
+
+static size_t ZSTD_buildSeqStore(ZSTD_CCtx* zc, const void* src, size_t srcSize)
+{
+ ZSTD_matchState_t* const ms = &zc->blockState.matchState;
+ DEBUGLOG(5, "ZSTD_buildSeqStore (srcSize=%zu)", srcSize);
+ assert(srcSize <= ZSTD_BLOCKSIZE_MAX);
+ /* Assert that we have correctly flushed the ctx params into the ms's copy */
+ ZSTD_assertEqualCParams(zc->appliedParams.cParams, ms->cParams);
+ if (srcSize < MIN_CBLOCK_SIZE+ZSTD_blockHeaderSize+1) {
+ if (zc->appliedParams.cParams.strategy >= ZSTD_btopt) {
+ ZSTD_ldm_skipRawSeqStoreBytes(&zc->externSeqStore, srcSize);
+ } else {
+ ZSTD_ldm_skipSequences(&zc->externSeqStore, srcSize, zc->appliedParams.cParams.minMatch);
+ }
+ return ZSTDbss_noCompress; /* don't even attempt compression below a certain srcSize */
+ }
+ ZSTD_resetSeqStore(&(zc->seqStore));
+ /* required for optimal parser to read stats from dictionary */
+ ms->opt.symbolCosts = &zc->blockState.prevCBlock->entropy;
+ /* tell the optimal parser how we expect to compress literals */
+ ms->opt.literalCompressionMode = zc->appliedParams.literalCompressionMode;
+ /* a gap between an attached dict and the current window is not safe,
+ * they must remain adjacent,
+ * and when that stops being the case, the dict must be unset */
+ assert(ms->dictMatchState == NULL || ms->loadedDictEnd == ms->window.dictLimit);
+
+ /* limited update after a very long match */
+ { const BYTE* const base = ms->window.base;
+ const BYTE* const istart = (const BYTE*)src;
+ const U32 curr = (U32)(istart-base);
+ if (sizeof(ptrdiff_t)==8) assert(istart - base < (ptrdiff_t)(U32)(-1)); /* ensure no overflow */
+ if (curr > ms->nextToUpdate + 384)
+ ms->nextToUpdate = curr - MIN(192, (U32)(curr - ms->nextToUpdate - 384));
+ }
+
+ /* select and store sequences */
+ { ZSTD_dictMode_e const dictMode = ZSTD_matchState_dictMode(ms);
+ size_t lastLLSize;
+ { int i;
+ for (i = 0; i < ZSTD_REP_NUM; ++i)
+ zc->blockState.nextCBlock->rep[i] = zc->blockState.prevCBlock->rep[i];
+ }
+ if (zc->externSeqStore.pos < zc->externSeqStore.size) {
+ assert(zc->appliedParams.ldmParams.enableLdm == ZSTD_ps_disable);
+ /* Updates ldmSeqStore.pos */
+ lastLLSize =
+ ZSTD_ldm_blockCompress(&zc->externSeqStore,
+ ms, &zc->seqStore,
+ zc->blockState.nextCBlock->rep,
+ zc->appliedParams.useRowMatchFinder,
+ src, srcSize);
+ assert(zc->externSeqStore.pos <= zc->externSeqStore.size);
+ } else if (zc->appliedParams.ldmParams.enableLdm == ZSTD_ps_enable) {
+ rawSeqStore_t ldmSeqStore = kNullRawSeqStore;
+
+ ldmSeqStore.seq = zc->ldmSequences;
+ ldmSeqStore.capacity = zc->maxNbLdmSequences;
+ /* Updates ldmSeqStore.size */
+ FORWARD_IF_ERROR(ZSTD_ldm_generateSequences(&zc->ldmState, &ldmSeqStore,
+ &zc->appliedParams.ldmParams,
+ src, srcSize), "");
+ /* Updates ldmSeqStore.pos */
+ lastLLSize =
+ ZSTD_ldm_blockCompress(&ldmSeqStore,
+ ms, &zc->seqStore,
+ zc->blockState.nextCBlock->rep,
+ zc->appliedParams.useRowMatchFinder,
+ src, srcSize);
+ assert(ldmSeqStore.pos == ldmSeqStore.size);
+ } else { /* not long range mode */
+ ZSTD_blockCompressor const blockCompressor = ZSTD_selectBlockCompressor(zc->appliedParams.cParams.strategy,
+ zc->appliedParams.useRowMatchFinder,
+ dictMode);
+ ms->ldmSeqStore = NULL;
+ lastLLSize = blockCompressor(ms, &zc->seqStore, zc->blockState.nextCBlock->rep, src, srcSize);
+ }
+ { const BYTE* const lastLiterals = (const BYTE*)src + srcSize - lastLLSize;
+ ZSTD_storeLastLiterals(&zc->seqStore, lastLiterals, lastLLSize);
+ } }
+ return ZSTDbss_compress;
+}
+
+static void ZSTD_copyBlockSequences(ZSTD_CCtx* zc)
+{
+ const seqStore_t* seqStore = ZSTD_getSeqStore(zc);
+ const seqDef* seqStoreSeqs = seqStore->sequencesStart;
+ size_t seqStoreSeqSize = seqStore->sequences - seqStoreSeqs;
+ size_t seqStoreLiteralsSize = (size_t)(seqStore->lit - seqStore->litStart);
+ size_t literalsRead = 0;
+ size_t lastLLSize;
+
+ ZSTD_Sequence* outSeqs = &zc->seqCollector.seqStart[zc->seqCollector.seqIndex];
+ size_t i;
+ repcodes_t updatedRepcodes;
+
+ assert(zc->seqCollector.seqIndex + 1 < zc->seqCollector.maxSequences);
+ /* Ensure we have enough space for last literals "sequence" */
+ assert(zc->seqCollector.maxSequences >= seqStoreSeqSize + 1);
+ ZSTD_memcpy(updatedRepcodes.rep, zc->blockState.prevCBlock->rep, sizeof(repcodes_t));
+ for (i = 0; i < seqStoreSeqSize; ++i) {
+ U32 rawOffset = seqStoreSeqs[i].offBase - ZSTD_REP_NUM;
+ outSeqs[i].litLength = seqStoreSeqs[i].litLength;
+ outSeqs[i].matchLength = seqStoreSeqs[i].mlBase + MINMATCH;
+ outSeqs[i].rep = 0;
+
+ if (i == seqStore->longLengthPos) {
+ if (seqStore->longLengthType == ZSTD_llt_literalLength) {
+ outSeqs[i].litLength += 0x10000;
+ } else if (seqStore->longLengthType == ZSTD_llt_matchLength) {
+ outSeqs[i].matchLength += 0x10000;
+ }
+ }
+
+ if (seqStoreSeqs[i].offBase <= ZSTD_REP_NUM) {
+ /* Derive the correct offset corresponding to a repcode */
+ outSeqs[i].rep = seqStoreSeqs[i].offBase;
+ if (outSeqs[i].litLength != 0) {
+ rawOffset = updatedRepcodes.rep[outSeqs[i].rep - 1];
+ } else {
+ if (outSeqs[i].rep == 3) {
+ rawOffset = updatedRepcodes.rep[0] - 1;
+ } else {
+ rawOffset = updatedRepcodes.rep[outSeqs[i].rep];
+ }
+ }
+ }
+ outSeqs[i].offset = rawOffset;
+ /* seqStoreSeqs[i].offset == offCode+1, and ZSTD_updateRep() expects offCode
+ so we provide seqStoreSeqs[i].offset - 1 */
+ ZSTD_updateRep(updatedRepcodes.rep,
+ seqStoreSeqs[i].offBase - 1,
+ seqStoreSeqs[i].litLength == 0);
+ literalsRead += outSeqs[i].litLength;
+ }
+ /* Insert last literals (if any exist) in the block as a sequence with ml == off == 0.
+ * If there are no last literals, then we'll emit (of: 0, ml: 0, ll: 0), which is a marker
+ * for the block boundary, according to the API.
+ */
+ assert(seqStoreLiteralsSize >= literalsRead);
+ lastLLSize = seqStoreLiteralsSize - literalsRead;
+ outSeqs[i].litLength = (U32)lastLLSize;
+ outSeqs[i].matchLength = outSeqs[i].offset = outSeqs[i].rep = 0;
+ seqStoreSeqSize++;
+ zc->seqCollector.seqIndex += seqStoreSeqSize;
+}
+
+size_t ZSTD_generateSequences(ZSTD_CCtx* zc, ZSTD_Sequence* outSeqs,
+ size_t outSeqsSize, const void* src, size_t srcSize)
+{
+ const size_t dstCapacity = ZSTD_compressBound(srcSize);
+ void* dst = ZSTD_customMalloc(dstCapacity, ZSTD_defaultCMem);
+ SeqCollector seqCollector;
+
+ RETURN_ERROR_IF(dst == NULL, memory_allocation, "NULL pointer!");
+
+ seqCollector.collectSequences = 1;
+ seqCollector.seqStart = outSeqs;
+ seqCollector.seqIndex = 0;
+ seqCollector.maxSequences = outSeqsSize;
+ zc->seqCollector = seqCollector;
+
+ ZSTD_compress2(zc, dst, dstCapacity, src, srcSize);
+ ZSTD_customFree(dst, ZSTD_defaultCMem);
+ return zc->seqCollector.seqIndex;
+}
+
+size_t ZSTD_mergeBlockDelimiters(ZSTD_Sequence* sequences, size_t seqsSize) {
+ size_t in = 0;
+ size_t out = 0;
+ for (; in < seqsSize; ++in) {
+ if (sequences[in].offset == 0 && sequences[in].matchLength == 0) {
+ if (in != seqsSize - 1) {
+ sequences[in+1].litLength += sequences[in].litLength;
+ }
+ } else {
+ sequences[out] = sequences[in];
+ ++out;
+ }
+ }
+ return out;
+}
+
+/* Unrolled loop to read four size_ts of input at a time. Returns 1 if is RLE, 0 if not. */
+static int ZSTD_isRLE(const BYTE* src, size_t length) {
+ const BYTE* ip = src;
+ const BYTE value = ip[0];
+ const size_t valueST = (size_t)((U64)value * 0x0101010101010101ULL);
+ const size_t unrollSize = sizeof(size_t) * 4;
+ const size_t unrollMask = unrollSize - 1;
+ const size_t prefixLength = length & unrollMask;
+ size_t i;
+ size_t u;
+ if (length == 1) return 1;
+ /* Check if prefix is RLE first before using unrolled loop */
+ if (prefixLength && ZSTD_count(ip+1, ip, ip+prefixLength) != prefixLength-1) {
+ return 0;
+ }
+ for (i = prefixLength; i != length; i += unrollSize) {
+ for (u = 0; u < unrollSize; u += sizeof(size_t)) {
+ if (MEM_readST(ip + i + u) != valueST) {
+ return 0;
+ }
+ }
+ }
+ return 1;
+}
+
+/* Returns true if the given block may be RLE.
+ * This is just a heuristic based on the compressibility.
+ * It may return both false positives and false negatives.
+ */
+static int ZSTD_maybeRLE(seqStore_t const* seqStore)
+{
+ size_t const nbSeqs = (size_t)(seqStore->sequences - seqStore->sequencesStart);
+ size_t const nbLits = (size_t)(seqStore->lit - seqStore->litStart);
+
+ return nbSeqs < 4 && nbLits < 10;
+}
+
+static void ZSTD_blockState_confirmRepcodesAndEntropyTables(ZSTD_blockState_t* const bs)
+{
+ ZSTD_compressedBlockState_t* const tmp = bs->prevCBlock;
+ bs->prevCBlock = bs->nextCBlock;
+ bs->nextCBlock = tmp;
+}
+
+/* Writes the block header */
+static void writeBlockHeader(void* op, size_t cSize, size_t blockSize, U32 lastBlock) {
+ U32 const cBlockHeader = cSize == 1 ?
+ lastBlock + (((U32)bt_rle)<<1) + (U32)(blockSize << 3) :
+ lastBlock + (((U32)bt_compressed)<<1) + (U32)(cSize << 3);
+ MEM_writeLE24(op, cBlockHeader);
+ DEBUGLOG(3, "writeBlockHeader: cSize: %zu blockSize: %zu lastBlock: %u", cSize, blockSize, lastBlock);
+}
+
+/* ZSTD_buildBlockEntropyStats_literals() :
+ * Builds entropy for the literals.
+ * Stores literals block type (raw, rle, compressed, repeat) and
+ * huffman description table to hufMetadata.
+ * Requires ENTROPY_WORKSPACE_SIZE workspace
+ * @return : size of huffman description table or error code */
+static size_t ZSTD_buildBlockEntropyStats_literals(void* const src, size_t srcSize,
+ const ZSTD_hufCTables_t* prevHuf,
+ ZSTD_hufCTables_t* nextHuf,
+ ZSTD_hufCTablesMetadata_t* hufMetadata,
+ const int literalsCompressionIsDisabled,
+ void* workspace, size_t wkspSize)
+{
+ BYTE* const wkspStart = (BYTE*)workspace;
+ BYTE* const wkspEnd = wkspStart + wkspSize;
+ BYTE* const countWkspStart = wkspStart;
+ unsigned* const countWksp = (unsigned*)workspace;
+ const size_t countWkspSize = (HUF_SYMBOLVALUE_MAX + 1) * sizeof(unsigned);
+ BYTE* const nodeWksp = countWkspStart + countWkspSize;
+ const size_t nodeWkspSize = wkspEnd-nodeWksp;
+ unsigned maxSymbolValue = HUF_SYMBOLVALUE_MAX;
+ unsigned huffLog = HUF_TABLELOG_DEFAULT;
+ HUF_repeat repeat = prevHuf->repeatMode;
+ DEBUGLOG(5, "ZSTD_buildBlockEntropyStats_literals (srcSize=%zu)", srcSize);
+
+ /* Prepare nextEntropy assuming reusing the existing table */
+ ZSTD_memcpy(nextHuf, prevHuf, sizeof(*prevHuf));
+
+ if (literalsCompressionIsDisabled) {
+ DEBUGLOG(5, "set_basic - disabled");
+ hufMetadata->hType = set_basic;
+ return 0;
+ }
+
+ /* small ? don't even attempt compression (speed opt) */
+#ifndef COMPRESS_LITERALS_SIZE_MIN
+#define COMPRESS_LITERALS_SIZE_MIN 63
+#endif
+ { size_t const minLitSize = (prevHuf->repeatMode == HUF_repeat_valid) ? 6 : COMPRESS_LITERALS_SIZE_MIN;
+ if (srcSize <= minLitSize) {
+ DEBUGLOG(5, "set_basic - too small");
+ hufMetadata->hType = set_basic;
+ return 0;
+ }
+ }
+
+ /* Scan input and build symbol stats */
+ { size_t const largest = HIST_count_wksp (countWksp, &maxSymbolValue, (const BYTE*)src, srcSize, workspace, wkspSize);
+ FORWARD_IF_ERROR(largest, "HIST_count_wksp failed");
+ if (largest == srcSize) {
+ DEBUGLOG(5, "set_rle");
+ hufMetadata->hType = set_rle;
+ return 0;
+ }
+ if (largest <= (srcSize >> 7)+4) {
+ DEBUGLOG(5, "set_basic - no gain");
+ hufMetadata->hType = set_basic;
+ return 0;
+ }
+ }
+
+ /* Validate the previous Huffman table */
+ if (repeat == HUF_repeat_check && !HUF_validateCTable((HUF_CElt const*)prevHuf->CTable, countWksp, maxSymbolValue)) {
+ repeat = HUF_repeat_none;
+ }
+
+ /* Build Huffman Tree */
+ ZSTD_memset(nextHuf->CTable, 0, sizeof(nextHuf->CTable));
+ huffLog = HUF_optimalTableLog(huffLog, srcSize, maxSymbolValue);
+ { size_t const maxBits = HUF_buildCTable_wksp((HUF_CElt*)nextHuf->CTable, countWksp,
+ maxSymbolValue, huffLog,
+ nodeWksp, nodeWkspSize);
+ FORWARD_IF_ERROR(maxBits, "HUF_buildCTable_wksp");
+ huffLog = (U32)maxBits;
+ { /* Build and write the CTable */
+ size_t const newCSize = HUF_estimateCompressedSize(
+ (HUF_CElt*)nextHuf->CTable, countWksp, maxSymbolValue);
+ size_t const hSize = HUF_writeCTable_wksp(
+ hufMetadata->hufDesBuffer, sizeof(hufMetadata->hufDesBuffer),
+ (HUF_CElt*)nextHuf->CTable, maxSymbolValue, huffLog,
+ nodeWksp, nodeWkspSize);
+ /* Check against repeating the previous CTable */
+ if (repeat != HUF_repeat_none) {
+ size_t const oldCSize = HUF_estimateCompressedSize(
+ (HUF_CElt const*)prevHuf->CTable, countWksp, maxSymbolValue);
+ if (oldCSize < srcSize && (oldCSize <= hSize + newCSize || hSize + 12 >= srcSize)) {
+ DEBUGLOG(5, "set_repeat - smaller");
+ ZSTD_memcpy(nextHuf, prevHuf, sizeof(*prevHuf));
+ hufMetadata->hType = set_repeat;
+ return 0;
+ }
+ }
+ if (newCSize + hSize >= srcSize) {
+ DEBUGLOG(5, "set_basic - no gains");
+ ZSTD_memcpy(nextHuf, prevHuf, sizeof(*prevHuf));
+ hufMetadata->hType = set_basic;
+ return 0;
+ }
+ DEBUGLOG(5, "set_compressed (hSize=%u)", (U32)hSize);
+ hufMetadata->hType = set_compressed;
+ nextHuf->repeatMode = HUF_repeat_check;
+ return hSize;
+ }
+ }
+}
+
+
+/* ZSTD_buildDummySequencesStatistics():
+ * Returns a ZSTD_symbolEncodingTypeStats_t with all encoding types as set_basic,
+ * and updates nextEntropy to the appropriate repeatMode.
+ */
+static ZSTD_symbolEncodingTypeStats_t
+ZSTD_buildDummySequencesStatistics(ZSTD_fseCTables_t* nextEntropy) {
+ ZSTD_symbolEncodingTypeStats_t stats = {set_basic, set_basic, set_basic, 0, 0};
+ nextEntropy->litlength_repeatMode = FSE_repeat_none;
+ nextEntropy->offcode_repeatMode = FSE_repeat_none;
+ nextEntropy->matchlength_repeatMode = FSE_repeat_none;
+ return stats;
+}
+
+/* ZSTD_buildBlockEntropyStats_sequences() :
+ * Builds entropy for the sequences.
+ * Stores symbol compression modes and fse table to fseMetadata.
+ * Requires ENTROPY_WORKSPACE_SIZE wksp.
+ * @return : size of fse tables or error code */
+static size_t ZSTD_buildBlockEntropyStats_sequences(seqStore_t* seqStorePtr,
+ const ZSTD_fseCTables_t* prevEntropy,
+ ZSTD_fseCTables_t* nextEntropy,
+ const ZSTD_CCtx_params* cctxParams,
+ ZSTD_fseCTablesMetadata_t* fseMetadata,
+ void* workspace, size_t wkspSize)
+{
+ ZSTD_strategy const strategy = cctxParams->cParams.strategy;
+ size_t const nbSeq = seqStorePtr->sequences - seqStorePtr->sequencesStart;
+ BYTE* const ostart = fseMetadata->fseTablesBuffer;
+ BYTE* const oend = ostart + sizeof(fseMetadata->fseTablesBuffer);
+ BYTE* op = ostart;
+ unsigned* countWorkspace = (unsigned*)workspace;
+ unsigned* entropyWorkspace = countWorkspace + (MaxSeq + 1);
+ size_t entropyWorkspaceSize = wkspSize - (MaxSeq + 1) * sizeof(*countWorkspace);
+ ZSTD_symbolEncodingTypeStats_t stats;
+
+ DEBUGLOG(5, "ZSTD_buildBlockEntropyStats_sequences (nbSeq=%zu)", nbSeq);
+ stats = nbSeq != 0 ? ZSTD_buildSequencesStatistics(seqStorePtr, nbSeq,
+ prevEntropy, nextEntropy, op, oend,
+ strategy, countWorkspace,
+ entropyWorkspace, entropyWorkspaceSize)
+ : ZSTD_buildDummySequencesStatistics(nextEntropy);
+ FORWARD_IF_ERROR(stats.size, "ZSTD_buildSequencesStatistics failed!");
+ fseMetadata->llType = (symbolEncodingType_e) stats.LLtype;
+ fseMetadata->ofType = (symbolEncodingType_e) stats.Offtype;
+ fseMetadata->mlType = (symbolEncodingType_e) stats.MLtype;
+ fseMetadata->lastCountSize = stats.lastCountSize;
+ return stats.size;
+}
+
+
+/* ZSTD_buildBlockEntropyStats() :
+ * Builds entropy for the block.
+ * Requires workspace size ENTROPY_WORKSPACE_SIZE
+ *
+ * @return : 0 on success or error code
+ */
+size_t ZSTD_buildBlockEntropyStats(seqStore_t* seqStorePtr,
+ const ZSTD_entropyCTables_t* prevEntropy,
+ ZSTD_entropyCTables_t* nextEntropy,
+ const ZSTD_CCtx_params* cctxParams,
+ ZSTD_entropyCTablesMetadata_t* entropyMetadata,
+ void* workspace, size_t wkspSize)
+{
+ size_t const litSize = seqStorePtr->lit - seqStorePtr->litStart;
+ entropyMetadata->hufMetadata.hufDesSize =
+ ZSTD_buildBlockEntropyStats_literals(seqStorePtr->litStart, litSize,
+ &prevEntropy->huf, &nextEntropy->huf,
+ &entropyMetadata->hufMetadata,
+ ZSTD_literalsCompressionIsDisabled(cctxParams),
+ workspace, wkspSize);
+ FORWARD_IF_ERROR(entropyMetadata->hufMetadata.hufDesSize, "ZSTD_buildBlockEntropyStats_literals failed");
+ entropyMetadata->fseMetadata.fseTablesSize =
+ ZSTD_buildBlockEntropyStats_sequences(seqStorePtr,
+ &prevEntropy->fse, &nextEntropy->fse,
+ cctxParams,
+ &entropyMetadata->fseMetadata,
+ workspace, wkspSize);
+ FORWARD_IF_ERROR(entropyMetadata->fseMetadata.fseTablesSize, "ZSTD_buildBlockEntropyStats_sequences failed");
+ return 0;
+}
+
+/* Returns the size estimate for the literals section (header + content) of a block */
+static size_t ZSTD_estimateBlockSize_literal(const BYTE* literals, size_t litSize,
+ const ZSTD_hufCTables_t* huf,
+ const ZSTD_hufCTablesMetadata_t* hufMetadata,
+ void* workspace, size_t wkspSize,
+ int writeEntropy)
+{
+ unsigned* const countWksp = (unsigned*)workspace;
+ unsigned maxSymbolValue = HUF_SYMBOLVALUE_MAX;
+ size_t literalSectionHeaderSize = 3 + (litSize >= 1 KB) + (litSize >= 16 KB);
+ U32 singleStream = litSize < 256;
+
+ if (hufMetadata->hType == set_basic) return litSize;
+ else if (hufMetadata->hType == set_rle) return 1;
+ else if (hufMetadata->hType == set_compressed || hufMetadata->hType == set_repeat) {
+ size_t const largest = HIST_count_wksp (countWksp, &maxSymbolValue, (const BYTE*)literals, litSize, workspace, wkspSize);
+ if (ZSTD_isError(largest)) return litSize;
+ { size_t cLitSizeEstimate = HUF_estimateCompressedSize((const HUF_CElt*)huf->CTable, countWksp, maxSymbolValue);
+ if (writeEntropy) cLitSizeEstimate += hufMetadata->hufDesSize;
+ if (!singleStream) cLitSizeEstimate += 6; /* multi-stream huffman uses 6-byte jump table */
+ return cLitSizeEstimate + literalSectionHeaderSize;
+ } }
+ assert(0); /* impossible */
+ return 0;
+}
+
+/* Returns the size estimate for the FSE-compressed symbols (of, ml, ll) of a block */
+static size_t ZSTD_estimateBlockSize_symbolType(symbolEncodingType_e type,
+ const BYTE* codeTable, size_t nbSeq, unsigned maxCode,
+ const FSE_CTable* fseCTable,
+ const U8* additionalBits,
+ short const* defaultNorm, U32 defaultNormLog, U32 defaultMax,
+ void* workspace, size_t wkspSize)
+{
+ unsigned* const countWksp = (unsigned*)workspace;
+ const BYTE* ctp = codeTable;
+ const BYTE* const ctStart = ctp;
+ const BYTE* const ctEnd = ctStart + nbSeq;
+ size_t cSymbolTypeSizeEstimateInBits = 0;
+ unsigned max = maxCode;
+
+ HIST_countFast_wksp(countWksp, &max, codeTable, nbSeq, workspace, wkspSize); /* can't fail */
+ if (type == set_basic) {
+ /* We selected this encoding type, so it must be valid. */
+ assert(max <= defaultMax);
+ (void)defaultMax;
+ cSymbolTypeSizeEstimateInBits = ZSTD_crossEntropyCost(defaultNorm, defaultNormLog, countWksp, max);
+ } else if (type == set_rle) {
+ cSymbolTypeSizeEstimateInBits = 0;
+ } else if (type == set_compressed || type == set_repeat) {
+ cSymbolTypeSizeEstimateInBits = ZSTD_fseBitCost(fseCTable, countWksp, max);
+ }
+ if (ZSTD_isError(cSymbolTypeSizeEstimateInBits)) {
+ return nbSeq * 10;
+ }
+ while (ctp < ctEnd) {
+ if (additionalBits) cSymbolTypeSizeEstimateInBits += additionalBits[*ctp];
+ else cSymbolTypeSizeEstimateInBits += *ctp; /* for offset, offset code is also the number of additional bits */
+ ctp++;
+ }
+ return cSymbolTypeSizeEstimateInBits >> 3;
+}
+
+/* Returns the size estimate for the sequences section (header + content) of a block */
+static size_t ZSTD_estimateBlockSize_sequences(const BYTE* ofCodeTable,
+ const BYTE* llCodeTable,
+ const BYTE* mlCodeTable,
+ size_t nbSeq,
+ const ZSTD_fseCTables_t* fseTables,
+ const ZSTD_fseCTablesMetadata_t* fseMetadata,
+ void* workspace, size_t wkspSize,
+ int writeEntropy)
+{
+ size_t sequencesSectionHeaderSize = 1 /* seqHead */ + 1 /* min seqSize size */ + (nbSeq >= 128) + (nbSeq >= LONGNBSEQ);
+ size_t cSeqSizeEstimate = 0;
+ cSeqSizeEstimate += ZSTD_estimateBlockSize_symbolType(fseMetadata->ofType, ofCodeTable, nbSeq, MaxOff,
+ fseTables->offcodeCTable, NULL,
+ OF_defaultNorm, OF_defaultNormLog, DefaultMaxOff,
+ workspace, wkspSize);
+ cSeqSizeEstimate += ZSTD_estimateBlockSize_symbolType(fseMetadata->llType, llCodeTable, nbSeq, MaxLL,
+ fseTables->litlengthCTable, LL_bits,
+ LL_defaultNorm, LL_defaultNormLog, MaxLL,
+ workspace, wkspSize);
+ cSeqSizeEstimate += ZSTD_estimateBlockSize_symbolType(fseMetadata->mlType, mlCodeTable, nbSeq, MaxML,
+ fseTables->matchlengthCTable, ML_bits,
+ ML_defaultNorm, ML_defaultNormLog, MaxML,
+ workspace, wkspSize);
+ if (writeEntropy) cSeqSizeEstimate += fseMetadata->fseTablesSize;
+ return cSeqSizeEstimate + sequencesSectionHeaderSize;
+}
+
+/* Returns the size estimate for a given stream of literals, of, ll, ml */
+static size_t ZSTD_estimateBlockSize(const BYTE* literals, size_t litSize,
+ const BYTE* ofCodeTable,
+ const BYTE* llCodeTable,
+ const BYTE* mlCodeTable,
+ size_t nbSeq,
+ const ZSTD_entropyCTables_t* entropy,
+ const ZSTD_entropyCTablesMetadata_t* entropyMetadata,
+ void* workspace, size_t wkspSize,
+ int writeLitEntropy, int writeSeqEntropy) {
+ size_t const literalsSize = ZSTD_estimateBlockSize_literal(literals, litSize,
+ &entropy->huf, &entropyMetadata->hufMetadata,
+ workspace, wkspSize, writeLitEntropy);
+ size_t const seqSize = ZSTD_estimateBlockSize_sequences(ofCodeTable, llCodeTable, mlCodeTable,
+ nbSeq, &entropy->fse, &entropyMetadata->fseMetadata,
+ workspace, wkspSize, writeSeqEntropy);
+ return seqSize + literalsSize + ZSTD_blockHeaderSize;
+}
+
+/* Builds entropy statistics and uses them for blocksize estimation.
+ *
+ * Returns the estimated compressed size of the seqStore, or a zstd error.
+ */
+static size_t ZSTD_buildEntropyStatisticsAndEstimateSubBlockSize(seqStore_t* seqStore, ZSTD_CCtx* zc) {
+ ZSTD_entropyCTablesMetadata_t* entropyMetadata = &zc->blockSplitCtx.entropyMetadata;
+ DEBUGLOG(6, "ZSTD_buildEntropyStatisticsAndEstimateSubBlockSize()");
+ FORWARD_IF_ERROR(ZSTD_buildBlockEntropyStats(seqStore,
+ &zc->blockState.prevCBlock->entropy,
+ &zc->blockState.nextCBlock->entropy,
+ &zc->appliedParams,
+ entropyMetadata,
+ zc->entropyWorkspace, ENTROPY_WORKSPACE_SIZE /* statically allocated in resetCCtx */), "");
+ return ZSTD_estimateBlockSize(seqStore->litStart, (size_t)(seqStore->lit - seqStore->litStart),
+ seqStore->ofCode, seqStore->llCode, seqStore->mlCode,
+ (size_t)(seqStore->sequences - seqStore->sequencesStart),
+ &zc->blockState.nextCBlock->entropy, entropyMetadata, zc->entropyWorkspace, ENTROPY_WORKSPACE_SIZE,
+ (int)(entropyMetadata->hufMetadata.hType == set_compressed), 1);
+}
+
+/* Returns literals bytes represented in a seqStore */
+static size_t ZSTD_countSeqStoreLiteralsBytes(const seqStore_t* const seqStore) {
+ size_t literalsBytes = 0;
+ size_t const nbSeqs = seqStore->sequences - seqStore->sequencesStart;
+ size_t i;
+ for (i = 0; i < nbSeqs; ++i) {
+ seqDef seq = seqStore->sequencesStart[i];
+ literalsBytes += seq.litLength;
+ if (i == seqStore->longLengthPos && seqStore->longLengthType == ZSTD_llt_literalLength) {
+ literalsBytes += 0x10000;
+ }
+ }
+ return literalsBytes;
+}
+
+/* Returns match bytes represented in a seqStore */
+static size_t ZSTD_countSeqStoreMatchBytes(const seqStore_t* const seqStore) {
+ size_t matchBytes = 0;
+ size_t const nbSeqs = seqStore->sequences - seqStore->sequencesStart;
+ size_t i;
+ for (i = 0; i < nbSeqs; ++i) {
+ seqDef seq = seqStore->sequencesStart[i];
+ matchBytes += seq.mlBase + MINMATCH;
+ if (i == seqStore->longLengthPos && seqStore->longLengthType == ZSTD_llt_matchLength) {
+ matchBytes += 0x10000;
+ }
+ }
+ return matchBytes;
+}
+
+/* Derives the seqStore that is a chunk of the originalSeqStore from [startIdx, endIdx).
+ * Stores the result in resultSeqStore.
+ */
+static void ZSTD_deriveSeqStoreChunk(seqStore_t* resultSeqStore,
+ const seqStore_t* originalSeqStore,
+ size_t startIdx, size_t endIdx) {
+ BYTE* const litEnd = originalSeqStore->lit;
+ size_t literalsBytes;
+ size_t literalsBytesPreceding = 0;
+
+ *resultSeqStore = *originalSeqStore;
+ if (startIdx > 0) {
+ resultSeqStore->sequences = originalSeqStore->sequencesStart + startIdx;
+ literalsBytesPreceding = ZSTD_countSeqStoreLiteralsBytes(resultSeqStore);
+ }
+
+ /* Move longLengthPos into the correct position if necessary */
+ if (originalSeqStore->longLengthType != ZSTD_llt_none) {
+ if (originalSeqStore->longLengthPos < startIdx || originalSeqStore->longLengthPos > endIdx) {
+ resultSeqStore->longLengthType = ZSTD_llt_none;
+ } else {
+ resultSeqStore->longLengthPos -= (U32)startIdx;
+ }
+ }
+ resultSeqStore->sequencesStart = originalSeqStore->sequencesStart + startIdx;
+ resultSeqStore->sequences = originalSeqStore->sequencesStart + endIdx;
+ literalsBytes = ZSTD_countSeqStoreLiteralsBytes(resultSeqStore);
+ resultSeqStore->litStart += literalsBytesPreceding;
+ if (endIdx == (size_t)(originalSeqStore->sequences - originalSeqStore->sequencesStart)) {
+ /* This accounts for possible last literals if the derived chunk reaches the end of the block */
+ resultSeqStore->lit = litEnd;
+ } else {
+ resultSeqStore->lit = resultSeqStore->litStart+literalsBytes;
+ }
+ resultSeqStore->llCode += startIdx;
+ resultSeqStore->mlCode += startIdx;
+ resultSeqStore->ofCode += startIdx;
+}
+
+/*
+ * Returns the raw offset represented by the combination of offCode, ll0, and repcode history.
+ * offCode must represent a repcode in the numeric representation of ZSTD_storeSeq().
+ */
+static U32
+ZSTD_resolveRepcodeToRawOffset(const U32 rep[ZSTD_REP_NUM], const U32 offCode, const U32 ll0)
+{
+ U32 const adjustedOffCode = STORED_REPCODE(offCode) - 1 + ll0; /* [ 0 - 3 ] */
+ assert(STORED_IS_REPCODE(offCode));
+ if (adjustedOffCode == ZSTD_REP_NUM) {
+ /* litlength == 0 and offCode == 2 implies selection of first repcode - 1 */
+ assert(rep[0] > 0);
+ return rep[0] - 1;
+ }
+ return rep[adjustedOffCode];
+}
+
+/*
+ * ZSTD_seqStore_resolveOffCodes() reconciles any possible divergences in offset history that may arise
+ * due to emission of RLE/raw blocks that disturb the offset history,
+ * and replaces any repcodes within the seqStore that may be invalid.
+ *
+ * dRepcodes are updated as would be on the decompression side.
+ * cRepcodes are updated exactly in accordance with the seqStore.
+ *
+ * Note : this function assumes seq->offBase respects the following numbering scheme :
+ * 0 : invalid
+ * 1-3 : repcode 1-3
+ * 4+ : real_offset+3
+ */
+static void ZSTD_seqStore_resolveOffCodes(repcodes_t* const dRepcodes, repcodes_t* const cRepcodes,
+ seqStore_t* const seqStore, U32 const nbSeq) {
+ U32 idx = 0;
+ for (; idx < nbSeq; ++idx) {
+ seqDef* const seq = seqStore->sequencesStart + idx;
+ U32 const ll0 = (seq->litLength == 0);
+ U32 const offCode = OFFBASE_TO_STORED(seq->offBase);
+ assert(seq->offBase > 0);
+ if (STORED_IS_REPCODE(offCode)) {
+ U32 const dRawOffset = ZSTD_resolveRepcodeToRawOffset(dRepcodes->rep, offCode, ll0);
+ U32 const cRawOffset = ZSTD_resolveRepcodeToRawOffset(cRepcodes->rep, offCode, ll0);
+ /* Adjust simulated decompression repcode history if we come across a mismatch. Replace
+ * the repcode with the offset it actually references, determined by the compression
+ * repcode history.
+ */
+ if (dRawOffset != cRawOffset) {
+ seq->offBase = cRawOffset + ZSTD_REP_NUM;
+ }
+ }
+ /* Compression repcode history is always updated with values directly from the unmodified seqStore.
+ * Decompression repcode history may use modified seq->offset value taken from compression repcode history.
+ */
+ ZSTD_updateRep(dRepcodes->rep, OFFBASE_TO_STORED(seq->offBase), ll0);
+ ZSTD_updateRep(cRepcodes->rep, offCode, ll0);
+ }
+}
+
+/* ZSTD_compressSeqStore_singleBlock():
+ * Compresses a seqStore into a block with a block header, into the buffer dst.
+ *
+ * Returns the total size of that block (including header) or a ZSTD error code.
+ */
+static size_t
+ZSTD_compressSeqStore_singleBlock(ZSTD_CCtx* zc, seqStore_t* const seqStore,
+ repcodes_t* const dRep, repcodes_t* const cRep,
+ void* dst, size_t dstCapacity,
+ const void* src, size_t srcSize,
+ U32 lastBlock, U32 isPartition)
+{
+ const U32 rleMaxLength = 25;
+ BYTE* op = (BYTE*)dst;
+ const BYTE* ip = (const BYTE*)src;
+ size_t cSize;
+ size_t cSeqsSize;
+
+ /* In case of an RLE or raw block, the simulated decompression repcode history must be reset */
+ repcodes_t const dRepOriginal = *dRep;
+ DEBUGLOG(5, "ZSTD_compressSeqStore_singleBlock");
+ if (isPartition)
+ ZSTD_seqStore_resolveOffCodes(dRep, cRep, seqStore, (U32)(seqStore->sequences - seqStore->sequencesStart));
+
+ RETURN_ERROR_IF(dstCapacity < ZSTD_blockHeaderSize, dstSize_tooSmall, "Block header doesn't fit");
+ cSeqsSize = ZSTD_entropyCompressSeqStore(seqStore,
+ &zc->blockState.prevCBlock->entropy, &zc->blockState.nextCBlock->entropy,
+ &zc->appliedParams,
+ op + ZSTD_blockHeaderSize, dstCapacity - ZSTD_blockHeaderSize,
+ srcSize,
+ zc->entropyWorkspace, ENTROPY_WORKSPACE_SIZE /* statically allocated in resetCCtx */,
+ zc->bmi2);
+ FORWARD_IF_ERROR(cSeqsSize, "ZSTD_entropyCompressSeqStore failed!");
+
+ if (!zc->isFirstBlock &&
+ cSeqsSize < rleMaxLength &&
+ ZSTD_isRLE((BYTE const*)src, srcSize)) {
+ /* We don't want to emit our first block as a RLE even if it qualifies because
+ * doing so will cause the decoder (cli only) to throw a "should consume all input error."
+ * This is only an issue for zstd <= v1.4.3
+ */
+ cSeqsSize = 1;
+ }
+
+ if (zc->seqCollector.collectSequences) {
+ ZSTD_copyBlockSequences(zc);
+ ZSTD_blockState_confirmRepcodesAndEntropyTables(&zc->blockState);
+ return 0;
+ }
+
+ if (cSeqsSize == 0) {
+ cSize = ZSTD_noCompressBlock(op, dstCapacity, ip, srcSize, lastBlock);
+ FORWARD_IF_ERROR(cSize, "Nocompress block failed");
+ DEBUGLOG(4, "Writing out nocompress block, size: %zu", cSize);
+ *dRep = dRepOriginal; /* reset simulated decompression repcode history */
+ } else if (cSeqsSize == 1) {
+ cSize = ZSTD_rleCompressBlock(op, dstCapacity, *ip, srcSize, lastBlock);
+ FORWARD_IF_ERROR(cSize, "RLE compress block failed");
+ DEBUGLOG(4, "Writing out RLE block, size: %zu", cSize);
+ *dRep = dRepOriginal; /* reset simulated decompression repcode history */
+ } else {
+ ZSTD_blockState_confirmRepcodesAndEntropyTables(&zc->blockState);
+ writeBlockHeader(op, cSeqsSize, srcSize, lastBlock);
+ cSize = ZSTD_blockHeaderSize + cSeqsSize;
+ DEBUGLOG(4, "Writing out compressed block, size: %zu", cSize);
+ }
+
+ if (zc->blockState.prevCBlock->entropy.fse.offcode_repeatMode == FSE_repeat_valid)
+ zc->blockState.prevCBlock->entropy.fse.offcode_repeatMode = FSE_repeat_check;
+
+ return cSize;
+}
+
+/* Struct to keep track of where we are in our recursive calls. */
+typedef struct {
+ U32* splitLocations; /* Array of split indices */
+ size_t idx; /* The current index within splitLocations being worked on */
+} seqStoreSplits;
+
+#define MIN_SEQUENCES_BLOCK_SPLITTING 300
+
+/* Helper function to perform the recursive search for block splits.
+ * Estimates the cost of seqStore prior to split, and estimates the cost of splitting the sequences in half.
+ * If advantageous to split, then we recurse down the two sub-blocks. If not, or if an error occurred in estimation, then
+ * we do not recurse.
+ *
+ * Note: The recursion depth is capped by a heuristic minimum number of sequences, defined by MIN_SEQUENCES_BLOCK_SPLITTING.
+ * In theory, this means the absolute largest recursion depth is 10 == log2(maxNbSeqInBlock/MIN_SEQUENCES_BLOCK_SPLITTING).
+ * In practice, recursion depth usually doesn't go beyond 4.
+ *
+ * Furthermore, the number of splits is capped by ZSTD_MAX_NB_BLOCK_SPLITS. At ZSTD_MAX_NB_BLOCK_SPLITS == 196 with the current existing blockSize
+ * maximum of 128 KB, this value is actually impossible to reach.
+ */
+static void
+ZSTD_deriveBlockSplitsHelper(seqStoreSplits* splits, size_t startIdx, size_t endIdx,
+ ZSTD_CCtx* zc, const seqStore_t* origSeqStore)
+{
+ seqStore_t* fullSeqStoreChunk = &zc->blockSplitCtx.fullSeqStoreChunk;
+ seqStore_t* firstHalfSeqStore = &zc->blockSplitCtx.firstHalfSeqStore;
+ seqStore_t* secondHalfSeqStore = &zc->blockSplitCtx.secondHalfSeqStore;
+ size_t estimatedOriginalSize;
+ size_t estimatedFirstHalfSize;
+ size_t estimatedSecondHalfSize;
+ size_t midIdx = (startIdx + endIdx)/2;
+
+ if (endIdx - startIdx < MIN_SEQUENCES_BLOCK_SPLITTING || splits->idx >= ZSTD_MAX_NB_BLOCK_SPLITS) {
+ DEBUGLOG(6, "ZSTD_deriveBlockSplitsHelper: Too few sequences");
+ return;
+ }
+ DEBUGLOG(4, "ZSTD_deriveBlockSplitsHelper: startIdx=%zu endIdx=%zu", startIdx, endIdx);
+ ZSTD_deriveSeqStoreChunk(fullSeqStoreChunk, origSeqStore, startIdx, endIdx);
+ ZSTD_deriveSeqStoreChunk(firstHalfSeqStore, origSeqStore, startIdx, midIdx);
+ ZSTD_deriveSeqStoreChunk(secondHalfSeqStore, origSeqStore, midIdx, endIdx);
+ estimatedOriginalSize = ZSTD_buildEntropyStatisticsAndEstimateSubBlockSize(fullSeqStoreChunk, zc);
+ estimatedFirstHalfSize = ZSTD_buildEntropyStatisticsAndEstimateSubBlockSize(firstHalfSeqStore, zc);
+ estimatedSecondHalfSize = ZSTD_buildEntropyStatisticsAndEstimateSubBlockSize(secondHalfSeqStore, zc);
+ DEBUGLOG(4, "Estimated original block size: %zu -- First half split: %zu -- Second half split: %zu",
+ estimatedOriginalSize, estimatedFirstHalfSize, estimatedSecondHalfSize);
+ if (ZSTD_isError(estimatedOriginalSize) || ZSTD_isError(estimatedFirstHalfSize) || ZSTD_isError(estimatedSecondHalfSize)) {
+ return;
+ }
+ if (estimatedFirstHalfSize + estimatedSecondHalfSize < estimatedOriginalSize) {
+ ZSTD_deriveBlockSplitsHelper(splits, startIdx, midIdx, zc, origSeqStore);
+ splits->splitLocations[splits->idx] = (U32)midIdx;
+ splits->idx++;
+ ZSTD_deriveBlockSplitsHelper(splits, midIdx, endIdx, zc, origSeqStore);
+ }
+}
+
+/* Base recursive function. Populates a table with intra-block partition indices that can improve compression ratio.
+ *
+ * Returns the number of splits made (which equals the size of the partition table - 1).
+ */
+static size_t ZSTD_deriveBlockSplits(ZSTD_CCtx* zc, U32 partitions[], U32 nbSeq) {
+ seqStoreSplits splits = {partitions, 0};
+ if (nbSeq <= 4) {
+ DEBUGLOG(4, "ZSTD_deriveBlockSplits: Too few sequences to split");
+ /* Refuse to try and split anything with less than 4 sequences */
+ return 0;
+ }
+ ZSTD_deriveBlockSplitsHelper(&splits, 0, nbSeq, zc, &zc->seqStore);
+ splits.splitLocations[splits.idx] = nbSeq;
+ DEBUGLOG(5, "ZSTD_deriveBlockSplits: final nb partitions: %zu", splits.idx+1);
+ return splits.idx;
+}
+
+/* ZSTD_compressBlock_splitBlock():
+ * Attempts to split a given block into multiple blocks to improve compression ratio.
+ *
+ * Returns combined size of all blocks (which includes headers), or a ZSTD error code.
+ */
+static size_t
+ZSTD_compressBlock_splitBlock_internal(ZSTD_CCtx* zc, void* dst, size_t dstCapacity,
+ const void* src, size_t blockSize, U32 lastBlock, U32 nbSeq)
+{
+ size_t cSize = 0;
+ const BYTE* ip = (const BYTE*)src;
+ BYTE* op = (BYTE*)dst;
+ size_t i = 0;
+ size_t srcBytesTotal = 0;
+ U32* partitions = zc->blockSplitCtx.partitions; /* size == ZSTD_MAX_NB_BLOCK_SPLITS */
+ seqStore_t* nextSeqStore = &zc->blockSplitCtx.nextSeqStore;
+ seqStore_t* currSeqStore = &zc->blockSplitCtx.currSeqStore;
+ size_t numSplits = ZSTD_deriveBlockSplits(zc, partitions, nbSeq);
+
+ /* If a block is split and some partitions are emitted as RLE/uncompressed, then repcode history
+ * may become invalid. In order to reconcile potentially invalid repcodes, we keep track of two
+ * separate repcode histories that simulate repcode history on compression and decompression side,
+ * and use the histories to determine whether we must replace a particular repcode with its raw offset.
+ *
+ * 1) cRep gets updated for each partition, regardless of whether the block was emitted as uncompressed
+ * or RLE. This allows us to retrieve the offset value that an invalid repcode references within
+ * a nocompress/RLE block.
+ * 2) dRep gets updated only for compressed partitions, and when a repcode gets replaced, will use
+ * the replacement offset value rather than the original repcode to update the repcode history.
+ * dRep also will be the final repcode history sent to the next block.
+ *
+ * See ZSTD_seqStore_resolveOffCodes() for more details.
+ */
+ repcodes_t dRep;
+ repcodes_t cRep;
+ ZSTD_memcpy(dRep.rep, zc->blockState.prevCBlock->rep, sizeof(repcodes_t));
+ ZSTD_memcpy(cRep.rep, zc->blockState.prevCBlock->rep, sizeof(repcodes_t));
+ ZSTD_memset(nextSeqStore, 0, sizeof(seqStore_t));
+
+ DEBUGLOG(4, "ZSTD_compressBlock_splitBlock_internal (dstCapacity=%u, dictLimit=%u, nextToUpdate=%u)",
+ (unsigned)dstCapacity, (unsigned)zc->blockState.matchState.window.dictLimit,
+ (unsigned)zc->blockState.matchState.nextToUpdate);
+
+ if (numSplits == 0) {
+ size_t cSizeSingleBlock = ZSTD_compressSeqStore_singleBlock(zc, &zc->seqStore,
+ &dRep, &cRep,
+ op, dstCapacity,
+ ip, blockSize,
+ lastBlock, 0 /* isPartition */);
+ FORWARD_IF_ERROR(cSizeSingleBlock, "Compressing single block from splitBlock_internal() failed!");
+ DEBUGLOG(5, "ZSTD_compressBlock_splitBlock_internal: No splits");
+ assert(cSizeSingleBlock <= ZSTD_BLOCKSIZE_MAX + ZSTD_blockHeaderSize);
+ return cSizeSingleBlock;
+ }
+
+ ZSTD_deriveSeqStoreChunk(currSeqStore, &zc->seqStore, 0, partitions[0]);
+ for (i = 0; i <= numSplits; ++i) {
+ size_t srcBytes;
+ size_t cSizeChunk;
+ U32 const lastPartition = (i == numSplits);
+ U32 lastBlockEntireSrc = 0;
+
+ srcBytes = ZSTD_countSeqStoreLiteralsBytes(currSeqStore) + ZSTD_countSeqStoreMatchBytes(currSeqStore);
+ srcBytesTotal += srcBytes;
+ if (lastPartition) {
+ /* This is the final partition, need to account for possible last literals */
+ srcBytes += blockSize - srcBytesTotal;
+ lastBlockEntireSrc = lastBlock;
+ } else {
+ ZSTD_deriveSeqStoreChunk(nextSeqStore, &zc->seqStore, partitions[i], partitions[i+1]);
+ }
+
+ cSizeChunk = ZSTD_compressSeqStore_singleBlock(zc, currSeqStore,
+ &dRep, &cRep,
+ op, dstCapacity,
+ ip, srcBytes,
+ lastBlockEntireSrc, 1 /* isPartition */);
+ DEBUGLOG(5, "Estimated size: %zu actual size: %zu", ZSTD_buildEntropyStatisticsAndEstimateSubBlockSize(currSeqStore, zc), cSizeChunk);
+ FORWARD_IF_ERROR(cSizeChunk, "Compressing chunk failed!");
+
+ ip += srcBytes;
+ op += cSizeChunk;
+ dstCapacity -= cSizeChunk;
+ cSize += cSizeChunk;
+ *currSeqStore = *nextSeqStore;
+ assert(cSizeChunk <= ZSTD_BLOCKSIZE_MAX + ZSTD_blockHeaderSize);
+ }
+ /* cRep and dRep may have diverged during the compression. If so, we use the dRep repcodes
+ * for the next block.
+ */
+ ZSTD_memcpy(zc->blockState.prevCBlock->rep, dRep.rep, sizeof(repcodes_t));
+ return cSize;
+}
+
+static size_t
+ZSTD_compressBlock_splitBlock(ZSTD_CCtx* zc,
+ void* dst, size_t dstCapacity,
+ const void* src, size_t srcSize, U32 lastBlock)
+{
+ const BYTE* ip = (const BYTE*)src;
+ BYTE* op = (BYTE*)dst;
+ U32 nbSeq;
+ size_t cSize;
+ DEBUGLOG(4, "ZSTD_compressBlock_splitBlock");
+ assert(zc->appliedParams.useBlockSplitter == ZSTD_ps_enable);
+
+ { const size_t bss = ZSTD_buildSeqStore(zc, src, srcSize);
+ FORWARD_IF_ERROR(bss, "ZSTD_buildSeqStore failed");
+ if (bss == ZSTDbss_noCompress) {
+ if (zc->blockState.prevCBlock->entropy.fse.offcode_repeatMode == FSE_repeat_valid)
+ zc->blockState.prevCBlock->entropy.fse.offcode_repeatMode = FSE_repeat_check;
+ cSize = ZSTD_noCompressBlock(op, dstCapacity, ip, srcSize, lastBlock);
+ FORWARD_IF_ERROR(cSize, "ZSTD_noCompressBlock failed");
+ DEBUGLOG(4, "ZSTD_compressBlock_splitBlock: Nocompress block");
+ return cSize;
+ }
+ nbSeq = (U32)(zc->seqStore.sequences - zc->seqStore.sequencesStart);
+ }
+
+ cSize = ZSTD_compressBlock_splitBlock_internal(zc, dst, dstCapacity, src, srcSize, lastBlock, nbSeq);
+ FORWARD_IF_ERROR(cSize, "Splitting blocks failed!");
+ return cSize;
+}
+
+static size_t
+ZSTD_compressBlock_internal(ZSTD_CCtx* zc,
+ void* dst, size_t dstCapacity,
+ const void* src, size_t srcSize, U32 frame)
+{
+ /* This the upper bound for the length of an rle block.
+ * This isn't the actual upper bound. Finding the real threshold
+ * needs further investigation.
+ */
+ const U32 rleMaxLength = 25;
+ size_t cSize;
+ const BYTE* ip = (const BYTE*)src;
+ BYTE* op = (BYTE*)dst;
+ DEBUGLOG(5, "ZSTD_compressBlock_internal (dstCapacity=%u, dictLimit=%u, nextToUpdate=%u)",
+ (unsigned)dstCapacity, (unsigned)zc->blockState.matchState.window.dictLimit,
+ (unsigned)zc->blockState.matchState.nextToUpdate);
+
+ { const size_t bss = ZSTD_buildSeqStore(zc, src, srcSize);
+ FORWARD_IF_ERROR(bss, "ZSTD_buildSeqStore failed");
+ if (bss == ZSTDbss_noCompress) { cSize = 0; goto out; }
+ }
+
+ if (zc->seqCollector.collectSequences) {
+ ZSTD_copyBlockSequences(zc);
+ ZSTD_blockState_confirmRepcodesAndEntropyTables(&zc->blockState);
+ return 0;
+ }
+
+ /* encode sequences and literals */
+ cSize = ZSTD_entropyCompressSeqStore(&zc->seqStore,
+ &zc->blockState.prevCBlock->entropy, &zc->blockState.nextCBlock->entropy,
+ &zc->appliedParams,
+ dst, dstCapacity,
+ srcSize,
+ zc->entropyWorkspace, ENTROPY_WORKSPACE_SIZE /* statically allocated in resetCCtx */,
+ zc->bmi2);
+
+ if (frame &&
+ /* We don't want to emit our first block as a RLE even if it qualifies because
+ * doing so will cause the decoder (cli only) to throw a "should consume all input error."
+ * This is only an issue for zstd <= v1.4.3
+ */
+ !zc->isFirstBlock &&
+ cSize < rleMaxLength &&
+ ZSTD_isRLE(ip, srcSize))
+ {
+ cSize = 1;
+ op[0] = ip[0];
+ }
+
+out:
+ if (!ZSTD_isError(cSize) && cSize > 1) {
+ ZSTD_blockState_confirmRepcodesAndEntropyTables(&zc->blockState);
+ }
+ /* We check that dictionaries have offset codes available for the first
+ * block. After the first block, the offcode table might not have large
+ * enough codes to represent the offsets in the data.
+ */
+ if (zc->blockState.prevCBlock->entropy.fse.offcode_repeatMode == FSE_repeat_valid)
+ zc->blockState.prevCBlock->entropy.fse.offcode_repeatMode = FSE_repeat_check;
+
+ return cSize;
+}
+
+static size_t ZSTD_compressBlock_targetCBlockSize_body(ZSTD_CCtx* zc,
+ void* dst, size_t dstCapacity,
+ const void* src, size_t srcSize,
+ const size_t bss, U32 lastBlock)
+{
+ DEBUGLOG(6, "Attempting ZSTD_compressSuperBlock()");
+ if (bss == ZSTDbss_compress) {
+ if (/* We don't want to emit our first block as a RLE even if it qualifies because
+ * doing so will cause the decoder (cli only) to throw a "should consume all input error."
+ * This is only an issue for zstd <= v1.4.3
+ */
+ !zc->isFirstBlock &&
+ ZSTD_maybeRLE(&zc->seqStore) &&
+ ZSTD_isRLE((BYTE const*)src, srcSize))
+ {
+ return ZSTD_rleCompressBlock(dst, dstCapacity, *(BYTE const*)src, srcSize, lastBlock);
+ }
+ /* Attempt superblock compression.
+ *
+ * Note that compressed size of ZSTD_compressSuperBlock() is not bound by the
+ * standard ZSTD_compressBound(). This is a problem, because even if we have
+ * space now, taking an extra byte now could cause us to run out of space later
+ * and violate ZSTD_compressBound().
+ *
+ * Define blockBound(blockSize) = blockSize + ZSTD_blockHeaderSize.
+ *
+ * In order to respect ZSTD_compressBound() we must attempt to emit a raw
+ * uncompressed block in these cases:
+ * * cSize == 0: Return code for an uncompressed block.
+ * * cSize == dstSize_tooSmall: We may have expanded beyond blockBound(srcSize).
+ * ZSTD_noCompressBlock() will return dstSize_tooSmall if we are really out of
+ * output space.
+ * * cSize >= blockBound(srcSize): We have expanded the block too much so
+ * emit an uncompressed block.
+ */
+ {
+ size_t const cSize = ZSTD_compressSuperBlock(zc, dst, dstCapacity, src, srcSize, lastBlock);
+ if (cSize != ERROR(dstSize_tooSmall)) {
+ size_t const maxCSize = srcSize - ZSTD_minGain(srcSize, zc->appliedParams.cParams.strategy);
+ FORWARD_IF_ERROR(cSize, "ZSTD_compressSuperBlock failed");
+ if (cSize != 0 && cSize < maxCSize + ZSTD_blockHeaderSize) {
+ ZSTD_blockState_confirmRepcodesAndEntropyTables(&zc->blockState);
+ return cSize;
+ }
+ }
+ }
+ }
+
+ DEBUGLOG(6, "Resorting to ZSTD_noCompressBlock()");
+ /* Superblock compression failed, attempt to emit a single no compress block.
+ * The decoder will be able to stream this block since it is uncompressed.
+ */
+ return ZSTD_noCompressBlock(dst, dstCapacity, src, srcSize, lastBlock);
+}
+
+static size_t ZSTD_compressBlock_targetCBlockSize(ZSTD_CCtx* zc,
+ void* dst, size_t dstCapacity,
+ const void* src, size_t srcSize,
+ U32 lastBlock)
+{
+ size_t cSize = 0;
+ const size_t bss = ZSTD_buildSeqStore(zc, src, srcSize);
+ DEBUGLOG(5, "ZSTD_compressBlock_targetCBlockSize (dstCapacity=%u, dictLimit=%u, nextToUpdate=%u, srcSize=%zu)",
+ (unsigned)dstCapacity, (unsigned)zc->blockState.matchState.window.dictLimit, (unsigned)zc->blockState.matchState.nextToUpdate, srcSize);
+ FORWARD_IF_ERROR(bss, "ZSTD_buildSeqStore failed");
+
+ cSize = ZSTD_compressBlock_targetCBlockSize_body(zc, dst, dstCapacity, src, srcSize, bss, lastBlock);
+ FORWARD_IF_ERROR(cSize, "ZSTD_compressBlock_targetCBlockSize_body failed");
+
+ if (zc->blockState.prevCBlock->entropy.fse.offcode_repeatMode == FSE_repeat_valid)
+ zc->blockState.prevCBlock->entropy.fse.offcode_repeatMode = FSE_repeat_check;
+
+ return cSize;
+}
+
+static void ZSTD_overflowCorrectIfNeeded(ZSTD_matchState_t* ms,
+ ZSTD_cwksp* ws,
+ ZSTD_CCtx_params const* params,
+ void const* ip,
+ void const* iend)
+{
+ U32 const cycleLog = ZSTD_cycleLog(params->cParams.chainLog, params->cParams.strategy);
+ U32 const maxDist = (U32)1 << params->cParams.windowLog;
+ if (ZSTD_window_needOverflowCorrection(ms->window, cycleLog, maxDist, ms->loadedDictEnd, ip, iend)) {
+ U32 const correction = ZSTD_window_correctOverflow(&ms->window, cycleLog, maxDist, ip);
+ ZSTD_STATIC_ASSERT(ZSTD_CHAINLOG_MAX <= 30);
+ ZSTD_STATIC_ASSERT(ZSTD_WINDOWLOG_MAX_32 <= 30);
+ ZSTD_STATIC_ASSERT(ZSTD_WINDOWLOG_MAX <= 31);
+ ZSTD_cwksp_mark_tables_dirty(ws);
+ ZSTD_reduceIndex(ms, params, correction);
+ ZSTD_cwksp_mark_tables_clean(ws);
+ if (ms->nextToUpdate < correction) ms->nextToUpdate = 0;
+ else ms->nextToUpdate -= correction;
+ /* invalidate dictionaries on overflow correction */
+ ms->loadedDictEnd = 0;
+ ms->dictMatchState = NULL;
+ }
+}
+
+/*! ZSTD_compress_frameChunk() :
+* Compress a chunk of data into one or multiple blocks.
+* All blocks will be terminated, all input will be consumed.
+* Function will issue an error if there is not enough `dstCapacity` to hold the compressed content.
+* Frame is supposed already started (header already produced)
+* @return : compressed size, or an error code
+*/
+static size_t ZSTD_compress_frameChunk(ZSTD_CCtx* cctx,
+ void* dst, size_t dstCapacity,
+ const void* src, size_t srcSize,
+ U32 lastFrameChunk)
+{
+ size_t blockSize = cctx->blockSize;
+ size_t remaining = srcSize;
+ const BYTE* ip = (const BYTE*)src;
+ BYTE* const ostart = (BYTE*)dst;
+ BYTE* op = ostart;
+ U32 const maxDist = (U32)1 << cctx->appliedParams.cParams.windowLog;
+
+ assert(cctx->appliedParams.cParams.windowLog <= ZSTD_WINDOWLOG_MAX);
+
+ DEBUGLOG(4, "ZSTD_compress_frameChunk (blockSize=%u)", (unsigned)blockSize);
+ if (cctx->appliedParams.fParams.checksumFlag && srcSize)
+ xxh64_update(&cctx->xxhState, src, srcSize);
+
+ while (remaining) {
+ ZSTD_matchState_t* const ms = &cctx->blockState.matchState;
+ U32 const lastBlock = lastFrameChunk & (blockSize >= remaining);
+
+ RETURN_ERROR_IF(dstCapacity < ZSTD_blockHeaderSize + MIN_CBLOCK_SIZE,
+ dstSize_tooSmall,
+ "not enough space to store compressed block");
+ if (remaining < blockSize) blockSize = remaining;
+
+ ZSTD_overflowCorrectIfNeeded(
+ ms, &cctx->workspace, &cctx->appliedParams, ip, ip + blockSize);
+ ZSTD_checkDictValidity(&ms->window, ip + blockSize, maxDist, &ms->loadedDictEnd, &ms->dictMatchState);
+ ZSTD_window_enforceMaxDist(&ms->window, ip, maxDist, &ms->loadedDictEnd, &ms->dictMatchState);
+
+ /* Ensure hash/chain table insertion resumes no sooner than lowlimit */
+ if (ms->nextToUpdate < ms->window.lowLimit) ms->nextToUpdate = ms->window.lowLimit;
+
+ { size_t cSize;
+ if (ZSTD_useTargetCBlockSize(&cctx->appliedParams)) {
+ cSize = ZSTD_compressBlock_targetCBlockSize(cctx, op, dstCapacity, ip, blockSize, lastBlock);
+ FORWARD_IF_ERROR(cSize, "ZSTD_compressBlock_targetCBlockSize failed");
+ assert(cSize > 0);
+ assert(cSize <= blockSize + ZSTD_blockHeaderSize);
+ } else if (ZSTD_blockSplitterEnabled(&cctx->appliedParams)) {
+ cSize = ZSTD_compressBlock_splitBlock(cctx, op, dstCapacity, ip, blockSize, lastBlock);
+ FORWARD_IF_ERROR(cSize, "ZSTD_compressBlock_splitBlock failed");
+ assert(cSize > 0 || cctx->seqCollector.collectSequences == 1);
+ } else {
+ cSize = ZSTD_compressBlock_internal(cctx,
+ op+ZSTD_blockHeaderSize, dstCapacity-ZSTD_blockHeaderSize,
+ ip, blockSize, 1 /* frame */);
+ FORWARD_IF_ERROR(cSize, "ZSTD_compressBlock_internal failed");
+
+ if (cSize == 0) { /* block is not compressible */
+ cSize = ZSTD_noCompressBlock(op, dstCapacity, ip, blockSize, lastBlock);
+ FORWARD_IF_ERROR(cSize, "ZSTD_noCompressBlock failed");
+ } else {
+ U32 const cBlockHeader = cSize == 1 ?
+ lastBlock + (((U32)bt_rle)<<1) + (U32)(blockSize << 3) :
+ lastBlock + (((U32)bt_compressed)<<1) + (U32)(cSize << 3);
+ MEM_writeLE24(op, cBlockHeader);
+ cSize += ZSTD_blockHeaderSize;
+ }
+ }
+
+
+ ip += blockSize;
+ assert(remaining >= blockSize);
+ remaining -= blockSize;
+ op += cSize;
+ assert(dstCapacity >= cSize);
+ dstCapacity -= cSize;
+ cctx->isFirstBlock = 0;
+ DEBUGLOG(5, "ZSTD_compress_frameChunk: adding a block of size %u",
+ (unsigned)cSize);
+ } }
+
+ if (lastFrameChunk && (op>ostart)) cctx->stage = ZSTDcs_ending;
+ return (size_t)(op-ostart);
+}
+
+
+static size_t ZSTD_writeFrameHeader(void* dst, size_t dstCapacity,
+ const ZSTD_CCtx_params* params, U64 pledgedSrcSize, U32 dictID)
+{ BYTE* const op = (BYTE*)dst;
+ U32 const dictIDSizeCodeLength = (dictID>0) + (dictID>=256) + (dictID>=65536); /* 0-3 */
+ U32 const dictIDSizeCode = params->fParams.noDictIDFlag ? 0 : dictIDSizeCodeLength; /* 0-3 */
+ U32 const checksumFlag = params->fParams.checksumFlag>0;
+ U32 const windowSize = (U32)1 << params->cParams.windowLog;
+ U32 const singleSegment = params->fParams.contentSizeFlag && (windowSize >= pledgedSrcSize);
+ BYTE const windowLogByte = (BYTE)((params->cParams.windowLog - ZSTD_WINDOWLOG_ABSOLUTEMIN) << 3);
+ U32 const fcsCode = params->fParams.contentSizeFlag ?
+ (pledgedSrcSize>=256) + (pledgedSrcSize>=65536+256) + (pledgedSrcSize>=0xFFFFFFFFU) : 0; /* 0-3 */
+ BYTE const frameHeaderDescriptionByte = (BYTE)(dictIDSizeCode + (checksumFlag<<2) + (singleSegment<<5) + (fcsCode<<6) );
+ size_t pos=0;
+
+ assert(!(params->fParams.contentSizeFlag && pledgedSrcSize == ZSTD_CONTENTSIZE_UNKNOWN));
+ RETURN_ERROR_IF(dstCapacity < ZSTD_FRAMEHEADERSIZE_MAX, dstSize_tooSmall,
+ "dst buf is too small to fit worst-case frame header size.");
+ DEBUGLOG(4, "ZSTD_writeFrameHeader : dictIDFlag : %u ; dictID : %u ; dictIDSizeCode : %u",
+ !params->fParams.noDictIDFlag, (unsigned)dictID, (unsigned)dictIDSizeCode);
+ if (params->format == ZSTD_f_zstd1) {
+ MEM_writeLE32(dst, ZSTD_MAGICNUMBER);
+ pos = 4;
+ }
+ op[pos++] = frameHeaderDescriptionByte;
+ if (!singleSegment) op[pos++] = windowLogByte;
+ switch(dictIDSizeCode)
+ {
+ default:
+ assert(0); /* impossible */
+ ZSTD_FALLTHROUGH;
+ case 0 : break;
+ case 1 : op[pos] = (BYTE)(dictID); pos++; break;
+ case 2 : MEM_writeLE16(op+pos, (U16)dictID); pos+=2; break;
+ case 3 : MEM_writeLE32(op+pos, dictID); pos+=4; break;
+ }
+ switch(fcsCode)
+ {
+ default:
+ assert(0); /* impossible */
+ ZSTD_FALLTHROUGH;
+ case 0 : if (singleSegment) op[pos++] = (BYTE)(pledgedSrcSize); break;
+ case 1 : MEM_writeLE16(op+pos, (U16)(pledgedSrcSize-256)); pos+=2; break;
+ case 2 : MEM_writeLE32(op+pos, (U32)(pledgedSrcSize)); pos+=4; break;
+ case 3 : MEM_writeLE64(op+pos, (U64)(pledgedSrcSize)); pos+=8; break;
+ }
+ return pos;
+}
+
+/* ZSTD_writeSkippableFrame_advanced() :
+ * Writes out a skippable frame with the specified magic number variant (16 are supported),
+ * from ZSTD_MAGIC_SKIPPABLE_START to ZSTD_MAGIC_SKIPPABLE_START+15, and the desired source data.
+ *
+ * Returns the total number of bytes written, or a ZSTD error code.
+ */
+size_t ZSTD_writeSkippableFrame(void* dst, size_t dstCapacity,
+ const void* src, size_t srcSize, unsigned magicVariant) {
+ BYTE* op = (BYTE*)dst;
+ RETURN_ERROR_IF(dstCapacity < srcSize + ZSTD_SKIPPABLEHEADERSIZE /* Skippable frame overhead */,
+ dstSize_tooSmall, "Not enough room for skippable frame");
+ RETURN_ERROR_IF(srcSize > (unsigned)0xFFFFFFFF, srcSize_wrong, "Src size too large for skippable frame");
+ RETURN_ERROR_IF(magicVariant > 15, parameter_outOfBound, "Skippable frame magic number variant not supported");
+
+ MEM_writeLE32(op, (U32)(ZSTD_MAGIC_SKIPPABLE_START + magicVariant));
+ MEM_writeLE32(op+4, (U32)srcSize);
+ ZSTD_memcpy(op+8, src, srcSize);
+ return srcSize + ZSTD_SKIPPABLEHEADERSIZE;
+}
+
+/* ZSTD_writeLastEmptyBlock() :
+ * output an empty Block with end-of-frame mark to complete a frame
+ * @return : size of data written into `dst` (== ZSTD_blockHeaderSize (defined in zstd_internal.h))
+ * or an error code if `dstCapacity` is too small (<ZSTD_blockHeaderSize)
+ */
+size_t ZSTD_writeLastEmptyBlock(void* dst, size_t dstCapacity)
+{
+ RETURN_ERROR_IF(dstCapacity < ZSTD_blockHeaderSize, dstSize_tooSmall,
+ "dst buf is too small to write frame trailer empty block.");
+ { U32 const cBlockHeader24 = 1 /*lastBlock*/ + (((U32)bt_raw)<<1); /* 0 size */
+ MEM_writeLE24(dst, cBlockHeader24);
+ return ZSTD_blockHeaderSize;
+ }
+}
+
+size_t ZSTD_referenceExternalSequences(ZSTD_CCtx* cctx, rawSeq* seq, size_t nbSeq)
+{
+ RETURN_ERROR_IF(cctx->stage != ZSTDcs_init, stage_wrong,
+ "wrong cctx stage");
+ RETURN_ERROR_IF(cctx->appliedParams.ldmParams.enableLdm == ZSTD_ps_enable,
+ parameter_unsupported,
+ "incompatible with ldm");
+ cctx->externSeqStore.seq = seq;
+ cctx->externSeqStore.size = nbSeq;
+ cctx->externSeqStore.capacity = nbSeq;
+ cctx->externSeqStore.pos = 0;
+ cctx->externSeqStore.posInSequence = 0;
+ return 0;
+}
+
+
+static size_t ZSTD_compressContinue_internal (ZSTD_CCtx* cctx,
+ void* dst, size_t dstCapacity,
+ const void* src, size_t srcSize,
+ U32 frame, U32 lastFrameChunk)
+{
+ ZSTD_matchState_t* const ms = &cctx->blockState.matchState;
+ size_t fhSize = 0;
+
+ DEBUGLOG(5, "ZSTD_compressContinue_internal, stage: %u, srcSize: %u",
+ cctx->stage, (unsigned)srcSize);
+ RETURN_ERROR_IF(cctx->stage==ZSTDcs_created, stage_wrong,
+ "missing init (ZSTD_compressBegin)");
+
+ if (frame && (cctx->stage==ZSTDcs_init)) {
+ fhSize = ZSTD_writeFrameHeader(dst, dstCapacity, &cctx->appliedParams,
+ cctx->pledgedSrcSizePlusOne-1, cctx->dictID);
+ FORWARD_IF_ERROR(fhSize, "ZSTD_writeFrameHeader failed");
+ assert(fhSize <= dstCapacity);
+ dstCapacity -= fhSize;
+ dst = (char*)dst + fhSize;
+ cctx->stage = ZSTDcs_ongoing;
+ }
+
+ if (!srcSize) return fhSize; /* do not generate an empty block if no input */
+
+ if (!ZSTD_window_update(&ms->window, src, srcSize, ms->forceNonContiguous)) {
+ ms->forceNonContiguous = 0;
+ ms->nextToUpdate = ms->window.dictLimit;
+ }
+ if (cctx->appliedParams.ldmParams.enableLdm == ZSTD_ps_enable) {
+ ZSTD_window_update(&cctx->ldmState.window, src, srcSize, /* forceNonContiguous */ 0);
+ }
+
+ if (!frame) {
+ /* overflow check and correction for block mode */
+ ZSTD_overflowCorrectIfNeeded(
+ ms, &cctx->workspace, &cctx->appliedParams,
+ src, (BYTE const*)src + srcSize);
+ }
+
+ DEBUGLOG(5, "ZSTD_compressContinue_internal (blockSize=%u)", (unsigned)cctx->blockSize);
+ { size_t const cSize = frame ?
+ ZSTD_compress_frameChunk (cctx, dst, dstCapacity, src, srcSize, lastFrameChunk) :
+ ZSTD_compressBlock_internal (cctx, dst, dstCapacity, src, srcSize, 0 /* frame */);
+ FORWARD_IF_ERROR(cSize, "%s", frame ? "ZSTD_compress_frameChunk failed" : "ZSTD_compressBlock_internal failed");
+ cctx->consumedSrcSize += srcSize;
+ cctx->producedCSize += (cSize + fhSize);
+ assert(!(cctx->appliedParams.fParams.contentSizeFlag && cctx->pledgedSrcSizePlusOne == 0));
+ if (cctx->pledgedSrcSizePlusOne != 0) { /* control src size */
+ ZSTD_STATIC_ASSERT(ZSTD_CONTENTSIZE_UNKNOWN == (unsigned long long)-1);
+ RETURN_ERROR_IF(
+ cctx->consumedSrcSize+1 > cctx->pledgedSrcSizePlusOne,
+ srcSize_wrong,
+ "error : pledgedSrcSize = %u, while realSrcSize >= %u",
+ (unsigned)cctx->pledgedSrcSizePlusOne-1,
+ (unsigned)cctx->consumedSrcSize);
+ }
+ return cSize + fhSize;
+ }
+}
+
+size_t ZSTD_compressContinue (ZSTD_CCtx* cctx,
+ void* dst, size_t dstCapacity,
+ const void* src, size_t srcSize)
+{
+ DEBUGLOG(5, "ZSTD_compressContinue (srcSize=%u)", (unsigned)srcSize);
+ return ZSTD_compressContinue_internal(cctx, dst, dstCapacity, src, srcSize, 1 /* frame mode */, 0 /* last chunk */);
+}
+
+
+size_t ZSTD_getBlockSize(const ZSTD_CCtx* cctx)
+{
+ ZSTD_compressionParameters const cParams = cctx->appliedParams.cParams;
+ assert(!ZSTD_checkCParams(cParams));
+ return MIN (ZSTD_BLOCKSIZE_MAX, (U32)1 << cParams.windowLog);
+}
+
+size_t ZSTD_compressBlock(ZSTD_CCtx* cctx, void* dst, size_t dstCapacity, const void* src, size_t srcSize)
+{
+ DEBUGLOG(5, "ZSTD_compressBlock: srcSize = %u", (unsigned)srcSize);
+ { size_t const blockSizeMax = ZSTD_getBlockSize(cctx);
+ RETURN_ERROR_IF(srcSize > blockSizeMax, srcSize_wrong, "input is larger than a block"); }
+
+ return ZSTD_compressContinue_internal(cctx, dst, dstCapacity, src, srcSize, 0 /* frame mode */, 0 /* last chunk */);
+}
+
+/*! ZSTD_loadDictionaryContent() :
+ * @return : 0, or an error code
+ */
+static size_t ZSTD_loadDictionaryContent(ZSTD_matchState_t* ms,
+ ldmState_t* ls,
+ ZSTD_cwksp* ws,
+ ZSTD_CCtx_params const* params,
+ const void* src, size_t srcSize,
+ ZSTD_dictTableLoadMethod_e dtlm)
+{
+ const BYTE* ip = (const BYTE*) src;
+ const BYTE* const iend = ip + srcSize;
+ int const loadLdmDict = params->ldmParams.enableLdm == ZSTD_ps_enable && ls != NULL;
+
+ /* Assert that we the ms params match the params we're being given */
+ ZSTD_assertEqualCParams(params->cParams, ms->cParams);
+
+ if (srcSize > ZSTD_CHUNKSIZE_MAX) {
+ /* Allow the dictionary to set indices up to exactly ZSTD_CURRENT_MAX.
+ * Dictionaries right at the edge will immediately trigger overflow
+ * correction, but I don't want to insert extra constraints here.
+ */
+ U32 const maxDictSize = ZSTD_CURRENT_MAX - 1;
+ /* We must have cleared our windows when our source is this large. */
+ assert(ZSTD_window_isEmpty(ms->window));
+ if (loadLdmDict)
+ assert(ZSTD_window_isEmpty(ls->window));
+ /* If the dictionary is too large, only load the suffix of the dictionary. */
+ if (srcSize > maxDictSize) {
+ ip = iend - maxDictSize;
+ src = ip;
+ srcSize = maxDictSize;
+ }
+ }
+
+ DEBUGLOG(4, "ZSTD_loadDictionaryContent(): useRowMatchFinder=%d", (int)params->useRowMatchFinder);
+ ZSTD_window_update(&ms->window, src, srcSize, /* forceNonContiguous */ 0);
+ ms->loadedDictEnd = params->forceWindow ? 0 : (U32)(iend - ms->window.base);
+ ms->forceNonContiguous = params->deterministicRefPrefix;
+
+ if (loadLdmDict) {
+ ZSTD_window_update(&ls->window, src, srcSize, /* forceNonContiguous */ 0);
+ ls->loadedDictEnd = params->forceWindow ? 0 : (U32)(iend - ls->window.base);
+ }
+
+ if (srcSize <= HASH_READ_SIZE) return 0;
+
+ ZSTD_overflowCorrectIfNeeded(ms, ws, params, ip, iend);
+
+ if (loadLdmDict)
+ ZSTD_ldm_fillHashTable(ls, ip, iend, &params->ldmParams);
+
+ switch(params->cParams.strategy)
+ {
+ case ZSTD_fast:
+ ZSTD_fillHashTable(ms, iend, dtlm);
+ break;
+ case ZSTD_dfast:
+ ZSTD_fillDoubleHashTable(ms, iend, dtlm);
+ break;
+
+ case ZSTD_greedy:
+ case ZSTD_lazy:
+ case ZSTD_lazy2:
+ assert(srcSize >= HASH_READ_SIZE);
+ if (ms->dedicatedDictSearch) {
+ assert(ms->chainTable != NULL);
+ ZSTD_dedicatedDictSearch_lazy_loadDictionary(ms, iend-HASH_READ_SIZE);
+ } else {
+ assert(params->useRowMatchFinder != ZSTD_ps_auto);
+ if (params->useRowMatchFinder == ZSTD_ps_enable) {
+ size_t const tagTableSize = ((size_t)1 << params->cParams.hashLog) * sizeof(U16);
+ ZSTD_memset(ms->tagTable, 0, tagTableSize);
+ ZSTD_row_update(ms, iend-HASH_READ_SIZE);
+ DEBUGLOG(4, "Using row-based hash table for lazy dict");
+ } else {
+ ZSTD_insertAndFindFirstIndex(ms, iend-HASH_READ_SIZE);
+ DEBUGLOG(4, "Using chain-based hash table for lazy dict");
+ }
+ }
+ break;
+
+ case ZSTD_btlazy2: /* we want the dictionary table fully sorted */
+ case ZSTD_btopt:
+ case ZSTD_btultra:
+ case ZSTD_btultra2:
+ assert(srcSize >= HASH_READ_SIZE);
+ ZSTD_updateTree(ms, iend-HASH_READ_SIZE, iend);
+ break;
+
+ default:
+ assert(0); /* not possible : not a valid strategy id */
+ }
+
+ ms->nextToUpdate = (U32)(iend - ms->window.base);
+ return 0;
+}
+
+
+/* Dictionaries that assign zero probability to symbols that show up causes problems
+ * when FSE encoding. Mark dictionaries with zero probability symbols as FSE_repeat_check
+ * and only dictionaries with 100% valid symbols can be assumed valid.
+ */
+static FSE_repeat ZSTD_dictNCountRepeat(short* normalizedCounter, unsigned dictMaxSymbolValue, unsigned maxSymbolValue)
+{
+ U32 s;
+ if (dictMaxSymbolValue < maxSymbolValue) {
+ return FSE_repeat_check;
+ }
+ for (s = 0; s <= maxSymbolValue; ++s) {
+ if (normalizedCounter[s] == 0) {
+ return FSE_repeat_check;
+ }
+ }
+ return FSE_repeat_valid;
+}
+
+size_t ZSTD_loadCEntropy(ZSTD_compressedBlockState_t* bs, void* workspace,
+ const void* const dict, size_t dictSize)
+{
+ short offcodeNCount[MaxOff+1];
+ unsigned offcodeMaxValue = MaxOff;
+ const BYTE* dictPtr = (const BYTE*)dict; /* skip magic num and dict ID */
+ const BYTE* const dictEnd = dictPtr + dictSize;
+ dictPtr += 8;
+ bs->entropy.huf.repeatMode = HUF_repeat_check;
+
+ { unsigned maxSymbolValue = 255;
+ unsigned hasZeroWeights = 1;
+ size_t const hufHeaderSize = HUF_readCTable((HUF_CElt*)bs->entropy.huf.CTable, &maxSymbolValue, dictPtr,
+ dictEnd-dictPtr, &hasZeroWeights);
+
+ /* We only set the loaded table as valid if it contains all non-zero
+ * weights. Otherwise, we set it to check */
+ if (!hasZeroWeights)
+ bs->entropy.huf.repeatMode = HUF_repeat_valid;
+
+ RETURN_ERROR_IF(HUF_isError(hufHeaderSize), dictionary_corrupted, "");
+ RETURN_ERROR_IF(maxSymbolValue < 255, dictionary_corrupted, "");
+ dictPtr += hufHeaderSize;
+ }
+
+ { unsigned offcodeLog;
+ size_t const offcodeHeaderSize = FSE_readNCount(offcodeNCount, &offcodeMaxValue, &offcodeLog, dictPtr, dictEnd-dictPtr);
+ RETURN_ERROR_IF(FSE_isError(offcodeHeaderSize), dictionary_corrupted, "");
+ RETURN_ERROR_IF(offcodeLog > OffFSELog, dictionary_corrupted, "");
+ /* fill all offset symbols to avoid garbage at end of table */
+ RETURN_ERROR_IF(FSE_isError(FSE_buildCTable_wksp(
+ bs->entropy.fse.offcodeCTable,
+ offcodeNCount, MaxOff, offcodeLog,
+ workspace, HUF_WORKSPACE_SIZE)),
+ dictionary_corrupted, "");
+ /* Defer checking offcodeMaxValue because we need to know the size of the dictionary content */
+ dictPtr += offcodeHeaderSize;
+ }
+
+ { short matchlengthNCount[MaxML+1];
+ unsigned matchlengthMaxValue = MaxML, matchlengthLog;
+ size_t const matchlengthHeaderSize = FSE_readNCount(matchlengthNCount, &matchlengthMaxValue, &matchlengthLog, dictPtr, dictEnd-dictPtr);
+ RETURN_ERROR_IF(FSE_isError(matchlengthHeaderSize), dictionary_corrupted, "");
+ RETURN_ERROR_IF(matchlengthLog > MLFSELog, dictionary_corrupted, "");
+ RETURN_ERROR_IF(FSE_isError(FSE_buildCTable_wksp(
+ bs->entropy.fse.matchlengthCTable,
+ matchlengthNCount, matchlengthMaxValue, matchlengthLog,
+ workspace, HUF_WORKSPACE_SIZE)),
+ dictionary_corrupted, "");
+ bs->entropy.fse.matchlength_repeatMode = ZSTD_dictNCountRepeat(matchlengthNCount, matchlengthMaxValue, MaxML);
+ dictPtr += matchlengthHeaderSize;
+ }
+
+ { short litlengthNCount[MaxLL+1];
+ unsigned litlengthMaxValue = MaxLL, litlengthLog;
+ size_t const litlengthHeaderSize = FSE_readNCount(litlengthNCount, &litlengthMaxValue, &litlengthLog, dictPtr, dictEnd-dictPtr);
+ RETURN_ERROR_IF(FSE_isError(litlengthHeaderSize), dictionary_corrupted, "");
+ RETURN_ERROR_IF(litlengthLog > LLFSELog, dictionary_corrupted, "");
+ RETURN_ERROR_IF(FSE_isError(FSE_buildCTable_wksp(
+ bs->entropy.fse.litlengthCTable,
+ litlengthNCount, litlengthMaxValue, litlengthLog,
+ workspace, HUF_WORKSPACE_SIZE)),
+ dictionary_corrupted, "");
+ bs->entropy.fse.litlength_repeatMode = ZSTD_dictNCountRepeat(litlengthNCount, litlengthMaxValue, MaxLL);
+ dictPtr += litlengthHeaderSize;
+ }
+
+ RETURN_ERROR_IF(dictPtr+12 > dictEnd, dictionary_corrupted, "");
+ bs->rep[0] = MEM_readLE32(dictPtr+0);
+ bs->rep[1] = MEM_readLE32(dictPtr+4);
+ bs->rep[2] = MEM_readLE32(dictPtr+8);
+ dictPtr += 12;
+
+ { size_t const dictContentSize = (size_t)(dictEnd - dictPtr);
+ U32 offcodeMax = MaxOff;
+ if (dictContentSize <= ((U32)-1) - 128 KB) {
+ U32 const maxOffset = (U32)dictContentSize + 128 KB; /* The maximum offset that must be supported */
+ offcodeMax = ZSTD_highbit32(maxOffset); /* Calculate minimum offset code required to represent maxOffset */
+ }
+ /* All offset values <= dictContentSize + 128 KB must be representable for a valid table */
+ bs->entropy.fse.offcode_repeatMode = ZSTD_dictNCountRepeat(offcodeNCount, offcodeMaxValue, MIN(offcodeMax, MaxOff));
+
+ /* All repCodes must be <= dictContentSize and != 0 */
+ { U32 u;
+ for (u=0; u<3; u++) {
+ RETURN_ERROR_IF(bs->rep[u] == 0, dictionary_corrupted, "");
+ RETURN_ERROR_IF(bs->rep[u] > dictContentSize, dictionary_corrupted, "");
+ } } }
+
+ return dictPtr - (const BYTE*)dict;
+}
+
+/* Dictionary format :
+ * See :
+ * https://github.com/facebook/zstd/blob/release/doc/zstd_compression_format.md#dictionary-format
+ */
+/*! ZSTD_loadZstdDictionary() :
+ * @return : dictID, or an error code
+ * assumptions : magic number supposed already checked
+ * dictSize supposed >= 8
+ */
+static size_t ZSTD_loadZstdDictionary(ZSTD_compressedBlockState_t* bs,
+ ZSTD_matchState_t* ms,
+ ZSTD_cwksp* ws,
+ ZSTD_CCtx_params const* params,
+ const void* dict, size_t dictSize,
+ ZSTD_dictTableLoadMethod_e dtlm,
+ void* workspace)
+{
+ const BYTE* dictPtr = (const BYTE*)dict;
+ const BYTE* const dictEnd = dictPtr + dictSize;
+ size_t dictID;
+ size_t eSize;
+ ZSTD_STATIC_ASSERT(HUF_WORKSPACE_SIZE >= (1<<MAX(MLFSELog,LLFSELog)));
+ assert(dictSize >= 8);
+ assert(MEM_readLE32(dictPtr) == ZSTD_MAGIC_DICTIONARY);
+
+ dictID = params->fParams.noDictIDFlag ? 0 : MEM_readLE32(dictPtr + 4 /* skip magic number */ );
+ eSize = ZSTD_loadCEntropy(bs, workspace, dict, dictSize);
+ FORWARD_IF_ERROR(eSize, "ZSTD_loadCEntropy failed");
+ dictPtr += eSize;
+
+ {
+ size_t const dictContentSize = (size_t)(dictEnd - dictPtr);
+ FORWARD_IF_ERROR(ZSTD_loadDictionaryContent(
+ ms, NULL, ws, params, dictPtr, dictContentSize, dtlm), "");
+ }
+ return dictID;
+}
+
+/* ZSTD_compress_insertDictionary() :
+* @return : dictID, or an error code */
+static size_t
+ZSTD_compress_insertDictionary(ZSTD_compressedBlockState_t* bs,
+ ZSTD_matchState_t* ms,
+ ldmState_t* ls,
+ ZSTD_cwksp* ws,
+ const ZSTD_CCtx_params* params,
+ const void* dict, size_t dictSize,
+ ZSTD_dictContentType_e dictContentType,
+ ZSTD_dictTableLoadMethod_e dtlm,
+ void* workspace)
+{
+ DEBUGLOG(4, "ZSTD_compress_insertDictionary (dictSize=%u)", (U32)dictSize);
+ if ((dict==NULL) || (dictSize<8)) {
+ RETURN_ERROR_IF(dictContentType == ZSTD_dct_fullDict, dictionary_wrong, "");
+ return 0;
+ }
+
+ ZSTD_reset_compressedBlockState(bs);
+
+ /* dict restricted modes */
+ if (dictContentType == ZSTD_dct_rawContent)
+ return ZSTD_loadDictionaryContent(ms, ls, ws, params, dict, dictSize, dtlm);
+
+ if (MEM_readLE32(dict) != ZSTD_MAGIC_DICTIONARY) {
+ if (dictContentType == ZSTD_dct_auto) {
+ DEBUGLOG(4, "raw content dictionary detected");
+ return ZSTD_loadDictionaryContent(
+ ms, ls, ws, params, dict, dictSize, dtlm);
+ }
+ RETURN_ERROR_IF(dictContentType == ZSTD_dct_fullDict, dictionary_wrong, "");
+ assert(0); /* impossible */
+ }
+
+ /* dict as full zstd dictionary */
+ return ZSTD_loadZstdDictionary(
+ bs, ms, ws, params, dict, dictSize, dtlm, workspace);
+}
+
+#define ZSTD_USE_CDICT_PARAMS_SRCSIZE_CUTOFF (128 KB)
+#define ZSTD_USE_CDICT_PARAMS_DICTSIZE_MULTIPLIER (6ULL)
+
+/*! ZSTD_compressBegin_internal() :
+ * @return : 0, or an error code */
+static size_t ZSTD_compressBegin_internal(ZSTD_CCtx* cctx,
+ const void* dict, size_t dictSize,
+ ZSTD_dictContentType_e dictContentType,
+ ZSTD_dictTableLoadMethod_e dtlm,
+ const ZSTD_CDict* cdict,
+ const ZSTD_CCtx_params* params, U64 pledgedSrcSize,
+ ZSTD_buffered_policy_e zbuff)
+{
+ size_t const dictContentSize = cdict ? cdict->dictContentSize : dictSize;
+ DEBUGLOG(4, "ZSTD_compressBegin_internal: wlog=%u", params->cParams.windowLog);
+ /* params are supposed to be fully validated at this point */
+ assert(!ZSTD_isError(ZSTD_checkCParams(params->cParams)));
+ assert(!((dict) && (cdict))); /* either dict or cdict, not both */
+ if ( (cdict)
+ && (cdict->dictContentSize > 0)
+ && ( pledgedSrcSize < ZSTD_USE_CDICT_PARAMS_SRCSIZE_CUTOFF
+ || pledgedSrcSize < cdict->dictContentSize * ZSTD_USE_CDICT_PARAMS_DICTSIZE_MULTIPLIER
+ || pledgedSrcSize == ZSTD_CONTENTSIZE_UNKNOWN
+ || cdict->compressionLevel == 0)
+ && (params->attachDictPref != ZSTD_dictForceLoad) ) {
+ return ZSTD_resetCCtx_usingCDict(cctx, cdict, params, pledgedSrcSize, zbuff);
+ }
+
+ FORWARD_IF_ERROR( ZSTD_resetCCtx_internal(cctx, params, pledgedSrcSize,
+ dictContentSize,
+ ZSTDcrp_makeClean, zbuff) , "");
+ { size_t const dictID = cdict ?
+ ZSTD_compress_insertDictionary(
+ cctx->blockState.prevCBlock, &cctx->blockState.matchState,
+ &cctx->ldmState, &cctx->workspace, &cctx->appliedParams, cdict->dictContent,
+ cdict->dictContentSize, cdict->dictContentType, dtlm,
+ cctx->entropyWorkspace)
+ : ZSTD_compress_insertDictionary(
+ cctx->blockState.prevCBlock, &cctx->blockState.matchState,
+ &cctx->ldmState, &cctx->workspace, &cctx->appliedParams, dict, dictSize,
+ dictContentType, dtlm, cctx->entropyWorkspace);
+ FORWARD_IF_ERROR(dictID, "ZSTD_compress_insertDictionary failed");
+ assert(dictID <= UINT_MAX);
+ cctx->dictID = (U32)dictID;
+ cctx->dictContentSize = dictContentSize;
+ }
+ return 0;
+}
+
+size_t ZSTD_compressBegin_advanced_internal(ZSTD_CCtx* cctx,
+ const void* dict, size_t dictSize,
+ ZSTD_dictContentType_e dictContentType,
+ ZSTD_dictTableLoadMethod_e dtlm,
+ const ZSTD_CDict* cdict,
+ const ZSTD_CCtx_params* params,
+ unsigned long long pledgedSrcSize)
+{
+ DEBUGLOG(4, "ZSTD_compressBegin_advanced_internal: wlog=%u", params->cParams.windowLog);
+ /* compression parameters verification and optimization */
+ FORWARD_IF_ERROR( ZSTD_checkCParams(params->cParams) , "");
+ return ZSTD_compressBegin_internal(cctx,
+ dict, dictSize, dictContentType, dtlm,
+ cdict,
+ params, pledgedSrcSize,
+ ZSTDb_not_buffered);
+}
+
+/*! ZSTD_compressBegin_advanced() :
+* @return : 0, or an error code */
+size_t ZSTD_compressBegin_advanced(ZSTD_CCtx* cctx,
+ const void* dict, size_t dictSize,
+ ZSTD_parameters params, unsigned long long pledgedSrcSize)
+{
+ ZSTD_CCtx_params cctxParams;
+ ZSTD_CCtxParams_init_internal(&cctxParams, &params, ZSTD_NO_CLEVEL);
+ return ZSTD_compressBegin_advanced_internal(cctx,
+ dict, dictSize, ZSTD_dct_auto, ZSTD_dtlm_fast,
+ NULL /*cdict*/,
+ &cctxParams, pledgedSrcSize);
+}
+
+size_t ZSTD_compressBegin_usingDict(ZSTD_CCtx* cctx, const void* dict, size_t dictSize, int compressionLevel)
+{
+ ZSTD_CCtx_params cctxParams;
+ {
+ ZSTD_parameters const params = ZSTD_getParams_internal(compressionLevel, ZSTD_CONTENTSIZE_UNKNOWN, dictSize, ZSTD_cpm_noAttachDict);
+ ZSTD_CCtxParams_init_internal(&cctxParams, &params, (compressionLevel == 0) ? ZSTD_CLEVEL_DEFAULT : compressionLevel);
+ }
+ DEBUGLOG(4, "ZSTD_compressBegin_usingDict (dictSize=%u)", (unsigned)dictSize);
+ return ZSTD_compressBegin_internal(cctx, dict, dictSize, ZSTD_dct_auto, ZSTD_dtlm_fast, NULL,
+ &cctxParams, ZSTD_CONTENTSIZE_UNKNOWN, ZSTDb_not_buffered);
+}
+
+size_t ZSTD_compressBegin(ZSTD_CCtx* cctx, int compressionLevel)
+{
+ return ZSTD_compressBegin_usingDict(cctx, NULL, 0, compressionLevel);
+}
+
+
+/*! ZSTD_writeEpilogue() :
+* Ends a frame.
+* @return : nb of bytes written into dst (or an error code) */
+static size_t ZSTD_writeEpilogue(ZSTD_CCtx* cctx, void* dst, size_t dstCapacity)
+{
+ BYTE* const ostart = (BYTE*)dst;
+ BYTE* op = ostart;
+ size_t fhSize = 0;
+
+ DEBUGLOG(4, "ZSTD_writeEpilogue");
+ RETURN_ERROR_IF(cctx->stage == ZSTDcs_created, stage_wrong, "init missing");
+
+ /* special case : empty frame */
+ if (cctx->stage == ZSTDcs_init) {
+ fhSize = ZSTD_writeFrameHeader(dst, dstCapacity, &cctx->appliedParams, 0, 0);
+ FORWARD_IF_ERROR(fhSize, "ZSTD_writeFrameHeader failed");
+ dstCapacity -= fhSize;
+ op += fhSize;
+ cctx->stage = ZSTDcs_ongoing;
+ }
+
+ if (cctx->stage != ZSTDcs_ending) {
+ /* write one last empty block, make it the "last" block */
+ U32 const cBlockHeader24 = 1 /* last block */ + (((U32)bt_raw)<<1) + 0;
+ RETURN_ERROR_IF(dstCapacity<4, dstSize_tooSmall, "no room for epilogue");
+ MEM_writeLE32(op, cBlockHeader24);
+ op += ZSTD_blockHeaderSize;
+ dstCapacity -= ZSTD_blockHeaderSize;
+ }
+
+ if (cctx->appliedParams.fParams.checksumFlag) {
+ U32 const checksum = (U32) xxh64_digest(&cctx->xxhState);
+ RETURN_ERROR_IF(dstCapacity<4, dstSize_tooSmall, "no room for checksum");
+ DEBUGLOG(4, "ZSTD_writeEpilogue: write checksum : %08X", (unsigned)checksum);
+ MEM_writeLE32(op, checksum);
+ op += 4;
+ }
+
+ cctx->stage = ZSTDcs_created; /* return to "created but no init" status */
+ return op-ostart;
+}
+
+void ZSTD_CCtx_trace(ZSTD_CCtx* cctx, size_t extraCSize)
+{
+ (void)cctx;
+ (void)extraCSize;
+}
+
+size_t ZSTD_compressEnd (ZSTD_CCtx* cctx,
+ void* dst, size_t dstCapacity,
+ const void* src, size_t srcSize)
+{
+ size_t endResult;
+ size_t const cSize = ZSTD_compressContinue_internal(cctx,
+ dst, dstCapacity, src, srcSize,
+ 1 /* frame mode */, 1 /* last chunk */);
+ FORWARD_IF_ERROR(cSize, "ZSTD_compressContinue_internal failed");
+ endResult = ZSTD_writeEpilogue(cctx, (char*)dst + cSize, dstCapacity-cSize);
+ FORWARD_IF_ERROR(endResult, "ZSTD_writeEpilogue failed");
+ assert(!(cctx->appliedParams.fParams.contentSizeFlag && cctx->pledgedSrcSizePlusOne == 0));
+ if (cctx->pledgedSrcSizePlusOne != 0) { /* control src size */
+ ZSTD_STATIC_ASSERT(ZSTD_CONTENTSIZE_UNKNOWN == (unsigned long long)-1);
+ DEBUGLOG(4, "end of frame : controlling src size");
+ RETURN_ERROR_IF(
+ cctx->pledgedSrcSizePlusOne != cctx->consumedSrcSize+1,
+ srcSize_wrong,
+ "error : pledgedSrcSize = %u, while realSrcSize = %u",
+ (unsigned)cctx->pledgedSrcSizePlusOne-1,
+ (unsigned)cctx->consumedSrcSize);
+ }
+ ZSTD_CCtx_trace(cctx, endResult);
+ return cSize + endResult;
+}
+
+size_t ZSTD_compress_advanced (ZSTD_CCtx* cctx,
+ void* dst, size_t dstCapacity,
+ const void* src, size_t srcSize,
+ const void* dict,size_t dictSize,
+ ZSTD_parameters params)
+{
+ DEBUGLOG(4, "ZSTD_compress_advanced");
+ FORWARD_IF_ERROR(ZSTD_checkCParams(params.cParams), "");
+ ZSTD_CCtxParams_init_internal(&cctx->simpleApiParams, &params, ZSTD_NO_CLEVEL);
+ return ZSTD_compress_advanced_internal(cctx,
+ dst, dstCapacity,
+ src, srcSize,
+ dict, dictSize,
+ &cctx->simpleApiParams);
+}
+
+/* Internal */
+size_t ZSTD_compress_advanced_internal(
+ ZSTD_CCtx* cctx,
+ void* dst, size_t dstCapacity,
+ const void* src, size_t srcSize,
+ const void* dict,size_t dictSize,
+ const ZSTD_CCtx_params* params)
+{
+ DEBUGLOG(4, "ZSTD_compress_advanced_internal (srcSize:%u)", (unsigned)srcSize);
+ FORWARD_IF_ERROR( ZSTD_compressBegin_internal(cctx,
+ dict, dictSize, ZSTD_dct_auto, ZSTD_dtlm_fast, NULL,
+ params, srcSize, ZSTDb_not_buffered) , "");
+ return ZSTD_compressEnd(cctx, dst, dstCapacity, src, srcSize);
+}
+
+size_t ZSTD_compress_usingDict(ZSTD_CCtx* cctx,
+ void* dst, size_t dstCapacity,
+ const void* src, size_t srcSize,
+ const void* dict, size_t dictSize,
+ int compressionLevel)
+{
+ {
+ ZSTD_parameters const params = ZSTD_getParams_internal(compressionLevel, srcSize, dict ? dictSize : 0, ZSTD_cpm_noAttachDict);
+ assert(params.fParams.contentSizeFlag == 1);
+ ZSTD_CCtxParams_init_internal(&cctx->simpleApiParams, &params, (compressionLevel == 0) ? ZSTD_CLEVEL_DEFAULT: compressionLevel);
+ }
+ DEBUGLOG(4, "ZSTD_compress_usingDict (srcSize=%u)", (unsigned)srcSize);
+ return ZSTD_compress_advanced_internal(cctx, dst, dstCapacity, src, srcSize, dict, dictSize, &cctx->simpleApiParams);
+}
+
+size_t ZSTD_compressCCtx(ZSTD_CCtx* cctx,
+ void* dst, size_t dstCapacity,
+ const void* src, size_t srcSize,
+ int compressionLevel)
+{
+ DEBUGLOG(4, "ZSTD_compressCCtx (srcSize=%u)", (unsigned)srcSize);
+ assert(cctx != NULL);
+ return ZSTD_compress_usingDict(cctx, dst, dstCapacity, src, srcSize, NULL, 0, compressionLevel);
+}
+
+size_t ZSTD_compress(void* dst, size_t dstCapacity,
+ const void* src, size_t srcSize,
+ int compressionLevel)
+{
+ size_t result;
+ ZSTD_CCtx* cctx = ZSTD_createCCtx();
+ RETURN_ERROR_IF(!cctx, memory_allocation, "ZSTD_createCCtx failed");
+ result = ZSTD_compressCCtx(cctx, dst, dstCapacity, src, srcSize, compressionLevel);
+ ZSTD_freeCCtx(cctx);
+ return result;
+}
+
+
+/* ===== Dictionary API ===== */
+
+/*! ZSTD_estimateCDictSize_advanced() :
+ * Estimate amount of memory that will be needed to create a dictionary with following arguments */
+size_t ZSTD_estimateCDictSize_advanced(
+ size_t dictSize, ZSTD_compressionParameters cParams,
+ ZSTD_dictLoadMethod_e dictLoadMethod)
+{
+ DEBUGLOG(5, "sizeof(ZSTD_CDict) : %u", (unsigned)sizeof(ZSTD_CDict));
+ return ZSTD_cwksp_alloc_size(sizeof(ZSTD_CDict))
+ + ZSTD_cwksp_alloc_size(HUF_WORKSPACE_SIZE)
+ /* enableDedicatedDictSearch == 1 ensures that CDict estimation will not be too small
+ * in case we are using DDS with row-hash. */
+ + ZSTD_sizeof_matchState(&cParams, ZSTD_resolveRowMatchFinderMode(ZSTD_ps_auto, &cParams),
+ /* enableDedicatedDictSearch */ 1, /* forCCtx */ 0)
+ + (dictLoadMethod == ZSTD_dlm_byRef ? 0
+ : ZSTD_cwksp_alloc_size(ZSTD_cwksp_align(dictSize, sizeof(void *))));
+}
+
+size_t ZSTD_estimateCDictSize(size_t dictSize, int compressionLevel)
+{
+ ZSTD_compressionParameters const cParams = ZSTD_getCParams_internal(compressionLevel, ZSTD_CONTENTSIZE_UNKNOWN, dictSize, ZSTD_cpm_createCDict);
+ return ZSTD_estimateCDictSize_advanced(dictSize, cParams, ZSTD_dlm_byCopy);
+}
+
+size_t ZSTD_sizeof_CDict(const ZSTD_CDict* cdict)
+{
+ if (cdict==NULL) return 0; /* support sizeof on NULL */
+ DEBUGLOG(5, "sizeof(*cdict) : %u", (unsigned)sizeof(*cdict));
+ /* cdict may be in the workspace */
+ return (cdict->workspace.workspace == cdict ? 0 : sizeof(*cdict))
+ + ZSTD_cwksp_sizeof(&cdict->workspace);
+}
+
+static size_t ZSTD_initCDict_internal(
+ ZSTD_CDict* cdict,
+ const void* dictBuffer, size_t dictSize,
+ ZSTD_dictLoadMethod_e dictLoadMethod,
+ ZSTD_dictContentType_e dictContentType,
+ ZSTD_CCtx_params params)
+{
+ DEBUGLOG(3, "ZSTD_initCDict_internal (dictContentType:%u)", (unsigned)dictContentType);
+ assert(!ZSTD_checkCParams(params.cParams));
+ cdict->matchState.cParams = params.cParams;
+ cdict->matchState.dedicatedDictSearch = params.enableDedicatedDictSearch;
+ if ((dictLoadMethod == ZSTD_dlm_byRef) || (!dictBuffer) || (!dictSize)) {
+ cdict->dictContent = dictBuffer;
+ } else {
+ void *internalBuffer = ZSTD_cwksp_reserve_object(&cdict->workspace, ZSTD_cwksp_align(dictSize, sizeof(void*)));
+ RETURN_ERROR_IF(!internalBuffer, memory_allocation, "NULL pointer!");
+ cdict->dictContent = internalBuffer;
+ ZSTD_memcpy(internalBuffer, dictBuffer, dictSize);
+ }
+ cdict->dictContentSize = dictSize;
+ cdict->dictContentType = dictContentType;
+
+ cdict->entropyWorkspace = (U32*)ZSTD_cwksp_reserve_object(&cdict->workspace, HUF_WORKSPACE_SIZE);
+
+
+ /* Reset the state to no dictionary */
+ ZSTD_reset_compressedBlockState(&cdict->cBlockState);
+ FORWARD_IF_ERROR(ZSTD_reset_matchState(
+ &cdict->matchState,
+ &cdict->workspace,
+ &params.cParams,
+ params.useRowMatchFinder,
+ ZSTDcrp_makeClean,
+ ZSTDirp_reset,
+ ZSTD_resetTarget_CDict), "");
+ /* (Maybe) load the dictionary
+ * Skips loading the dictionary if it is < 8 bytes.
+ */
+ { params.compressionLevel = ZSTD_CLEVEL_DEFAULT;
+ params.fParams.contentSizeFlag = 1;
+ { size_t const dictID = ZSTD_compress_insertDictionary(
+ &cdict->cBlockState, &cdict->matchState, NULL, &cdict->workspace,
+ &params, cdict->dictContent, cdict->dictContentSize,
+ dictContentType, ZSTD_dtlm_full, cdict->entropyWorkspace);
+ FORWARD_IF_ERROR(dictID, "ZSTD_compress_insertDictionary failed");
+ assert(dictID <= (size_t)(U32)-1);
+ cdict->dictID = (U32)dictID;
+ }
+ }
+
+ return 0;
+}
+
+static ZSTD_CDict* ZSTD_createCDict_advanced_internal(size_t dictSize,
+ ZSTD_dictLoadMethod_e dictLoadMethod,
+ ZSTD_compressionParameters cParams,
+ ZSTD_paramSwitch_e useRowMatchFinder,
+ U32 enableDedicatedDictSearch,
+ ZSTD_customMem customMem)
+{
+ if ((!customMem.customAlloc) ^ (!customMem.customFree)) return NULL;
+
+ { size_t const workspaceSize =
+ ZSTD_cwksp_alloc_size(sizeof(ZSTD_CDict)) +
+ ZSTD_cwksp_alloc_size(HUF_WORKSPACE_SIZE) +
+ ZSTD_sizeof_matchState(&cParams, useRowMatchFinder, enableDedicatedDictSearch, /* forCCtx */ 0) +
+ (dictLoadMethod == ZSTD_dlm_byRef ? 0
+ : ZSTD_cwksp_alloc_size(ZSTD_cwksp_align(dictSize, sizeof(void*))));
+ void* const workspace = ZSTD_customMalloc(workspaceSize, customMem);
+ ZSTD_cwksp ws;
+ ZSTD_CDict* cdict;
+
+ if (!workspace) {
+ ZSTD_customFree(workspace, customMem);
+ return NULL;
+ }
+
+ ZSTD_cwksp_init(&ws, workspace, workspaceSize, ZSTD_cwksp_dynamic_alloc);
+
+ cdict = (ZSTD_CDict*)ZSTD_cwksp_reserve_object(&ws, sizeof(ZSTD_CDict));
+ assert(cdict != NULL);
+ ZSTD_cwksp_move(&cdict->workspace, &ws);
+ cdict->customMem = customMem;
+ cdict->compressionLevel = ZSTD_NO_CLEVEL; /* signals advanced API usage */
+ cdict->useRowMatchFinder = useRowMatchFinder;
+ return cdict;
+ }
+}
+
+ZSTD_CDict* ZSTD_createCDict_advanced(const void* dictBuffer, size_t dictSize,
+ ZSTD_dictLoadMethod_e dictLoadMethod,
+ ZSTD_dictContentType_e dictContentType,
+ ZSTD_compressionParameters cParams,
+ ZSTD_customMem customMem)
+{
+ ZSTD_CCtx_params cctxParams;
+ ZSTD_memset(&cctxParams, 0, sizeof(cctxParams));
+ ZSTD_CCtxParams_init(&cctxParams, 0);
+ cctxParams.cParams = cParams;
+ cctxParams.customMem = customMem;
+ return ZSTD_createCDict_advanced2(
+ dictBuffer, dictSize,
+ dictLoadMethod, dictContentType,
+ &cctxParams, customMem);
+}
+
+ZSTD_CDict* ZSTD_createCDict_advanced2(
+ const void* dict, size_t dictSize,
+ ZSTD_dictLoadMethod_e dictLoadMethod,
+ ZSTD_dictContentType_e dictContentType,
+ const ZSTD_CCtx_params* originalCctxParams,
+ ZSTD_customMem customMem)
+{
+ ZSTD_CCtx_params cctxParams = *originalCctxParams;
+ ZSTD_compressionParameters cParams;
+ ZSTD_CDict* cdict;
+
+ DEBUGLOG(3, "ZSTD_createCDict_advanced2, mode %u", (unsigned)dictContentType);
+ if (!customMem.customAlloc ^ !customMem.customFree) return NULL;
+
+ if (cctxParams.enableDedicatedDictSearch) {
+ cParams = ZSTD_dedicatedDictSearch_getCParams(
+ cctxParams.compressionLevel, dictSize);
+ ZSTD_overrideCParams(&cParams, &cctxParams.cParams);
+ } else {
+ cParams = ZSTD_getCParamsFromCCtxParams(
+ &cctxParams, ZSTD_CONTENTSIZE_UNKNOWN, dictSize, ZSTD_cpm_createCDict);
+ }
+
+ if (!ZSTD_dedicatedDictSearch_isSupported(&cParams)) {
+ /* Fall back to non-DDSS params */
+ cctxParams.enableDedicatedDictSearch = 0;
+ cParams = ZSTD_getCParamsFromCCtxParams(
+ &cctxParams, ZSTD_CONTENTSIZE_UNKNOWN, dictSize, ZSTD_cpm_createCDict);
+ }
+
+ DEBUGLOG(3, "ZSTD_createCDict_advanced2: DDS: %u", cctxParams.enableDedicatedDictSearch);
+ cctxParams.cParams = cParams;
+ cctxParams.useRowMatchFinder = ZSTD_resolveRowMatchFinderMode(cctxParams.useRowMatchFinder, &cParams);
+
+ cdict = ZSTD_createCDict_advanced_internal(dictSize,
+ dictLoadMethod, cctxParams.cParams,
+ cctxParams.useRowMatchFinder, cctxParams.enableDedicatedDictSearch,
+ customMem);
+
+ if (ZSTD_isError( ZSTD_initCDict_internal(cdict,
+ dict, dictSize,
+ dictLoadMethod, dictContentType,
+ cctxParams) )) {
+ ZSTD_freeCDict(cdict);
+ return NULL;
+ }
+
+ return cdict;
+}
+
+ZSTD_CDict* ZSTD_createCDict(const void* dict, size_t dictSize, int compressionLevel)
+{
+ ZSTD_compressionParameters cParams = ZSTD_getCParams_internal(compressionLevel, ZSTD_CONTENTSIZE_UNKNOWN, dictSize, ZSTD_cpm_createCDict);
+ ZSTD_CDict* const cdict = ZSTD_createCDict_advanced(dict, dictSize,
+ ZSTD_dlm_byCopy, ZSTD_dct_auto,
+ cParams, ZSTD_defaultCMem);
+ if (cdict)
+ cdict->compressionLevel = (compressionLevel == 0) ? ZSTD_CLEVEL_DEFAULT : compressionLevel;
+ return cdict;
+}
+
+ZSTD_CDict* ZSTD_createCDict_byReference(const void* dict, size_t dictSize, int compressionLevel)
+{
+ ZSTD_compressionParameters cParams = ZSTD_getCParams_internal(compressionLevel, ZSTD_CONTENTSIZE_UNKNOWN, dictSize, ZSTD_cpm_createCDict);
+ ZSTD_CDict* const cdict = ZSTD_createCDict_advanced(dict, dictSize,
+ ZSTD_dlm_byRef, ZSTD_dct_auto,
+ cParams, ZSTD_defaultCMem);
+ if (cdict)
+ cdict->compressionLevel = (compressionLevel == 0) ? ZSTD_CLEVEL_DEFAULT : compressionLevel;
+ return cdict;
+}
+
+size_t ZSTD_freeCDict(ZSTD_CDict* cdict)
+{
+ if (cdict==NULL) return 0; /* support free on NULL */
+ { ZSTD_customMem const cMem = cdict->customMem;
+ int cdictInWorkspace = ZSTD_cwksp_owns_buffer(&cdict->workspace, cdict);
+ ZSTD_cwksp_free(&cdict->workspace, cMem);
+ if (!cdictInWorkspace) {
+ ZSTD_customFree(cdict, cMem);
+ }
+ return 0;
+ }
+}
+
+/*! ZSTD_initStaticCDict_advanced() :
+ * Generate a digested dictionary in provided memory area.
+ * workspace: The memory area to emplace the dictionary into.
+ * Provided pointer must 8-bytes aligned.
+ * It must outlive dictionary usage.
+ * workspaceSize: Use ZSTD_estimateCDictSize()
+ * to determine how large workspace must be.
+ * cParams : use ZSTD_getCParams() to transform a compression level
+ * into its relevants cParams.
+ * @return : pointer to ZSTD_CDict*, or NULL if error (size too small)
+ * Note : there is no corresponding "free" function.
+ * Since workspace was allocated externally, it must be freed externally.
+ */
+const ZSTD_CDict* ZSTD_initStaticCDict(
+ void* workspace, size_t workspaceSize,
+ const void* dict, size_t dictSize,
+ ZSTD_dictLoadMethod_e dictLoadMethod,
+ ZSTD_dictContentType_e dictContentType,
+ ZSTD_compressionParameters cParams)
+{
+ ZSTD_paramSwitch_e const useRowMatchFinder = ZSTD_resolveRowMatchFinderMode(ZSTD_ps_auto, &cParams);
+ /* enableDedicatedDictSearch == 1 ensures matchstate is not too small in case this CDict will be used for DDS + row hash */
+ size_t const matchStateSize = ZSTD_sizeof_matchState(&cParams, useRowMatchFinder, /* enableDedicatedDictSearch */ 1, /* forCCtx */ 0);
+ size_t const neededSize = ZSTD_cwksp_alloc_size(sizeof(ZSTD_CDict))
+ + (dictLoadMethod == ZSTD_dlm_byRef ? 0
+ : ZSTD_cwksp_alloc_size(ZSTD_cwksp_align(dictSize, sizeof(void*))))
+ + ZSTD_cwksp_alloc_size(HUF_WORKSPACE_SIZE)
+ + matchStateSize;
+ ZSTD_CDict* cdict;
+ ZSTD_CCtx_params params;
+
+ if ((size_t)workspace & 7) return NULL; /* 8-aligned */
+
+ {
+ ZSTD_cwksp ws;
+ ZSTD_cwksp_init(&ws, workspace, workspaceSize, ZSTD_cwksp_static_alloc);
+ cdict = (ZSTD_CDict*)ZSTD_cwksp_reserve_object(&ws, sizeof(ZSTD_CDict));
+ if (cdict == NULL) return NULL;
+ ZSTD_cwksp_move(&cdict->workspace, &ws);
+ }
+
+ DEBUGLOG(4, "(workspaceSize < neededSize) : (%u < %u) => %u",
+ (unsigned)workspaceSize, (unsigned)neededSize, (unsigned)(workspaceSize < neededSize));
+ if (workspaceSize < neededSize) return NULL;
+
+ ZSTD_CCtxParams_init(&params, 0);
+ params.cParams = cParams;
+ params.useRowMatchFinder = useRowMatchFinder;
+ cdict->useRowMatchFinder = useRowMatchFinder;
+
+ if (ZSTD_isError( ZSTD_initCDict_internal(cdict,
+ dict, dictSize,
+ dictLoadMethod, dictContentType,
+ params) ))
+ return NULL;
+
+ return cdict;
+}
+
+ZSTD_compressionParameters ZSTD_getCParamsFromCDict(const ZSTD_CDict* cdict)
+{
+ assert(cdict != NULL);
+ return cdict->matchState.cParams;
+}
+
+/*! ZSTD_getDictID_fromCDict() :
+ * Provides the dictID of the dictionary loaded into `cdict`.
+ * If @return == 0, the dictionary is not conformant to Zstandard specification, or empty.
+ * Non-conformant dictionaries can still be loaded, but as content-only dictionaries. */
+unsigned ZSTD_getDictID_fromCDict(const ZSTD_CDict* cdict)
+{
+ if (cdict==NULL) return 0;
+ return cdict->dictID;
+}
+
+/* ZSTD_compressBegin_usingCDict_internal() :
+ * Implementation of various ZSTD_compressBegin_usingCDict* functions.
+ */
+static size_t ZSTD_compressBegin_usingCDict_internal(
+ ZSTD_CCtx* const cctx, const ZSTD_CDict* const cdict,
+ ZSTD_frameParameters const fParams, unsigned long long const pledgedSrcSize)
+{
+ ZSTD_CCtx_params cctxParams;
+ DEBUGLOG(4, "ZSTD_compressBegin_usingCDict_internal");
+ RETURN_ERROR_IF(cdict==NULL, dictionary_wrong, "NULL pointer!");
+ /* Initialize the cctxParams from the cdict */
+ {
+ ZSTD_parameters params;
+ params.fParams = fParams;
+ params.cParams = ( pledgedSrcSize < ZSTD_USE_CDICT_PARAMS_SRCSIZE_CUTOFF
+ || pledgedSrcSize < cdict->dictContentSize * ZSTD_USE_CDICT_PARAMS_DICTSIZE_MULTIPLIER
+ || pledgedSrcSize == ZSTD_CONTENTSIZE_UNKNOWN
+ || cdict->compressionLevel == 0 ) ?
+ ZSTD_getCParamsFromCDict(cdict)
+ : ZSTD_getCParams(cdict->compressionLevel,
+ pledgedSrcSize,
+ cdict->dictContentSize);
+ ZSTD_CCtxParams_init_internal(&cctxParams, &params, cdict->compressionLevel);
+ }
+ /* Increase window log to fit the entire dictionary and source if the
+ * source size is known. Limit the increase to 19, which is the
+ * window log for compression level 1 with the largest source size.
+ */
+ if (pledgedSrcSize != ZSTD_CONTENTSIZE_UNKNOWN) {
+ U32 const limitedSrcSize = (U32)MIN(pledgedSrcSize, 1U << 19);
+ U32 const limitedSrcLog = limitedSrcSize > 1 ? ZSTD_highbit32(limitedSrcSize - 1) + 1 : 1;
+ cctxParams.cParams.windowLog = MAX(cctxParams.cParams.windowLog, limitedSrcLog);
+ }
+ return ZSTD_compressBegin_internal(cctx,
+ NULL, 0, ZSTD_dct_auto, ZSTD_dtlm_fast,
+ cdict,
+ &cctxParams, pledgedSrcSize,
+ ZSTDb_not_buffered);
+}
+
+
+/* ZSTD_compressBegin_usingCDict_advanced() :
+ * This function is DEPRECATED.
+ * cdict must be != NULL */
+size_t ZSTD_compressBegin_usingCDict_advanced(
+ ZSTD_CCtx* const cctx, const ZSTD_CDict* const cdict,
+ ZSTD_frameParameters const fParams, unsigned long long const pledgedSrcSize)
+{
+ return ZSTD_compressBegin_usingCDict_internal(cctx, cdict, fParams, pledgedSrcSize);
+}
+
+/* ZSTD_compressBegin_usingCDict() :
+ * cdict must be != NULL */
+size_t ZSTD_compressBegin_usingCDict(ZSTD_CCtx* cctx, const ZSTD_CDict* cdict)
+{
+ ZSTD_frameParameters const fParams = { 0 /*content*/, 0 /*checksum*/, 0 /*noDictID*/ };
+ return ZSTD_compressBegin_usingCDict_internal(cctx, cdict, fParams, ZSTD_CONTENTSIZE_UNKNOWN);
+}
+
+/*! ZSTD_compress_usingCDict_internal():
+ * Implementation of various ZSTD_compress_usingCDict* functions.
+ */
+static size_t ZSTD_compress_usingCDict_internal(ZSTD_CCtx* cctx,
+ void* dst, size_t dstCapacity,
+ const void* src, size_t srcSize,
+ const ZSTD_CDict* cdict, ZSTD_frameParameters fParams)
+{
+ FORWARD_IF_ERROR(ZSTD_compressBegin_usingCDict_internal(cctx, cdict, fParams, srcSize), ""); /* will check if cdict != NULL */
+ return ZSTD_compressEnd(cctx, dst, dstCapacity, src, srcSize);
+}
+
+/*! ZSTD_compress_usingCDict_advanced():
+ * This function is DEPRECATED.
+ */
+size_t ZSTD_compress_usingCDict_advanced(ZSTD_CCtx* cctx,
+ void* dst, size_t dstCapacity,
+ const void* src, size_t srcSize,
+ const ZSTD_CDict* cdict, ZSTD_frameParameters fParams)
+{
+ return ZSTD_compress_usingCDict_internal(cctx, dst, dstCapacity, src, srcSize, cdict, fParams);
+}
+
+/*! ZSTD_compress_usingCDict() :
+ * Compression using a digested Dictionary.
+ * Faster startup than ZSTD_compress_usingDict(), recommended when same dictionary is used multiple times.
+ * Note that compression parameters are decided at CDict creation time
+ * while frame parameters are hardcoded */
+size_t ZSTD_compress_usingCDict(ZSTD_CCtx* cctx,
+ void* dst, size_t dstCapacity,
+ const void* src, size_t srcSize,
+ const ZSTD_CDict* cdict)
+{
+ ZSTD_frameParameters const fParams = { 1 /*content*/, 0 /*checksum*/, 0 /*noDictID*/ };
+ return ZSTD_compress_usingCDict_internal(cctx, dst, dstCapacity, src, srcSize, cdict, fParams);
+}
+
+
+
+/* ******************************************************************
+* Streaming
+********************************************************************/
+
+ZSTD_CStream* ZSTD_createCStream(void)
+{
+ DEBUGLOG(3, "ZSTD_createCStream");
+ return ZSTD_createCStream_advanced(ZSTD_defaultCMem);
+}
+
+ZSTD_CStream* ZSTD_initStaticCStream(void *workspace, size_t workspaceSize)
+{
+ return ZSTD_initStaticCCtx(workspace, workspaceSize);
+}
+
+ZSTD_CStream* ZSTD_createCStream_advanced(ZSTD_customMem customMem)
+{ /* CStream and CCtx are now same object */
+ return ZSTD_createCCtx_advanced(customMem);
+}
+
+size_t ZSTD_freeCStream(ZSTD_CStream* zcs)
+{
+ return ZSTD_freeCCtx(zcs); /* same object */
+}
+
+
+
+/*====== Initialization ======*/
+
+size_t ZSTD_CStreamInSize(void) { return ZSTD_BLOCKSIZE_MAX; }
+
+size_t ZSTD_CStreamOutSize(void)
+{
+ return ZSTD_compressBound(ZSTD_BLOCKSIZE_MAX) + ZSTD_blockHeaderSize + 4 /* 32-bits hash */ ;
+}
+
+static ZSTD_cParamMode_e ZSTD_getCParamMode(ZSTD_CDict const* cdict, ZSTD_CCtx_params const* params, U64 pledgedSrcSize)
+{
+ if (cdict != NULL && ZSTD_shouldAttachDict(cdict, params, pledgedSrcSize))
+ return ZSTD_cpm_attachDict;
+ else
+ return ZSTD_cpm_noAttachDict;
+}
+
+/* ZSTD_resetCStream():
+ * pledgedSrcSize == 0 means "unknown" */
+size_t ZSTD_resetCStream(ZSTD_CStream* zcs, unsigned long long pss)
+{
+ /* temporary : 0 interpreted as "unknown" during transition period.
+ * Users willing to specify "unknown" **must** use ZSTD_CONTENTSIZE_UNKNOWN.
+ * 0 will be interpreted as "empty" in the future.
+ */
+ U64 const pledgedSrcSize = (pss==0) ? ZSTD_CONTENTSIZE_UNKNOWN : pss;
+ DEBUGLOG(4, "ZSTD_resetCStream: pledgedSrcSize = %u", (unsigned)pledgedSrcSize);
+ FORWARD_IF_ERROR( ZSTD_CCtx_reset(zcs, ZSTD_reset_session_only) , "");
+ FORWARD_IF_ERROR( ZSTD_CCtx_setPledgedSrcSize(zcs, pledgedSrcSize) , "");
+ return 0;
+}
+
+/*! ZSTD_initCStream_internal() :
+ * Note : for lib/compress only. Used by zstdmt_compress.c.
+ * Assumption 1 : params are valid
+ * Assumption 2 : either dict, or cdict, is defined, not both */
+size_t ZSTD_initCStream_internal(ZSTD_CStream* zcs,
+ const void* dict, size_t dictSize, const ZSTD_CDict* cdict,
+ const ZSTD_CCtx_params* params,
+ unsigned long long pledgedSrcSize)
+{
+ DEBUGLOG(4, "ZSTD_initCStream_internal");
+ FORWARD_IF_ERROR( ZSTD_CCtx_reset(zcs, ZSTD_reset_session_only) , "");
+ FORWARD_IF_ERROR( ZSTD_CCtx_setPledgedSrcSize(zcs, pledgedSrcSize) , "");
+ assert(!ZSTD_isError(ZSTD_checkCParams(params->cParams)));
+ zcs->requestedParams = *params;
+ assert(!((dict) && (cdict))); /* either dict or cdict, not both */
+ if (dict) {
+ FORWARD_IF_ERROR( ZSTD_CCtx_loadDictionary(zcs, dict, dictSize) , "");
+ } else {
+ /* Dictionary is cleared if !cdict */
+ FORWARD_IF_ERROR( ZSTD_CCtx_refCDict(zcs, cdict) , "");
+ }
+ return 0;
+}
+
+/* ZSTD_initCStream_usingCDict_advanced() :
+ * same as ZSTD_initCStream_usingCDict(), with control over frame parameters */
+size_t ZSTD_initCStream_usingCDict_advanced(ZSTD_CStream* zcs,
+ const ZSTD_CDict* cdict,
+ ZSTD_frameParameters fParams,
+ unsigned long long pledgedSrcSize)
+{
+ DEBUGLOG(4, "ZSTD_initCStream_usingCDict_advanced");
+ FORWARD_IF_ERROR( ZSTD_CCtx_reset(zcs, ZSTD_reset_session_only) , "");
+ FORWARD_IF_ERROR( ZSTD_CCtx_setPledgedSrcSize(zcs, pledgedSrcSize) , "");
+ zcs->requestedParams.fParams = fParams;
+ FORWARD_IF_ERROR( ZSTD_CCtx_refCDict(zcs, cdict) , "");
+ return 0;
+}
+
+/* note : cdict must outlive compression session */
+size_t ZSTD_initCStream_usingCDict(ZSTD_CStream* zcs, const ZSTD_CDict* cdict)
+{
+ DEBUGLOG(4, "ZSTD_initCStream_usingCDict");
+ FORWARD_IF_ERROR( ZSTD_CCtx_reset(zcs, ZSTD_reset_session_only) , "");
+ FORWARD_IF_ERROR( ZSTD_CCtx_refCDict(zcs, cdict) , "");
+ return 0;
+}
+
+
+/* ZSTD_initCStream_advanced() :
+ * pledgedSrcSize must be exact.
+ * if srcSize is not known at init time, use value ZSTD_CONTENTSIZE_UNKNOWN.
+ * dict is loaded with default parameters ZSTD_dct_auto and ZSTD_dlm_byCopy. */
+size_t ZSTD_initCStream_advanced(ZSTD_CStream* zcs,
+ const void* dict, size_t dictSize,
+ ZSTD_parameters params, unsigned long long pss)
+{
+ /* for compatibility with older programs relying on this behavior.
+ * Users should now specify ZSTD_CONTENTSIZE_UNKNOWN.
+ * This line will be removed in the future.
+ */
+ U64 const pledgedSrcSize = (pss==0 && params.fParams.contentSizeFlag==0) ? ZSTD_CONTENTSIZE_UNKNOWN : pss;
+ DEBUGLOG(4, "ZSTD_initCStream_advanced");
+ FORWARD_IF_ERROR( ZSTD_CCtx_reset(zcs, ZSTD_reset_session_only) , "");
+ FORWARD_IF_ERROR( ZSTD_CCtx_setPledgedSrcSize(zcs, pledgedSrcSize) , "");
+ FORWARD_IF_ERROR( ZSTD_checkCParams(params.cParams) , "");
+ ZSTD_CCtxParams_setZstdParams(&zcs->requestedParams, &params);
+ FORWARD_IF_ERROR( ZSTD_CCtx_loadDictionary(zcs, dict, dictSize) , "");
+ return 0;
+}
+
+size_t ZSTD_initCStream_usingDict(ZSTD_CStream* zcs, const void* dict, size_t dictSize, int compressionLevel)
+{
+ DEBUGLOG(4, "ZSTD_initCStream_usingDict");
+ FORWARD_IF_ERROR( ZSTD_CCtx_reset(zcs, ZSTD_reset_session_only) , "");
+ FORWARD_IF_ERROR( ZSTD_CCtx_setParameter(zcs, ZSTD_c_compressionLevel, compressionLevel) , "");
+ FORWARD_IF_ERROR( ZSTD_CCtx_loadDictionary(zcs, dict, dictSize) , "");
+ return 0;
+}
+
+size_t ZSTD_initCStream_srcSize(ZSTD_CStream* zcs, int compressionLevel, unsigned long long pss)
+{
+ /* temporary : 0 interpreted as "unknown" during transition period.
+ * Users willing to specify "unknown" **must** use ZSTD_CONTENTSIZE_UNKNOWN.
+ * 0 will be interpreted as "empty" in the future.
+ */
+ U64 const pledgedSrcSize = (pss==0) ? ZSTD_CONTENTSIZE_UNKNOWN : pss;
+ DEBUGLOG(4, "ZSTD_initCStream_srcSize");
+ FORWARD_IF_ERROR( ZSTD_CCtx_reset(zcs, ZSTD_reset_session_only) , "");
+ FORWARD_IF_ERROR( ZSTD_CCtx_refCDict(zcs, NULL) , "");
+ FORWARD_IF_ERROR( ZSTD_CCtx_setParameter(zcs, ZSTD_c_compressionLevel, compressionLevel) , "");
+ FORWARD_IF_ERROR( ZSTD_CCtx_setPledgedSrcSize(zcs, pledgedSrcSize) , "");
+ return 0;
+}
+
+size_t ZSTD_initCStream(ZSTD_CStream* zcs, int compressionLevel)
+{
+ DEBUGLOG(4, "ZSTD_initCStream");
+ FORWARD_IF_ERROR( ZSTD_CCtx_reset(zcs, ZSTD_reset_session_only) , "");
+ FORWARD_IF_ERROR( ZSTD_CCtx_refCDict(zcs, NULL) , "");
+ FORWARD_IF_ERROR( ZSTD_CCtx_setParameter(zcs, ZSTD_c_compressionLevel, compressionLevel) , "");
+ return 0;
+}
+
+/*====== Compression ======*/
+
+static size_t ZSTD_nextInputSizeHint(const ZSTD_CCtx* cctx)
+{
+ size_t hintInSize = cctx->inBuffTarget - cctx->inBuffPos;
+ if (hintInSize==0) hintInSize = cctx->blockSize;
+ return hintInSize;
+}
+
+/* ZSTD_compressStream_generic():
+ * internal function for all *compressStream*() variants
+ * non-static, because can be called from zstdmt_compress.c
+ * @return : hint size for next input */
+static size_t ZSTD_compressStream_generic(ZSTD_CStream* zcs,
+ ZSTD_outBuffer* output,
+ ZSTD_inBuffer* input,
+ ZSTD_EndDirective const flushMode)
+{
+ const char* const istart = (const char*)input->src;
+ const char* const iend = input->size != 0 ? istart + input->size : istart;
+ const char* ip = input->pos != 0 ? istart + input->pos : istart;
+ char* const ostart = (char*)output->dst;
+ char* const oend = output->size != 0 ? ostart + output->size : ostart;
+ char* op = output->pos != 0 ? ostart + output->pos : ostart;
+ U32 someMoreWork = 1;
+
+ /* check expectations */
+ DEBUGLOG(5, "ZSTD_compressStream_generic, flush=%u", (unsigned)flushMode);
+ if (zcs->appliedParams.inBufferMode == ZSTD_bm_buffered) {
+ assert(zcs->inBuff != NULL);
+ assert(zcs->inBuffSize > 0);
+ }
+ if (zcs->appliedParams.outBufferMode == ZSTD_bm_buffered) {
+ assert(zcs->outBuff != NULL);
+ assert(zcs->outBuffSize > 0);
+ }
+ assert(output->pos <= output->size);
+ assert(input->pos <= input->size);
+ assert((U32)flushMode <= (U32)ZSTD_e_end);
+
+ while (someMoreWork) {
+ switch(zcs->streamStage)
+ {
+ case zcss_init:
+ RETURN_ERROR(init_missing, "call ZSTD_initCStream() first!");
+
+ case zcss_load:
+ if ( (flushMode == ZSTD_e_end)
+ && ( (size_t)(oend-op) >= ZSTD_compressBound(iend-ip) /* Enough output space */
+ || zcs->appliedParams.outBufferMode == ZSTD_bm_stable) /* OR we are allowed to return dstSizeTooSmall */
+ && (zcs->inBuffPos == 0) ) {
+ /* shortcut to compression pass directly into output buffer */
+ size_t const cSize = ZSTD_compressEnd(zcs,
+ op, oend-op, ip, iend-ip);
+ DEBUGLOG(4, "ZSTD_compressEnd : cSize=%u", (unsigned)cSize);
+ FORWARD_IF_ERROR(cSize, "ZSTD_compressEnd failed");
+ ip = iend;
+ op += cSize;
+ zcs->frameEnded = 1;
+ ZSTD_CCtx_reset(zcs, ZSTD_reset_session_only);
+ someMoreWork = 0; break;
+ }
+ /* complete loading into inBuffer in buffered mode */
+ if (zcs->appliedParams.inBufferMode == ZSTD_bm_buffered) {
+ size_t const toLoad = zcs->inBuffTarget - zcs->inBuffPos;
+ size_t const loaded = ZSTD_limitCopy(
+ zcs->inBuff + zcs->inBuffPos, toLoad,
+ ip, iend-ip);
+ zcs->inBuffPos += loaded;
+ if (loaded != 0)
+ ip += loaded;
+ if ( (flushMode == ZSTD_e_continue)
+ && (zcs->inBuffPos < zcs->inBuffTarget) ) {
+ /* not enough input to fill full block : stop here */
+ someMoreWork = 0; break;
+ }
+ if ( (flushMode == ZSTD_e_flush)
+ && (zcs->inBuffPos == zcs->inToCompress) ) {
+ /* empty */
+ someMoreWork = 0; break;
+ }
+ }
+ /* compress current block (note : this stage cannot be stopped in the middle) */
+ DEBUGLOG(5, "stream compression stage (flushMode==%u)", flushMode);
+ { int const inputBuffered = (zcs->appliedParams.inBufferMode == ZSTD_bm_buffered);
+ void* cDst;
+ size_t cSize;
+ size_t oSize = oend-op;
+ size_t const iSize = inputBuffered
+ ? zcs->inBuffPos - zcs->inToCompress
+ : MIN((size_t)(iend - ip), zcs->blockSize);
+ if (oSize >= ZSTD_compressBound(iSize) || zcs->appliedParams.outBufferMode == ZSTD_bm_stable)
+ cDst = op; /* compress into output buffer, to skip flush stage */
+ else
+ cDst = zcs->outBuff, oSize = zcs->outBuffSize;
+ if (inputBuffered) {
+ unsigned const lastBlock = (flushMode == ZSTD_e_end) && (ip==iend);
+ cSize = lastBlock ?
+ ZSTD_compressEnd(zcs, cDst, oSize,
+ zcs->inBuff + zcs->inToCompress, iSize) :
+ ZSTD_compressContinue(zcs, cDst, oSize,
+ zcs->inBuff + zcs->inToCompress, iSize);
+ FORWARD_IF_ERROR(cSize, "%s", lastBlock ? "ZSTD_compressEnd failed" : "ZSTD_compressContinue failed");
+ zcs->frameEnded = lastBlock;
+ /* prepare next block */
+ zcs->inBuffTarget = zcs->inBuffPos + zcs->blockSize;
+ if (zcs->inBuffTarget > zcs->inBuffSize)
+ zcs->inBuffPos = 0, zcs->inBuffTarget = zcs->blockSize;
+ DEBUGLOG(5, "inBuffTarget:%u / inBuffSize:%u",
+ (unsigned)zcs->inBuffTarget, (unsigned)zcs->inBuffSize);
+ if (!lastBlock)
+ assert(zcs->inBuffTarget <= zcs->inBuffSize);
+ zcs->inToCompress = zcs->inBuffPos;
+ } else {
+ unsigned const lastBlock = (ip + iSize == iend);
+ assert(flushMode == ZSTD_e_end /* Already validated */);
+ cSize = lastBlock ?
+ ZSTD_compressEnd(zcs, cDst, oSize, ip, iSize) :
+ ZSTD_compressContinue(zcs, cDst, oSize, ip, iSize);
+ /* Consume the input prior to error checking to mirror buffered mode. */
+ if (iSize > 0)
+ ip += iSize;
+ FORWARD_IF_ERROR(cSize, "%s", lastBlock ? "ZSTD_compressEnd failed" : "ZSTD_compressContinue failed");
+ zcs->frameEnded = lastBlock;
+ if (lastBlock)
+ assert(ip == iend);
+ }
+ if (cDst == op) { /* no need to flush */
+ op += cSize;
+ if (zcs->frameEnded) {
+ DEBUGLOG(5, "Frame completed directly in outBuffer");
+ someMoreWork = 0;
+ ZSTD_CCtx_reset(zcs, ZSTD_reset_session_only);
+ }
+ break;
+ }
+ zcs->outBuffContentSize = cSize;
+ zcs->outBuffFlushedSize = 0;
+ zcs->streamStage = zcss_flush; /* pass-through to flush stage */
+ }
+ ZSTD_FALLTHROUGH;
+ case zcss_flush:
+ DEBUGLOG(5, "flush stage");
+ assert(zcs->appliedParams.outBufferMode == ZSTD_bm_buffered);
+ { size_t const toFlush = zcs->outBuffContentSize - zcs->outBuffFlushedSize;
+ size_t const flushed = ZSTD_limitCopy(op, (size_t)(oend-op),
+ zcs->outBuff + zcs->outBuffFlushedSize, toFlush);
+ DEBUGLOG(5, "toFlush: %u into %u ==> flushed: %u",
+ (unsigned)toFlush, (unsigned)(oend-op), (unsigned)flushed);
+ if (flushed)
+ op += flushed;
+ zcs->outBuffFlushedSize += flushed;
+ if (toFlush!=flushed) {
+ /* flush not fully completed, presumably because dst is too small */
+ assert(op==oend);
+ someMoreWork = 0;
+ break;
+ }
+ zcs->outBuffContentSize = zcs->outBuffFlushedSize = 0;
+ if (zcs->frameEnded) {
+ DEBUGLOG(5, "Frame completed on flush");
+ someMoreWork = 0;
+ ZSTD_CCtx_reset(zcs, ZSTD_reset_session_only);
+ break;
+ }
+ zcs->streamStage = zcss_load;
+ break;
+ }
+
+ default: /* impossible */
+ assert(0);
+ }
+ }
+
+ input->pos = ip - istart;
+ output->pos = op - ostart;
+ if (zcs->frameEnded) return 0;
+ return ZSTD_nextInputSizeHint(zcs);
+}
+
+static size_t ZSTD_nextInputSizeHint_MTorST(const ZSTD_CCtx* cctx)
+{
+ return ZSTD_nextInputSizeHint(cctx);
+
+}
+
+size_t ZSTD_compressStream(ZSTD_CStream* zcs, ZSTD_outBuffer* output, ZSTD_inBuffer* input)
+{
+ FORWARD_IF_ERROR( ZSTD_compressStream2(zcs, output, input, ZSTD_e_continue) , "");
+ return ZSTD_nextInputSizeHint_MTorST(zcs);
+}
+
+/* After a compression call set the expected input/output buffer.
+ * This is validated at the start of the next compression call.
+ */
+static void ZSTD_setBufferExpectations(ZSTD_CCtx* cctx, ZSTD_outBuffer const* output, ZSTD_inBuffer const* input)
+{
+ if (cctx->appliedParams.inBufferMode == ZSTD_bm_stable) {
+ cctx->expectedInBuffer = *input;
+ }
+ if (cctx->appliedParams.outBufferMode == ZSTD_bm_stable) {
+ cctx->expectedOutBufferSize = output->size - output->pos;
+ }
+}
+
+/* Validate that the input/output buffers match the expectations set by
+ * ZSTD_setBufferExpectations.
+ */
+static size_t ZSTD_checkBufferStability(ZSTD_CCtx const* cctx,
+ ZSTD_outBuffer const* output,
+ ZSTD_inBuffer const* input,
+ ZSTD_EndDirective endOp)
+{
+ if (cctx->appliedParams.inBufferMode == ZSTD_bm_stable) {
+ ZSTD_inBuffer const expect = cctx->expectedInBuffer;
+ if (expect.src != input->src || expect.pos != input->pos || expect.size != input->size)
+ RETURN_ERROR(srcBuffer_wrong, "ZSTD_c_stableInBuffer enabled but input differs!");
+ if (endOp != ZSTD_e_end)
+ RETURN_ERROR(srcBuffer_wrong, "ZSTD_c_stableInBuffer can only be used with ZSTD_e_end!");
+ }
+ if (cctx->appliedParams.outBufferMode == ZSTD_bm_stable) {
+ size_t const outBufferSize = output->size - output->pos;
+ if (cctx->expectedOutBufferSize != outBufferSize)
+ RETURN_ERROR(dstBuffer_wrong, "ZSTD_c_stableOutBuffer enabled but output size differs!");
+ }
+ return 0;
+}
+
+static size_t ZSTD_CCtx_init_compressStream2(ZSTD_CCtx* cctx,
+ ZSTD_EndDirective endOp,
+ size_t inSize) {
+ ZSTD_CCtx_params params = cctx->requestedParams;
+ ZSTD_prefixDict const prefixDict = cctx->prefixDict;
+ FORWARD_IF_ERROR( ZSTD_initLocalDict(cctx) , ""); /* Init the local dict if present. */
+ ZSTD_memset(&cctx->prefixDict, 0, sizeof(cctx->prefixDict)); /* single usage */
+ assert(prefixDict.dict==NULL || cctx->cdict==NULL); /* only one can be set */
+ if (cctx->cdict && !cctx->localDict.cdict) {
+ /* Let the cdict's compression level take priority over the requested params.
+ * But do not take the cdict's compression level if the "cdict" is actually a localDict
+ * generated from ZSTD_initLocalDict().
+ */
+ params.compressionLevel = cctx->cdict->compressionLevel;
+ }
+ DEBUGLOG(4, "ZSTD_compressStream2 : transparent init stage");
+ if (endOp == ZSTD_e_end) cctx->pledgedSrcSizePlusOne = inSize + 1; /* auto-fix pledgedSrcSize */
+ {
+ size_t const dictSize = prefixDict.dict
+ ? prefixDict.dictSize
+ : (cctx->cdict ? cctx->cdict->dictContentSize : 0);
+ ZSTD_cParamMode_e const mode = ZSTD_getCParamMode(cctx->cdict, &params, cctx->pledgedSrcSizePlusOne - 1);
+ params.cParams = ZSTD_getCParamsFromCCtxParams(
+ &params, cctx->pledgedSrcSizePlusOne-1,
+ dictSize, mode);
+ }
+
+ params.useBlockSplitter = ZSTD_resolveBlockSplitterMode(params.useBlockSplitter, &params.cParams);
+ params.ldmParams.enableLdm = ZSTD_resolveEnableLdm(params.ldmParams.enableLdm, &params.cParams);
+ params.useRowMatchFinder = ZSTD_resolveRowMatchFinderMode(params.useRowMatchFinder, &params.cParams);
+
+ { U64 const pledgedSrcSize = cctx->pledgedSrcSizePlusOne - 1;
+ assert(!ZSTD_isError(ZSTD_checkCParams(params.cParams)));
+ FORWARD_IF_ERROR( ZSTD_compressBegin_internal(cctx,
+ prefixDict.dict, prefixDict.dictSize, prefixDict.dictContentType, ZSTD_dtlm_fast,
+ cctx->cdict,
+ &params, pledgedSrcSize,
+ ZSTDb_buffered) , "");
+ assert(cctx->appliedParams.nbWorkers == 0);
+ cctx->inToCompress = 0;
+ cctx->inBuffPos = 0;
+ if (cctx->appliedParams.inBufferMode == ZSTD_bm_buffered) {
+ /* for small input: avoid automatic flush on reaching end of block, since
+ * it would require to add a 3-bytes null block to end frame
+ */
+ cctx->inBuffTarget = cctx->blockSize + (cctx->blockSize == pledgedSrcSize);
+ } else {
+ cctx->inBuffTarget = 0;
+ }
+ cctx->outBuffContentSize = cctx->outBuffFlushedSize = 0;
+ cctx->streamStage = zcss_load;
+ cctx->frameEnded = 0;
+ }
+ return 0;
+}
+
+size_t ZSTD_compressStream2( ZSTD_CCtx* cctx,
+ ZSTD_outBuffer* output,
+ ZSTD_inBuffer* input,
+ ZSTD_EndDirective endOp)
+{
+ DEBUGLOG(5, "ZSTD_compressStream2, endOp=%u ", (unsigned)endOp);
+ /* check conditions */
+ RETURN_ERROR_IF(output->pos > output->size, dstSize_tooSmall, "invalid output buffer");
+ RETURN_ERROR_IF(input->pos > input->size, srcSize_wrong, "invalid input buffer");
+ RETURN_ERROR_IF((U32)endOp > (U32)ZSTD_e_end, parameter_outOfBound, "invalid endDirective");
+ assert(cctx != NULL);
+
+ /* transparent initialization stage */
+ if (cctx->streamStage == zcss_init) {
+ FORWARD_IF_ERROR(ZSTD_CCtx_init_compressStream2(cctx, endOp, input->size), "CompressStream2 initialization failed");
+ ZSTD_setBufferExpectations(cctx, output, input); /* Set initial buffer expectations now that we've initialized */
+ }
+ /* end of transparent initialization stage */
+
+ FORWARD_IF_ERROR(ZSTD_checkBufferStability(cctx, output, input, endOp), "invalid buffers");
+ /* compression stage */
+ FORWARD_IF_ERROR( ZSTD_compressStream_generic(cctx, output, input, endOp) , "");
+ DEBUGLOG(5, "completed ZSTD_compressStream2");
+ ZSTD_setBufferExpectations(cctx, output, input);
+ return cctx->outBuffContentSize - cctx->outBuffFlushedSize; /* remaining to flush */
+}
+
+size_t ZSTD_compressStream2_simpleArgs (
+ ZSTD_CCtx* cctx,
+ void* dst, size_t dstCapacity, size_t* dstPos,
+ const void* src, size_t srcSize, size_t* srcPos,
+ ZSTD_EndDirective endOp)
+{
+ ZSTD_outBuffer output = { dst, dstCapacity, *dstPos };
+ ZSTD_inBuffer input = { src, srcSize, *srcPos };
+ /* ZSTD_compressStream2() will check validity of dstPos and srcPos */
+ size_t const cErr = ZSTD_compressStream2(cctx, &output, &input, endOp);
+ *dstPos = output.pos;
+ *srcPos = input.pos;
+ return cErr;
+}
+
+size_t ZSTD_compress2(ZSTD_CCtx* cctx,
+ void* dst, size_t dstCapacity,
+ const void* src, size_t srcSize)
+{
+ ZSTD_bufferMode_e const originalInBufferMode = cctx->requestedParams.inBufferMode;
+ ZSTD_bufferMode_e const originalOutBufferMode = cctx->requestedParams.outBufferMode;
+ DEBUGLOG(4, "ZSTD_compress2 (srcSize=%u)", (unsigned)srcSize);
+ ZSTD_CCtx_reset(cctx, ZSTD_reset_session_only);
+ /* Enable stable input/output buffers. */
+ cctx->requestedParams.inBufferMode = ZSTD_bm_stable;
+ cctx->requestedParams.outBufferMode = ZSTD_bm_stable;
+ { size_t oPos = 0;
+ size_t iPos = 0;
+ size_t const result = ZSTD_compressStream2_simpleArgs(cctx,
+ dst, dstCapacity, &oPos,
+ src, srcSize, &iPos,
+ ZSTD_e_end);
+ /* Reset to the original values. */
+ cctx->requestedParams.inBufferMode = originalInBufferMode;
+ cctx->requestedParams.outBufferMode = originalOutBufferMode;
+ FORWARD_IF_ERROR(result, "ZSTD_compressStream2_simpleArgs failed");
+ if (result != 0) { /* compression not completed, due to lack of output space */
+ assert(oPos == dstCapacity);
+ RETURN_ERROR(dstSize_tooSmall, "");
+ }
+ assert(iPos == srcSize); /* all input is expected consumed */
+ return oPos;
+ }
+}
+
+typedef struct {
+ U32 idx; /* Index in array of ZSTD_Sequence */
+ U32 posInSequence; /* Position within sequence at idx */
+ size_t posInSrc; /* Number of bytes given by sequences provided so far */
+} ZSTD_sequencePosition;
+
+/* ZSTD_validateSequence() :
+ * @offCode : is presumed to follow format required by ZSTD_storeSeq()
+ * @returns a ZSTD error code if sequence is not valid
+ */
+static size_t
+ZSTD_validateSequence(U32 offCode, U32 matchLength,
+ size_t posInSrc, U32 windowLog, size_t dictSize)
+{
+ U32 const windowSize = 1 << windowLog;
+ /* posInSrc represents the amount of data the decoder would decode up to this point.
+ * As long as the amount of data decoded is less than or equal to window size, offsets may be
+ * larger than the total length of output decoded in order to reference the dict, even larger than
+ * window size. After output surpasses windowSize, we're limited to windowSize offsets again.
+ */
+ size_t const offsetBound = posInSrc > windowSize ? (size_t)windowSize : posInSrc + (size_t)dictSize;
+ RETURN_ERROR_IF(offCode > STORE_OFFSET(offsetBound), corruption_detected, "Offset too large!");
+ RETURN_ERROR_IF(matchLength < MINMATCH, corruption_detected, "Matchlength too small");
+ return 0;
+}
+
+/* Returns an offset code, given a sequence's raw offset, the ongoing repcode array, and whether litLength == 0 */
+static U32 ZSTD_finalizeOffCode(U32 rawOffset, const U32 rep[ZSTD_REP_NUM], U32 ll0)
+{
+ U32 offCode = STORE_OFFSET(rawOffset);
+
+ if (!ll0 && rawOffset == rep[0]) {
+ offCode = STORE_REPCODE_1;
+ } else if (rawOffset == rep[1]) {
+ offCode = STORE_REPCODE(2 - ll0);
+ } else if (rawOffset == rep[2]) {
+ offCode = STORE_REPCODE(3 - ll0);
+ } else if (ll0 && rawOffset == rep[0] - 1) {
+ offCode = STORE_REPCODE_3;
+ }
+ return offCode;
+}
+
+/* Returns 0 on success, and a ZSTD_error otherwise. This function scans through an array of
+ * ZSTD_Sequence, storing the sequences it finds, until it reaches a block delimiter.
+ */
+static size_t
+ZSTD_copySequencesToSeqStoreExplicitBlockDelim(ZSTD_CCtx* cctx,
+ ZSTD_sequencePosition* seqPos,
+ const ZSTD_Sequence* const inSeqs, size_t inSeqsSize,
+ const void* src, size_t blockSize)
+{
+ U32 idx = seqPos->idx;
+ BYTE const* ip = (BYTE const*)(src);
+ const BYTE* const iend = ip + blockSize;
+ repcodes_t updatedRepcodes;
+ U32 dictSize;
+
+ if (cctx->cdict) {
+ dictSize = (U32)cctx->cdict->dictContentSize;
+ } else if (cctx->prefixDict.dict) {
+ dictSize = (U32)cctx->prefixDict.dictSize;
+ } else {
+ dictSize = 0;
+ }
+ ZSTD_memcpy(updatedRepcodes.rep, cctx->blockState.prevCBlock->rep, sizeof(repcodes_t));
+ for (; (inSeqs[idx].matchLength != 0 || inSeqs[idx].offset != 0) && idx < inSeqsSize; ++idx) {
+ U32 const litLength = inSeqs[idx].litLength;
+ U32 const ll0 = (litLength == 0);
+ U32 const matchLength = inSeqs[idx].matchLength;
+ U32 const offCode = ZSTD_finalizeOffCode(inSeqs[idx].offset, updatedRepcodes.rep, ll0);
+ ZSTD_updateRep(updatedRepcodes.rep, offCode, ll0);
+
+ DEBUGLOG(6, "Storing sequence: (of: %u, ml: %u, ll: %u)", offCode, matchLength, litLength);
+ if (cctx->appliedParams.validateSequences) {
+ seqPos->posInSrc += litLength + matchLength;
+ FORWARD_IF_ERROR(ZSTD_validateSequence(offCode, matchLength, seqPos->posInSrc,
+ cctx->appliedParams.cParams.windowLog, dictSize),
+ "Sequence validation failed");
+ }
+ RETURN_ERROR_IF(idx - seqPos->idx > cctx->seqStore.maxNbSeq, memory_allocation,
+ "Not enough memory allocated. Try adjusting ZSTD_c_minMatch.");
+ ZSTD_storeSeq(&cctx->seqStore, litLength, ip, iend, offCode, matchLength);
+ ip += matchLength + litLength;
+ }
+ ZSTD_memcpy(cctx->blockState.nextCBlock->rep, updatedRepcodes.rep, sizeof(repcodes_t));
+
+ if (inSeqs[idx].litLength) {
+ DEBUGLOG(6, "Storing last literals of size: %u", inSeqs[idx].litLength);
+ ZSTD_storeLastLiterals(&cctx->seqStore, ip, inSeqs[idx].litLength);
+ ip += inSeqs[idx].litLength;
+ seqPos->posInSrc += inSeqs[idx].litLength;
+ }
+ RETURN_ERROR_IF(ip != iend, corruption_detected, "Blocksize doesn't agree with block delimiter!");
+ seqPos->idx = idx+1;
+ return 0;
+}
+
+/* Returns the number of bytes to move the current read position back by. Only non-zero
+ * if we ended up splitting a sequence. Otherwise, it may return a ZSTD error if something
+ * went wrong.
+ *
+ * This function will attempt to scan through blockSize bytes represented by the sequences
+ * in inSeqs, storing any (partial) sequences.
+ *
+ * Occasionally, we may want to change the actual number of bytes we consumed from inSeqs to
+ * avoid splitting a match, or to avoid splitting a match such that it would produce a match
+ * smaller than MINMATCH. In this case, we return the number of bytes that we didn't read from this block.
+ */
+static size_t
+ZSTD_copySequencesToSeqStoreNoBlockDelim(ZSTD_CCtx* cctx, ZSTD_sequencePosition* seqPos,
+ const ZSTD_Sequence* const inSeqs, size_t inSeqsSize,
+ const void* src, size_t blockSize)
+{
+ U32 idx = seqPos->idx;
+ U32 startPosInSequence = seqPos->posInSequence;
+ U32 endPosInSequence = seqPos->posInSequence + (U32)blockSize;
+ size_t dictSize;
+ BYTE const* ip = (BYTE const*)(src);
+ BYTE const* iend = ip + blockSize; /* May be adjusted if we decide to process fewer than blockSize bytes */
+ repcodes_t updatedRepcodes;
+ U32 bytesAdjustment = 0;
+ U32 finalMatchSplit = 0;
+
+ if (cctx->cdict) {
+ dictSize = cctx->cdict->dictContentSize;
+ } else if (cctx->prefixDict.dict) {
+ dictSize = cctx->prefixDict.dictSize;
+ } else {
+ dictSize = 0;
+ }
+ DEBUGLOG(5, "ZSTD_copySequencesToSeqStore: idx: %u PIS: %u blockSize: %zu", idx, startPosInSequence, blockSize);
+ DEBUGLOG(5, "Start seq: idx: %u (of: %u ml: %u ll: %u)", idx, inSeqs[idx].offset, inSeqs[idx].matchLength, inSeqs[idx].litLength);
+ ZSTD_memcpy(updatedRepcodes.rep, cctx->blockState.prevCBlock->rep, sizeof(repcodes_t));
+ while (endPosInSequence && idx < inSeqsSize && !finalMatchSplit) {
+ const ZSTD_Sequence currSeq = inSeqs[idx];
+ U32 litLength = currSeq.litLength;
+ U32 matchLength = currSeq.matchLength;
+ U32 const rawOffset = currSeq.offset;
+ U32 offCode;
+
+ /* Modify the sequence depending on where endPosInSequence lies */
+ if (endPosInSequence >= currSeq.litLength + currSeq.matchLength) {
+ if (startPosInSequence >= litLength) {
+ startPosInSequence -= litLength;
+ litLength = 0;
+ matchLength -= startPosInSequence;
+ } else {
+ litLength -= startPosInSequence;
+ }
+ /* Move to the next sequence */
+ endPosInSequence -= currSeq.litLength + currSeq.matchLength;
+ startPosInSequence = 0;
+ idx++;
+ } else {
+ /* This is the final (partial) sequence we're adding from inSeqs, and endPosInSequence
+ does not reach the end of the match. So, we have to split the sequence */
+ DEBUGLOG(6, "Require a split: diff: %u, idx: %u PIS: %u",
+ currSeq.litLength + currSeq.matchLength - endPosInSequence, idx, endPosInSequence);
+ if (endPosInSequence > litLength) {
+ U32 firstHalfMatchLength;
+ litLength = startPosInSequence >= litLength ? 0 : litLength - startPosInSequence;
+ firstHalfMatchLength = endPosInSequence - startPosInSequence - litLength;
+ if (matchLength > blockSize && firstHalfMatchLength >= cctx->appliedParams.cParams.minMatch) {
+ /* Only ever split the match if it is larger than the block size */
+ U32 secondHalfMatchLength = currSeq.matchLength + currSeq.litLength - endPosInSequence;
+ if (secondHalfMatchLength < cctx->appliedParams.cParams.minMatch) {
+ /* Move the endPosInSequence backward so that it creates match of minMatch length */
+ endPosInSequence -= cctx->appliedParams.cParams.minMatch - secondHalfMatchLength;
+ bytesAdjustment = cctx->appliedParams.cParams.minMatch - secondHalfMatchLength;
+ firstHalfMatchLength -= bytesAdjustment;
+ }
+ matchLength = firstHalfMatchLength;
+ /* Flag that we split the last match - after storing the sequence, exit the loop,
+ but keep the value of endPosInSequence */
+ finalMatchSplit = 1;
+ } else {
+ /* Move the position in sequence backwards so that we don't split match, and break to store
+ * the last literals. We use the original currSeq.litLength as a marker for where endPosInSequence
+ * should go. We prefer to do this whenever it is not necessary to split the match, or if doing so
+ * would cause the first half of the match to be too small
+ */
+ bytesAdjustment = endPosInSequence - currSeq.litLength;
+ endPosInSequence = currSeq.litLength;
+ break;
+ }
+ } else {
+ /* This sequence ends inside the literals, break to store the last literals */
+ break;
+ }
+ }
+ /* Check if this offset can be represented with a repcode */
+ { U32 const ll0 = (litLength == 0);
+ offCode = ZSTD_finalizeOffCode(rawOffset, updatedRepcodes.rep, ll0);
+ ZSTD_updateRep(updatedRepcodes.rep, offCode, ll0);
+ }
+
+ if (cctx->appliedParams.validateSequences) {
+ seqPos->posInSrc += litLength + matchLength;
+ FORWARD_IF_ERROR(ZSTD_validateSequence(offCode, matchLength, seqPos->posInSrc,
+ cctx->appliedParams.cParams.windowLog, dictSize),
+ "Sequence validation failed");
+ }
+ DEBUGLOG(6, "Storing sequence: (of: %u, ml: %u, ll: %u)", offCode, matchLength, litLength);
+ RETURN_ERROR_IF(idx - seqPos->idx > cctx->seqStore.maxNbSeq, memory_allocation,
+ "Not enough memory allocated. Try adjusting ZSTD_c_minMatch.");
+ ZSTD_storeSeq(&cctx->seqStore, litLength, ip, iend, offCode, matchLength);
+ ip += matchLength + litLength;
+ }
+ DEBUGLOG(5, "Ending seq: idx: %u (of: %u ml: %u ll: %u)", idx, inSeqs[idx].offset, inSeqs[idx].matchLength, inSeqs[idx].litLength);
+ assert(idx == inSeqsSize || endPosInSequence <= inSeqs[idx].litLength + inSeqs[idx].matchLength);
+ seqPos->idx = idx;
+ seqPos->posInSequence = endPosInSequence;
+ ZSTD_memcpy(cctx->blockState.nextCBlock->rep, updatedRepcodes.rep, sizeof(repcodes_t));
+
+ iend -= bytesAdjustment;
+ if (ip != iend) {
+ /* Store any last literals */
+ U32 lastLLSize = (U32)(iend - ip);
+ assert(ip <= iend);
+ DEBUGLOG(6, "Storing last literals of size: %u", lastLLSize);
+ ZSTD_storeLastLiterals(&cctx->seqStore, ip, lastLLSize);
+ seqPos->posInSrc += lastLLSize;
+ }
+
+ return bytesAdjustment;
+}
+
+typedef size_t (*ZSTD_sequenceCopier) (ZSTD_CCtx* cctx, ZSTD_sequencePosition* seqPos,
+ const ZSTD_Sequence* const inSeqs, size_t inSeqsSize,
+ const void* src, size_t blockSize);
+static ZSTD_sequenceCopier ZSTD_selectSequenceCopier(ZSTD_sequenceFormat_e mode)
+{
+ ZSTD_sequenceCopier sequenceCopier = NULL;
+ assert(ZSTD_cParam_withinBounds(ZSTD_c_blockDelimiters, mode));
+ if (mode == ZSTD_sf_explicitBlockDelimiters) {
+ return ZSTD_copySequencesToSeqStoreExplicitBlockDelim;
+ } else if (mode == ZSTD_sf_noBlockDelimiters) {
+ return ZSTD_copySequencesToSeqStoreNoBlockDelim;
+ }
+ assert(sequenceCopier != NULL);
+ return sequenceCopier;
+}
+
+/* Compress, block-by-block, all of the sequences given.
+ *
+ * Returns the cumulative size of all compressed blocks (including their headers),
+ * otherwise a ZSTD error.
+ */
+static size_t
+ZSTD_compressSequences_internal(ZSTD_CCtx* cctx,
+ void* dst, size_t dstCapacity,
+ const ZSTD_Sequence* inSeqs, size_t inSeqsSize,
+ const void* src, size_t srcSize)
+{
+ size_t cSize = 0;
+ U32 lastBlock;
+ size_t blockSize;
+ size_t compressedSeqsSize;
+ size_t remaining = srcSize;
+ ZSTD_sequencePosition seqPos = {0, 0, 0};
+
+ BYTE const* ip = (BYTE const*)src;
+ BYTE* op = (BYTE*)dst;
+ ZSTD_sequenceCopier const sequenceCopier = ZSTD_selectSequenceCopier(cctx->appliedParams.blockDelimiters);
+
+ DEBUGLOG(4, "ZSTD_compressSequences_internal srcSize: %zu, inSeqsSize: %zu", srcSize, inSeqsSize);
+ /* Special case: empty frame */
+ if (remaining == 0) {
+ U32 const cBlockHeader24 = 1 /* last block */ + (((U32)bt_raw)<<1);
+ RETURN_ERROR_IF(dstCapacity<4, dstSize_tooSmall, "No room for empty frame block header");
+ MEM_writeLE32(op, cBlockHeader24);
+ op += ZSTD_blockHeaderSize;
+ dstCapacity -= ZSTD_blockHeaderSize;
+ cSize += ZSTD_blockHeaderSize;
+ }
+
+ while (remaining) {
+ size_t cBlockSize;
+ size_t additionalByteAdjustment;
+ lastBlock = remaining <= cctx->blockSize;
+ blockSize = lastBlock ? (U32)remaining : (U32)cctx->blockSize;
+ ZSTD_resetSeqStore(&cctx->seqStore);
+ DEBUGLOG(4, "Working on new block. Blocksize: %zu", blockSize);
+
+ additionalByteAdjustment = sequenceCopier(cctx, &seqPos, inSeqs, inSeqsSize, ip, blockSize);
+ FORWARD_IF_ERROR(additionalByteAdjustment, "Bad sequence copy");
+ blockSize -= additionalByteAdjustment;
+
+ /* If blocks are too small, emit as a nocompress block */
+ if (blockSize < MIN_CBLOCK_SIZE+ZSTD_blockHeaderSize+1) {
+ cBlockSize = ZSTD_noCompressBlock(op, dstCapacity, ip, blockSize, lastBlock);
+ FORWARD_IF_ERROR(cBlockSize, "Nocompress block failed");
+ DEBUGLOG(4, "Block too small, writing out nocompress block: cSize: %zu", cBlockSize);
+ cSize += cBlockSize;
+ ip += blockSize;
+ op += cBlockSize;
+ remaining -= blockSize;
+ dstCapacity -= cBlockSize;
+ continue;
+ }
+
+ compressedSeqsSize = ZSTD_entropyCompressSeqStore(&cctx->seqStore,
+ &cctx->blockState.prevCBlock->entropy, &cctx->blockState.nextCBlock->entropy,
+ &cctx->appliedParams,
+ op + ZSTD_blockHeaderSize /* Leave space for block header */, dstCapacity - ZSTD_blockHeaderSize,
+ blockSize,
+ cctx->entropyWorkspace, ENTROPY_WORKSPACE_SIZE /* statically allocated in resetCCtx */,
+ cctx->bmi2);
+ FORWARD_IF_ERROR(compressedSeqsSize, "Compressing sequences of block failed");
+ DEBUGLOG(4, "Compressed sequences size: %zu", compressedSeqsSize);
+
+ if (!cctx->isFirstBlock &&
+ ZSTD_maybeRLE(&cctx->seqStore) &&
+ ZSTD_isRLE((BYTE const*)src, srcSize)) {
+ /* We don't want to emit our first block as a RLE even if it qualifies because
+ * doing so will cause the decoder (cli only) to throw a "should consume all input error."
+ * This is only an issue for zstd <= v1.4.3
+ */
+ compressedSeqsSize = 1;
+ }
+
+ if (compressedSeqsSize == 0) {
+ /* ZSTD_noCompressBlock writes the block header as well */
+ cBlockSize = ZSTD_noCompressBlock(op, dstCapacity, ip, blockSize, lastBlock);
+ FORWARD_IF_ERROR(cBlockSize, "Nocompress block failed");
+ DEBUGLOG(4, "Writing out nocompress block, size: %zu", cBlockSize);
+ } else if (compressedSeqsSize == 1) {
+ cBlockSize = ZSTD_rleCompressBlock(op, dstCapacity, *ip, blockSize, lastBlock);
+ FORWARD_IF_ERROR(cBlockSize, "RLE compress block failed");
+ DEBUGLOG(4, "Writing out RLE block, size: %zu", cBlockSize);
+ } else {
+ U32 cBlockHeader;
+ /* Error checking and repcodes update */
+ ZSTD_blockState_confirmRepcodesAndEntropyTables(&cctx->blockState);
+ if (cctx->blockState.prevCBlock->entropy.fse.offcode_repeatMode == FSE_repeat_valid)
+ cctx->blockState.prevCBlock->entropy.fse.offcode_repeatMode = FSE_repeat_check;
+
+ /* Write block header into beginning of block*/
+ cBlockHeader = lastBlock + (((U32)bt_compressed)<<1) + (U32)(compressedSeqsSize << 3);
+ MEM_writeLE24(op, cBlockHeader);
+ cBlockSize = ZSTD_blockHeaderSize + compressedSeqsSize;
+ DEBUGLOG(4, "Writing out compressed block, size: %zu", cBlockSize);
+ }
+
+ cSize += cBlockSize;
+ DEBUGLOG(4, "cSize running total: %zu", cSize);
+
+ if (lastBlock) {
+ break;
+ } else {
+ ip += blockSize;
+ op += cBlockSize;
+ remaining -= blockSize;
+ dstCapacity -= cBlockSize;
+ cctx->isFirstBlock = 0;
+ }
+ }
+
+ return cSize;
+}
+
+size_t ZSTD_compressSequences(ZSTD_CCtx* const cctx, void* dst, size_t dstCapacity,
+ const ZSTD_Sequence* inSeqs, size_t inSeqsSize,
+ const void* src, size_t srcSize)
+{
+ BYTE* op = (BYTE*)dst;
+ size_t cSize = 0;
+ size_t compressedBlocksSize = 0;
+ size_t frameHeaderSize = 0;
+
+ /* Transparent initialization stage, same as compressStream2() */
+ DEBUGLOG(3, "ZSTD_compressSequences()");
+ assert(cctx != NULL);
+ FORWARD_IF_ERROR(ZSTD_CCtx_init_compressStream2(cctx, ZSTD_e_end, srcSize), "CCtx initialization failed");
+ /* Begin writing output, starting with frame header */
+ frameHeaderSize = ZSTD_writeFrameHeader(op, dstCapacity, &cctx->appliedParams, srcSize, cctx->dictID);
+ op += frameHeaderSize;
+ dstCapacity -= frameHeaderSize;
+ cSize += frameHeaderSize;
+ if (cctx->appliedParams.fParams.checksumFlag && srcSize) {
+ xxh64_update(&cctx->xxhState, src, srcSize);
+ }
+ /* cSize includes block header size and compressed sequences size */
+ compressedBlocksSize = ZSTD_compressSequences_internal(cctx,
+ op, dstCapacity,
+ inSeqs, inSeqsSize,
+ src, srcSize);
+ FORWARD_IF_ERROR(compressedBlocksSize, "Compressing blocks failed!");
+ cSize += compressedBlocksSize;
+ dstCapacity -= compressedBlocksSize;
+
+ if (cctx->appliedParams.fParams.checksumFlag) {
+ U32 const checksum = (U32) xxh64_digest(&cctx->xxhState);
+ RETURN_ERROR_IF(dstCapacity<4, dstSize_tooSmall, "no room for checksum");
+ DEBUGLOG(4, "Write checksum : %08X", (unsigned)checksum);
+ MEM_writeLE32((char*)dst + cSize, checksum);
+ cSize += 4;
+ }
+
+ DEBUGLOG(3, "Final compressed size: %zu", cSize);
+ return cSize;
+}
+
+/*====== Finalize ======*/
+
+/*! ZSTD_flushStream() :
+ * @return : amount of data remaining to flush */
+size_t ZSTD_flushStream(ZSTD_CStream* zcs, ZSTD_outBuffer* output)
+{
+ ZSTD_inBuffer input = { NULL, 0, 0 };
+ return ZSTD_compressStream2(zcs, output, &input, ZSTD_e_flush);
+}
+
+
+size_t ZSTD_endStream(ZSTD_CStream* zcs, ZSTD_outBuffer* output)
+{
+ ZSTD_inBuffer input = { NULL, 0, 0 };
+ size_t const remainingToFlush = ZSTD_compressStream2(zcs, output, &input, ZSTD_e_end);
+ FORWARD_IF_ERROR( remainingToFlush , "ZSTD_compressStream2 failed");
+ if (zcs->appliedParams.nbWorkers > 0) return remainingToFlush; /* minimal estimation */
+ /* single thread mode : attempt to calculate remaining to flush more precisely */
+ { size_t const lastBlockSize = zcs->frameEnded ? 0 : ZSTD_BLOCKHEADERSIZE;
+ size_t const checksumSize = (size_t)(zcs->frameEnded ? 0 : zcs->appliedParams.fParams.checksumFlag * 4);
+ size_t const toFlush = remainingToFlush + lastBlockSize + checksumSize;
+ DEBUGLOG(4, "ZSTD_endStream : remaining to flush : %u", (unsigned)toFlush);
+ return toFlush;
+ }
+}
+
+
+/*-===== Pre-defined compression levels =====-*/
+#include "clevels.h"
+
+int ZSTD_maxCLevel(void) { return ZSTD_MAX_CLEVEL; }
+int ZSTD_minCLevel(void) { return (int)-ZSTD_TARGETLENGTH_MAX; }
+int ZSTD_defaultCLevel(void) { return ZSTD_CLEVEL_DEFAULT; }
+
+static ZSTD_compressionParameters ZSTD_dedicatedDictSearch_getCParams(int const compressionLevel, size_t const dictSize)
+{
+ ZSTD_compressionParameters cParams = ZSTD_getCParams_internal(compressionLevel, 0, dictSize, ZSTD_cpm_createCDict);
+ switch (cParams.strategy) {
+ case ZSTD_fast:
+ case ZSTD_dfast:
+ break;
+ case ZSTD_greedy:
+ case ZSTD_lazy:
+ case ZSTD_lazy2:
+ cParams.hashLog += ZSTD_LAZY_DDSS_BUCKET_LOG;
+ break;
+ case ZSTD_btlazy2:
+ case ZSTD_btopt:
+ case ZSTD_btultra:
+ case ZSTD_btultra2:
+ break;
+ }
+ return cParams;
+}
+
+static int ZSTD_dedicatedDictSearch_isSupported(
+ ZSTD_compressionParameters const* cParams)
+{
+ return (cParams->strategy >= ZSTD_greedy)
+ && (cParams->strategy <= ZSTD_lazy2)
+ && (cParams->hashLog > cParams->chainLog)
+ && (cParams->chainLog <= 24);
+}
+
+/*
+ * Reverses the adjustment applied to cparams when enabling dedicated dict
+ * search. This is used to recover the params set to be used in the working
+ * context. (Otherwise, those tables would also grow.)
+ */
+static void ZSTD_dedicatedDictSearch_revertCParams(
+ ZSTD_compressionParameters* cParams) {
+ switch (cParams->strategy) {
+ case ZSTD_fast:
+ case ZSTD_dfast:
+ break;
+ case ZSTD_greedy:
+ case ZSTD_lazy:
+ case ZSTD_lazy2:
+ cParams->hashLog -= ZSTD_LAZY_DDSS_BUCKET_LOG;
+ if (cParams->hashLog < ZSTD_HASHLOG_MIN) {
+ cParams->hashLog = ZSTD_HASHLOG_MIN;
+ }
+ break;
+ case ZSTD_btlazy2:
+ case ZSTD_btopt:
+ case ZSTD_btultra:
+ case ZSTD_btultra2:
+ break;
+ }
+}
+
+static U64 ZSTD_getCParamRowSize(U64 srcSizeHint, size_t dictSize, ZSTD_cParamMode_e mode)
+{
+ switch (mode) {
+ case ZSTD_cpm_unknown:
+ case ZSTD_cpm_noAttachDict:
+ case ZSTD_cpm_createCDict:
+ break;
+ case ZSTD_cpm_attachDict:
+ dictSize = 0;
+ break;
+ default:
+ assert(0);
+ break;
+ }
+ { int const unknown = srcSizeHint == ZSTD_CONTENTSIZE_UNKNOWN;
+ size_t const addedSize = unknown && dictSize > 0 ? 500 : 0;
+ return unknown && dictSize == 0 ? ZSTD_CONTENTSIZE_UNKNOWN : srcSizeHint+dictSize+addedSize;
+ }
+}
+
+/*! ZSTD_getCParams_internal() :
+ * @return ZSTD_compressionParameters structure for a selected compression level, srcSize and dictSize.
+ * Note: srcSizeHint 0 means 0, use ZSTD_CONTENTSIZE_UNKNOWN for unknown.
+ * Use dictSize == 0 for unknown or unused.
+ * Note: `mode` controls how we treat the `dictSize`. See docs for `ZSTD_cParamMode_e`. */
+static ZSTD_compressionParameters ZSTD_getCParams_internal(int compressionLevel, unsigned long long srcSizeHint, size_t dictSize, ZSTD_cParamMode_e mode)
+{
+ U64 const rSize = ZSTD_getCParamRowSize(srcSizeHint, dictSize, mode);
+ U32 const tableID = (rSize <= 256 KB) + (rSize <= 128 KB) + (rSize <= 16 KB);
+ int row;
+ DEBUGLOG(5, "ZSTD_getCParams_internal (cLevel=%i)", compressionLevel);
+
+ /* row */
+ if (compressionLevel == 0) row = ZSTD_CLEVEL_DEFAULT; /* 0 == default */
+ else if (compressionLevel < 0) row = 0; /* entry 0 is baseline for fast mode */
+ else if (compressionLevel > ZSTD_MAX_CLEVEL) row = ZSTD_MAX_CLEVEL;
+ else row = compressionLevel;
+
+ { ZSTD_compressionParameters cp = ZSTD_defaultCParameters[tableID][row];
+ DEBUGLOG(5, "ZSTD_getCParams_internal selected tableID: %u row: %u strat: %u", tableID, row, (U32)cp.strategy);
+ /* acceleration factor */
+ if (compressionLevel < 0) {
+ int const clampedCompressionLevel = MAX(ZSTD_minCLevel(), compressionLevel);
+ cp.targetLength = (unsigned)(-clampedCompressionLevel);
+ }
+ /* refine parameters based on srcSize & dictSize */
+ return ZSTD_adjustCParams_internal(cp, srcSizeHint, dictSize, mode);
+ }
+}
+
+/*! ZSTD_getCParams() :
+ * @return ZSTD_compressionParameters structure for a selected compression level, srcSize and dictSize.
+ * Size values are optional, provide 0 if not known or unused */
+ZSTD_compressionParameters ZSTD_getCParams(int compressionLevel, unsigned long long srcSizeHint, size_t dictSize)
+{
+ if (srcSizeHint == 0) srcSizeHint = ZSTD_CONTENTSIZE_UNKNOWN;
+ return ZSTD_getCParams_internal(compressionLevel, srcSizeHint, dictSize, ZSTD_cpm_unknown);
+}
+
+/*! ZSTD_getParams() :
+ * same idea as ZSTD_getCParams()
+ * @return a `ZSTD_parameters` structure (instead of `ZSTD_compressionParameters`).
+ * Fields of `ZSTD_frameParameters` are set to default values */
+static ZSTD_parameters ZSTD_getParams_internal(int compressionLevel, unsigned long long srcSizeHint, size_t dictSize, ZSTD_cParamMode_e mode) {
+ ZSTD_parameters params;
+ ZSTD_compressionParameters const cParams = ZSTD_getCParams_internal(compressionLevel, srcSizeHint, dictSize, mode);
+ DEBUGLOG(5, "ZSTD_getParams (cLevel=%i)", compressionLevel);
+ ZSTD_memset(&params, 0, sizeof(params));
+ params.cParams = cParams;
+ params.fParams.contentSizeFlag = 1;
+ return params;
+}
+
+/*! ZSTD_getParams() :
+ * same idea as ZSTD_getCParams()
+ * @return a `ZSTD_parameters` structure (instead of `ZSTD_compressionParameters`).
+ * Fields of `ZSTD_frameParameters` are set to default values */
+ZSTD_parameters ZSTD_getParams(int compressionLevel, unsigned long long srcSizeHint, size_t dictSize) {
+ if (srcSizeHint == 0) srcSizeHint = ZSTD_CONTENTSIZE_UNKNOWN;
+ return ZSTD_getParams_internal(compressionLevel, srcSizeHint, dictSize, ZSTD_cpm_unknown);
+}
diff --git a/lib/zstd/compress/zstd_compress_internal.h b/lib/zstd/compress/zstd_compress_internal.h
new file mode 100644
index 0000000000..71697a11ae
--- /dev/null
+++ b/lib/zstd/compress/zstd_compress_internal.h
@@ -0,0 +1,1399 @@
+/*
+ * Copyright (c) Yann Collet, Facebook, Inc.
+ * All rights reserved.
+ *
+ * This source code is licensed under both the BSD-style license (found in the
+ * LICENSE file in the root directory of this source tree) and the GPLv2 (found
+ * in the COPYING file in the root directory of this source tree).
+ * You may select, at your option, one of the above-listed licenses.
+ */
+
+/* This header contains definitions
+ * that shall **only** be used by modules within lib/compress.
+ */
+
+#ifndef ZSTD_COMPRESS_H
+#define ZSTD_COMPRESS_H
+
+/*-*************************************
+* Dependencies
+***************************************/
+#include "../common/zstd_internal.h"
+#include "zstd_cwksp.h"
+
+
+/*-*************************************
+* Constants
+***************************************/
+#define kSearchStrength 8
+#define HASH_READ_SIZE 8
+#define ZSTD_DUBT_UNSORTED_MARK 1 /* For btlazy2 strategy, index ZSTD_DUBT_UNSORTED_MARK==1 means "unsorted".
+ It could be confused for a real successor at index "1", if sorted as larger than its predecessor.
+ It's not a big deal though : candidate will just be sorted again.
+ Additionally, candidate position 1 will be lost.
+ But candidate 1 cannot hide a large tree of candidates, so it's a minimal loss.
+ The benefit is that ZSTD_DUBT_UNSORTED_MARK cannot be mishandled after table re-use with a different strategy.
+ This constant is required by ZSTD_compressBlock_btlazy2() and ZSTD_reduceTable_internal() */
+
+
+/*-*************************************
+* Context memory management
+***************************************/
+typedef enum { ZSTDcs_created=0, ZSTDcs_init, ZSTDcs_ongoing, ZSTDcs_ending } ZSTD_compressionStage_e;
+typedef enum { zcss_init=0, zcss_load, zcss_flush } ZSTD_cStreamStage;
+
+typedef struct ZSTD_prefixDict_s {
+ const void* dict;
+ size_t dictSize;
+ ZSTD_dictContentType_e dictContentType;
+} ZSTD_prefixDict;
+
+typedef struct {
+ void* dictBuffer;
+ void const* dict;
+ size_t dictSize;
+ ZSTD_dictContentType_e dictContentType;
+ ZSTD_CDict* cdict;
+} ZSTD_localDict;
+
+typedef struct {
+ HUF_CElt CTable[HUF_CTABLE_SIZE_ST(255)];
+ HUF_repeat repeatMode;
+} ZSTD_hufCTables_t;
+
+typedef struct {
+ FSE_CTable offcodeCTable[FSE_CTABLE_SIZE_U32(OffFSELog, MaxOff)];
+ FSE_CTable matchlengthCTable[FSE_CTABLE_SIZE_U32(MLFSELog, MaxML)];
+ FSE_CTable litlengthCTable[FSE_CTABLE_SIZE_U32(LLFSELog, MaxLL)];
+ FSE_repeat offcode_repeatMode;
+ FSE_repeat matchlength_repeatMode;
+ FSE_repeat litlength_repeatMode;
+} ZSTD_fseCTables_t;
+
+typedef struct {
+ ZSTD_hufCTables_t huf;
+ ZSTD_fseCTables_t fse;
+} ZSTD_entropyCTables_t;
+
+/* *********************************************
+* Entropy buffer statistics structs and funcs *
+***********************************************/
+/* ZSTD_hufCTablesMetadata_t :
+ * Stores Literals Block Type for a super-block in hType, and
+ * huffman tree description in hufDesBuffer.
+ * hufDesSize refers to the size of huffman tree description in bytes.
+ * This metadata is populated in ZSTD_buildBlockEntropyStats_literals() */
+typedef struct {
+ symbolEncodingType_e hType;
+ BYTE hufDesBuffer[ZSTD_MAX_HUF_HEADER_SIZE];
+ size_t hufDesSize;
+} ZSTD_hufCTablesMetadata_t;
+
+/* ZSTD_fseCTablesMetadata_t :
+ * Stores symbol compression modes for a super-block in {ll, ol, ml}Type, and
+ * fse tables in fseTablesBuffer.
+ * fseTablesSize refers to the size of fse tables in bytes.
+ * This metadata is populated in ZSTD_buildBlockEntropyStats_sequences() */
+typedef struct {
+ symbolEncodingType_e llType;
+ symbolEncodingType_e ofType;
+ symbolEncodingType_e mlType;
+ BYTE fseTablesBuffer[ZSTD_MAX_FSE_HEADERS_SIZE];
+ size_t fseTablesSize;
+ size_t lastCountSize; /* This is to account for bug in 1.3.4. More detail in ZSTD_entropyCompressSeqStore_internal() */
+} ZSTD_fseCTablesMetadata_t;
+
+typedef struct {
+ ZSTD_hufCTablesMetadata_t hufMetadata;
+ ZSTD_fseCTablesMetadata_t fseMetadata;
+} ZSTD_entropyCTablesMetadata_t;
+
+/* ZSTD_buildBlockEntropyStats() :
+ * Builds entropy for the block.
+ * @return : 0 on success or error code */
+size_t ZSTD_buildBlockEntropyStats(seqStore_t* seqStorePtr,
+ const ZSTD_entropyCTables_t* prevEntropy,
+ ZSTD_entropyCTables_t* nextEntropy,
+ const ZSTD_CCtx_params* cctxParams,
+ ZSTD_entropyCTablesMetadata_t* entropyMetadata,
+ void* workspace, size_t wkspSize);
+
+/* *******************************
+* Compression internals structs *
+*********************************/
+
+typedef struct {
+ U32 off; /* Offset sumtype code for the match, using ZSTD_storeSeq() format */
+ U32 len; /* Raw length of match */
+} ZSTD_match_t;
+
+typedef struct {
+ U32 offset; /* Offset of sequence */
+ U32 litLength; /* Length of literals prior to match */
+ U32 matchLength; /* Raw length of match */
+} rawSeq;
+
+typedef struct {
+ rawSeq* seq; /* The start of the sequences */
+ size_t pos; /* The index in seq where reading stopped. pos <= size. */
+ size_t posInSequence; /* The position within the sequence at seq[pos] where reading
+ stopped. posInSequence <= seq[pos].litLength + seq[pos].matchLength */
+ size_t size; /* The number of sequences. <= capacity. */
+ size_t capacity; /* The capacity starting from `seq` pointer */
+} rawSeqStore_t;
+
+UNUSED_ATTR static const rawSeqStore_t kNullRawSeqStore = {NULL, 0, 0, 0, 0};
+
+typedef struct {
+ int price;
+ U32 off;
+ U32 mlen;
+ U32 litlen;
+ U32 rep[ZSTD_REP_NUM];
+} ZSTD_optimal_t;
+
+typedef enum { zop_dynamic=0, zop_predef } ZSTD_OptPrice_e;
+
+typedef struct {
+ /* All tables are allocated inside cctx->workspace by ZSTD_resetCCtx_internal() */
+ unsigned* litFreq; /* table of literals statistics, of size 256 */
+ unsigned* litLengthFreq; /* table of litLength statistics, of size (MaxLL+1) */
+ unsigned* matchLengthFreq; /* table of matchLength statistics, of size (MaxML+1) */
+ unsigned* offCodeFreq; /* table of offCode statistics, of size (MaxOff+1) */
+ ZSTD_match_t* matchTable; /* list of found matches, of size ZSTD_OPT_NUM+1 */
+ ZSTD_optimal_t* priceTable; /* All positions tracked by optimal parser, of size ZSTD_OPT_NUM+1 */
+
+ U32 litSum; /* nb of literals */
+ U32 litLengthSum; /* nb of litLength codes */
+ U32 matchLengthSum; /* nb of matchLength codes */
+ U32 offCodeSum; /* nb of offset codes */
+ U32 litSumBasePrice; /* to compare to log2(litfreq) */
+ U32 litLengthSumBasePrice; /* to compare to log2(llfreq) */
+ U32 matchLengthSumBasePrice;/* to compare to log2(mlfreq) */
+ U32 offCodeSumBasePrice; /* to compare to log2(offreq) */
+ ZSTD_OptPrice_e priceType; /* prices can be determined dynamically, or follow a pre-defined cost structure */
+ const ZSTD_entropyCTables_t* symbolCosts; /* pre-calculated dictionary statistics */
+ ZSTD_paramSwitch_e literalCompressionMode;
+} optState_t;
+
+typedef struct {
+ ZSTD_entropyCTables_t entropy;
+ U32 rep[ZSTD_REP_NUM];
+} ZSTD_compressedBlockState_t;
+
+typedef struct {
+ BYTE const* nextSrc; /* next block here to continue on current prefix */
+ BYTE const* base; /* All regular indexes relative to this position */
+ BYTE const* dictBase; /* extDict indexes relative to this position */
+ U32 dictLimit; /* below that point, need extDict */
+ U32 lowLimit; /* below that point, no more valid data */
+ U32 nbOverflowCorrections; /* Number of times overflow correction has run since
+ * ZSTD_window_init(). Useful for debugging coredumps
+ * and for ZSTD_WINDOW_OVERFLOW_CORRECT_FREQUENTLY.
+ */
+} ZSTD_window_t;
+
+#define ZSTD_WINDOW_START_INDEX 2
+
+typedef struct ZSTD_matchState_t ZSTD_matchState_t;
+
+#define ZSTD_ROW_HASH_CACHE_SIZE 8 /* Size of prefetching hash cache for row-based matchfinder */
+
+struct ZSTD_matchState_t {
+ ZSTD_window_t window; /* State for window round buffer management */
+ U32 loadedDictEnd; /* index of end of dictionary, within context's referential.
+ * When loadedDictEnd != 0, a dictionary is in use, and still valid.
+ * This relies on a mechanism to set loadedDictEnd=0 when dictionary is no longer within distance.
+ * Such mechanism is provided within ZSTD_window_enforceMaxDist() and ZSTD_checkDictValidity().
+ * When dict referential is copied into active context (i.e. not attached),
+ * loadedDictEnd == dictSize, since referential starts from zero.
+ */
+ U32 nextToUpdate; /* index from which to continue table update */
+ U32 hashLog3; /* dispatch table for matches of len==3 : larger == faster, more memory */
+
+ U32 rowHashLog; /* For row-based matchfinder: Hashlog based on nb of rows in the hashTable.*/
+ U16* tagTable; /* For row-based matchFinder: A row-based table containing the hashes and head index. */
+ U32 hashCache[ZSTD_ROW_HASH_CACHE_SIZE]; /* For row-based matchFinder: a cache of hashes to improve speed */
+
+ U32* hashTable;
+ U32* hashTable3;
+ U32* chainTable;
+
+ U32 forceNonContiguous; /* Non-zero if we should force non-contiguous load for the next window update. */
+
+ int dedicatedDictSearch; /* Indicates whether this matchState is using the
+ * dedicated dictionary search structure.
+ */
+ optState_t opt; /* optimal parser state */
+ const ZSTD_matchState_t* dictMatchState;
+ ZSTD_compressionParameters cParams;
+ const rawSeqStore_t* ldmSeqStore;
+};
+
+typedef struct {
+ ZSTD_compressedBlockState_t* prevCBlock;
+ ZSTD_compressedBlockState_t* nextCBlock;
+ ZSTD_matchState_t matchState;
+} ZSTD_blockState_t;
+
+typedef struct {
+ U32 offset;
+ U32 checksum;
+} ldmEntry_t;
+
+typedef struct {
+ BYTE const* split;
+ U32 hash;
+ U32 checksum;
+ ldmEntry_t* bucket;
+} ldmMatchCandidate_t;
+
+#define LDM_BATCH_SIZE 64
+
+typedef struct {
+ ZSTD_window_t window; /* State for the window round buffer management */
+ ldmEntry_t* hashTable;
+ U32 loadedDictEnd;
+ BYTE* bucketOffsets; /* Next position in bucket to insert entry */
+ size_t splitIndices[LDM_BATCH_SIZE];
+ ldmMatchCandidate_t matchCandidates[LDM_BATCH_SIZE];
+} ldmState_t;
+
+typedef struct {
+ ZSTD_paramSwitch_e enableLdm; /* ZSTD_ps_enable to enable LDM. ZSTD_ps_auto by default */
+ U32 hashLog; /* Log size of hashTable */
+ U32 bucketSizeLog; /* Log bucket size for collision resolution, at most 8 */
+ U32 minMatchLength; /* Minimum match length */
+ U32 hashRateLog; /* Log number of entries to skip */
+ U32 windowLog; /* Window log for the LDM */
+} ldmParams_t;
+
+typedef struct {
+ int collectSequences;
+ ZSTD_Sequence* seqStart;
+ size_t seqIndex;
+ size_t maxSequences;
+} SeqCollector;
+
+struct ZSTD_CCtx_params_s {
+ ZSTD_format_e format;
+ ZSTD_compressionParameters cParams;
+ ZSTD_frameParameters fParams;
+
+ int compressionLevel;
+ int forceWindow; /* force back-references to respect limit of
+ * 1<<wLog, even for dictionary */
+ size_t targetCBlockSize; /* Tries to fit compressed block size to be around targetCBlockSize.
+ * No target when targetCBlockSize == 0.
+ * There is no guarantee on compressed block size */
+ int srcSizeHint; /* User's best guess of source size.
+ * Hint is not valid when srcSizeHint == 0.
+ * There is no guarantee that hint is close to actual source size */
+
+ ZSTD_dictAttachPref_e attachDictPref;
+ ZSTD_paramSwitch_e literalCompressionMode;
+
+ /* Multithreading: used to pass parameters to mtctx */
+ int nbWorkers;
+ size_t jobSize;
+ int overlapLog;
+ int rsyncable;
+
+ /* Long distance matching parameters */
+ ldmParams_t ldmParams;
+
+ /* Dedicated dict search algorithm trigger */
+ int enableDedicatedDictSearch;
+
+ /* Input/output buffer modes */
+ ZSTD_bufferMode_e inBufferMode;
+ ZSTD_bufferMode_e outBufferMode;
+
+ /* Sequence compression API */
+ ZSTD_sequenceFormat_e blockDelimiters;
+ int validateSequences;
+
+ /* Block splitting */
+ ZSTD_paramSwitch_e useBlockSplitter;
+
+ /* Param for deciding whether to use row-based matchfinder */
+ ZSTD_paramSwitch_e useRowMatchFinder;
+
+ /* Always load a dictionary in ext-dict mode (not prefix mode)? */
+ int deterministicRefPrefix;
+
+ /* Internal use, for createCCtxParams() and freeCCtxParams() only */
+ ZSTD_customMem customMem;
+}; /* typedef'd to ZSTD_CCtx_params within "zstd.h" */
+
+#define COMPRESS_SEQUENCES_WORKSPACE_SIZE (sizeof(unsigned) * (MaxSeq + 2))
+#define ENTROPY_WORKSPACE_SIZE (HUF_WORKSPACE_SIZE + COMPRESS_SEQUENCES_WORKSPACE_SIZE)
+
+/*
+ * Indicates whether this compression proceeds directly from user-provided
+ * source buffer to user-provided destination buffer (ZSTDb_not_buffered), or
+ * whether the context needs to buffer the input/output (ZSTDb_buffered).
+ */
+typedef enum {
+ ZSTDb_not_buffered,
+ ZSTDb_buffered
+} ZSTD_buffered_policy_e;
+
+/*
+ * Struct that contains all elements of block splitter that should be allocated
+ * in a wksp.
+ */
+#define ZSTD_MAX_NB_BLOCK_SPLITS 196
+typedef struct {
+ seqStore_t fullSeqStoreChunk;
+ seqStore_t firstHalfSeqStore;
+ seqStore_t secondHalfSeqStore;
+ seqStore_t currSeqStore;
+ seqStore_t nextSeqStore;
+
+ U32 partitions[ZSTD_MAX_NB_BLOCK_SPLITS];
+ ZSTD_entropyCTablesMetadata_t entropyMetadata;
+} ZSTD_blockSplitCtx;
+
+struct ZSTD_CCtx_s {
+ ZSTD_compressionStage_e stage;
+ int cParamsChanged; /* == 1 if cParams(except wlog) or compression level are changed in requestedParams. Triggers transmission of new params to ZSTDMT (if available) then reset to 0. */
+ int bmi2; /* == 1 if the CPU supports BMI2 and 0 otherwise. CPU support is determined dynamically once per context lifetime. */
+ ZSTD_CCtx_params requestedParams;
+ ZSTD_CCtx_params appliedParams;
+ ZSTD_CCtx_params simpleApiParams; /* Param storage used by the simple API - not sticky. Must only be used in top-level simple API functions for storage. */
+ U32 dictID;
+ size_t dictContentSize;
+
+ ZSTD_cwksp workspace; /* manages buffer for dynamic allocations */
+ size_t blockSize;
+ unsigned long long pledgedSrcSizePlusOne; /* this way, 0 (default) == unknown */
+ unsigned long long consumedSrcSize;
+ unsigned long long producedCSize;
+ struct xxh64_state xxhState;
+ ZSTD_customMem customMem;
+ ZSTD_threadPool* pool;
+ size_t staticSize;
+ SeqCollector seqCollector;
+ int isFirstBlock;
+ int initialized;
+
+ seqStore_t seqStore; /* sequences storage ptrs */
+ ldmState_t ldmState; /* long distance matching state */
+ rawSeq* ldmSequences; /* Storage for the ldm output sequences */
+ size_t maxNbLdmSequences;
+ rawSeqStore_t externSeqStore; /* Mutable reference to external sequences */
+ ZSTD_blockState_t blockState;
+ U32* entropyWorkspace; /* entropy workspace of ENTROPY_WORKSPACE_SIZE bytes */
+
+ /* Whether we are streaming or not */
+ ZSTD_buffered_policy_e bufferedPolicy;
+
+ /* streaming */
+ char* inBuff;
+ size_t inBuffSize;
+ size_t inToCompress;
+ size_t inBuffPos;
+ size_t inBuffTarget;
+ char* outBuff;
+ size_t outBuffSize;
+ size_t outBuffContentSize;
+ size_t outBuffFlushedSize;
+ ZSTD_cStreamStage streamStage;
+ U32 frameEnded;
+
+ /* Stable in/out buffer verification */
+ ZSTD_inBuffer expectedInBuffer;
+ size_t expectedOutBufferSize;
+
+ /* Dictionary */
+ ZSTD_localDict localDict;
+ const ZSTD_CDict* cdict;
+ ZSTD_prefixDict prefixDict; /* single-usage dictionary */
+
+ /* Multi-threading */
+
+ /* Tracing */
+
+ /* Workspace for block splitter */
+ ZSTD_blockSplitCtx blockSplitCtx;
+};
+
+typedef enum { ZSTD_dtlm_fast, ZSTD_dtlm_full } ZSTD_dictTableLoadMethod_e;
+
+typedef enum {
+ ZSTD_noDict = 0,
+ ZSTD_extDict = 1,
+ ZSTD_dictMatchState = 2,
+ ZSTD_dedicatedDictSearch = 3
+} ZSTD_dictMode_e;
+
+typedef enum {
+ ZSTD_cpm_noAttachDict = 0, /* Compression with ZSTD_noDict or ZSTD_extDict.
+ * In this mode we use both the srcSize and the dictSize
+ * when selecting and adjusting parameters.
+ */
+ ZSTD_cpm_attachDict = 1, /* Compression with ZSTD_dictMatchState or ZSTD_dedicatedDictSearch.
+ * In this mode we only take the srcSize into account when selecting
+ * and adjusting parameters.
+ */
+ ZSTD_cpm_createCDict = 2, /* Creating a CDict.
+ * In this mode we take both the source size and the dictionary size
+ * into account when selecting and adjusting the parameters.
+ */
+ ZSTD_cpm_unknown = 3, /* ZSTD_getCParams, ZSTD_getParams, ZSTD_adjustParams.
+ * We don't know what these parameters are for. We default to the legacy
+ * behavior of taking both the source size and the dict size into account
+ * when selecting and adjusting parameters.
+ */
+} ZSTD_cParamMode_e;
+
+typedef size_t (*ZSTD_blockCompressor) (
+ ZSTD_matchState_t* bs, seqStore_t* seqStore, U32 rep[ZSTD_REP_NUM],
+ void const* src, size_t srcSize);
+ZSTD_blockCompressor ZSTD_selectBlockCompressor(ZSTD_strategy strat, ZSTD_paramSwitch_e rowMatchfinderMode, ZSTD_dictMode_e dictMode);
+
+
+MEM_STATIC U32 ZSTD_LLcode(U32 litLength)
+{
+ static const BYTE LL_Code[64] = { 0, 1, 2, 3, 4, 5, 6, 7,
+ 8, 9, 10, 11, 12, 13, 14, 15,
+ 16, 16, 17, 17, 18, 18, 19, 19,
+ 20, 20, 20, 20, 21, 21, 21, 21,
+ 22, 22, 22, 22, 22, 22, 22, 22,
+ 23, 23, 23, 23, 23, 23, 23, 23,
+ 24, 24, 24, 24, 24, 24, 24, 24,
+ 24, 24, 24, 24, 24, 24, 24, 24 };
+ static const U32 LL_deltaCode = 19;
+ return (litLength > 63) ? ZSTD_highbit32(litLength) + LL_deltaCode : LL_Code[litLength];
+}
+
+/* ZSTD_MLcode() :
+ * note : mlBase = matchLength - MINMATCH;
+ * because it's the format it's stored in seqStore->sequences */
+MEM_STATIC U32 ZSTD_MLcode(U32 mlBase)
+{
+ static const BYTE ML_Code[128] = { 0, 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15,
+ 16, 17, 18, 19, 20, 21, 22, 23, 24, 25, 26, 27, 28, 29, 30, 31,
+ 32, 32, 33, 33, 34, 34, 35, 35, 36, 36, 36, 36, 37, 37, 37, 37,
+ 38, 38, 38, 38, 38, 38, 38, 38, 39, 39, 39, 39, 39, 39, 39, 39,
+ 40, 40, 40, 40, 40, 40, 40, 40, 40, 40, 40, 40, 40, 40, 40, 40,
+ 41, 41, 41, 41, 41, 41, 41, 41, 41, 41, 41, 41, 41, 41, 41, 41,
+ 42, 42, 42, 42, 42, 42, 42, 42, 42, 42, 42, 42, 42, 42, 42, 42,
+ 42, 42, 42, 42, 42, 42, 42, 42, 42, 42, 42, 42, 42, 42, 42, 42 };
+ static const U32 ML_deltaCode = 36;
+ return (mlBase > 127) ? ZSTD_highbit32(mlBase) + ML_deltaCode : ML_Code[mlBase];
+}
+
+/* ZSTD_cParam_withinBounds:
+ * @return 1 if value is within cParam bounds,
+ * 0 otherwise */
+MEM_STATIC int ZSTD_cParam_withinBounds(ZSTD_cParameter cParam, int value)
+{
+ ZSTD_bounds const bounds = ZSTD_cParam_getBounds(cParam);
+ if (ZSTD_isError(bounds.error)) return 0;
+ if (value < bounds.lowerBound) return 0;
+ if (value > bounds.upperBound) return 0;
+ return 1;
+}
+
+/* ZSTD_noCompressBlock() :
+ * Writes uncompressed block to dst buffer from given src.
+ * Returns the size of the block */
+MEM_STATIC size_t ZSTD_noCompressBlock (void* dst, size_t dstCapacity, const void* src, size_t srcSize, U32 lastBlock)
+{
+ U32 const cBlockHeader24 = lastBlock + (((U32)bt_raw)<<1) + (U32)(srcSize << 3);
+ RETURN_ERROR_IF(srcSize + ZSTD_blockHeaderSize > dstCapacity,
+ dstSize_tooSmall, "dst buf too small for uncompressed block");
+ MEM_writeLE24(dst, cBlockHeader24);
+ ZSTD_memcpy((BYTE*)dst + ZSTD_blockHeaderSize, src, srcSize);
+ return ZSTD_blockHeaderSize + srcSize;
+}
+
+MEM_STATIC size_t ZSTD_rleCompressBlock (void* dst, size_t dstCapacity, BYTE src, size_t srcSize, U32 lastBlock)
+{
+ BYTE* const op = (BYTE*)dst;
+ U32 const cBlockHeader = lastBlock + (((U32)bt_rle)<<1) + (U32)(srcSize << 3);
+ RETURN_ERROR_IF(dstCapacity < 4, dstSize_tooSmall, "");
+ MEM_writeLE24(op, cBlockHeader);
+ op[3] = src;
+ return 4;
+}
+
+
+/* ZSTD_minGain() :
+ * minimum compression required
+ * to generate a compress block or a compressed literals section.
+ * note : use same formula for both situations */
+MEM_STATIC size_t ZSTD_minGain(size_t srcSize, ZSTD_strategy strat)
+{
+ U32 const minlog = (strat>=ZSTD_btultra) ? (U32)(strat) - 1 : 6;
+ ZSTD_STATIC_ASSERT(ZSTD_btultra == 8);
+ assert(ZSTD_cParam_withinBounds(ZSTD_c_strategy, strat));
+ return (srcSize >> minlog) + 2;
+}
+
+MEM_STATIC int ZSTD_literalsCompressionIsDisabled(const ZSTD_CCtx_params* cctxParams)
+{
+ switch (cctxParams->literalCompressionMode) {
+ case ZSTD_ps_enable:
+ return 0;
+ case ZSTD_ps_disable:
+ return 1;
+ default:
+ assert(0 /* impossible: pre-validated */);
+ ZSTD_FALLTHROUGH;
+ case ZSTD_ps_auto:
+ return (cctxParams->cParams.strategy == ZSTD_fast) && (cctxParams->cParams.targetLength > 0);
+ }
+}
+
+/*! ZSTD_safecopyLiterals() :
+ * memcpy() function that won't read beyond more than WILDCOPY_OVERLENGTH bytes past ilimit_w.
+ * Only called when the sequence ends past ilimit_w, so it only needs to be optimized for single
+ * large copies.
+ */
+static void
+ZSTD_safecopyLiterals(BYTE* op, BYTE const* ip, BYTE const* const iend, BYTE const* ilimit_w)
+{
+ assert(iend > ilimit_w);
+ if (ip <= ilimit_w) {
+ ZSTD_wildcopy(op, ip, ilimit_w - ip, ZSTD_no_overlap);
+ op += ilimit_w - ip;
+ ip = ilimit_w;
+ }
+ while (ip < iend) *op++ = *ip++;
+}
+
+#define ZSTD_REP_MOVE (ZSTD_REP_NUM-1)
+#define STORE_REPCODE_1 STORE_REPCODE(1)
+#define STORE_REPCODE_2 STORE_REPCODE(2)
+#define STORE_REPCODE_3 STORE_REPCODE(3)
+#define STORE_REPCODE(r) (assert((r)>=1), assert((r)<=3), (r)-1)
+#define STORE_OFFSET(o) (assert((o)>0), o + ZSTD_REP_MOVE)
+#define STORED_IS_OFFSET(o) ((o) > ZSTD_REP_MOVE)
+#define STORED_IS_REPCODE(o) ((o) <= ZSTD_REP_MOVE)
+#define STORED_OFFSET(o) (assert(STORED_IS_OFFSET(o)), (o)-ZSTD_REP_MOVE)
+#define STORED_REPCODE(o) (assert(STORED_IS_REPCODE(o)), (o)+1) /* returns ID 1,2,3 */
+#define STORED_TO_OFFBASE(o) ((o)+1)
+#define OFFBASE_TO_STORED(o) ((o)-1)
+
+/*! ZSTD_storeSeq() :
+ * Store a sequence (litlen, litPtr, offCode and matchLength) into seqStore_t.
+ * @offBase_minus1 : Users should use employ macros STORE_REPCODE_X and STORE_OFFSET().
+ * @matchLength : must be >= MINMATCH
+ * Allowed to overread literals up to litLimit.
+*/
+HINT_INLINE UNUSED_ATTR void
+ZSTD_storeSeq(seqStore_t* seqStorePtr,
+ size_t litLength, const BYTE* literals, const BYTE* litLimit,
+ U32 offBase_minus1,
+ size_t matchLength)
+{
+ BYTE const* const litLimit_w = litLimit - WILDCOPY_OVERLENGTH;
+ BYTE const* const litEnd = literals + litLength;
+#if defined(DEBUGLEVEL) && (DEBUGLEVEL >= 6)
+ static const BYTE* g_start = NULL;
+ if (g_start==NULL) g_start = (const BYTE*)literals; /* note : index only works for compression within a single segment */
+ { U32 const pos = (U32)((const BYTE*)literals - g_start);
+ DEBUGLOG(6, "Cpos%7u :%3u literals, match%4u bytes at offCode%7u",
+ pos, (U32)litLength, (U32)matchLength, (U32)offBase_minus1);
+ }
+#endif
+ assert((size_t)(seqStorePtr->sequences - seqStorePtr->sequencesStart) < seqStorePtr->maxNbSeq);
+ /* copy Literals */
+ assert(seqStorePtr->maxNbLit <= 128 KB);
+ assert(seqStorePtr->lit + litLength <= seqStorePtr->litStart + seqStorePtr->maxNbLit);
+ assert(literals + litLength <= litLimit);
+ if (litEnd <= litLimit_w) {
+ /* Common case we can use wildcopy.
+ * First copy 16 bytes, because literals are likely short.
+ */
+ assert(WILDCOPY_OVERLENGTH >= 16);
+ ZSTD_copy16(seqStorePtr->lit, literals);
+ if (litLength > 16) {
+ ZSTD_wildcopy(seqStorePtr->lit+16, literals+16, (ptrdiff_t)litLength-16, ZSTD_no_overlap);
+ }
+ } else {
+ ZSTD_safecopyLiterals(seqStorePtr->lit, literals, litEnd, litLimit_w);
+ }
+ seqStorePtr->lit += litLength;
+
+ /* literal Length */
+ if (litLength>0xFFFF) {
+ assert(seqStorePtr->longLengthType == ZSTD_llt_none); /* there can only be a single long length */
+ seqStorePtr->longLengthType = ZSTD_llt_literalLength;
+ seqStorePtr->longLengthPos = (U32)(seqStorePtr->sequences - seqStorePtr->sequencesStart);
+ }
+ seqStorePtr->sequences[0].litLength = (U16)litLength;
+
+ /* match offset */
+ seqStorePtr->sequences[0].offBase = STORED_TO_OFFBASE(offBase_minus1);
+
+ /* match Length */
+ assert(matchLength >= MINMATCH);
+ { size_t const mlBase = matchLength - MINMATCH;
+ if (mlBase>0xFFFF) {
+ assert(seqStorePtr->longLengthType == ZSTD_llt_none); /* there can only be a single long length */
+ seqStorePtr->longLengthType = ZSTD_llt_matchLength;
+ seqStorePtr->longLengthPos = (U32)(seqStorePtr->sequences - seqStorePtr->sequencesStart);
+ }
+ seqStorePtr->sequences[0].mlBase = (U16)mlBase;
+ }
+
+ seqStorePtr->sequences++;
+}
+
+/* ZSTD_updateRep() :
+ * updates in-place @rep (array of repeat offsets)
+ * @offBase_minus1 : sum-type, with same numeric representation as ZSTD_storeSeq()
+ */
+MEM_STATIC void
+ZSTD_updateRep(U32 rep[ZSTD_REP_NUM], U32 const offBase_minus1, U32 const ll0)
+{
+ if (STORED_IS_OFFSET(offBase_minus1)) { /* full offset */
+ rep[2] = rep[1];
+ rep[1] = rep[0];
+ rep[0] = STORED_OFFSET(offBase_minus1);
+ } else { /* repcode */
+ U32 const repCode = STORED_REPCODE(offBase_minus1) - 1 + ll0;
+ if (repCode > 0) { /* note : if repCode==0, no change */
+ U32 const currentOffset = (repCode==ZSTD_REP_NUM) ? (rep[0] - 1) : rep[repCode];
+ rep[2] = (repCode >= 2) ? rep[1] : rep[2];
+ rep[1] = rep[0];
+ rep[0] = currentOffset;
+ } else { /* repCode == 0 */
+ /* nothing to do */
+ }
+ }
+}
+
+typedef struct repcodes_s {
+ U32 rep[3];
+} repcodes_t;
+
+MEM_STATIC repcodes_t
+ZSTD_newRep(U32 const rep[ZSTD_REP_NUM], U32 const offBase_minus1, U32 const ll0)
+{
+ repcodes_t newReps;
+ ZSTD_memcpy(&newReps, rep, sizeof(newReps));
+ ZSTD_updateRep(newReps.rep, offBase_minus1, ll0);
+ return newReps;
+}
+
+
+/*-*************************************
+* Match length counter
+***************************************/
+static unsigned ZSTD_NbCommonBytes (size_t val)
+{
+ if (MEM_isLittleEndian()) {
+ if (MEM_64bits()) {
+# if (__GNUC__ >= 4)
+ return (__builtin_ctzll((U64)val) >> 3);
+# else
+ static const int DeBruijnBytePos[64] = { 0, 0, 0, 0, 0, 1, 1, 2,
+ 0, 3, 1, 3, 1, 4, 2, 7,
+ 0, 2, 3, 6, 1, 5, 3, 5,
+ 1, 3, 4, 4, 2, 5, 6, 7,
+ 7, 0, 1, 2, 3, 3, 4, 6,
+ 2, 6, 5, 5, 3, 4, 5, 6,
+ 7, 1, 2, 4, 6, 4, 4, 5,
+ 7, 2, 6, 5, 7, 6, 7, 7 };
+ return DeBruijnBytePos[((U64)((val & -(long long)val) * 0x0218A392CDABBD3FULL)) >> 58];
+# endif
+ } else { /* 32 bits */
+# if (__GNUC__ >= 3)
+ return (__builtin_ctz((U32)val) >> 3);
+# else
+ static const int DeBruijnBytePos[32] = { 0, 0, 3, 0, 3, 1, 3, 0,
+ 3, 2, 2, 1, 3, 2, 0, 1,
+ 3, 3, 1, 2, 2, 2, 2, 0,
+ 3, 1, 2, 0, 1, 0, 1, 1 };
+ return DeBruijnBytePos[((U32)((val & -(S32)val) * 0x077CB531U)) >> 27];
+# endif
+ }
+ } else { /* Big Endian CPU */
+ if (MEM_64bits()) {
+# if (__GNUC__ >= 4)
+ return (__builtin_clzll(val) >> 3);
+# else
+ unsigned r;
+ const unsigned n32 = sizeof(size_t)*4; /* calculate this way due to compiler complaining in 32-bits mode */
+ if (!(val>>n32)) { r=4; } else { r=0; val>>=n32; }
+ if (!(val>>16)) { r+=2; val>>=8; } else { val>>=24; }
+ r += (!val);
+ return r;
+# endif
+ } else { /* 32 bits */
+# if (__GNUC__ >= 3)
+ return (__builtin_clz((U32)val) >> 3);
+# else
+ unsigned r;
+ if (!(val>>16)) { r=2; val>>=8; } else { r=0; val>>=24; }
+ r += (!val);
+ return r;
+# endif
+ } }
+}
+
+
+MEM_STATIC size_t ZSTD_count(const BYTE* pIn, const BYTE* pMatch, const BYTE* const pInLimit)
+{
+ const BYTE* const pStart = pIn;
+ const BYTE* const pInLoopLimit = pInLimit - (sizeof(size_t)-1);
+
+ if (pIn < pInLoopLimit) {
+ { size_t const diff = MEM_readST(pMatch) ^ MEM_readST(pIn);
+ if (diff) return ZSTD_NbCommonBytes(diff); }
+ pIn+=sizeof(size_t); pMatch+=sizeof(size_t);
+ while (pIn < pInLoopLimit) {
+ size_t const diff = MEM_readST(pMatch) ^ MEM_readST(pIn);
+ if (!diff) { pIn+=sizeof(size_t); pMatch+=sizeof(size_t); continue; }
+ pIn += ZSTD_NbCommonBytes(diff);
+ return (size_t)(pIn - pStart);
+ } }
+ if (MEM_64bits() && (pIn<(pInLimit-3)) && (MEM_read32(pMatch) == MEM_read32(pIn))) { pIn+=4; pMatch+=4; }
+ if ((pIn<(pInLimit-1)) && (MEM_read16(pMatch) == MEM_read16(pIn))) { pIn+=2; pMatch+=2; }
+ if ((pIn<pInLimit) && (*pMatch == *pIn)) pIn++;
+ return (size_t)(pIn - pStart);
+}
+
+/* ZSTD_count_2segments() :
+ * can count match length with `ip` & `match` in 2 different segments.
+ * convention : on reaching mEnd, match count continue starting from iStart
+ */
+MEM_STATIC size_t
+ZSTD_count_2segments(const BYTE* ip, const BYTE* match,
+ const BYTE* iEnd, const BYTE* mEnd, const BYTE* iStart)
+{
+ const BYTE* const vEnd = MIN( ip + (mEnd - match), iEnd);
+ size_t const matchLength = ZSTD_count(ip, match, vEnd);
+ if (match + matchLength != mEnd) return matchLength;
+ DEBUGLOG(7, "ZSTD_count_2segments: found a 2-parts match (current length==%zu)", matchLength);
+ DEBUGLOG(7, "distance from match beginning to end dictionary = %zi", mEnd - match);
+ DEBUGLOG(7, "distance from current pos to end buffer = %zi", iEnd - ip);
+ DEBUGLOG(7, "next byte : ip==%02X, istart==%02X", ip[matchLength], *iStart);
+ DEBUGLOG(7, "final match length = %zu", matchLength + ZSTD_count(ip+matchLength, iStart, iEnd));
+ return matchLength + ZSTD_count(ip+matchLength, iStart, iEnd);
+}
+
+
+/*-*************************************
+ * Hashes
+ ***************************************/
+static const U32 prime3bytes = 506832829U;
+static U32 ZSTD_hash3(U32 u, U32 h) { return ((u << (32-24)) * prime3bytes) >> (32-h) ; }
+MEM_STATIC size_t ZSTD_hash3Ptr(const void* ptr, U32 h) { return ZSTD_hash3(MEM_readLE32(ptr), h); } /* only in zstd_opt.h */
+
+static const U32 prime4bytes = 2654435761U;
+static U32 ZSTD_hash4(U32 u, U32 h) { return (u * prime4bytes) >> (32-h) ; }
+static size_t ZSTD_hash4Ptr(const void* ptr, U32 h) { return ZSTD_hash4(MEM_read32(ptr), h); }
+
+static const U64 prime5bytes = 889523592379ULL;
+static size_t ZSTD_hash5(U64 u, U32 h) { return (size_t)(((u << (64-40)) * prime5bytes) >> (64-h)) ; }
+static size_t ZSTD_hash5Ptr(const void* p, U32 h) { return ZSTD_hash5(MEM_readLE64(p), h); }
+
+static const U64 prime6bytes = 227718039650203ULL;
+static size_t ZSTD_hash6(U64 u, U32 h) { return (size_t)(((u << (64-48)) * prime6bytes) >> (64-h)) ; }
+static size_t ZSTD_hash6Ptr(const void* p, U32 h) { return ZSTD_hash6(MEM_readLE64(p), h); }
+
+static const U64 prime7bytes = 58295818150454627ULL;
+static size_t ZSTD_hash7(U64 u, U32 h) { return (size_t)(((u << (64-56)) * prime7bytes) >> (64-h)) ; }
+static size_t ZSTD_hash7Ptr(const void* p, U32 h) { return ZSTD_hash7(MEM_readLE64(p), h); }
+
+static const U64 prime8bytes = 0xCF1BBCDCB7A56463ULL;
+static size_t ZSTD_hash8(U64 u, U32 h) { return (size_t)(((u) * prime8bytes) >> (64-h)) ; }
+static size_t ZSTD_hash8Ptr(const void* p, U32 h) { return ZSTD_hash8(MEM_readLE64(p), h); }
+
+MEM_STATIC FORCE_INLINE_ATTR
+size_t ZSTD_hashPtr(const void* p, U32 hBits, U32 mls)
+{
+ switch(mls)
+ {
+ default:
+ case 4: return ZSTD_hash4Ptr(p, hBits);
+ case 5: return ZSTD_hash5Ptr(p, hBits);
+ case 6: return ZSTD_hash6Ptr(p, hBits);
+ case 7: return ZSTD_hash7Ptr(p, hBits);
+ case 8: return ZSTD_hash8Ptr(p, hBits);
+ }
+}
+
+/* ZSTD_ipow() :
+ * Return base^exponent.
+ */
+static U64 ZSTD_ipow(U64 base, U64 exponent)
+{
+ U64 power = 1;
+ while (exponent) {
+ if (exponent & 1) power *= base;
+ exponent >>= 1;
+ base *= base;
+ }
+ return power;
+}
+
+#define ZSTD_ROLL_HASH_CHAR_OFFSET 10
+
+/* ZSTD_rollingHash_append() :
+ * Add the buffer to the hash value.
+ */
+static U64 ZSTD_rollingHash_append(U64 hash, void const* buf, size_t size)
+{
+ BYTE const* istart = (BYTE const*)buf;
+ size_t pos;
+ for (pos = 0; pos < size; ++pos) {
+ hash *= prime8bytes;
+ hash += istart[pos] + ZSTD_ROLL_HASH_CHAR_OFFSET;
+ }
+ return hash;
+}
+
+/* ZSTD_rollingHash_compute() :
+ * Compute the rolling hash value of the buffer.
+ */
+MEM_STATIC U64 ZSTD_rollingHash_compute(void const* buf, size_t size)
+{
+ return ZSTD_rollingHash_append(0, buf, size);
+}
+
+/* ZSTD_rollingHash_primePower() :
+ * Compute the primePower to be passed to ZSTD_rollingHash_rotate() for a hash
+ * over a window of length bytes.
+ */
+MEM_STATIC U64 ZSTD_rollingHash_primePower(U32 length)
+{
+ return ZSTD_ipow(prime8bytes, length - 1);
+}
+
+/* ZSTD_rollingHash_rotate() :
+ * Rotate the rolling hash by one byte.
+ */
+MEM_STATIC U64 ZSTD_rollingHash_rotate(U64 hash, BYTE toRemove, BYTE toAdd, U64 primePower)
+{
+ hash -= (toRemove + ZSTD_ROLL_HASH_CHAR_OFFSET) * primePower;
+ hash *= prime8bytes;
+ hash += toAdd + ZSTD_ROLL_HASH_CHAR_OFFSET;
+ return hash;
+}
+
+/*-*************************************
+* Round buffer management
+***************************************/
+#if (ZSTD_WINDOWLOG_MAX_64 > 31)
+# error "ZSTD_WINDOWLOG_MAX is too large : would overflow ZSTD_CURRENT_MAX"
+#endif
+/* Max current allowed */
+#define ZSTD_CURRENT_MAX ((3U << 29) + (1U << ZSTD_WINDOWLOG_MAX))
+/* Maximum chunk size before overflow correction needs to be called again */
+#define ZSTD_CHUNKSIZE_MAX \
+ ( ((U32)-1) /* Maximum ending current index */ \
+ - ZSTD_CURRENT_MAX) /* Maximum beginning lowLimit */
+
+/*
+ * ZSTD_window_clear():
+ * Clears the window containing the history by simply setting it to empty.
+ */
+MEM_STATIC void ZSTD_window_clear(ZSTD_window_t* window)
+{
+ size_t const endT = (size_t)(window->nextSrc - window->base);
+ U32 const end = (U32)endT;
+
+ window->lowLimit = end;
+ window->dictLimit = end;
+}
+
+MEM_STATIC U32 ZSTD_window_isEmpty(ZSTD_window_t const window)
+{
+ return window.dictLimit == ZSTD_WINDOW_START_INDEX &&
+ window.lowLimit == ZSTD_WINDOW_START_INDEX &&
+ (window.nextSrc - window.base) == ZSTD_WINDOW_START_INDEX;
+}
+
+/*
+ * ZSTD_window_hasExtDict():
+ * Returns non-zero if the window has a non-empty extDict.
+ */
+MEM_STATIC U32 ZSTD_window_hasExtDict(ZSTD_window_t const window)
+{
+ return window.lowLimit < window.dictLimit;
+}
+
+/*
+ * ZSTD_matchState_dictMode():
+ * Inspects the provided matchState and figures out what dictMode should be
+ * passed to the compressor.
+ */
+MEM_STATIC ZSTD_dictMode_e ZSTD_matchState_dictMode(const ZSTD_matchState_t *ms)
+{
+ return ZSTD_window_hasExtDict(ms->window) ?
+ ZSTD_extDict :
+ ms->dictMatchState != NULL ?
+ (ms->dictMatchState->dedicatedDictSearch ? ZSTD_dedicatedDictSearch : ZSTD_dictMatchState) :
+ ZSTD_noDict;
+}
+
+/* Defining this macro to non-zero tells zstd to run the overflow correction
+ * code much more frequently. This is very inefficient, and should only be
+ * used for tests and fuzzers.
+ */
+#ifndef ZSTD_WINDOW_OVERFLOW_CORRECT_FREQUENTLY
+# ifdef FUZZING_BUILD_MODE_UNSAFE_FOR_PRODUCTION
+# define ZSTD_WINDOW_OVERFLOW_CORRECT_FREQUENTLY 1
+# else
+# define ZSTD_WINDOW_OVERFLOW_CORRECT_FREQUENTLY 0
+# endif
+#endif
+
+/*
+ * ZSTD_window_canOverflowCorrect():
+ * Returns non-zero if the indices are large enough for overflow correction
+ * to work correctly without impacting compression ratio.
+ */
+MEM_STATIC U32 ZSTD_window_canOverflowCorrect(ZSTD_window_t const window,
+ U32 cycleLog,
+ U32 maxDist,
+ U32 loadedDictEnd,
+ void const* src)
+{
+ U32 const cycleSize = 1u << cycleLog;
+ U32 const curr = (U32)((BYTE const*)src - window.base);
+ U32 const minIndexToOverflowCorrect = cycleSize
+ + MAX(maxDist, cycleSize)
+ + ZSTD_WINDOW_START_INDEX;
+
+ /* Adjust the min index to backoff the overflow correction frequency,
+ * so we don't waste too much CPU in overflow correction. If this
+ * computation overflows we don't really care, we just need to make
+ * sure it is at least minIndexToOverflowCorrect.
+ */
+ U32 const adjustment = window.nbOverflowCorrections + 1;
+ U32 const adjustedIndex = MAX(minIndexToOverflowCorrect * adjustment,
+ minIndexToOverflowCorrect);
+ U32 const indexLargeEnough = curr > adjustedIndex;
+
+ /* Only overflow correct early if the dictionary is invalidated already,
+ * so we don't hurt compression ratio.
+ */
+ U32 const dictionaryInvalidated = curr > maxDist + loadedDictEnd;
+
+ return indexLargeEnough && dictionaryInvalidated;
+}
+
+/*
+ * ZSTD_window_needOverflowCorrection():
+ * Returns non-zero if the indices are getting too large and need overflow
+ * protection.
+ */
+MEM_STATIC U32 ZSTD_window_needOverflowCorrection(ZSTD_window_t const window,
+ U32 cycleLog,
+ U32 maxDist,
+ U32 loadedDictEnd,
+ void const* src,
+ void const* srcEnd)
+{
+ U32 const curr = (U32)((BYTE const*)srcEnd - window.base);
+ if (ZSTD_WINDOW_OVERFLOW_CORRECT_FREQUENTLY) {
+ if (ZSTD_window_canOverflowCorrect(window, cycleLog, maxDist, loadedDictEnd, src)) {
+ return 1;
+ }
+ }
+ return curr > ZSTD_CURRENT_MAX;
+}
+
+/*
+ * ZSTD_window_correctOverflow():
+ * Reduces the indices to protect from index overflow.
+ * Returns the correction made to the indices, which must be applied to every
+ * stored index.
+ *
+ * The least significant cycleLog bits of the indices must remain the same,
+ * which may be 0. Every index up to maxDist in the past must be valid.
+ */
+MEM_STATIC U32 ZSTD_window_correctOverflow(ZSTD_window_t* window, U32 cycleLog,
+ U32 maxDist, void const* src)
+{
+ /* preemptive overflow correction:
+ * 1. correction is large enough:
+ * lowLimit > (3<<29) ==> current > 3<<29 + 1<<windowLog
+ * 1<<windowLog <= newCurrent < 1<<chainLog + 1<<windowLog
+ *
+ * current - newCurrent
+ * > (3<<29 + 1<<windowLog) - (1<<windowLog + 1<<chainLog)
+ * > (3<<29) - (1<<chainLog)
+ * > (3<<29) - (1<<30) (NOTE: chainLog <= 30)
+ * > 1<<29
+ *
+ * 2. (ip+ZSTD_CHUNKSIZE_MAX - cctx->base) doesn't overflow:
+ * After correction, current is less than (1<<chainLog + 1<<windowLog).
+ * In 64-bit mode we are safe, because we have 64-bit ptrdiff_t.
+ * In 32-bit mode we are safe, because (chainLog <= 29), so
+ * ip+ZSTD_CHUNKSIZE_MAX - cctx->base < 1<<32.
+ * 3. (cctx->lowLimit + 1<<windowLog) < 1<<32:
+ * windowLog <= 31 ==> 3<<29 + 1<<windowLog < 7<<29 < 1<<32.
+ */
+ U32 const cycleSize = 1u << cycleLog;
+ U32 const cycleMask = cycleSize - 1;
+ U32 const curr = (U32)((BYTE const*)src - window->base);
+ U32 const currentCycle = curr & cycleMask;
+ /* Ensure newCurrent - maxDist >= ZSTD_WINDOW_START_INDEX. */
+ U32 const currentCycleCorrection = currentCycle < ZSTD_WINDOW_START_INDEX
+ ? MAX(cycleSize, ZSTD_WINDOW_START_INDEX)
+ : 0;
+ U32 const newCurrent = currentCycle
+ + currentCycleCorrection
+ + MAX(maxDist, cycleSize);
+ U32 const correction = curr - newCurrent;
+ /* maxDist must be a power of two so that:
+ * (newCurrent & cycleMask) == (curr & cycleMask)
+ * This is required to not corrupt the chains / binary tree.
+ */
+ assert((maxDist & (maxDist - 1)) == 0);
+ assert((curr & cycleMask) == (newCurrent & cycleMask));
+ assert(curr > newCurrent);
+ if (!ZSTD_WINDOW_OVERFLOW_CORRECT_FREQUENTLY) {
+ /* Loose bound, should be around 1<<29 (see above) */
+ assert(correction > 1<<28);
+ }
+
+ window->base += correction;
+ window->dictBase += correction;
+ if (window->lowLimit < correction + ZSTD_WINDOW_START_INDEX) {
+ window->lowLimit = ZSTD_WINDOW_START_INDEX;
+ } else {
+ window->lowLimit -= correction;
+ }
+ if (window->dictLimit < correction + ZSTD_WINDOW_START_INDEX) {
+ window->dictLimit = ZSTD_WINDOW_START_INDEX;
+ } else {
+ window->dictLimit -= correction;
+ }
+
+ /* Ensure we can still reference the full window. */
+ assert(newCurrent >= maxDist);
+ assert(newCurrent - maxDist >= ZSTD_WINDOW_START_INDEX);
+ /* Ensure that lowLimit and dictLimit didn't underflow. */
+ assert(window->lowLimit <= newCurrent);
+ assert(window->dictLimit <= newCurrent);
+
+ ++window->nbOverflowCorrections;
+
+ DEBUGLOG(4, "Correction of 0x%x bytes to lowLimit=0x%x", correction,
+ window->lowLimit);
+ return correction;
+}
+
+/*
+ * ZSTD_window_enforceMaxDist():
+ * Updates lowLimit so that:
+ * (srcEnd - base) - lowLimit == maxDist + loadedDictEnd
+ *
+ * It ensures index is valid as long as index >= lowLimit.
+ * This must be called before a block compression call.
+ *
+ * loadedDictEnd is only defined if a dictionary is in use for current compression.
+ * As the name implies, loadedDictEnd represents the index at end of dictionary.
+ * The value lies within context's referential, it can be directly compared to blockEndIdx.
+ *
+ * If loadedDictEndPtr is NULL, no dictionary is in use, and we use loadedDictEnd == 0.
+ * If loadedDictEndPtr is not NULL, we set it to zero after updating lowLimit.
+ * This is because dictionaries are allowed to be referenced fully
+ * as long as the last byte of the dictionary is in the window.
+ * Once input has progressed beyond window size, dictionary cannot be referenced anymore.
+ *
+ * In normal dict mode, the dictionary lies between lowLimit and dictLimit.
+ * In dictMatchState mode, lowLimit and dictLimit are the same,
+ * and the dictionary is below them.
+ * forceWindow and dictMatchState are therefore incompatible.
+ */
+MEM_STATIC void
+ZSTD_window_enforceMaxDist(ZSTD_window_t* window,
+ const void* blockEnd,
+ U32 maxDist,
+ U32* loadedDictEndPtr,
+ const ZSTD_matchState_t** dictMatchStatePtr)
+{
+ U32 const blockEndIdx = (U32)((BYTE const*)blockEnd - window->base);
+ U32 const loadedDictEnd = (loadedDictEndPtr != NULL) ? *loadedDictEndPtr : 0;
+ DEBUGLOG(5, "ZSTD_window_enforceMaxDist: blockEndIdx=%u, maxDist=%u, loadedDictEnd=%u",
+ (unsigned)blockEndIdx, (unsigned)maxDist, (unsigned)loadedDictEnd);
+
+ /* - When there is no dictionary : loadedDictEnd == 0.
+ In which case, the test (blockEndIdx > maxDist) is merely to avoid
+ overflowing next operation `newLowLimit = blockEndIdx - maxDist`.
+ - When there is a standard dictionary :
+ Index referential is copied from the dictionary,
+ which means it starts from 0.
+ In which case, loadedDictEnd == dictSize,
+ and it makes sense to compare `blockEndIdx > maxDist + dictSize`
+ since `blockEndIdx` also starts from zero.
+ - When there is an attached dictionary :
+ loadedDictEnd is expressed within the referential of the context,
+ so it can be directly compared against blockEndIdx.
+ */
+ if (blockEndIdx > maxDist + loadedDictEnd) {
+ U32 const newLowLimit = blockEndIdx - maxDist;
+ if (window->lowLimit < newLowLimit) window->lowLimit = newLowLimit;
+ if (window->dictLimit < window->lowLimit) {
+ DEBUGLOG(5, "Update dictLimit to match lowLimit, from %u to %u",
+ (unsigned)window->dictLimit, (unsigned)window->lowLimit);
+ window->dictLimit = window->lowLimit;
+ }
+ /* On reaching window size, dictionaries are invalidated */
+ if (loadedDictEndPtr) *loadedDictEndPtr = 0;
+ if (dictMatchStatePtr) *dictMatchStatePtr = NULL;
+ }
+}
+
+/* Similar to ZSTD_window_enforceMaxDist(),
+ * but only invalidates dictionary
+ * when input progresses beyond window size.
+ * assumption : loadedDictEndPtr and dictMatchStatePtr are valid (non NULL)
+ * loadedDictEnd uses same referential as window->base
+ * maxDist is the window size */
+MEM_STATIC void
+ZSTD_checkDictValidity(const ZSTD_window_t* window,
+ const void* blockEnd,
+ U32 maxDist,
+ U32* loadedDictEndPtr,
+ const ZSTD_matchState_t** dictMatchStatePtr)
+{
+ assert(loadedDictEndPtr != NULL);
+ assert(dictMatchStatePtr != NULL);
+ { U32 const blockEndIdx = (U32)((BYTE const*)blockEnd - window->base);
+ U32 const loadedDictEnd = *loadedDictEndPtr;
+ DEBUGLOG(5, "ZSTD_checkDictValidity: blockEndIdx=%u, maxDist=%u, loadedDictEnd=%u",
+ (unsigned)blockEndIdx, (unsigned)maxDist, (unsigned)loadedDictEnd);
+ assert(blockEndIdx >= loadedDictEnd);
+
+ if (blockEndIdx > loadedDictEnd + maxDist) {
+ /* On reaching window size, dictionaries are invalidated.
+ * For simplification, if window size is reached anywhere within next block,
+ * the dictionary is invalidated for the full block.
+ */
+ DEBUGLOG(6, "invalidating dictionary for current block (distance > windowSize)");
+ *loadedDictEndPtr = 0;
+ *dictMatchStatePtr = NULL;
+ } else {
+ if (*loadedDictEndPtr != 0) {
+ DEBUGLOG(6, "dictionary considered valid for current block");
+ } } }
+}
+
+MEM_STATIC void ZSTD_window_init(ZSTD_window_t* window) {
+ ZSTD_memset(window, 0, sizeof(*window));
+ window->base = (BYTE const*)" ";
+ window->dictBase = (BYTE const*)" ";
+ ZSTD_STATIC_ASSERT(ZSTD_DUBT_UNSORTED_MARK < ZSTD_WINDOW_START_INDEX); /* Start above ZSTD_DUBT_UNSORTED_MARK */
+ window->dictLimit = ZSTD_WINDOW_START_INDEX; /* start from >0, so that 1st position is valid */
+ window->lowLimit = ZSTD_WINDOW_START_INDEX; /* it ensures first and later CCtx usages compress the same */
+ window->nextSrc = window->base + ZSTD_WINDOW_START_INDEX; /* see issue #1241 */
+ window->nbOverflowCorrections = 0;
+}
+
+/*
+ * ZSTD_window_update():
+ * Updates the window by appending [src, src + srcSize) to the window.
+ * If it is not contiguous, the current prefix becomes the extDict, and we
+ * forget about the extDict. Handles overlap of the prefix and extDict.
+ * Returns non-zero if the segment is contiguous.
+ */
+MEM_STATIC U32 ZSTD_window_update(ZSTD_window_t* window,
+ void const* src, size_t srcSize,
+ int forceNonContiguous)
+{
+ BYTE const* const ip = (BYTE const*)src;
+ U32 contiguous = 1;
+ DEBUGLOG(5, "ZSTD_window_update");
+ if (srcSize == 0)
+ return contiguous;
+ assert(window->base != NULL);
+ assert(window->dictBase != NULL);
+ /* Check if blocks follow each other */
+ if (src != window->nextSrc || forceNonContiguous) {
+ /* not contiguous */
+ size_t const distanceFromBase = (size_t)(window->nextSrc - window->base);
+ DEBUGLOG(5, "Non contiguous blocks, new segment starts at %u", window->dictLimit);
+ window->lowLimit = window->dictLimit;
+ assert(distanceFromBase == (size_t)(U32)distanceFromBase); /* should never overflow */
+ window->dictLimit = (U32)distanceFromBase;
+ window->dictBase = window->base;
+ window->base = ip - distanceFromBase;
+ /* ms->nextToUpdate = window->dictLimit; */
+ if (window->dictLimit - window->lowLimit < HASH_READ_SIZE) window->lowLimit = window->dictLimit; /* too small extDict */
+ contiguous = 0;
+ }
+ window->nextSrc = ip + srcSize;
+ /* if input and dictionary overlap : reduce dictionary (area presumed modified by input) */
+ if ( (ip+srcSize > window->dictBase + window->lowLimit)
+ & (ip < window->dictBase + window->dictLimit)) {
+ ptrdiff_t const highInputIdx = (ip + srcSize) - window->dictBase;
+ U32 const lowLimitMax = (highInputIdx > (ptrdiff_t)window->dictLimit) ? window->dictLimit : (U32)highInputIdx;
+ window->lowLimit = lowLimitMax;
+ DEBUGLOG(5, "Overlapping extDict and input : new lowLimit = %u", window->lowLimit);
+ }
+ return contiguous;
+}
+
+/*
+ * Returns the lowest allowed match index. It may either be in the ext-dict or the prefix.
+ */
+MEM_STATIC U32 ZSTD_getLowestMatchIndex(const ZSTD_matchState_t* ms, U32 curr, unsigned windowLog)
+{
+ U32 const maxDistance = 1U << windowLog;
+ U32 const lowestValid = ms->window.lowLimit;
+ U32 const withinWindow = (curr - lowestValid > maxDistance) ? curr - maxDistance : lowestValid;
+ U32 const isDictionary = (ms->loadedDictEnd != 0);
+ /* When using a dictionary the entire dictionary is valid if a single byte of the dictionary
+ * is within the window. We invalidate the dictionary (and set loadedDictEnd to 0) when it isn't
+ * valid for the entire block. So this check is sufficient to find the lowest valid match index.
+ */
+ U32 const matchLowest = isDictionary ? lowestValid : withinWindow;
+ return matchLowest;
+}
+
+/*
+ * Returns the lowest allowed match index in the prefix.
+ */
+MEM_STATIC U32 ZSTD_getLowestPrefixIndex(const ZSTD_matchState_t* ms, U32 curr, unsigned windowLog)
+{
+ U32 const maxDistance = 1U << windowLog;
+ U32 const lowestValid = ms->window.dictLimit;
+ U32 const withinWindow = (curr - lowestValid > maxDistance) ? curr - maxDistance : lowestValid;
+ U32 const isDictionary = (ms->loadedDictEnd != 0);
+ /* When computing the lowest prefix index we need to take the dictionary into account to handle
+ * the edge case where the dictionary and the source are contiguous in memory.
+ */
+ U32 const matchLowest = isDictionary ? lowestValid : withinWindow;
+ return matchLowest;
+}
+
+
+
+/* debug functions */
+#if (DEBUGLEVEL>=2)
+
+MEM_STATIC double ZSTD_fWeight(U32 rawStat)
+{
+ U32 const fp_accuracy = 8;
+ U32 const fp_multiplier = (1 << fp_accuracy);
+ U32 const newStat = rawStat + 1;
+ U32 const hb = ZSTD_highbit32(newStat);
+ U32 const BWeight = hb * fp_multiplier;
+ U32 const FWeight = (newStat << fp_accuracy) >> hb;
+ U32 const weight = BWeight + FWeight;
+ assert(hb + fp_accuracy < 31);
+ return (double)weight / fp_multiplier;
+}
+
+/* display a table content,
+ * listing each element, its frequency, and its predicted bit cost */
+MEM_STATIC void ZSTD_debugTable(const U32* table, U32 max)
+{
+ unsigned u, sum;
+ for (u=0, sum=0; u<=max; u++) sum += table[u];
+ DEBUGLOG(2, "total nb elts: %u", sum);
+ for (u=0; u<=max; u++) {
+ DEBUGLOG(2, "%2u: %5u (%.2f)",
+ u, table[u], ZSTD_fWeight(sum) - ZSTD_fWeight(table[u]) );
+ }
+}
+
+#endif
+
+
+
+/* ===============================================================
+ * Shared internal declarations
+ * These prototypes may be called from sources not in lib/compress
+ * =============================================================== */
+
+/* ZSTD_loadCEntropy() :
+ * dict : must point at beginning of a valid zstd dictionary.
+ * return : size of dictionary header (size of magic number + dict ID + entropy tables)
+ * assumptions : magic number supposed already checked
+ * and dictSize >= 8 */
+size_t ZSTD_loadCEntropy(ZSTD_compressedBlockState_t* bs, void* workspace,
+ const void* const dict, size_t dictSize);
+
+void ZSTD_reset_compressedBlockState(ZSTD_compressedBlockState_t* bs);
+
+/* ==============================================================
+ * Private declarations
+ * These prototypes shall only be called from within lib/compress
+ * ============================================================== */
+
+/* ZSTD_getCParamsFromCCtxParams() :
+ * cParams are built depending on compressionLevel, src size hints,
+ * LDM and manually set compression parameters.
+ * Note: srcSizeHint == 0 means 0!
+ */
+ZSTD_compressionParameters ZSTD_getCParamsFromCCtxParams(
+ const ZSTD_CCtx_params* CCtxParams, U64 srcSizeHint, size_t dictSize, ZSTD_cParamMode_e mode);
+
+/*! ZSTD_initCStream_internal() :
+ * Private use only. Init streaming operation.
+ * expects params to be valid.
+ * must receive dict, or cdict, or none, but not both.
+ * @return : 0, or an error code */
+size_t ZSTD_initCStream_internal(ZSTD_CStream* zcs,
+ const void* dict, size_t dictSize,
+ const ZSTD_CDict* cdict,
+ const ZSTD_CCtx_params* params, unsigned long long pledgedSrcSize);
+
+void ZSTD_resetSeqStore(seqStore_t* ssPtr);
+
+/*! ZSTD_getCParamsFromCDict() :
+ * as the name implies */
+ZSTD_compressionParameters ZSTD_getCParamsFromCDict(const ZSTD_CDict* cdict);
+
+/* ZSTD_compressBegin_advanced_internal() :
+ * Private use only. To be called from zstdmt_compress.c. */
+size_t ZSTD_compressBegin_advanced_internal(ZSTD_CCtx* cctx,
+ const void* dict, size_t dictSize,
+ ZSTD_dictContentType_e dictContentType,
+ ZSTD_dictTableLoadMethod_e dtlm,
+ const ZSTD_CDict* cdict,
+ const ZSTD_CCtx_params* params,
+ unsigned long long pledgedSrcSize);
+
+/* ZSTD_compress_advanced_internal() :
+ * Private use only. To be called from zstdmt_compress.c. */
+size_t ZSTD_compress_advanced_internal(ZSTD_CCtx* cctx,
+ void* dst, size_t dstCapacity,
+ const void* src, size_t srcSize,
+ const void* dict,size_t dictSize,
+ const ZSTD_CCtx_params* params);
+
+
+/* ZSTD_writeLastEmptyBlock() :
+ * output an empty Block with end-of-frame mark to complete a frame
+ * @return : size of data written into `dst` (== ZSTD_blockHeaderSize (defined in zstd_internal.h))
+ * or an error code if `dstCapacity` is too small (<ZSTD_blockHeaderSize)
+ */
+size_t ZSTD_writeLastEmptyBlock(void* dst, size_t dstCapacity);
+
+
+/* ZSTD_referenceExternalSequences() :
+ * Must be called before starting a compression operation.
+ * seqs must parse a prefix of the source.
+ * This cannot be used when long range matching is enabled.
+ * Zstd will use these sequences, and pass the literals to a secondary block
+ * compressor.
+ * @return : An error code on failure.
+ * NOTE: seqs are not verified! Invalid sequences can cause out-of-bounds memory
+ * access and data corruption.
+ */
+size_t ZSTD_referenceExternalSequences(ZSTD_CCtx* cctx, rawSeq* seq, size_t nbSeq);
+
+/* ZSTD_cycleLog() :
+ * condition for correct operation : hashLog > 1 */
+U32 ZSTD_cycleLog(U32 hashLog, ZSTD_strategy strat);
+
+/* ZSTD_CCtx_trace() :
+ * Trace the end of a compression call.
+ */
+void ZSTD_CCtx_trace(ZSTD_CCtx* cctx, size_t extraCSize);
+
+#endif /* ZSTD_COMPRESS_H */
diff --git a/lib/zstd/compress/zstd_compress_literals.c b/lib/zstd/compress/zstd_compress_literals.c
new file mode 100644
index 0000000000..52b0a8059a
--- /dev/null
+++ b/lib/zstd/compress/zstd_compress_literals.c
@@ -0,0 +1,159 @@
+/*
+ * Copyright (c) Yann Collet, Facebook, Inc.
+ * All rights reserved.
+ *
+ * This source code is licensed under both the BSD-style license (found in the
+ * LICENSE file in the root directory of this source tree) and the GPLv2 (found
+ * in the COPYING file in the root directory of this source tree).
+ * You may select, at your option, one of the above-listed licenses.
+ */
+
+ /*-*************************************
+ * Dependencies
+ ***************************************/
+#include "zstd_compress_literals.h"
+
+size_t ZSTD_noCompressLiterals (void* dst, size_t dstCapacity, const void* src, size_t srcSize)
+{
+ BYTE* const ostart = (BYTE*)dst;
+ U32 const flSize = 1 + (srcSize>31) + (srcSize>4095);
+
+ RETURN_ERROR_IF(srcSize + flSize > dstCapacity, dstSize_tooSmall, "");
+
+ switch(flSize)
+ {
+ case 1: /* 2 - 1 - 5 */
+ ostart[0] = (BYTE)((U32)set_basic + (srcSize<<3));
+ break;
+ case 2: /* 2 - 2 - 12 */
+ MEM_writeLE16(ostart, (U16)((U32)set_basic + (1<<2) + (srcSize<<4)));
+ break;
+ case 3: /* 2 - 2 - 20 */
+ MEM_writeLE32(ostart, (U32)((U32)set_basic + (3<<2) + (srcSize<<4)));
+ break;
+ default: /* not necessary : flSize is {1,2,3} */
+ assert(0);
+ }
+
+ ZSTD_memcpy(ostart + flSize, src, srcSize);
+ DEBUGLOG(5, "Raw literals: %u -> %u", (U32)srcSize, (U32)(srcSize + flSize));
+ return srcSize + flSize;
+}
+
+size_t ZSTD_compressRleLiteralsBlock (void* dst, size_t dstCapacity, const void* src, size_t srcSize)
+{
+ BYTE* const ostart = (BYTE*)dst;
+ U32 const flSize = 1 + (srcSize>31) + (srcSize>4095);
+
+ (void)dstCapacity; /* dstCapacity already guaranteed to be >=4, hence large enough */
+
+ switch(flSize)
+ {
+ case 1: /* 2 - 1 - 5 */
+ ostart[0] = (BYTE)((U32)set_rle + (srcSize<<3));
+ break;
+ case 2: /* 2 - 2 - 12 */
+ MEM_writeLE16(ostart, (U16)((U32)set_rle + (1<<2) + (srcSize<<4)));
+ break;
+ case 3: /* 2 - 2 - 20 */
+ MEM_writeLE32(ostart, (U32)((U32)set_rle + (3<<2) + (srcSize<<4)));
+ break;
+ default: /* not necessary : flSize is {1,2,3} */
+ assert(0);
+ }
+
+ ostart[flSize] = *(const BYTE*)src;
+ DEBUGLOG(5, "RLE literals: %u -> %u", (U32)srcSize, (U32)flSize + 1);
+ return flSize+1;
+}
+
+size_t ZSTD_compressLiterals (ZSTD_hufCTables_t const* prevHuf,
+ ZSTD_hufCTables_t* nextHuf,
+ ZSTD_strategy strategy, int disableLiteralCompression,
+ void* dst, size_t dstCapacity,
+ const void* src, size_t srcSize,
+ void* entropyWorkspace, size_t entropyWorkspaceSize,
+ const int bmi2,
+ unsigned suspectUncompressible)
+{
+ size_t const minGain = ZSTD_minGain(srcSize, strategy);
+ size_t const lhSize = 3 + (srcSize >= 1 KB) + (srcSize >= 16 KB);
+ BYTE* const ostart = (BYTE*)dst;
+ U32 singleStream = srcSize < 256;
+ symbolEncodingType_e hType = set_compressed;
+ size_t cLitSize;
+
+ DEBUGLOG(5,"ZSTD_compressLiterals (disableLiteralCompression=%i srcSize=%u)",
+ disableLiteralCompression, (U32)srcSize);
+
+ /* Prepare nextEntropy assuming reusing the existing table */
+ ZSTD_memcpy(nextHuf, prevHuf, sizeof(*prevHuf));
+
+ if (disableLiteralCompression)
+ return ZSTD_noCompressLiterals(dst, dstCapacity, src, srcSize);
+
+ /* small ? don't even attempt compression (speed opt) */
+# define COMPRESS_LITERALS_SIZE_MIN 63
+ { size_t const minLitSize = (prevHuf->repeatMode == HUF_repeat_valid) ? 6 : COMPRESS_LITERALS_SIZE_MIN;
+ if (srcSize <= minLitSize) return ZSTD_noCompressLiterals(dst, dstCapacity, src, srcSize);
+ }
+
+ RETURN_ERROR_IF(dstCapacity < lhSize+1, dstSize_tooSmall, "not enough space for compression");
+ { HUF_repeat repeat = prevHuf->repeatMode;
+ int const preferRepeat = strategy < ZSTD_lazy ? srcSize <= 1024 : 0;
+ if (repeat == HUF_repeat_valid && lhSize == 3) singleStream = 1;
+ cLitSize = singleStream ?
+ HUF_compress1X_repeat(
+ ostart+lhSize, dstCapacity-lhSize, src, srcSize,
+ HUF_SYMBOLVALUE_MAX, HUF_TABLELOG_DEFAULT, entropyWorkspace, entropyWorkspaceSize,
+ (HUF_CElt*)nextHuf->CTable, &repeat, preferRepeat, bmi2, suspectUncompressible) :
+ HUF_compress4X_repeat(
+ ostart+lhSize, dstCapacity-lhSize, src, srcSize,
+ HUF_SYMBOLVALUE_MAX, HUF_TABLELOG_DEFAULT, entropyWorkspace, entropyWorkspaceSize,
+ (HUF_CElt*)nextHuf->CTable, &repeat, preferRepeat, bmi2, suspectUncompressible);
+ if (repeat != HUF_repeat_none) {
+ /* reused the existing table */
+ DEBUGLOG(5, "Reusing previous huffman table");
+ hType = set_repeat;
+ }
+ }
+
+ if ((cLitSize==0) || (cLitSize >= srcSize - minGain) || ERR_isError(cLitSize)) {
+ ZSTD_memcpy(nextHuf, prevHuf, sizeof(*prevHuf));
+ return ZSTD_noCompressLiterals(dst, dstCapacity, src, srcSize);
+ }
+ if (cLitSize==1) {
+ ZSTD_memcpy(nextHuf, prevHuf, sizeof(*prevHuf));
+ return ZSTD_compressRleLiteralsBlock(dst, dstCapacity, src, srcSize);
+ }
+
+ if (hType == set_compressed) {
+ /* using a newly constructed table */
+ nextHuf->repeatMode = HUF_repeat_check;
+ }
+
+ /* Build header */
+ switch(lhSize)
+ {
+ case 3: /* 2 - 2 - 10 - 10 */
+ { U32 const lhc = hType + ((!singleStream) << 2) + ((U32)srcSize<<4) + ((U32)cLitSize<<14);
+ MEM_writeLE24(ostart, lhc);
+ break;
+ }
+ case 4: /* 2 - 2 - 14 - 14 */
+ { U32 const lhc = hType + (2 << 2) + ((U32)srcSize<<4) + ((U32)cLitSize<<18);
+ MEM_writeLE32(ostart, lhc);
+ break;
+ }
+ case 5: /* 2 - 2 - 18 - 18 */
+ { U32 const lhc = hType + (3 << 2) + ((U32)srcSize<<4) + ((U32)cLitSize<<22);
+ MEM_writeLE32(ostart, lhc);
+ ostart[4] = (BYTE)(cLitSize >> 10);
+ break;
+ }
+ default: /* not possible : lhSize is {3,4,5} */
+ assert(0);
+ }
+ DEBUGLOG(5, "Compressed literals: %u -> %u", (U32)srcSize, (U32)(lhSize+cLitSize));
+ return lhSize+cLitSize;
+}
diff --git a/lib/zstd/compress/zstd_compress_literals.h b/lib/zstd/compress/zstd_compress_literals.h
new file mode 100644
index 0000000000..9775fb97cb
--- /dev/null
+++ b/lib/zstd/compress/zstd_compress_literals.h
@@ -0,0 +1,31 @@
+/*
+ * Copyright (c) Yann Collet, Facebook, Inc.
+ * All rights reserved.
+ *
+ * This source code is licensed under both the BSD-style license (found in the
+ * LICENSE file in the root directory of this source tree) and the GPLv2 (found
+ * in the COPYING file in the root directory of this source tree).
+ * You may select, at your option, one of the above-listed licenses.
+ */
+
+#ifndef ZSTD_COMPRESS_LITERALS_H
+#define ZSTD_COMPRESS_LITERALS_H
+
+#include "zstd_compress_internal.h" /* ZSTD_hufCTables_t, ZSTD_minGain() */
+
+
+size_t ZSTD_noCompressLiterals (void* dst, size_t dstCapacity, const void* src, size_t srcSize);
+
+size_t ZSTD_compressRleLiteralsBlock (void* dst, size_t dstCapacity, const void* src, size_t srcSize);
+
+/* If suspectUncompressible then some sampling checks will be run to potentially skip huffman coding */
+size_t ZSTD_compressLiterals (ZSTD_hufCTables_t const* prevHuf,
+ ZSTD_hufCTables_t* nextHuf,
+ ZSTD_strategy strategy, int disableLiteralCompression,
+ void* dst, size_t dstCapacity,
+ const void* src, size_t srcSize,
+ void* entropyWorkspace, size_t entropyWorkspaceSize,
+ const int bmi2,
+ unsigned suspectUncompressible);
+
+#endif /* ZSTD_COMPRESS_LITERALS_H */
diff --git a/lib/zstd/compress/zstd_compress_sequences.c b/lib/zstd/compress/zstd_compress_sequences.c
new file mode 100644
index 0000000000..21ddc1b37a
--- /dev/null
+++ b/lib/zstd/compress/zstd_compress_sequences.c
@@ -0,0 +1,442 @@
+/*
+ * Copyright (c) Yann Collet, Facebook, Inc.
+ * All rights reserved.
+ *
+ * This source code is licensed under both the BSD-style license (found in the
+ * LICENSE file in the root directory of this source tree) and the GPLv2 (found
+ * in the COPYING file in the root directory of this source tree).
+ * You may select, at your option, one of the above-listed licenses.
+ */
+
+ /*-*************************************
+ * Dependencies
+ ***************************************/
+#include "zstd_compress_sequences.h"
+
+/*
+ * -log2(x / 256) lookup table for x in [0, 256).
+ * If x == 0: Return 0
+ * Else: Return floor(-log2(x / 256) * 256)
+ */
+static unsigned const kInverseProbabilityLog256[256] = {
+ 0, 2048, 1792, 1642, 1536, 1453, 1386, 1329, 1280, 1236, 1197, 1162,
+ 1130, 1100, 1073, 1047, 1024, 1001, 980, 960, 941, 923, 906, 889,
+ 874, 859, 844, 830, 817, 804, 791, 779, 768, 756, 745, 734,
+ 724, 714, 704, 694, 685, 676, 667, 658, 650, 642, 633, 626,
+ 618, 610, 603, 595, 588, 581, 574, 567, 561, 554, 548, 542,
+ 535, 529, 523, 517, 512, 506, 500, 495, 489, 484, 478, 473,
+ 468, 463, 458, 453, 448, 443, 438, 434, 429, 424, 420, 415,
+ 411, 407, 402, 398, 394, 390, 386, 382, 377, 373, 370, 366,
+ 362, 358, 354, 350, 347, 343, 339, 336, 332, 329, 325, 322,
+ 318, 315, 311, 308, 305, 302, 298, 295, 292, 289, 286, 282,
+ 279, 276, 273, 270, 267, 264, 261, 258, 256, 253, 250, 247,
+ 244, 241, 239, 236, 233, 230, 228, 225, 222, 220, 217, 215,
+ 212, 209, 207, 204, 202, 199, 197, 194, 192, 190, 187, 185,
+ 182, 180, 178, 175, 173, 171, 168, 166, 164, 162, 159, 157,
+ 155, 153, 151, 149, 146, 144, 142, 140, 138, 136, 134, 132,
+ 130, 128, 126, 123, 121, 119, 117, 115, 114, 112, 110, 108,
+ 106, 104, 102, 100, 98, 96, 94, 93, 91, 89, 87, 85,
+ 83, 82, 80, 78, 76, 74, 73, 71, 69, 67, 66, 64,
+ 62, 61, 59, 57, 55, 54, 52, 50, 49, 47, 46, 44,
+ 42, 41, 39, 37, 36, 34, 33, 31, 30, 28, 26, 25,
+ 23, 22, 20, 19, 17, 16, 14, 13, 11, 10, 8, 7,
+ 5, 4, 2, 1,
+};
+
+static unsigned ZSTD_getFSEMaxSymbolValue(FSE_CTable const* ctable) {
+ void const* ptr = ctable;
+ U16 const* u16ptr = (U16 const*)ptr;
+ U32 const maxSymbolValue = MEM_read16(u16ptr + 1);
+ return maxSymbolValue;
+}
+
+/*
+ * Returns true if we should use ncount=-1 else we should
+ * use ncount=1 for low probability symbols instead.
+ */
+static unsigned ZSTD_useLowProbCount(size_t const nbSeq)
+{
+ /* Heuristic: This should cover most blocks <= 16K and
+ * start to fade out after 16K to about 32K depending on
+ * comprssibility.
+ */
+ return nbSeq >= 2048;
+}
+
+/*
+ * Returns the cost in bytes of encoding the normalized count header.
+ * Returns an error if any of the helper functions return an error.
+ */
+static size_t ZSTD_NCountCost(unsigned const* count, unsigned const max,
+ size_t const nbSeq, unsigned const FSELog)
+{
+ BYTE wksp[FSE_NCOUNTBOUND];
+ S16 norm[MaxSeq + 1];
+ const U32 tableLog = FSE_optimalTableLog(FSELog, nbSeq, max);
+ FORWARD_IF_ERROR(FSE_normalizeCount(norm, tableLog, count, nbSeq, max, ZSTD_useLowProbCount(nbSeq)), "");
+ return FSE_writeNCount(wksp, sizeof(wksp), norm, max, tableLog);
+}
+
+/*
+ * Returns the cost in bits of encoding the distribution described by count
+ * using the entropy bound.
+ */
+static size_t ZSTD_entropyCost(unsigned const* count, unsigned const max, size_t const total)
+{
+ unsigned cost = 0;
+ unsigned s;
+
+ assert(total > 0);
+ for (s = 0; s <= max; ++s) {
+ unsigned norm = (unsigned)((256 * count[s]) / total);
+ if (count[s] != 0 && norm == 0)
+ norm = 1;
+ assert(count[s] < total);
+ cost += count[s] * kInverseProbabilityLog256[norm];
+ }
+ return cost >> 8;
+}
+
+/*
+ * Returns the cost in bits of encoding the distribution in count using ctable.
+ * Returns an error if ctable cannot represent all the symbols in count.
+ */
+size_t ZSTD_fseBitCost(
+ FSE_CTable const* ctable,
+ unsigned const* count,
+ unsigned const max)
+{
+ unsigned const kAccuracyLog = 8;
+ size_t cost = 0;
+ unsigned s;
+ FSE_CState_t cstate;
+ FSE_initCState(&cstate, ctable);
+ if (ZSTD_getFSEMaxSymbolValue(ctable) < max) {
+ DEBUGLOG(5, "Repeat FSE_CTable has maxSymbolValue %u < %u",
+ ZSTD_getFSEMaxSymbolValue(ctable), max);
+ return ERROR(GENERIC);
+ }
+ for (s = 0; s <= max; ++s) {
+ unsigned const tableLog = cstate.stateLog;
+ unsigned const badCost = (tableLog + 1) << kAccuracyLog;
+ unsigned const bitCost = FSE_bitCost(cstate.symbolTT, tableLog, s, kAccuracyLog);
+ if (count[s] == 0)
+ continue;
+ if (bitCost >= badCost) {
+ DEBUGLOG(5, "Repeat FSE_CTable has Prob[%u] == 0", s);
+ return ERROR(GENERIC);
+ }
+ cost += (size_t)count[s] * bitCost;
+ }
+ return cost >> kAccuracyLog;
+}
+
+/*
+ * Returns the cost in bits of encoding the distribution in count using the
+ * table described by norm. The max symbol support by norm is assumed >= max.
+ * norm must be valid for every symbol with non-zero probability in count.
+ */
+size_t ZSTD_crossEntropyCost(short const* norm, unsigned accuracyLog,
+ unsigned const* count, unsigned const max)
+{
+ unsigned const shift = 8 - accuracyLog;
+ size_t cost = 0;
+ unsigned s;
+ assert(accuracyLog <= 8);
+ for (s = 0; s <= max; ++s) {
+ unsigned const normAcc = (norm[s] != -1) ? (unsigned)norm[s] : 1;
+ unsigned const norm256 = normAcc << shift;
+ assert(norm256 > 0);
+ assert(norm256 < 256);
+ cost += count[s] * kInverseProbabilityLog256[norm256];
+ }
+ return cost >> 8;
+}
+
+symbolEncodingType_e
+ZSTD_selectEncodingType(
+ FSE_repeat* repeatMode, unsigned const* count, unsigned const max,
+ size_t const mostFrequent, size_t nbSeq, unsigned const FSELog,
+ FSE_CTable const* prevCTable,
+ short const* defaultNorm, U32 defaultNormLog,
+ ZSTD_defaultPolicy_e const isDefaultAllowed,
+ ZSTD_strategy const strategy)
+{
+ ZSTD_STATIC_ASSERT(ZSTD_defaultDisallowed == 0 && ZSTD_defaultAllowed != 0);
+ if (mostFrequent == nbSeq) {
+ *repeatMode = FSE_repeat_none;
+ if (isDefaultAllowed && nbSeq <= 2) {
+ /* Prefer set_basic over set_rle when there are 2 or less symbols,
+ * since RLE uses 1 byte, but set_basic uses 5-6 bits per symbol.
+ * If basic encoding isn't possible, always choose RLE.
+ */
+ DEBUGLOG(5, "Selected set_basic");
+ return set_basic;
+ }
+ DEBUGLOG(5, "Selected set_rle");
+ return set_rle;
+ }
+ if (strategy < ZSTD_lazy) {
+ if (isDefaultAllowed) {
+ size_t const staticFse_nbSeq_max = 1000;
+ size_t const mult = 10 - strategy;
+ size_t const baseLog = 3;
+ size_t const dynamicFse_nbSeq_min = (((size_t)1 << defaultNormLog) * mult) >> baseLog; /* 28-36 for offset, 56-72 for lengths */
+ assert(defaultNormLog >= 5 && defaultNormLog <= 6); /* xx_DEFAULTNORMLOG */
+ assert(mult <= 9 && mult >= 7);
+ if ( (*repeatMode == FSE_repeat_valid)
+ && (nbSeq < staticFse_nbSeq_max) ) {
+ DEBUGLOG(5, "Selected set_repeat");
+ return set_repeat;
+ }
+ if ( (nbSeq < dynamicFse_nbSeq_min)
+ || (mostFrequent < (nbSeq >> (defaultNormLog-1))) ) {
+ DEBUGLOG(5, "Selected set_basic");
+ /* The format allows default tables to be repeated, but it isn't useful.
+ * When using simple heuristics to select encoding type, we don't want
+ * to confuse these tables with dictionaries. When running more careful
+ * analysis, we don't need to waste time checking both repeating tables
+ * and default tables.
+ */
+ *repeatMode = FSE_repeat_none;
+ return set_basic;
+ }
+ }
+ } else {
+ size_t const basicCost = isDefaultAllowed ? ZSTD_crossEntropyCost(defaultNorm, defaultNormLog, count, max) : ERROR(GENERIC);
+ size_t const repeatCost = *repeatMode != FSE_repeat_none ? ZSTD_fseBitCost(prevCTable, count, max) : ERROR(GENERIC);
+ size_t const NCountCost = ZSTD_NCountCost(count, max, nbSeq, FSELog);
+ size_t const compressedCost = (NCountCost << 3) + ZSTD_entropyCost(count, max, nbSeq);
+
+ if (isDefaultAllowed) {
+ assert(!ZSTD_isError(basicCost));
+ assert(!(*repeatMode == FSE_repeat_valid && ZSTD_isError(repeatCost)));
+ }
+ assert(!ZSTD_isError(NCountCost));
+ assert(compressedCost < ERROR(maxCode));
+ DEBUGLOG(5, "Estimated bit costs: basic=%u\trepeat=%u\tcompressed=%u",
+ (unsigned)basicCost, (unsigned)repeatCost, (unsigned)compressedCost);
+ if (basicCost <= repeatCost && basicCost <= compressedCost) {
+ DEBUGLOG(5, "Selected set_basic");
+ assert(isDefaultAllowed);
+ *repeatMode = FSE_repeat_none;
+ return set_basic;
+ }
+ if (repeatCost <= compressedCost) {
+ DEBUGLOG(5, "Selected set_repeat");
+ assert(!ZSTD_isError(repeatCost));
+ return set_repeat;
+ }
+ assert(compressedCost < basicCost && compressedCost < repeatCost);
+ }
+ DEBUGLOG(5, "Selected set_compressed");
+ *repeatMode = FSE_repeat_check;
+ return set_compressed;
+}
+
+typedef struct {
+ S16 norm[MaxSeq + 1];
+ U32 wksp[FSE_BUILD_CTABLE_WORKSPACE_SIZE_U32(MaxSeq, MaxFSELog)];
+} ZSTD_BuildCTableWksp;
+
+size_t
+ZSTD_buildCTable(void* dst, size_t dstCapacity,
+ FSE_CTable* nextCTable, U32 FSELog, symbolEncodingType_e type,
+ unsigned* count, U32 max,
+ const BYTE* codeTable, size_t nbSeq,
+ const S16* defaultNorm, U32 defaultNormLog, U32 defaultMax,
+ const FSE_CTable* prevCTable, size_t prevCTableSize,
+ void* entropyWorkspace, size_t entropyWorkspaceSize)
+{
+ BYTE* op = (BYTE*)dst;
+ const BYTE* const oend = op + dstCapacity;
+ DEBUGLOG(6, "ZSTD_buildCTable (dstCapacity=%u)", (unsigned)dstCapacity);
+
+ switch (type) {
+ case set_rle:
+ FORWARD_IF_ERROR(FSE_buildCTable_rle(nextCTable, (BYTE)max), "");
+ RETURN_ERROR_IF(dstCapacity==0, dstSize_tooSmall, "not enough space");
+ *op = codeTable[0];
+ return 1;
+ case set_repeat:
+ ZSTD_memcpy(nextCTable, prevCTable, prevCTableSize);
+ return 0;
+ case set_basic:
+ FORWARD_IF_ERROR(FSE_buildCTable_wksp(nextCTable, defaultNorm, defaultMax, defaultNormLog, entropyWorkspace, entropyWorkspaceSize), ""); /* note : could be pre-calculated */
+ return 0;
+ case set_compressed: {
+ ZSTD_BuildCTableWksp* wksp = (ZSTD_BuildCTableWksp*)entropyWorkspace;
+ size_t nbSeq_1 = nbSeq;
+ const U32 tableLog = FSE_optimalTableLog(FSELog, nbSeq, max);
+ if (count[codeTable[nbSeq-1]] > 1) {
+ count[codeTable[nbSeq-1]]--;
+ nbSeq_1--;
+ }
+ assert(nbSeq_1 > 1);
+ assert(entropyWorkspaceSize >= sizeof(ZSTD_BuildCTableWksp));
+ (void)entropyWorkspaceSize;
+ FORWARD_IF_ERROR(FSE_normalizeCount(wksp->norm, tableLog, count, nbSeq_1, max, ZSTD_useLowProbCount(nbSeq_1)), "FSE_normalizeCount failed");
+ assert(oend >= op);
+ { size_t const NCountSize = FSE_writeNCount(op, (size_t)(oend - op), wksp->norm, max, tableLog); /* overflow protected */
+ FORWARD_IF_ERROR(NCountSize, "FSE_writeNCount failed");
+ FORWARD_IF_ERROR(FSE_buildCTable_wksp(nextCTable, wksp->norm, max, tableLog, wksp->wksp, sizeof(wksp->wksp)), "FSE_buildCTable_wksp failed");
+ return NCountSize;
+ }
+ }
+ default: assert(0); RETURN_ERROR(GENERIC, "impossible to reach");
+ }
+}
+
+FORCE_INLINE_TEMPLATE size_t
+ZSTD_encodeSequences_body(
+ void* dst, size_t dstCapacity,
+ FSE_CTable const* CTable_MatchLength, BYTE const* mlCodeTable,
+ FSE_CTable const* CTable_OffsetBits, BYTE const* ofCodeTable,
+ FSE_CTable const* CTable_LitLength, BYTE const* llCodeTable,
+ seqDef const* sequences, size_t nbSeq, int longOffsets)
+{
+ BIT_CStream_t blockStream;
+ FSE_CState_t stateMatchLength;
+ FSE_CState_t stateOffsetBits;
+ FSE_CState_t stateLitLength;
+
+ RETURN_ERROR_IF(
+ ERR_isError(BIT_initCStream(&blockStream, dst, dstCapacity)),
+ dstSize_tooSmall, "not enough space remaining");
+ DEBUGLOG(6, "available space for bitstream : %i (dstCapacity=%u)",
+ (int)(blockStream.endPtr - blockStream.startPtr),
+ (unsigned)dstCapacity);
+
+ /* first symbols */
+ FSE_initCState2(&stateMatchLength, CTable_MatchLength, mlCodeTable[nbSeq-1]);
+ FSE_initCState2(&stateOffsetBits, CTable_OffsetBits, ofCodeTable[nbSeq-1]);
+ FSE_initCState2(&stateLitLength, CTable_LitLength, llCodeTable[nbSeq-1]);
+ BIT_addBits(&blockStream, sequences[nbSeq-1].litLength, LL_bits[llCodeTable[nbSeq-1]]);
+ if (MEM_32bits()) BIT_flushBits(&blockStream);
+ BIT_addBits(&blockStream, sequences[nbSeq-1].mlBase, ML_bits[mlCodeTable[nbSeq-1]]);
+ if (MEM_32bits()) BIT_flushBits(&blockStream);
+ if (longOffsets) {
+ U32 const ofBits = ofCodeTable[nbSeq-1];
+ unsigned const extraBits = ofBits - MIN(ofBits, STREAM_ACCUMULATOR_MIN-1);
+ if (extraBits) {
+ BIT_addBits(&blockStream, sequences[nbSeq-1].offBase, extraBits);
+ BIT_flushBits(&blockStream);
+ }
+ BIT_addBits(&blockStream, sequences[nbSeq-1].offBase >> extraBits,
+ ofBits - extraBits);
+ } else {
+ BIT_addBits(&blockStream, sequences[nbSeq-1].offBase, ofCodeTable[nbSeq-1]);
+ }
+ BIT_flushBits(&blockStream);
+
+ { size_t n;
+ for (n=nbSeq-2 ; n<nbSeq ; n--) { /* intentional underflow */
+ BYTE const llCode = llCodeTable[n];
+ BYTE const ofCode = ofCodeTable[n];
+ BYTE const mlCode = mlCodeTable[n];
+ U32 const llBits = LL_bits[llCode];
+ U32 const ofBits = ofCode;
+ U32 const mlBits = ML_bits[mlCode];
+ DEBUGLOG(6, "encoding: litlen:%2u - matchlen:%2u - offCode:%7u",
+ (unsigned)sequences[n].litLength,
+ (unsigned)sequences[n].mlBase + MINMATCH,
+ (unsigned)sequences[n].offBase);
+ /* 32b*/ /* 64b*/
+ /* (7)*/ /* (7)*/
+ FSE_encodeSymbol(&blockStream, &stateOffsetBits, ofCode); /* 15 */ /* 15 */
+ FSE_encodeSymbol(&blockStream, &stateMatchLength, mlCode); /* 24 */ /* 24 */
+ if (MEM_32bits()) BIT_flushBits(&blockStream); /* (7)*/
+ FSE_encodeSymbol(&blockStream, &stateLitLength, llCode); /* 16 */ /* 33 */
+ if (MEM_32bits() || (ofBits+mlBits+llBits >= 64-7-(LLFSELog+MLFSELog+OffFSELog)))
+ BIT_flushBits(&blockStream); /* (7)*/
+ BIT_addBits(&blockStream, sequences[n].litLength, llBits);
+ if (MEM_32bits() && ((llBits+mlBits)>24)) BIT_flushBits(&blockStream);
+ BIT_addBits(&blockStream, sequences[n].mlBase, mlBits);
+ if (MEM_32bits() || (ofBits+mlBits+llBits > 56)) BIT_flushBits(&blockStream);
+ if (longOffsets) {
+ unsigned const extraBits = ofBits - MIN(ofBits, STREAM_ACCUMULATOR_MIN-1);
+ if (extraBits) {
+ BIT_addBits(&blockStream, sequences[n].offBase, extraBits);
+ BIT_flushBits(&blockStream); /* (7)*/
+ }
+ BIT_addBits(&blockStream, sequences[n].offBase >> extraBits,
+ ofBits - extraBits); /* 31 */
+ } else {
+ BIT_addBits(&blockStream, sequences[n].offBase, ofBits); /* 31 */
+ }
+ BIT_flushBits(&blockStream); /* (7)*/
+ DEBUGLOG(7, "remaining space : %i", (int)(blockStream.endPtr - blockStream.ptr));
+ } }
+
+ DEBUGLOG(6, "ZSTD_encodeSequences: flushing ML state with %u bits", stateMatchLength.stateLog);
+ FSE_flushCState(&blockStream, &stateMatchLength);
+ DEBUGLOG(6, "ZSTD_encodeSequences: flushing Off state with %u bits", stateOffsetBits.stateLog);
+ FSE_flushCState(&blockStream, &stateOffsetBits);
+ DEBUGLOG(6, "ZSTD_encodeSequences: flushing LL state with %u bits", stateLitLength.stateLog);
+ FSE_flushCState(&blockStream, &stateLitLength);
+
+ { size_t const streamSize = BIT_closeCStream(&blockStream);
+ RETURN_ERROR_IF(streamSize==0, dstSize_tooSmall, "not enough space");
+ return streamSize;
+ }
+}
+
+static size_t
+ZSTD_encodeSequences_default(
+ void* dst, size_t dstCapacity,
+ FSE_CTable const* CTable_MatchLength, BYTE const* mlCodeTable,
+ FSE_CTable const* CTable_OffsetBits, BYTE const* ofCodeTable,
+ FSE_CTable const* CTable_LitLength, BYTE const* llCodeTable,
+ seqDef const* sequences, size_t nbSeq, int longOffsets)
+{
+ return ZSTD_encodeSequences_body(dst, dstCapacity,
+ CTable_MatchLength, mlCodeTable,
+ CTable_OffsetBits, ofCodeTable,
+ CTable_LitLength, llCodeTable,
+ sequences, nbSeq, longOffsets);
+}
+
+
+#if DYNAMIC_BMI2
+
+static BMI2_TARGET_ATTRIBUTE size_t
+ZSTD_encodeSequences_bmi2(
+ void* dst, size_t dstCapacity,
+ FSE_CTable const* CTable_MatchLength, BYTE const* mlCodeTable,
+ FSE_CTable const* CTable_OffsetBits, BYTE const* ofCodeTable,
+ FSE_CTable const* CTable_LitLength, BYTE const* llCodeTable,
+ seqDef const* sequences, size_t nbSeq, int longOffsets)
+{
+ return ZSTD_encodeSequences_body(dst, dstCapacity,
+ CTable_MatchLength, mlCodeTable,
+ CTable_OffsetBits, ofCodeTable,
+ CTable_LitLength, llCodeTable,
+ sequences, nbSeq, longOffsets);
+}
+
+#endif
+
+size_t ZSTD_encodeSequences(
+ void* dst, size_t dstCapacity,
+ FSE_CTable const* CTable_MatchLength, BYTE const* mlCodeTable,
+ FSE_CTable const* CTable_OffsetBits, BYTE const* ofCodeTable,
+ FSE_CTable const* CTable_LitLength, BYTE const* llCodeTable,
+ seqDef const* sequences, size_t nbSeq, int longOffsets, int bmi2)
+{
+ DEBUGLOG(5, "ZSTD_encodeSequences: dstCapacity = %u", (unsigned)dstCapacity);
+#if DYNAMIC_BMI2
+ if (bmi2) {
+ return ZSTD_encodeSequences_bmi2(dst, dstCapacity,
+ CTable_MatchLength, mlCodeTable,
+ CTable_OffsetBits, ofCodeTable,
+ CTable_LitLength, llCodeTable,
+ sequences, nbSeq, longOffsets);
+ }
+#endif
+ (void)bmi2;
+ return ZSTD_encodeSequences_default(dst, dstCapacity,
+ CTable_MatchLength, mlCodeTable,
+ CTable_OffsetBits, ofCodeTable,
+ CTable_LitLength, llCodeTable,
+ sequences, nbSeq, longOffsets);
+}
diff --git a/lib/zstd/compress/zstd_compress_sequences.h b/lib/zstd/compress/zstd_compress_sequences.h
new file mode 100644
index 0000000000..7991364c2f
--- /dev/null
+++ b/lib/zstd/compress/zstd_compress_sequences.h
@@ -0,0 +1,54 @@
+/*
+ * Copyright (c) Yann Collet, Facebook, Inc.
+ * All rights reserved.
+ *
+ * This source code is licensed under both the BSD-style license (found in the
+ * LICENSE file in the root directory of this source tree) and the GPLv2 (found
+ * in the COPYING file in the root directory of this source tree).
+ * You may select, at your option, one of the above-listed licenses.
+ */
+
+#ifndef ZSTD_COMPRESS_SEQUENCES_H
+#define ZSTD_COMPRESS_SEQUENCES_H
+
+#include "../common/fse.h" /* FSE_repeat, FSE_CTable */
+#include "../common/zstd_internal.h" /* symbolEncodingType_e, ZSTD_strategy */
+
+typedef enum {
+ ZSTD_defaultDisallowed = 0,
+ ZSTD_defaultAllowed = 1
+} ZSTD_defaultPolicy_e;
+
+symbolEncodingType_e
+ZSTD_selectEncodingType(
+ FSE_repeat* repeatMode, unsigned const* count, unsigned const max,
+ size_t const mostFrequent, size_t nbSeq, unsigned const FSELog,
+ FSE_CTable const* prevCTable,
+ short const* defaultNorm, U32 defaultNormLog,
+ ZSTD_defaultPolicy_e const isDefaultAllowed,
+ ZSTD_strategy const strategy);
+
+size_t
+ZSTD_buildCTable(void* dst, size_t dstCapacity,
+ FSE_CTable* nextCTable, U32 FSELog, symbolEncodingType_e type,
+ unsigned* count, U32 max,
+ const BYTE* codeTable, size_t nbSeq,
+ const S16* defaultNorm, U32 defaultNormLog, U32 defaultMax,
+ const FSE_CTable* prevCTable, size_t prevCTableSize,
+ void* entropyWorkspace, size_t entropyWorkspaceSize);
+
+size_t ZSTD_encodeSequences(
+ void* dst, size_t dstCapacity,
+ FSE_CTable const* CTable_MatchLength, BYTE const* mlCodeTable,
+ FSE_CTable const* CTable_OffsetBits, BYTE const* ofCodeTable,
+ FSE_CTable const* CTable_LitLength, BYTE const* llCodeTable,
+ seqDef const* sequences, size_t nbSeq, int longOffsets, int bmi2);
+
+size_t ZSTD_fseBitCost(
+ FSE_CTable const* ctable,
+ unsigned const* count,
+ unsigned const max);
+
+size_t ZSTD_crossEntropyCost(short const* norm, unsigned accuracyLog,
+ unsigned const* count, unsigned const max);
+#endif /* ZSTD_COMPRESS_SEQUENCES_H */
diff --git a/lib/zstd/compress/zstd_compress_superblock.c b/lib/zstd/compress/zstd_compress_superblock.c
new file mode 100644
index 0000000000..17d836cc84
--- /dev/null
+++ b/lib/zstd/compress/zstd_compress_superblock.c
@@ -0,0 +1,573 @@
+/*
+ * Copyright (c) Yann Collet, Facebook, Inc.
+ * All rights reserved.
+ *
+ * This source code is licensed under both the BSD-style license (found in the
+ * LICENSE file in the root directory of this source tree) and the GPLv2 (found
+ * in the COPYING file in the root directory of this source tree).
+ * You may select, at your option, one of the above-listed licenses.
+ */
+
+ /*-*************************************
+ * Dependencies
+ ***************************************/
+#include "zstd_compress_superblock.h"
+
+#include "../common/zstd_internal.h" /* ZSTD_getSequenceLength */
+#include "hist.h" /* HIST_countFast_wksp */
+#include "zstd_compress_internal.h" /* ZSTD_[huf|fse|entropy]CTablesMetadata_t */
+#include "zstd_compress_sequences.h"
+#include "zstd_compress_literals.h"
+
+/* ZSTD_compressSubBlock_literal() :
+ * Compresses literals section for a sub-block.
+ * When we have to write the Huffman table we will sometimes choose a header
+ * size larger than necessary. This is because we have to pick the header size
+ * before we know the table size + compressed size, so we have a bound on the
+ * table size. If we guessed incorrectly, we fall back to uncompressed literals.
+ *
+ * We write the header when writeEntropy=1 and set entropyWritten=1 when we succeeded
+ * in writing the header, otherwise it is set to 0.
+ *
+ * hufMetadata->hType has literals block type info.
+ * If it is set_basic, all sub-blocks literals section will be Raw_Literals_Block.
+ * If it is set_rle, all sub-blocks literals section will be RLE_Literals_Block.
+ * If it is set_compressed, first sub-block's literals section will be Compressed_Literals_Block
+ * If it is set_compressed, first sub-block's literals section will be Treeless_Literals_Block
+ * and the following sub-blocks' literals sections will be Treeless_Literals_Block.
+ * @return : compressed size of literals section of a sub-block
+ * Or 0 if it unable to compress.
+ * Or error code */
+static size_t ZSTD_compressSubBlock_literal(const HUF_CElt* hufTable,
+ const ZSTD_hufCTablesMetadata_t* hufMetadata,
+ const BYTE* literals, size_t litSize,
+ void* dst, size_t dstSize,
+ const int bmi2, int writeEntropy, int* entropyWritten)
+{
+ size_t const header = writeEntropy ? 200 : 0;
+ size_t const lhSize = 3 + (litSize >= (1 KB - header)) + (litSize >= (16 KB - header));
+ BYTE* const ostart = (BYTE*)dst;
+ BYTE* const oend = ostart + dstSize;
+ BYTE* op = ostart + lhSize;
+ U32 const singleStream = lhSize == 3;
+ symbolEncodingType_e hType = writeEntropy ? hufMetadata->hType : set_repeat;
+ size_t cLitSize = 0;
+
+ (void)bmi2; /* TODO bmi2... */
+
+ DEBUGLOG(5, "ZSTD_compressSubBlock_literal (litSize=%zu, lhSize=%zu, writeEntropy=%d)", litSize, lhSize, writeEntropy);
+
+ *entropyWritten = 0;
+ if (litSize == 0 || hufMetadata->hType == set_basic) {
+ DEBUGLOG(5, "ZSTD_compressSubBlock_literal using raw literal");
+ return ZSTD_noCompressLiterals(dst, dstSize, literals, litSize);
+ } else if (hufMetadata->hType == set_rle) {
+ DEBUGLOG(5, "ZSTD_compressSubBlock_literal using rle literal");
+ return ZSTD_compressRleLiteralsBlock(dst, dstSize, literals, litSize);
+ }
+
+ assert(litSize > 0);
+ assert(hufMetadata->hType == set_compressed || hufMetadata->hType == set_repeat);
+
+ if (writeEntropy && hufMetadata->hType == set_compressed) {
+ ZSTD_memcpy(op, hufMetadata->hufDesBuffer, hufMetadata->hufDesSize);
+ op += hufMetadata->hufDesSize;
+ cLitSize += hufMetadata->hufDesSize;
+ DEBUGLOG(5, "ZSTD_compressSubBlock_literal (hSize=%zu)", hufMetadata->hufDesSize);
+ }
+
+ /* TODO bmi2 */
+ { const size_t cSize = singleStream ? HUF_compress1X_usingCTable(op, oend-op, literals, litSize, hufTable)
+ : HUF_compress4X_usingCTable(op, oend-op, literals, litSize, hufTable);
+ op += cSize;
+ cLitSize += cSize;
+ if (cSize == 0 || ERR_isError(cSize)) {
+ DEBUGLOG(5, "Failed to write entropy tables %s", ZSTD_getErrorName(cSize));
+ return 0;
+ }
+ /* If we expand and we aren't writing a header then emit uncompressed */
+ if (!writeEntropy && cLitSize >= litSize) {
+ DEBUGLOG(5, "ZSTD_compressSubBlock_literal using raw literal because uncompressible");
+ return ZSTD_noCompressLiterals(dst, dstSize, literals, litSize);
+ }
+ /* If we are writing headers then allow expansion that doesn't change our header size. */
+ if (lhSize < (size_t)(3 + (cLitSize >= 1 KB) + (cLitSize >= 16 KB))) {
+ assert(cLitSize > litSize);
+ DEBUGLOG(5, "Literals expanded beyond allowed header size");
+ return ZSTD_noCompressLiterals(dst, dstSize, literals, litSize);
+ }
+ DEBUGLOG(5, "ZSTD_compressSubBlock_literal (cSize=%zu)", cSize);
+ }
+
+ /* Build header */
+ switch(lhSize)
+ {
+ case 3: /* 2 - 2 - 10 - 10 */
+ { U32 const lhc = hType + ((!singleStream) << 2) + ((U32)litSize<<4) + ((U32)cLitSize<<14);
+ MEM_writeLE24(ostart, lhc);
+ break;
+ }
+ case 4: /* 2 - 2 - 14 - 14 */
+ { U32 const lhc = hType + (2 << 2) + ((U32)litSize<<4) + ((U32)cLitSize<<18);
+ MEM_writeLE32(ostart, lhc);
+ break;
+ }
+ case 5: /* 2 - 2 - 18 - 18 */
+ { U32 const lhc = hType + (3 << 2) + ((U32)litSize<<4) + ((U32)cLitSize<<22);
+ MEM_writeLE32(ostart, lhc);
+ ostart[4] = (BYTE)(cLitSize >> 10);
+ break;
+ }
+ default: /* not possible : lhSize is {3,4,5} */
+ assert(0);
+ }
+ *entropyWritten = 1;
+ DEBUGLOG(5, "Compressed literals: %u -> %u", (U32)litSize, (U32)(op-ostart));
+ return op-ostart;
+}
+
+static size_t ZSTD_seqDecompressedSize(seqStore_t const* seqStore, const seqDef* sequences, size_t nbSeq, size_t litSize, int lastSequence) {
+ const seqDef* const sstart = sequences;
+ const seqDef* const send = sequences + nbSeq;
+ const seqDef* sp = sstart;
+ size_t matchLengthSum = 0;
+ size_t litLengthSum = 0;
+ (void)(litLengthSum); /* suppress unused variable warning on some environments */
+ while (send-sp > 0) {
+ ZSTD_sequenceLength const seqLen = ZSTD_getSequenceLength(seqStore, sp);
+ litLengthSum += seqLen.litLength;
+ matchLengthSum += seqLen.matchLength;
+ sp++;
+ }
+ assert(litLengthSum <= litSize);
+ if (!lastSequence) {
+ assert(litLengthSum == litSize);
+ }
+ return matchLengthSum + litSize;
+}
+
+/* ZSTD_compressSubBlock_sequences() :
+ * Compresses sequences section for a sub-block.
+ * fseMetadata->llType, fseMetadata->ofType, and fseMetadata->mlType have
+ * symbol compression modes for the super-block.
+ * The first successfully compressed block will have these in its header.
+ * We set entropyWritten=1 when we succeed in compressing the sequences.
+ * The following sub-blocks will always have repeat mode.
+ * @return : compressed size of sequences section of a sub-block
+ * Or 0 if it is unable to compress
+ * Or error code. */
+static size_t ZSTD_compressSubBlock_sequences(const ZSTD_fseCTables_t* fseTables,
+ const ZSTD_fseCTablesMetadata_t* fseMetadata,
+ const seqDef* sequences, size_t nbSeq,
+ const BYTE* llCode, const BYTE* mlCode, const BYTE* ofCode,
+ const ZSTD_CCtx_params* cctxParams,
+ void* dst, size_t dstCapacity,
+ const int bmi2, int writeEntropy, int* entropyWritten)
+{
+ const int longOffsets = cctxParams->cParams.windowLog > STREAM_ACCUMULATOR_MIN;
+ BYTE* const ostart = (BYTE*)dst;
+ BYTE* const oend = ostart + dstCapacity;
+ BYTE* op = ostart;
+ BYTE* seqHead;
+
+ DEBUGLOG(5, "ZSTD_compressSubBlock_sequences (nbSeq=%zu, writeEntropy=%d, longOffsets=%d)", nbSeq, writeEntropy, longOffsets);
+
+ *entropyWritten = 0;
+ /* Sequences Header */
+ RETURN_ERROR_IF((oend-op) < 3 /*max nbSeq Size*/ + 1 /*seqHead*/,
+ dstSize_tooSmall, "");
+ if (nbSeq < 0x7F)
+ *op++ = (BYTE)nbSeq;
+ else if (nbSeq < LONGNBSEQ)
+ op[0] = (BYTE)((nbSeq>>8) + 0x80), op[1] = (BYTE)nbSeq, op+=2;
+ else
+ op[0]=0xFF, MEM_writeLE16(op+1, (U16)(nbSeq - LONGNBSEQ)), op+=3;
+ if (nbSeq==0) {
+ return op - ostart;
+ }
+
+ /* seqHead : flags for FSE encoding type */
+ seqHead = op++;
+
+ DEBUGLOG(5, "ZSTD_compressSubBlock_sequences (seqHeadSize=%u)", (unsigned)(op-ostart));
+
+ if (writeEntropy) {
+ const U32 LLtype = fseMetadata->llType;
+ const U32 Offtype = fseMetadata->ofType;
+ const U32 MLtype = fseMetadata->mlType;
+ DEBUGLOG(5, "ZSTD_compressSubBlock_sequences (fseTablesSize=%zu)", fseMetadata->fseTablesSize);
+ *seqHead = (BYTE)((LLtype<<6) + (Offtype<<4) + (MLtype<<2));
+ ZSTD_memcpy(op, fseMetadata->fseTablesBuffer, fseMetadata->fseTablesSize);
+ op += fseMetadata->fseTablesSize;
+ } else {
+ const U32 repeat = set_repeat;
+ *seqHead = (BYTE)((repeat<<6) + (repeat<<4) + (repeat<<2));
+ }
+
+ { size_t const bitstreamSize = ZSTD_encodeSequences(
+ op, oend - op,
+ fseTables->matchlengthCTable, mlCode,
+ fseTables->offcodeCTable, ofCode,
+ fseTables->litlengthCTable, llCode,
+ sequences, nbSeq,
+ longOffsets, bmi2);
+ FORWARD_IF_ERROR(bitstreamSize, "ZSTD_encodeSequences failed");
+ op += bitstreamSize;
+ /* zstd versions <= 1.3.4 mistakenly report corruption when
+ * FSE_readNCount() receives a buffer < 4 bytes.
+ * Fixed by https://github.com/facebook/zstd/pull/1146.
+ * This can happen when the last set_compressed table present is 2
+ * bytes and the bitstream is only one byte.
+ * In this exceedingly rare case, we will simply emit an uncompressed
+ * block, since it isn't worth optimizing.
+ */
+#ifndef FUZZING_BUILD_MODE_UNSAFE_FOR_PRODUCTION
+ if (writeEntropy && fseMetadata->lastCountSize && fseMetadata->lastCountSize + bitstreamSize < 4) {
+ /* NCountSize >= 2 && bitstreamSize > 0 ==> lastCountSize == 3 */
+ assert(fseMetadata->lastCountSize + bitstreamSize == 3);
+ DEBUGLOG(5, "Avoiding bug in zstd decoder in versions <= 1.3.4 by "
+ "emitting an uncompressed block.");
+ return 0;
+ }
+#endif
+ DEBUGLOG(5, "ZSTD_compressSubBlock_sequences (bitstreamSize=%zu)", bitstreamSize);
+ }
+
+ /* zstd versions <= 1.4.0 mistakenly report error when
+ * sequences section body size is less than 3 bytes.
+ * Fixed by https://github.com/facebook/zstd/pull/1664.
+ * This can happen when the previous sequences section block is compressed
+ * with rle mode and the current block's sequences section is compressed
+ * with repeat mode where sequences section body size can be 1 byte.
+ */
+#ifndef FUZZING_BUILD_MODE_UNSAFE_FOR_PRODUCTION
+ if (op-seqHead < 4) {
+ DEBUGLOG(5, "Avoiding bug in zstd decoder in versions <= 1.4.0 by emitting "
+ "an uncompressed block when sequences are < 4 bytes");
+ return 0;
+ }
+#endif
+
+ *entropyWritten = 1;
+ return op - ostart;
+}
+
+/* ZSTD_compressSubBlock() :
+ * Compresses a single sub-block.
+ * @return : compressed size of the sub-block
+ * Or 0 if it failed to compress. */
+static size_t ZSTD_compressSubBlock(const ZSTD_entropyCTables_t* entropy,
+ const ZSTD_entropyCTablesMetadata_t* entropyMetadata,
+ const seqDef* sequences, size_t nbSeq,
+ const BYTE* literals, size_t litSize,
+ const BYTE* llCode, const BYTE* mlCode, const BYTE* ofCode,
+ const ZSTD_CCtx_params* cctxParams,
+ void* dst, size_t dstCapacity,
+ const int bmi2,
+ int writeLitEntropy, int writeSeqEntropy,
+ int* litEntropyWritten, int* seqEntropyWritten,
+ U32 lastBlock)
+{
+ BYTE* const ostart = (BYTE*)dst;
+ BYTE* const oend = ostart + dstCapacity;
+ BYTE* op = ostart + ZSTD_blockHeaderSize;
+ DEBUGLOG(5, "ZSTD_compressSubBlock (litSize=%zu, nbSeq=%zu, writeLitEntropy=%d, writeSeqEntropy=%d, lastBlock=%d)",
+ litSize, nbSeq, writeLitEntropy, writeSeqEntropy, lastBlock);
+ { size_t cLitSize = ZSTD_compressSubBlock_literal((const HUF_CElt*)entropy->huf.CTable,
+ &entropyMetadata->hufMetadata, literals, litSize,
+ op, oend-op, bmi2, writeLitEntropy, litEntropyWritten);
+ FORWARD_IF_ERROR(cLitSize, "ZSTD_compressSubBlock_literal failed");
+ if (cLitSize == 0) return 0;
+ op += cLitSize;
+ }
+ { size_t cSeqSize = ZSTD_compressSubBlock_sequences(&entropy->fse,
+ &entropyMetadata->fseMetadata,
+ sequences, nbSeq,
+ llCode, mlCode, ofCode,
+ cctxParams,
+ op, oend-op,
+ bmi2, writeSeqEntropy, seqEntropyWritten);
+ FORWARD_IF_ERROR(cSeqSize, "ZSTD_compressSubBlock_sequences failed");
+ if (cSeqSize == 0) return 0;
+ op += cSeqSize;
+ }
+ /* Write block header */
+ { size_t cSize = (op-ostart)-ZSTD_blockHeaderSize;
+ U32 const cBlockHeader24 = lastBlock + (((U32)bt_compressed)<<1) + (U32)(cSize << 3);
+ MEM_writeLE24(ostart, cBlockHeader24);
+ }
+ return op-ostart;
+}
+
+static size_t ZSTD_estimateSubBlockSize_literal(const BYTE* literals, size_t litSize,
+ const ZSTD_hufCTables_t* huf,
+ const ZSTD_hufCTablesMetadata_t* hufMetadata,
+ void* workspace, size_t wkspSize,
+ int writeEntropy)
+{
+ unsigned* const countWksp = (unsigned*)workspace;
+ unsigned maxSymbolValue = 255;
+ size_t literalSectionHeaderSize = 3; /* Use hard coded size of 3 bytes */
+
+ if (hufMetadata->hType == set_basic) return litSize;
+ else if (hufMetadata->hType == set_rle) return 1;
+ else if (hufMetadata->hType == set_compressed || hufMetadata->hType == set_repeat) {
+ size_t const largest = HIST_count_wksp (countWksp, &maxSymbolValue, (const BYTE*)literals, litSize, workspace, wkspSize);
+ if (ZSTD_isError(largest)) return litSize;
+ { size_t cLitSizeEstimate = HUF_estimateCompressedSize((const HUF_CElt*)huf->CTable, countWksp, maxSymbolValue);
+ if (writeEntropy) cLitSizeEstimate += hufMetadata->hufDesSize;
+ return cLitSizeEstimate + literalSectionHeaderSize;
+ } }
+ assert(0); /* impossible */
+ return 0;
+}
+
+static size_t ZSTD_estimateSubBlockSize_symbolType(symbolEncodingType_e type,
+ const BYTE* codeTable, unsigned maxCode,
+ size_t nbSeq, const FSE_CTable* fseCTable,
+ const U8* additionalBits,
+ short const* defaultNorm, U32 defaultNormLog, U32 defaultMax,
+ void* workspace, size_t wkspSize)
+{
+ unsigned* const countWksp = (unsigned*)workspace;
+ const BYTE* ctp = codeTable;
+ const BYTE* const ctStart = ctp;
+ const BYTE* const ctEnd = ctStart + nbSeq;
+ size_t cSymbolTypeSizeEstimateInBits = 0;
+ unsigned max = maxCode;
+
+ HIST_countFast_wksp(countWksp, &max, codeTable, nbSeq, workspace, wkspSize); /* can't fail */
+ if (type == set_basic) {
+ /* We selected this encoding type, so it must be valid. */
+ assert(max <= defaultMax);
+ cSymbolTypeSizeEstimateInBits = max <= defaultMax
+ ? ZSTD_crossEntropyCost(defaultNorm, defaultNormLog, countWksp, max)
+ : ERROR(GENERIC);
+ } else if (type == set_rle) {
+ cSymbolTypeSizeEstimateInBits = 0;
+ } else if (type == set_compressed || type == set_repeat) {
+ cSymbolTypeSizeEstimateInBits = ZSTD_fseBitCost(fseCTable, countWksp, max);
+ }
+ if (ZSTD_isError(cSymbolTypeSizeEstimateInBits)) return nbSeq * 10;
+ while (ctp < ctEnd) {
+ if (additionalBits) cSymbolTypeSizeEstimateInBits += additionalBits[*ctp];
+ else cSymbolTypeSizeEstimateInBits += *ctp; /* for offset, offset code is also the number of additional bits */
+ ctp++;
+ }
+ return cSymbolTypeSizeEstimateInBits / 8;
+}
+
+static size_t ZSTD_estimateSubBlockSize_sequences(const BYTE* ofCodeTable,
+ const BYTE* llCodeTable,
+ const BYTE* mlCodeTable,
+ size_t nbSeq,
+ const ZSTD_fseCTables_t* fseTables,
+ const ZSTD_fseCTablesMetadata_t* fseMetadata,
+ void* workspace, size_t wkspSize,
+ int writeEntropy)
+{
+ size_t const sequencesSectionHeaderSize = 3; /* Use hard coded size of 3 bytes */
+ size_t cSeqSizeEstimate = 0;
+ if (nbSeq == 0) return sequencesSectionHeaderSize;
+ cSeqSizeEstimate += ZSTD_estimateSubBlockSize_symbolType(fseMetadata->ofType, ofCodeTable, MaxOff,
+ nbSeq, fseTables->offcodeCTable, NULL,
+ OF_defaultNorm, OF_defaultNormLog, DefaultMaxOff,
+ workspace, wkspSize);
+ cSeqSizeEstimate += ZSTD_estimateSubBlockSize_symbolType(fseMetadata->llType, llCodeTable, MaxLL,
+ nbSeq, fseTables->litlengthCTable, LL_bits,
+ LL_defaultNorm, LL_defaultNormLog, MaxLL,
+ workspace, wkspSize);
+ cSeqSizeEstimate += ZSTD_estimateSubBlockSize_symbolType(fseMetadata->mlType, mlCodeTable, MaxML,
+ nbSeq, fseTables->matchlengthCTable, ML_bits,
+ ML_defaultNorm, ML_defaultNormLog, MaxML,
+ workspace, wkspSize);
+ if (writeEntropy) cSeqSizeEstimate += fseMetadata->fseTablesSize;
+ return cSeqSizeEstimate + sequencesSectionHeaderSize;
+}
+
+static size_t ZSTD_estimateSubBlockSize(const BYTE* literals, size_t litSize,
+ const BYTE* ofCodeTable,
+ const BYTE* llCodeTable,
+ const BYTE* mlCodeTable,
+ size_t nbSeq,
+ const ZSTD_entropyCTables_t* entropy,
+ const ZSTD_entropyCTablesMetadata_t* entropyMetadata,
+ void* workspace, size_t wkspSize,
+ int writeLitEntropy, int writeSeqEntropy) {
+ size_t cSizeEstimate = 0;
+ cSizeEstimate += ZSTD_estimateSubBlockSize_literal(literals, litSize,
+ &entropy->huf, &entropyMetadata->hufMetadata,
+ workspace, wkspSize, writeLitEntropy);
+ cSizeEstimate += ZSTD_estimateSubBlockSize_sequences(ofCodeTable, llCodeTable, mlCodeTable,
+ nbSeq, &entropy->fse, &entropyMetadata->fseMetadata,
+ workspace, wkspSize, writeSeqEntropy);
+ return cSizeEstimate + ZSTD_blockHeaderSize;
+}
+
+static int ZSTD_needSequenceEntropyTables(ZSTD_fseCTablesMetadata_t const* fseMetadata)
+{
+ if (fseMetadata->llType == set_compressed || fseMetadata->llType == set_rle)
+ return 1;
+ if (fseMetadata->mlType == set_compressed || fseMetadata->mlType == set_rle)
+ return 1;
+ if (fseMetadata->ofType == set_compressed || fseMetadata->ofType == set_rle)
+ return 1;
+ return 0;
+}
+
+/* ZSTD_compressSubBlock_multi() :
+ * Breaks super-block into multiple sub-blocks and compresses them.
+ * Entropy will be written to the first block.
+ * The following blocks will use repeat mode to compress.
+ * All sub-blocks are compressed blocks (no raw or rle blocks).
+ * @return : compressed size of the super block (which is multiple ZSTD blocks)
+ * Or 0 if it failed to compress. */
+static size_t ZSTD_compressSubBlock_multi(const seqStore_t* seqStorePtr,
+ const ZSTD_compressedBlockState_t* prevCBlock,
+ ZSTD_compressedBlockState_t* nextCBlock,
+ const ZSTD_entropyCTablesMetadata_t* entropyMetadata,
+ const ZSTD_CCtx_params* cctxParams,
+ void* dst, size_t dstCapacity,
+ const void* src, size_t srcSize,
+ const int bmi2, U32 lastBlock,
+ void* workspace, size_t wkspSize)
+{
+ const seqDef* const sstart = seqStorePtr->sequencesStart;
+ const seqDef* const send = seqStorePtr->sequences;
+ const seqDef* sp = sstart;
+ const BYTE* const lstart = seqStorePtr->litStart;
+ const BYTE* const lend = seqStorePtr->lit;
+ const BYTE* lp = lstart;
+ BYTE const* ip = (BYTE const*)src;
+ BYTE const* const iend = ip + srcSize;
+ BYTE* const ostart = (BYTE*)dst;
+ BYTE* const oend = ostart + dstCapacity;
+ BYTE* op = ostart;
+ const BYTE* llCodePtr = seqStorePtr->llCode;
+ const BYTE* mlCodePtr = seqStorePtr->mlCode;
+ const BYTE* ofCodePtr = seqStorePtr->ofCode;
+ size_t targetCBlockSize = cctxParams->targetCBlockSize;
+ size_t litSize, seqCount;
+ int writeLitEntropy = entropyMetadata->hufMetadata.hType == set_compressed;
+ int writeSeqEntropy = 1;
+ int lastSequence = 0;
+
+ DEBUGLOG(5, "ZSTD_compressSubBlock_multi (litSize=%u, nbSeq=%u)",
+ (unsigned)(lend-lp), (unsigned)(send-sstart));
+
+ litSize = 0;
+ seqCount = 0;
+ do {
+ size_t cBlockSizeEstimate = 0;
+ if (sstart == send) {
+ lastSequence = 1;
+ } else {
+ const seqDef* const sequence = sp + seqCount;
+ lastSequence = sequence == send - 1;
+ litSize += ZSTD_getSequenceLength(seqStorePtr, sequence).litLength;
+ seqCount++;
+ }
+ if (lastSequence) {
+ assert(lp <= lend);
+ assert(litSize <= (size_t)(lend - lp));
+ litSize = (size_t)(lend - lp);
+ }
+ /* I think there is an optimization opportunity here.
+ * Calling ZSTD_estimateSubBlockSize for every sequence can be wasteful
+ * since it recalculates estimate from scratch.
+ * For example, it would recount literal distribution and symbol codes every time.
+ */
+ cBlockSizeEstimate = ZSTD_estimateSubBlockSize(lp, litSize, ofCodePtr, llCodePtr, mlCodePtr, seqCount,
+ &nextCBlock->entropy, entropyMetadata,
+ workspace, wkspSize, writeLitEntropy, writeSeqEntropy);
+ if (cBlockSizeEstimate > targetCBlockSize || lastSequence) {
+ int litEntropyWritten = 0;
+ int seqEntropyWritten = 0;
+ const size_t decompressedSize = ZSTD_seqDecompressedSize(seqStorePtr, sp, seqCount, litSize, lastSequence);
+ const size_t cSize = ZSTD_compressSubBlock(&nextCBlock->entropy, entropyMetadata,
+ sp, seqCount,
+ lp, litSize,
+ llCodePtr, mlCodePtr, ofCodePtr,
+ cctxParams,
+ op, oend-op,
+ bmi2, writeLitEntropy, writeSeqEntropy,
+ &litEntropyWritten, &seqEntropyWritten,
+ lastBlock && lastSequence);
+ FORWARD_IF_ERROR(cSize, "ZSTD_compressSubBlock failed");
+ if (cSize > 0 && cSize < decompressedSize) {
+ DEBUGLOG(5, "Committed the sub-block");
+ assert(ip + decompressedSize <= iend);
+ ip += decompressedSize;
+ sp += seqCount;
+ lp += litSize;
+ op += cSize;
+ llCodePtr += seqCount;
+ mlCodePtr += seqCount;
+ ofCodePtr += seqCount;
+ litSize = 0;
+ seqCount = 0;
+ /* Entropy only needs to be written once */
+ if (litEntropyWritten) {
+ writeLitEntropy = 0;
+ }
+ if (seqEntropyWritten) {
+ writeSeqEntropy = 0;
+ }
+ }
+ }
+ } while (!lastSequence);
+ if (writeLitEntropy) {
+ DEBUGLOG(5, "ZSTD_compressSubBlock_multi has literal entropy tables unwritten");
+ ZSTD_memcpy(&nextCBlock->entropy.huf, &prevCBlock->entropy.huf, sizeof(prevCBlock->entropy.huf));
+ }
+ if (writeSeqEntropy && ZSTD_needSequenceEntropyTables(&entropyMetadata->fseMetadata)) {
+ /* If we haven't written our entropy tables, then we've violated our contract and
+ * must emit an uncompressed block.
+ */
+ DEBUGLOG(5, "ZSTD_compressSubBlock_multi has sequence entropy tables unwritten");
+ return 0;
+ }
+ if (ip < iend) {
+ size_t const cSize = ZSTD_noCompressBlock(op, oend - op, ip, iend - ip, lastBlock);
+ DEBUGLOG(5, "ZSTD_compressSubBlock_multi last sub-block uncompressed, %zu bytes", (size_t)(iend - ip));
+ FORWARD_IF_ERROR(cSize, "ZSTD_noCompressBlock failed");
+ assert(cSize != 0);
+ op += cSize;
+ /* We have to regenerate the repcodes because we've skipped some sequences */
+ if (sp < send) {
+ seqDef const* seq;
+ repcodes_t rep;
+ ZSTD_memcpy(&rep, prevCBlock->rep, sizeof(rep));
+ for (seq = sstart; seq < sp; ++seq) {
+ ZSTD_updateRep(rep.rep, seq->offBase - 1, ZSTD_getSequenceLength(seqStorePtr, seq).litLength == 0);
+ }
+ ZSTD_memcpy(nextCBlock->rep, &rep, sizeof(rep));
+ }
+ }
+ DEBUGLOG(5, "ZSTD_compressSubBlock_multi compressed");
+ return op-ostart;
+}
+
+size_t ZSTD_compressSuperBlock(ZSTD_CCtx* zc,
+ void* dst, size_t dstCapacity,
+ void const* src, size_t srcSize,
+ unsigned lastBlock) {
+ ZSTD_entropyCTablesMetadata_t entropyMetadata;
+
+ FORWARD_IF_ERROR(ZSTD_buildBlockEntropyStats(&zc->seqStore,
+ &zc->blockState.prevCBlock->entropy,
+ &zc->blockState.nextCBlock->entropy,
+ &zc->appliedParams,
+ &entropyMetadata,
+ zc->entropyWorkspace, ENTROPY_WORKSPACE_SIZE /* statically allocated in resetCCtx */), "");
+
+ return ZSTD_compressSubBlock_multi(&zc->seqStore,
+ zc->blockState.prevCBlock,
+ zc->blockState.nextCBlock,
+ &entropyMetadata,
+ &zc->appliedParams,
+ dst, dstCapacity,
+ src, srcSize,
+ zc->bmi2, lastBlock,
+ zc->entropyWorkspace, ENTROPY_WORKSPACE_SIZE /* statically allocated in resetCCtx */);
+}
diff --git a/lib/zstd/compress/zstd_compress_superblock.h b/lib/zstd/compress/zstd_compress_superblock.h
new file mode 100644
index 0000000000..224ece7954
--- /dev/null
+++ b/lib/zstd/compress/zstd_compress_superblock.h
@@ -0,0 +1,32 @@
+/*
+ * Copyright (c) Yann Collet, Facebook, Inc.
+ * All rights reserved.
+ *
+ * This source code is licensed under both the BSD-style license (found in the
+ * LICENSE file in the root directory of this source tree) and the GPLv2 (found
+ * in the COPYING file in the root directory of this source tree).
+ * You may select, at your option, one of the above-listed licenses.
+ */
+
+#ifndef ZSTD_COMPRESS_ADVANCED_H
+#define ZSTD_COMPRESS_ADVANCED_H
+
+/*-*************************************
+* Dependencies
+***************************************/
+
+#include <linux/zstd.h> /* ZSTD_CCtx */
+
+/*-*************************************
+* Target Compressed Block Size
+***************************************/
+
+/* ZSTD_compressSuperBlock() :
+ * Used to compress a super block when targetCBlockSize is being used.
+ * The given block will be compressed into multiple sub blocks that are around targetCBlockSize. */
+size_t ZSTD_compressSuperBlock(ZSTD_CCtx* zc,
+ void* dst, size_t dstCapacity,
+ void const* src, size_t srcSize,
+ unsigned lastBlock);
+
+#endif /* ZSTD_COMPRESS_ADVANCED_H */
diff --git a/lib/zstd/compress/zstd_cwksp.h b/lib/zstd/compress/zstd_cwksp.h
new file mode 100644
index 0000000000..349fc923c3
--- /dev/null
+++ b/lib/zstd/compress/zstd_cwksp.h
@@ -0,0 +1,595 @@
+/*
+ * Copyright (c) Yann Collet, Facebook, Inc.
+ * All rights reserved.
+ *
+ * This source code is licensed under both the BSD-style license (found in the
+ * LICENSE file in the root directory of this source tree) and the GPLv2 (found
+ * in the COPYING file in the root directory of this source tree).
+ * You may select, at your option, one of the above-listed licenses.
+ */
+
+#ifndef ZSTD_CWKSP_H
+#define ZSTD_CWKSP_H
+
+/*-*************************************
+* Dependencies
+***************************************/
+#include "../common/zstd_internal.h"
+
+
+/*-*************************************
+* Constants
+***************************************/
+
+/* Since the workspace is effectively its own little malloc implementation /
+ * arena, when we run under ASAN, we should similarly insert redzones between
+ * each internal element of the workspace, so ASAN will catch overruns that
+ * reach outside an object but that stay inside the workspace.
+ *
+ * This defines the size of that redzone.
+ */
+#ifndef ZSTD_CWKSP_ASAN_REDZONE_SIZE
+#define ZSTD_CWKSP_ASAN_REDZONE_SIZE 128
+#endif
+
+
+/* Set our tables and aligneds to align by 64 bytes */
+#define ZSTD_CWKSP_ALIGNMENT_BYTES 64
+
+/*-*************************************
+* Structures
+***************************************/
+typedef enum {
+ ZSTD_cwksp_alloc_objects,
+ ZSTD_cwksp_alloc_buffers,
+ ZSTD_cwksp_alloc_aligned
+} ZSTD_cwksp_alloc_phase_e;
+
+/*
+ * Used to describe whether the workspace is statically allocated (and will not
+ * necessarily ever be freed), or if it's dynamically allocated and we can
+ * expect a well-formed caller to free this.
+ */
+typedef enum {
+ ZSTD_cwksp_dynamic_alloc,
+ ZSTD_cwksp_static_alloc
+} ZSTD_cwksp_static_alloc_e;
+
+/*
+ * Zstd fits all its internal datastructures into a single continuous buffer,
+ * so that it only needs to perform a single OS allocation (or so that a buffer
+ * can be provided to it and it can perform no allocations at all). This buffer
+ * is called the workspace.
+ *
+ * Several optimizations complicate that process of allocating memory ranges
+ * from this workspace for each internal datastructure:
+ *
+ * - These different internal datastructures have different setup requirements:
+ *
+ * - The static objects need to be cleared once and can then be trivially
+ * reused for each compression.
+ *
+ * - Various buffers don't need to be initialized at all--they are always
+ * written into before they're read.
+ *
+ * - The matchstate tables have a unique requirement that they don't need
+ * their memory to be totally cleared, but they do need the memory to have
+ * some bound, i.e., a guarantee that all values in the memory they've been
+ * allocated is less than some maximum value (which is the starting value
+ * for the indices that they will then use for compression). When this
+ * guarantee is provided to them, they can use the memory without any setup
+ * work. When it can't, they have to clear the area.
+ *
+ * - These buffers also have different alignment requirements.
+ *
+ * - We would like to reuse the objects in the workspace for multiple
+ * compressions without having to perform any expensive reallocation or
+ * reinitialization work.
+ *
+ * - We would like to be able to efficiently reuse the workspace across
+ * multiple compressions **even when the compression parameters change** and
+ * we need to resize some of the objects (where possible).
+ *
+ * To attempt to manage this buffer, given these constraints, the ZSTD_cwksp
+ * abstraction was created. It works as follows:
+ *
+ * Workspace Layout:
+ *
+ * [ ... workspace ... ]
+ * [objects][tables ... ->] free space [<- ... aligned][<- ... buffers]
+ *
+ * The various objects that live in the workspace are divided into the
+ * following categories, and are allocated separately:
+ *
+ * - Static objects: this is optionally the enclosing ZSTD_CCtx or ZSTD_CDict,
+ * so that literally everything fits in a single buffer. Note: if present,
+ * this must be the first object in the workspace, since ZSTD_customFree{CCtx,
+ * CDict}() rely on a pointer comparison to see whether one or two frees are
+ * required.
+ *
+ * - Fixed size objects: these are fixed-size, fixed-count objects that are
+ * nonetheless "dynamically" allocated in the workspace so that we can
+ * control how they're initialized separately from the broader ZSTD_CCtx.
+ * Examples:
+ * - Entropy Workspace
+ * - 2 x ZSTD_compressedBlockState_t
+ * - CDict dictionary contents
+ *
+ * - Tables: these are any of several different datastructures (hash tables,
+ * chain tables, binary trees) that all respect a common format: they are
+ * uint32_t arrays, all of whose values are between 0 and (nextSrc - base).
+ * Their sizes depend on the cparams. These tables are 64-byte aligned.
+ *
+ * - Aligned: these buffers are used for various purposes that require 4 byte
+ * alignment, but don't require any initialization before they're used. These
+ * buffers are each aligned to 64 bytes.
+ *
+ * - Buffers: these buffers are used for various purposes that don't require
+ * any alignment or initialization before they're used. This means they can
+ * be moved around at no cost for a new compression.
+ *
+ * Allocating Memory:
+ *
+ * The various types of objects must be allocated in order, so they can be
+ * correctly packed into the workspace buffer. That order is:
+ *
+ * 1. Objects
+ * 2. Buffers
+ * 3. Aligned/Tables
+ *
+ * Attempts to reserve objects of different types out of order will fail.
+ */
+typedef struct {
+ void* workspace;
+ void* workspaceEnd;
+
+ void* objectEnd;
+ void* tableEnd;
+ void* tableValidEnd;
+ void* allocStart;
+
+ BYTE allocFailed;
+ int workspaceOversizedDuration;
+ ZSTD_cwksp_alloc_phase_e phase;
+ ZSTD_cwksp_static_alloc_e isStatic;
+} ZSTD_cwksp;
+
+/*-*************************************
+* Functions
+***************************************/
+
+MEM_STATIC size_t ZSTD_cwksp_available_space(ZSTD_cwksp* ws);
+
+MEM_STATIC void ZSTD_cwksp_assert_internal_consistency(ZSTD_cwksp* ws) {
+ (void)ws;
+ assert(ws->workspace <= ws->objectEnd);
+ assert(ws->objectEnd <= ws->tableEnd);
+ assert(ws->objectEnd <= ws->tableValidEnd);
+ assert(ws->tableEnd <= ws->allocStart);
+ assert(ws->tableValidEnd <= ws->allocStart);
+ assert(ws->allocStart <= ws->workspaceEnd);
+}
+
+/*
+ * Align must be a power of 2.
+ */
+MEM_STATIC size_t ZSTD_cwksp_align(size_t size, size_t const align) {
+ size_t const mask = align - 1;
+ assert((align & mask) == 0);
+ return (size + mask) & ~mask;
+}
+
+/*
+ * Use this to determine how much space in the workspace we will consume to
+ * allocate this object. (Normally it should be exactly the size of the object,
+ * but under special conditions, like ASAN, where we pad each object, it might
+ * be larger.)
+ *
+ * Since tables aren't currently redzoned, you don't need to call through this
+ * to figure out how much space you need for the matchState tables. Everything
+ * else is though.
+ *
+ * Do not use for sizing aligned buffers. Instead, use ZSTD_cwksp_aligned_alloc_size().
+ */
+MEM_STATIC size_t ZSTD_cwksp_alloc_size(size_t size) {
+ if (size == 0)
+ return 0;
+ return size;
+}
+
+/*
+ * Returns an adjusted alloc size that is the nearest larger multiple of 64 bytes.
+ * Used to determine the number of bytes required for a given "aligned".
+ */
+MEM_STATIC size_t ZSTD_cwksp_aligned_alloc_size(size_t size) {
+ return ZSTD_cwksp_alloc_size(ZSTD_cwksp_align(size, ZSTD_CWKSP_ALIGNMENT_BYTES));
+}
+
+/*
+ * Returns the amount of additional space the cwksp must allocate
+ * for internal purposes (currently only alignment).
+ */
+MEM_STATIC size_t ZSTD_cwksp_slack_space_required(void) {
+ /* For alignment, the wksp will always allocate an additional n_1=[1, 64] bytes
+ * to align the beginning of tables section, as well as another n_2=[0, 63] bytes
+ * to align the beginning of the aligned section.
+ *
+ * n_1 + n_2 == 64 bytes if the cwksp is freshly allocated, due to tables and
+ * aligneds being sized in multiples of 64 bytes.
+ */
+ size_t const slackSpace = ZSTD_CWKSP_ALIGNMENT_BYTES;
+ return slackSpace;
+}
+
+
+/*
+ * Return the number of additional bytes required to align a pointer to the given number of bytes.
+ * alignBytes must be a power of two.
+ */
+MEM_STATIC size_t ZSTD_cwksp_bytes_to_align_ptr(void* ptr, const size_t alignBytes) {
+ size_t const alignBytesMask = alignBytes - 1;
+ size_t const bytes = (alignBytes - ((size_t)ptr & (alignBytesMask))) & alignBytesMask;
+ assert((alignBytes & alignBytesMask) == 0);
+ assert(bytes != ZSTD_CWKSP_ALIGNMENT_BYTES);
+ return bytes;
+}
+
+/*
+ * Internal function. Do not use directly.
+ * Reserves the given number of bytes within the aligned/buffer segment of the wksp,
+ * which counts from the end of the wksp (as opposed to the object/table segment).
+ *
+ * Returns a pointer to the beginning of that space.
+ */
+MEM_STATIC void*
+ZSTD_cwksp_reserve_internal_buffer_space(ZSTD_cwksp* ws, size_t const bytes)
+{
+ void* const alloc = (BYTE*)ws->allocStart - bytes;
+ void* const bottom = ws->tableEnd;
+ DEBUGLOG(5, "cwksp: reserving %p %zd bytes, %zd bytes remaining",
+ alloc, bytes, ZSTD_cwksp_available_space(ws) - bytes);
+ ZSTD_cwksp_assert_internal_consistency(ws);
+ assert(alloc >= bottom);
+ if (alloc < bottom) {
+ DEBUGLOG(4, "cwksp: alloc failed!");
+ ws->allocFailed = 1;
+ return NULL;
+ }
+ /* the area is reserved from the end of wksp.
+ * If it overlaps with tableValidEnd, it voids guarantees on values' range */
+ if (alloc < ws->tableValidEnd) {
+ ws->tableValidEnd = alloc;
+ }
+ ws->allocStart = alloc;
+ return alloc;
+}
+
+/*
+ * Moves the cwksp to the next phase, and does any necessary allocations.
+ * cwksp initialization must necessarily go through each phase in order.
+ * Returns a 0 on success, or zstd error
+ */
+MEM_STATIC size_t
+ZSTD_cwksp_internal_advance_phase(ZSTD_cwksp* ws, ZSTD_cwksp_alloc_phase_e phase)
+{
+ assert(phase >= ws->phase);
+ if (phase > ws->phase) {
+ /* Going from allocating objects to allocating buffers */
+ if (ws->phase < ZSTD_cwksp_alloc_buffers &&
+ phase >= ZSTD_cwksp_alloc_buffers) {
+ ws->tableValidEnd = ws->objectEnd;
+ }
+
+ /* Going from allocating buffers to allocating aligneds/tables */
+ if (ws->phase < ZSTD_cwksp_alloc_aligned &&
+ phase >= ZSTD_cwksp_alloc_aligned) {
+ { /* Align the start of the "aligned" to 64 bytes. Use [1, 64] bytes. */
+ size_t const bytesToAlign =
+ ZSTD_CWKSP_ALIGNMENT_BYTES - ZSTD_cwksp_bytes_to_align_ptr(ws->allocStart, ZSTD_CWKSP_ALIGNMENT_BYTES);
+ DEBUGLOG(5, "reserving aligned alignment addtl space: %zu", bytesToAlign);
+ ZSTD_STATIC_ASSERT((ZSTD_CWKSP_ALIGNMENT_BYTES & (ZSTD_CWKSP_ALIGNMENT_BYTES - 1)) == 0); /* power of 2 */
+ RETURN_ERROR_IF(!ZSTD_cwksp_reserve_internal_buffer_space(ws, bytesToAlign),
+ memory_allocation, "aligned phase - alignment initial allocation failed!");
+ }
+ { /* Align the start of the tables to 64 bytes. Use [0, 63] bytes */
+ void* const alloc = ws->objectEnd;
+ size_t const bytesToAlign = ZSTD_cwksp_bytes_to_align_ptr(alloc, ZSTD_CWKSP_ALIGNMENT_BYTES);
+ void* const objectEnd = (BYTE*)alloc + bytesToAlign;
+ DEBUGLOG(5, "reserving table alignment addtl space: %zu", bytesToAlign);
+ RETURN_ERROR_IF(objectEnd > ws->workspaceEnd, memory_allocation,
+ "table phase - alignment initial allocation failed!");
+ ws->objectEnd = objectEnd;
+ ws->tableEnd = objectEnd; /* table area starts being empty */
+ if (ws->tableValidEnd < ws->tableEnd) {
+ ws->tableValidEnd = ws->tableEnd;
+ } } }
+ ws->phase = phase;
+ ZSTD_cwksp_assert_internal_consistency(ws);
+ }
+ return 0;
+}
+
+/*
+ * Returns whether this object/buffer/etc was allocated in this workspace.
+ */
+MEM_STATIC int ZSTD_cwksp_owns_buffer(const ZSTD_cwksp* ws, const void* ptr)
+{
+ return (ptr != NULL) && (ws->workspace <= ptr) && (ptr <= ws->workspaceEnd);
+}
+
+/*
+ * Internal function. Do not use directly.
+ */
+MEM_STATIC void*
+ZSTD_cwksp_reserve_internal(ZSTD_cwksp* ws, size_t bytes, ZSTD_cwksp_alloc_phase_e phase)
+{
+ void* alloc;
+ if (ZSTD_isError(ZSTD_cwksp_internal_advance_phase(ws, phase)) || bytes == 0) {
+ return NULL;
+ }
+
+
+ alloc = ZSTD_cwksp_reserve_internal_buffer_space(ws, bytes);
+
+
+ return alloc;
+}
+
+/*
+ * Reserves and returns unaligned memory.
+ */
+MEM_STATIC BYTE* ZSTD_cwksp_reserve_buffer(ZSTD_cwksp* ws, size_t bytes)
+{
+ return (BYTE*)ZSTD_cwksp_reserve_internal(ws, bytes, ZSTD_cwksp_alloc_buffers);
+}
+
+/*
+ * Reserves and returns memory sized on and aligned on ZSTD_CWKSP_ALIGNMENT_BYTES (64 bytes).
+ */
+MEM_STATIC void* ZSTD_cwksp_reserve_aligned(ZSTD_cwksp* ws, size_t bytes)
+{
+ void* ptr = ZSTD_cwksp_reserve_internal(ws, ZSTD_cwksp_align(bytes, ZSTD_CWKSP_ALIGNMENT_BYTES),
+ ZSTD_cwksp_alloc_aligned);
+ assert(((size_t)ptr & (ZSTD_CWKSP_ALIGNMENT_BYTES-1))== 0);
+ return ptr;
+}
+
+/*
+ * Aligned on 64 bytes. These buffers have the special property that
+ * their values remain constrained, allowing us to re-use them without
+ * memset()-ing them.
+ */
+MEM_STATIC void* ZSTD_cwksp_reserve_table(ZSTD_cwksp* ws, size_t bytes)
+{
+ const ZSTD_cwksp_alloc_phase_e phase = ZSTD_cwksp_alloc_aligned;
+ void* alloc;
+ void* end;
+ void* top;
+
+ if (ZSTD_isError(ZSTD_cwksp_internal_advance_phase(ws, phase))) {
+ return NULL;
+ }
+ alloc = ws->tableEnd;
+ end = (BYTE *)alloc + bytes;
+ top = ws->allocStart;
+
+ DEBUGLOG(5, "cwksp: reserving %p table %zd bytes, %zd bytes remaining",
+ alloc, bytes, ZSTD_cwksp_available_space(ws) - bytes);
+ assert((bytes & (sizeof(U32)-1)) == 0);
+ ZSTD_cwksp_assert_internal_consistency(ws);
+ assert(end <= top);
+ if (end > top) {
+ DEBUGLOG(4, "cwksp: table alloc failed!");
+ ws->allocFailed = 1;
+ return NULL;
+ }
+ ws->tableEnd = end;
+
+
+ assert((bytes & (ZSTD_CWKSP_ALIGNMENT_BYTES-1)) == 0);
+ assert(((size_t)alloc & (ZSTD_CWKSP_ALIGNMENT_BYTES-1))== 0);
+ return alloc;
+}
+
+/*
+ * Aligned on sizeof(void*).
+ * Note : should happen only once, at workspace first initialization
+ */
+MEM_STATIC void* ZSTD_cwksp_reserve_object(ZSTD_cwksp* ws, size_t bytes)
+{
+ size_t const roundedBytes = ZSTD_cwksp_align(bytes, sizeof(void*));
+ void* alloc = ws->objectEnd;
+ void* end = (BYTE*)alloc + roundedBytes;
+
+
+ DEBUGLOG(4,
+ "cwksp: reserving %p object %zd bytes (rounded to %zd), %zd bytes remaining",
+ alloc, bytes, roundedBytes, ZSTD_cwksp_available_space(ws) - roundedBytes);
+ assert((size_t)alloc % ZSTD_ALIGNOF(void*) == 0);
+ assert(bytes % ZSTD_ALIGNOF(void*) == 0);
+ ZSTD_cwksp_assert_internal_consistency(ws);
+ /* we must be in the first phase, no advance is possible */
+ if (ws->phase != ZSTD_cwksp_alloc_objects || end > ws->workspaceEnd) {
+ DEBUGLOG(3, "cwksp: object alloc failed!");
+ ws->allocFailed = 1;
+ return NULL;
+ }
+ ws->objectEnd = end;
+ ws->tableEnd = end;
+ ws->tableValidEnd = end;
+
+
+ return alloc;
+}
+
+MEM_STATIC void ZSTD_cwksp_mark_tables_dirty(ZSTD_cwksp* ws)
+{
+ DEBUGLOG(4, "cwksp: ZSTD_cwksp_mark_tables_dirty");
+
+
+ assert(ws->tableValidEnd >= ws->objectEnd);
+ assert(ws->tableValidEnd <= ws->allocStart);
+ ws->tableValidEnd = ws->objectEnd;
+ ZSTD_cwksp_assert_internal_consistency(ws);
+}
+
+MEM_STATIC void ZSTD_cwksp_mark_tables_clean(ZSTD_cwksp* ws) {
+ DEBUGLOG(4, "cwksp: ZSTD_cwksp_mark_tables_clean");
+ assert(ws->tableValidEnd >= ws->objectEnd);
+ assert(ws->tableValidEnd <= ws->allocStart);
+ if (ws->tableValidEnd < ws->tableEnd) {
+ ws->tableValidEnd = ws->tableEnd;
+ }
+ ZSTD_cwksp_assert_internal_consistency(ws);
+}
+
+/*
+ * Zero the part of the allocated tables not already marked clean.
+ */
+MEM_STATIC void ZSTD_cwksp_clean_tables(ZSTD_cwksp* ws) {
+ DEBUGLOG(4, "cwksp: ZSTD_cwksp_clean_tables");
+ assert(ws->tableValidEnd >= ws->objectEnd);
+ assert(ws->tableValidEnd <= ws->allocStart);
+ if (ws->tableValidEnd < ws->tableEnd) {
+ ZSTD_memset(ws->tableValidEnd, 0, (BYTE*)ws->tableEnd - (BYTE*)ws->tableValidEnd);
+ }
+ ZSTD_cwksp_mark_tables_clean(ws);
+}
+
+/*
+ * Invalidates table allocations.
+ * All other allocations remain valid.
+ */
+MEM_STATIC void ZSTD_cwksp_clear_tables(ZSTD_cwksp* ws) {
+ DEBUGLOG(4, "cwksp: clearing tables!");
+
+
+ ws->tableEnd = ws->objectEnd;
+ ZSTD_cwksp_assert_internal_consistency(ws);
+}
+
+/*
+ * Invalidates all buffer, aligned, and table allocations.
+ * Object allocations remain valid.
+ */
+MEM_STATIC void ZSTD_cwksp_clear(ZSTD_cwksp* ws) {
+ DEBUGLOG(4, "cwksp: clearing!");
+
+
+
+ ws->tableEnd = ws->objectEnd;
+ ws->allocStart = ws->workspaceEnd;
+ ws->allocFailed = 0;
+ if (ws->phase > ZSTD_cwksp_alloc_buffers) {
+ ws->phase = ZSTD_cwksp_alloc_buffers;
+ }
+ ZSTD_cwksp_assert_internal_consistency(ws);
+}
+
+/*
+ * The provided workspace takes ownership of the buffer [start, start+size).
+ * Any existing values in the workspace are ignored (the previously managed
+ * buffer, if present, must be separately freed).
+ */
+MEM_STATIC void ZSTD_cwksp_init(ZSTD_cwksp* ws, void* start, size_t size, ZSTD_cwksp_static_alloc_e isStatic) {
+ DEBUGLOG(4, "cwksp: init'ing workspace with %zd bytes", size);
+ assert(((size_t)start & (sizeof(void*)-1)) == 0); /* ensure correct alignment */
+ ws->workspace = start;
+ ws->workspaceEnd = (BYTE*)start + size;
+ ws->objectEnd = ws->workspace;
+ ws->tableValidEnd = ws->objectEnd;
+ ws->phase = ZSTD_cwksp_alloc_objects;
+ ws->isStatic = isStatic;
+ ZSTD_cwksp_clear(ws);
+ ws->workspaceOversizedDuration = 0;
+ ZSTD_cwksp_assert_internal_consistency(ws);
+}
+
+MEM_STATIC size_t ZSTD_cwksp_create(ZSTD_cwksp* ws, size_t size, ZSTD_customMem customMem) {
+ void* workspace = ZSTD_customMalloc(size, customMem);
+ DEBUGLOG(4, "cwksp: creating new workspace with %zd bytes", size);
+ RETURN_ERROR_IF(workspace == NULL, memory_allocation, "NULL pointer!");
+ ZSTD_cwksp_init(ws, workspace, size, ZSTD_cwksp_dynamic_alloc);
+ return 0;
+}
+
+MEM_STATIC void ZSTD_cwksp_free(ZSTD_cwksp* ws, ZSTD_customMem customMem) {
+ void *ptr = ws->workspace;
+ DEBUGLOG(4, "cwksp: freeing workspace");
+ ZSTD_memset(ws, 0, sizeof(ZSTD_cwksp));
+ ZSTD_customFree(ptr, customMem);
+}
+
+/*
+ * Moves the management of a workspace from one cwksp to another. The src cwksp
+ * is left in an invalid state (src must be re-init()'ed before it's used again).
+ */
+MEM_STATIC void ZSTD_cwksp_move(ZSTD_cwksp* dst, ZSTD_cwksp* src) {
+ *dst = *src;
+ ZSTD_memset(src, 0, sizeof(ZSTD_cwksp));
+}
+
+MEM_STATIC size_t ZSTD_cwksp_sizeof(const ZSTD_cwksp* ws) {
+ return (size_t)((BYTE*)ws->workspaceEnd - (BYTE*)ws->workspace);
+}
+
+MEM_STATIC size_t ZSTD_cwksp_used(const ZSTD_cwksp* ws) {
+ return (size_t)((BYTE*)ws->tableEnd - (BYTE*)ws->workspace)
+ + (size_t)((BYTE*)ws->workspaceEnd - (BYTE*)ws->allocStart);
+}
+
+MEM_STATIC int ZSTD_cwksp_reserve_failed(const ZSTD_cwksp* ws) {
+ return ws->allocFailed;
+}
+
+/*-*************************************
+* Functions Checking Free Space
+***************************************/
+
+/* ZSTD_alignmentSpaceWithinBounds() :
+ * Returns if the estimated space needed for a wksp is within an acceptable limit of the
+ * actual amount of space used.
+ */
+MEM_STATIC int ZSTD_cwksp_estimated_space_within_bounds(const ZSTD_cwksp* const ws,
+ size_t const estimatedSpace, int resizedWorkspace) {
+ if (resizedWorkspace) {
+ /* Resized/newly allocated wksp should have exact bounds */
+ return ZSTD_cwksp_used(ws) == estimatedSpace;
+ } else {
+ /* Due to alignment, when reusing a workspace, we can actually consume 63 fewer or more bytes
+ * than estimatedSpace. See the comments in zstd_cwksp.h for details.
+ */
+ return (ZSTD_cwksp_used(ws) >= estimatedSpace - 63) && (ZSTD_cwksp_used(ws) <= estimatedSpace + 63);
+ }
+}
+
+
+MEM_STATIC size_t ZSTD_cwksp_available_space(ZSTD_cwksp* ws) {
+ return (size_t)((BYTE*)ws->allocStart - (BYTE*)ws->tableEnd);
+}
+
+MEM_STATIC int ZSTD_cwksp_check_available(ZSTD_cwksp* ws, size_t additionalNeededSpace) {
+ return ZSTD_cwksp_available_space(ws) >= additionalNeededSpace;
+}
+
+MEM_STATIC int ZSTD_cwksp_check_too_large(ZSTD_cwksp* ws, size_t additionalNeededSpace) {
+ return ZSTD_cwksp_check_available(
+ ws, additionalNeededSpace * ZSTD_WORKSPACETOOLARGE_FACTOR);
+}
+
+MEM_STATIC int ZSTD_cwksp_check_wasteful(ZSTD_cwksp* ws, size_t additionalNeededSpace) {
+ return ZSTD_cwksp_check_too_large(ws, additionalNeededSpace)
+ && ws->workspaceOversizedDuration > ZSTD_WORKSPACETOOLARGE_MAXDURATION;
+}
+
+MEM_STATIC void ZSTD_cwksp_bump_oversized_duration(
+ ZSTD_cwksp* ws, size_t additionalNeededSpace) {
+ if (ZSTD_cwksp_check_too_large(ws, additionalNeededSpace)) {
+ ws->workspaceOversizedDuration++;
+ } else {
+ ws->workspaceOversizedDuration = 0;
+ }
+}
+
+
+#endif /* ZSTD_CWKSP_H */
diff --git a/lib/zstd/compress/zstd_double_fast.c b/lib/zstd/compress/zstd_double_fast.c
new file mode 100644
index 0000000000..76933dea26
--- /dev/null
+++ b/lib/zstd/compress/zstd_double_fast.c
@@ -0,0 +1,696 @@
+/*
+ * Copyright (c) Yann Collet, Facebook, Inc.
+ * All rights reserved.
+ *
+ * This source code is licensed under both the BSD-style license (found in the
+ * LICENSE file in the root directory of this source tree) and the GPLv2 (found
+ * in the COPYING file in the root directory of this source tree).
+ * You may select, at your option, one of the above-listed licenses.
+ */
+
+#include "zstd_compress_internal.h"
+#include "zstd_double_fast.h"
+
+
+void ZSTD_fillDoubleHashTable(ZSTD_matchState_t* ms,
+ void const* end, ZSTD_dictTableLoadMethod_e dtlm)
+{
+ const ZSTD_compressionParameters* const cParams = &ms->cParams;
+ U32* const hashLarge = ms->hashTable;
+ U32 const hBitsL = cParams->hashLog;
+ U32 const mls = cParams->minMatch;
+ U32* const hashSmall = ms->chainTable;
+ U32 const hBitsS = cParams->chainLog;
+ const BYTE* const base = ms->window.base;
+ const BYTE* ip = base + ms->nextToUpdate;
+ const BYTE* const iend = ((const BYTE*)end) - HASH_READ_SIZE;
+ const U32 fastHashFillStep = 3;
+
+ /* Always insert every fastHashFillStep position into the hash tables.
+ * Insert the other positions into the large hash table if their entry
+ * is empty.
+ */
+ for (; ip + fastHashFillStep - 1 <= iend; ip += fastHashFillStep) {
+ U32 const curr = (U32)(ip - base);
+ U32 i;
+ for (i = 0; i < fastHashFillStep; ++i) {
+ size_t const smHash = ZSTD_hashPtr(ip + i, hBitsS, mls);
+ size_t const lgHash = ZSTD_hashPtr(ip + i, hBitsL, 8);
+ if (i == 0)
+ hashSmall[smHash] = curr + i;
+ if (i == 0 || hashLarge[lgHash] == 0)
+ hashLarge[lgHash] = curr + i;
+ /* Only load extra positions for ZSTD_dtlm_full */
+ if (dtlm == ZSTD_dtlm_fast)
+ break;
+ } }
+}
+
+
+FORCE_INLINE_TEMPLATE
+size_t ZSTD_compressBlock_doubleFast_noDict_generic(
+ ZSTD_matchState_t* ms, seqStore_t* seqStore, U32 rep[ZSTD_REP_NUM],
+ void const* src, size_t srcSize, U32 const mls /* template */)
+{
+ ZSTD_compressionParameters const* cParams = &ms->cParams;
+ U32* const hashLong = ms->hashTable;
+ const U32 hBitsL = cParams->hashLog;
+ U32* const hashSmall = ms->chainTable;
+ const U32 hBitsS = cParams->chainLog;
+ const BYTE* const base = ms->window.base;
+ const BYTE* const istart = (const BYTE*)src;
+ const BYTE* anchor = istart;
+ const U32 endIndex = (U32)((size_t)(istart - base) + srcSize);
+ /* presumes that, if there is a dictionary, it must be using Attach mode */
+ const U32 prefixLowestIndex = ZSTD_getLowestPrefixIndex(ms, endIndex, cParams->windowLog);
+ const BYTE* const prefixLowest = base + prefixLowestIndex;
+ const BYTE* const iend = istart + srcSize;
+ const BYTE* const ilimit = iend - HASH_READ_SIZE;
+ U32 offset_1=rep[0], offset_2=rep[1];
+ U32 offsetSaved = 0;
+
+ size_t mLength;
+ U32 offset;
+ U32 curr;
+
+ /* how many positions to search before increasing step size */
+ const size_t kStepIncr = 1 << kSearchStrength;
+ /* the position at which to increment the step size if no match is found */
+ const BYTE* nextStep;
+ size_t step; /* the current step size */
+
+ size_t hl0; /* the long hash at ip */
+ size_t hl1; /* the long hash at ip1 */
+
+ U32 idxl0; /* the long match index for ip */
+ U32 idxl1; /* the long match index for ip1 */
+
+ const BYTE* matchl0; /* the long match for ip */
+ const BYTE* matchs0; /* the short match for ip */
+ const BYTE* matchl1; /* the long match for ip1 */
+
+ const BYTE* ip = istart; /* the current position */
+ const BYTE* ip1; /* the next position */
+
+ DEBUGLOG(5, "ZSTD_compressBlock_doubleFast_noDict_generic");
+
+ /* init */
+ ip += ((ip - prefixLowest) == 0);
+ {
+ U32 const current = (U32)(ip - base);
+ U32 const windowLow = ZSTD_getLowestPrefixIndex(ms, current, cParams->windowLog);
+ U32 const maxRep = current - windowLow;
+ if (offset_2 > maxRep) offsetSaved = offset_2, offset_2 = 0;
+ if (offset_1 > maxRep) offsetSaved = offset_1, offset_1 = 0;
+ }
+
+ /* Outer Loop: one iteration per match found and stored */
+ while (1) {
+ step = 1;
+ nextStep = ip + kStepIncr;
+ ip1 = ip + step;
+
+ if (ip1 > ilimit) {
+ goto _cleanup;
+ }
+
+ hl0 = ZSTD_hashPtr(ip, hBitsL, 8);
+ idxl0 = hashLong[hl0];
+ matchl0 = base + idxl0;
+
+ /* Inner Loop: one iteration per search / position */
+ do {
+ const size_t hs0 = ZSTD_hashPtr(ip, hBitsS, mls);
+ const U32 idxs0 = hashSmall[hs0];
+ curr = (U32)(ip-base);
+ matchs0 = base + idxs0;
+
+ hashLong[hl0] = hashSmall[hs0] = curr; /* update hash tables */
+
+ /* check noDict repcode */
+ if ((offset_1 > 0) & (MEM_read32(ip+1-offset_1) == MEM_read32(ip+1))) {
+ mLength = ZSTD_count(ip+1+4, ip+1+4-offset_1, iend) + 4;
+ ip++;
+ ZSTD_storeSeq(seqStore, (size_t)(ip-anchor), anchor, iend, STORE_REPCODE_1, mLength);
+ goto _match_stored;
+ }
+
+ hl1 = ZSTD_hashPtr(ip1, hBitsL, 8);
+
+ if (idxl0 > prefixLowestIndex) {
+ /* check prefix long match */
+ if (MEM_read64(matchl0) == MEM_read64(ip)) {
+ mLength = ZSTD_count(ip+8, matchl0+8, iend) + 8;
+ offset = (U32)(ip-matchl0);
+ while (((ip>anchor) & (matchl0>prefixLowest)) && (ip[-1] == matchl0[-1])) { ip--; matchl0--; mLength++; } /* catch up */
+ goto _match_found;
+ }
+ }
+
+ idxl1 = hashLong[hl1];
+ matchl1 = base + idxl1;
+
+ if (idxs0 > prefixLowestIndex) {
+ /* check prefix short match */
+ if (MEM_read32(matchs0) == MEM_read32(ip)) {
+ goto _search_next_long;
+ }
+ }
+
+ if (ip1 >= nextStep) {
+ PREFETCH_L1(ip1 + 64);
+ PREFETCH_L1(ip1 + 128);
+ step++;
+ nextStep += kStepIncr;
+ }
+ ip = ip1;
+ ip1 += step;
+
+ hl0 = hl1;
+ idxl0 = idxl1;
+ matchl0 = matchl1;
+ #if defined(__aarch64__)
+ PREFETCH_L1(ip+256);
+ #endif
+ } while (ip1 <= ilimit);
+
+_cleanup:
+ /* save reps for next block */
+ rep[0] = offset_1 ? offset_1 : offsetSaved;
+ rep[1] = offset_2 ? offset_2 : offsetSaved;
+
+ /* Return the last literals size */
+ return (size_t)(iend - anchor);
+
+_search_next_long:
+
+ /* check prefix long +1 match */
+ if (idxl1 > prefixLowestIndex) {
+ if (MEM_read64(matchl1) == MEM_read64(ip1)) {
+ ip = ip1;
+ mLength = ZSTD_count(ip+8, matchl1+8, iend) + 8;
+ offset = (U32)(ip-matchl1);
+ while (((ip>anchor) & (matchl1>prefixLowest)) && (ip[-1] == matchl1[-1])) { ip--; matchl1--; mLength++; } /* catch up */
+ goto _match_found;
+ }
+ }
+
+ /* if no long +1 match, explore the short match we found */
+ mLength = ZSTD_count(ip+4, matchs0+4, iend) + 4;
+ offset = (U32)(ip - matchs0);
+ while (((ip>anchor) & (matchs0>prefixLowest)) && (ip[-1] == matchs0[-1])) { ip--; matchs0--; mLength++; } /* catch up */
+
+ /* fall-through */
+
+_match_found: /* requires ip, offset, mLength */
+ offset_2 = offset_1;
+ offset_1 = offset;
+
+ if (step < 4) {
+ /* It is unsafe to write this value back to the hashtable when ip1 is
+ * greater than or equal to the new ip we will have after we're done
+ * processing this match. Rather than perform that test directly
+ * (ip1 >= ip + mLength), which costs speed in practice, we do a simpler
+ * more predictable test. The minmatch even if we take a short match is
+ * 4 bytes, so as long as step, the distance between ip and ip1
+ * (initially) is less than 4, we know ip1 < new ip. */
+ hashLong[hl1] = (U32)(ip1 - base);
+ }
+
+ ZSTD_storeSeq(seqStore, (size_t)(ip-anchor), anchor, iend, STORE_OFFSET(offset), mLength);
+
+_match_stored:
+ /* match found */
+ ip += mLength;
+ anchor = ip;
+
+ if (ip <= ilimit) {
+ /* Complementary insertion */
+ /* done after iLimit test, as candidates could be > iend-8 */
+ { U32 const indexToInsert = curr+2;
+ hashLong[ZSTD_hashPtr(base+indexToInsert, hBitsL, 8)] = indexToInsert;
+ hashLong[ZSTD_hashPtr(ip-2, hBitsL, 8)] = (U32)(ip-2-base);
+ hashSmall[ZSTD_hashPtr(base+indexToInsert, hBitsS, mls)] = indexToInsert;
+ hashSmall[ZSTD_hashPtr(ip-1, hBitsS, mls)] = (U32)(ip-1-base);
+ }
+
+ /* check immediate repcode */
+ while ( (ip <= ilimit)
+ && ( (offset_2>0)
+ & (MEM_read32(ip) == MEM_read32(ip - offset_2)) )) {
+ /* store sequence */
+ size_t const rLength = ZSTD_count(ip+4, ip+4-offset_2, iend) + 4;
+ U32 const tmpOff = offset_2; offset_2 = offset_1; offset_1 = tmpOff; /* swap offset_2 <=> offset_1 */
+ hashSmall[ZSTD_hashPtr(ip, hBitsS, mls)] = (U32)(ip-base);
+ hashLong[ZSTD_hashPtr(ip, hBitsL, 8)] = (U32)(ip-base);
+ ZSTD_storeSeq(seqStore, 0, anchor, iend, STORE_REPCODE_1, rLength);
+ ip += rLength;
+ anchor = ip;
+ continue; /* faster when present ... (?) */
+ }
+ }
+ }
+}
+
+
+FORCE_INLINE_TEMPLATE
+size_t ZSTD_compressBlock_doubleFast_dictMatchState_generic(
+ ZSTD_matchState_t* ms, seqStore_t* seqStore, U32 rep[ZSTD_REP_NUM],
+ void const* src, size_t srcSize,
+ U32 const mls /* template */)
+{
+ ZSTD_compressionParameters const* cParams = &ms->cParams;
+ U32* const hashLong = ms->hashTable;
+ const U32 hBitsL = cParams->hashLog;
+ U32* const hashSmall = ms->chainTable;
+ const U32 hBitsS = cParams->chainLog;
+ const BYTE* const base = ms->window.base;
+ const BYTE* const istart = (const BYTE*)src;
+ const BYTE* ip = istart;
+ const BYTE* anchor = istart;
+ const U32 endIndex = (U32)((size_t)(istart - base) + srcSize);
+ /* presumes that, if there is a dictionary, it must be using Attach mode */
+ const U32 prefixLowestIndex = ZSTD_getLowestPrefixIndex(ms, endIndex, cParams->windowLog);
+ const BYTE* const prefixLowest = base + prefixLowestIndex;
+ const BYTE* const iend = istart + srcSize;
+ const BYTE* const ilimit = iend - HASH_READ_SIZE;
+ U32 offset_1=rep[0], offset_2=rep[1];
+ U32 offsetSaved = 0;
+
+ const ZSTD_matchState_t* const dms = ms->dictMatchState;
+ const ZSTD_compressionParameters* const dictCParams = &dms->cParams;
+ const U32* const dictHashLong = dms->hashTable;
+ const U32* const dictHashSmall = dms->chainTable;
+ const U32 dictStartIndex = dms->window.dictLimit;
+ const BYTE* const dictBase = dms->window.base;
+ const BYTE* const dictStart = dictBase + dictStartIndex;
+ const BYTE* const dictEnd = dms->window.nextSrc;
+ const U32 dictIndexDelta = prefixLowestIndex - (U32)(dictEnd - dictBase);
+ const U32 dictHBitsL = dictCParams->hashLog;
+ const U32 dictHBitsS = dictCParams->chainLog;
+ const U32 dictAndPrefixLength = (U32)((ip - prefixLowest) + (dictEnd - dictStart));
+
+ DEBUGLOG(5, "ZSTD_compressBlock_doubleFast_dictMatchState_generic");
+
+ /* if a dictionary is attached, it must be within window range */
+ assert(ms->window.dictLimit + (1U << cParams->windowLog) >= endIndex);
+
+ /* init */
+ ip += (dictAndPrefixLength == 0);
+
+ /* dictMatchState repCode checks don't currently handle repCode == 0
+ * disabling. */
+ assert(offset_1 <= dictAndPrefixLength);
+ assert(offset_2 <= dictAndPrefixLength);
+
+ /* Main Search Loop */
+ while (ip < ilimit) { /* < instead of <=, because repcode check at (ip+1) */
+ size_t mLength;
+ U32 offset;
+ size_t const h2 = ZSTD_hashPtr(ip, hBitsL, 8);
+ size_t const h = ZSTD_hashPtr(ip, hBitsS, mls);
+ size_t const dictHL = ZSTD_hashPtr(ip, dictHBitsL, 8);
+ size_t const dictHS = ZSTD_hashPtr(ip, dictHBitsS, mls);
+ U32 const curr = (U32)(ip-base);
+ U32 const matchIndexL = hashLong[h2];
+ U32 matchIndexS = hashSmall[h];
+ const BYTE* matchLong = base + matchIndexL;
+ const BYTE* match = base + matchIndexS;
+ const U32 repIndex = curr + 1 - offset_1;
+ const BYTE* repMatch = (repIndex < prefixLowestIndex) ?
+ dictBase + (repIndex - dictIndexDelta) :
+ base + repIndex;
+ hashLong[h2] = hashSmall[h] = curr; /* update hash tables */
+
+ /* check repcode */
+ if (((U32)((prefixLowestIndex-1) - repIndex) >= 3 /* intentional underflow */)
+ && (MEM_read32(repMatch) == MEM_read32(ip+1)) ) {
+ const BYTE* repMatchEnd = repIndex < prefixLowestIndex ? dictEnd : iend;
+ mLength = ZSTD_count_2segments(ip+1+4, repMatch+4, iend, repMatchEnd, prefixLowest) + 4;
+ ip++;
+ ZSTD_storeSeq(seqStore, (size_t)(ip-anchor), anchor, iend, STORE_REPCODE_1, mLength);
+ goto _match_stored;
+ }
+
+ if (matchIndexL > prefixLowestIndex) {
+ /* check prefix long match */
+ if (MEM_read64(matchLong) == MEM_read64(ip)) {
+ mLength = ZSTD_count(ip+8, matchLong+8, iend) + 8;
+ offset = (U32)(ip-matchLong);
+ while (((ip>anchor) & (matchLong>prefixLowest)) && (ip[-1] == matchLong[-1])) { ip--; matchLong--; mLength++; } /* catch up */
+ goto _match_found;
+ }
+ } else {
+ /* check dictMatchState long match */
+ U32 const dictMatchIndexL = dictHashLong[dictHL];
+ const BYTE* dictMatchL = dictBase + dictMatchIndexL;
+ assert(dictMatchL < dictEnd);
+
+ if (dictMatchL > dictStart && MEM_read64(dictMatchL) == MEM_read64(ip)) {
+ mLength = ZSTD_count_2segments(ip+8, dictMatchL+8, iend, dictEnd, prefixLowest) + 8;
+ offset = (U32)(curr - dictMatchIndexL - dictIndexDelta);
+ while (((ip>anchor) & (dictMatchL>dictStart)) && (ip[-1] == dictMatchL[-1])) { ip--; dictMatchL--; mLength++; } /* catch up */
+ goto _match_found;
+ } }
+
+ if (matchIndexS > prefixLowestIndex) {
+ /* check prefix short match */
+ if (MEM_read32(match) == MEM_read32(ip)) {
+ goto _search_next_long;
+ }
+ } else {
+ /* check dictMatchState short match */
+ U32 const dictMatchIndexS = dictHashSmall[dictHS];
+ match = dictBase + dictMatchIndexS;
+ matchIndexS = dictMatchIndexS + dictIndexDelta;
+
+ if (match > dictStart && MEM_read32(match) == MEM_read32(ip)) {
+ goto _search_next_long;
+ } }
+
+ ip += ((ip-anchor) >> kSearchStrength) + 1;
+#if defined(__aarch64__)
+ PREFETCH_L1(ip+256);
+#endif
+ continue;
+
+_search_next_long:
+
+ { size_t const hl3 = ZSTD_hashPtr(ip+1, hBitsL, 8);
+ size_t const dictHLNext = ZSTD_hashPtr(ip+1, dictHBitsL, 8);
+ U32 const matchIndexL3 = hashLong[hl3];
+ const BYTE* matchL3 = base + matchIndexL3;
+ hashLong[hl3] = curr + 1;
+
+ /* check prefix long +1 match */
+ if (matchIndexL3 > prefixLowestIndex) {
+ if (MEM_read64(matchL3) == MEM_read64(ip+1)) {
+ mLength = ZSTD_count(ip+9, matchL3+8, iend) + 8;
+ ip++;
+ offset = (U32)(ip-matchL3);
+ while (((ip>anchor) & (matchL3>prefixLowest)) && (ip[-1] == matchL3[-1])) { ip--; matchL3--; mLength++; } /* catch up */
+ goto _match_found;
+ }
+ } else {
+ /* check dict long +1 match */
+ U32 const dictMatchIndexL3 = dictHashLong[dictHLNext];
+ const BYTE* dictMatchL3 = dictBase + dictMatchIndexL3;
+ assert(dictMatchL3 < dictEnd);
+ if (dictMatchL3 > dictStart && MEM_read64(dictMatchL3) == MEM_read64(ip+1)) {
+ mLength = ZSTD_count_2segments(ip+1+8, dictMatchL3+8, iend, dictEnd, prefixLowest) + 8;
+ ip++;
+ offset = (U32)(curr + 1 - dictMatchIndexL3 - dictIndexDelta);
+ while (((ip>anchor) & (dictMatchL3>dictStart)) && (ip[-1] == dictMatchL3[-1])) { ip--; dictMatchL3--; mLength++; } /* catch up */
+ goto _match_found;
+ } } }
+
+ /* if no long +1 match, explore the short match we found */
+ if (matchIndexS < prefixLowestIndex) {
+ mLength = ZSTD_count_2segments(ip+4, match+4, iend, dictEnd, prefixLowest) + 4;
+ offset = (U32)(curr - matchIndexS);
+ while (((ip>anchor) & (match>dictStart)) && (ip[-1] == match[-1])) { ip--; match--; mLength++; } /* catch up */
+ } else {
+ mLength = ZSTD_count(ip+4, match+4, iend) + 4;
+ offset = (U32)(ip - match);
+ while (((ip>anchor) & (match>prefixLowest)) && (ip[-1] == match[-1])) { ip--; match--; mLength++; } /* catch up */
+ }
+
+_match_found:
+ offset_2 = offset_1;
+ offset_1 = offset;
+
+ ZSTD_storeSeq(seqStore, (size_t)(ip-anchor), anchor, iend, STORE_OFFSET(offset), mLength);
+
+_match_stored:
+ /* match found */
+ ip += mLength;
+ anchor = ip;
+
+ if (ip <= ilimit) {
+ /* Complementary insertion */
+ /* done after iLimit test, as candidates could be > iend-8 */
+ { U32 const indexToInsert = curr+2;
+ hashLong[ZSTD_hashPtr(base+indexToInsert, hBitsL, 8)] = indexToInsert;
+ hashLong[ZSTD_hashPtr(ip-2, hBitsL, 8)] = (U32)(ip-2-base);
+ hashSmall[ZSTD_hashPtr(base+indexToInsert, hBitsS, mls)] = indexToInsert;
+ hashSmall[ZSTD_hashPtr(ip-1, hBitsS, mls)] = (U32)(ip-1-base);
+ }
+
+ /* check immediate repcode */
+ while (ip <= ilimit) {
+ U32 const current2 = (U32)(ip-base);
+ U32 const repIndex2 = current2 - offset_2;
+ const BYTE* repMatch2 = repIndex2 < prefixLowestIndex ?
+ dictBase + repIndex2 - dictIndexDelta :
+ base + repIndex2;
+ if ( ((U32)((prefixLowestIndex-1) - (U32)repIndex2) >= 3 /* intentional overflow */)
+ && (MEM_read32(repMatch2) == MEM_read32(ip)) ) {
+ const BYTE* const repEnd2 = repIndex2 < prefixLowestIndex ? dictEnd : iend;
+ size_t const repLength2 = ZSTD_count_2segments(ip+4, repMatch2+4, iend, repEnd2, prefixLowest) + 4;
+ U32 tmpOffset = offset_2; offset_2 = offset_1; offset_1 = tmpOffset; /* swap offset_2 <=> offset_1 */
+ ZSTD_storeSeq(seqStore, 0, anchor, iend, STORE_REPCODE_1, repLength2);
+ hashSmall[ZSTD_hashPtr(ip, hBitsS, mls)] = current2;
+ hashLong[ZSTD_hashPtr(ip, hBitsL, 8)] = current2;
+ ip += repLength2;
+ anchor = ip;
+ continue;
+ }
+ break;
+ }
+ }
+ } /* while (ip < ilimit) */
+
+ /* save reps for next block */
+ rep[0] = offset_1 ? offset_1 : offsetSaved;
+ rep[1] = offset_2 ? offset_2 : offsetSaved;
+
+ /* Return the last literals size */
+ return (size_t)(iend - anchor);
+}
+
+#define ZSTD_GEN_DFAST_FN(dictMode, mls) \
+ static size_t ZSTD_compressBlock_doubleFast_##dictMode##_##mls( \
+ ZSTD_matchState_t* ms, seqStore_t* seqStore, U32 rep[ZSTD_REP_NUM], \
+ void const* src, size_t srcSize) \
+ { \
+ return ZSTD_compressBlock_doubleFast_##dictMode##_generic(ms, seqStore, rep, src, srcSize, mls); \
+ }
+
+ZSTD_GEN_DFAST_FN(noDict, 4)
+ZSTD_GEN_DFAST_FN(noDict, 5)
+ZSTD_GEN_DFAST_FN(noDict, 6)
+ZSTD_GEN_DFAST_FN(noDict, 7)
+
+ZSTD_GEN_DFAST_FN(dictMatchState, 4)
+ZSTD_GEN_DFAST_FN(dictMatchState, 5)
+ZSTD_GEN_DFAST_FN(dictMatchState, 6)
+ZSTD_GEN_DFAST_FN(dictMatchState, 7)
+
+
+size_t ZSTD_compressBlock_doubleFast(
+ ZSTD_matchState_t* ms, seqStore_t* seqStore, U32 rep[ZSTD_REP_NUM],
+ void const* src, size_t srcSize)
+{
+ const U32 mls = ms->cParams.minMatch;
+ switch(mls)
+ {
+ default: /* includes case 3 */
+ case 4 :
+ return ZSTD_compressBlock_doubleFast_noDict_4(ms, seqStore, rep, src, srcSize);
+ case 5 :
+ return ZSTD_compressBlock_doubleFast_noDict_5(ms, seqStore, rep, src, srcSize);
+ case 6 :
+ return ZSTD_compressBlock_doubleFast_noDict_6(ms, seqStore, rep, src, srcSize);
+ case 7 :
+ return ZSTD_compressBlock_doubleFast_noDict_7(ms, seqStore, rep, src, srcSize);
+ }
+}
+
+
+size_t ZSTD_compressBlock_doubleFast_dictMatchState(
+ ZSTD_matchState_t* ms, seqStore_t* seqStore, U32 rep[ZSTD_REP_NUM],
+ void const* src, size_t srcSize)
+{
+ const U32 mls = ms->cParams.minMatch;
+ switch(mls)
+ {
+ default: /* includes case 3 */
+ case 4 :
+ return ZSTD_compressBlock_doubleFast_dictMatchState_4(ms, seqStore, rep, src, srcSize);
+ case 5 :
+ return ZSTD_compressBlock_doubleFast_dictMatchState_5(ms, seqStore, rep, src, srcSize);
+ case 6 :
+ return ZSTD_compressBlock_doubleFast_dictMatchState_6(ms, seqStore, rep, src, srcSize);
+ case 7 :
+ return ZSTD_compressBlock_doubleFast_dictMatchState_7(ms, seqStore, rep, src, srcSize);
+ }
+}
+
+
+static size_t ZSTD_compressBlock_doubleFast_extDict_generic(
+ ZSTD_matchState_t* ms, seqStore_t* seqStore, U32 rep[ZSTD_REP_NUM],
+ void const* src, size_t srcSize,
+ U32 const mls /* template */)
+{
+ ZSTD_compressionParameters const* cParams = &ms->cParams;
+ U32* const hashLong = ms->hashTable;
+ U32 const hBitsL = cParams->hashLog;
+ U32* const hashSmall = ms->chainTable;
+ U32 const hBitsS = cParams->chainLog;
+ const BYTE* const istart = (const BYTE*)src;
+ const BYTE* ip = istart;
+ const BYTE* anchor = istart;
+ const BYTE* const iend = istart + srcSize;
+ const BYTE* const ilimit = iend - 8;
+ const BYTE* const base = ms->window.base;
+ const U32 endIndex = (U32)((size_t)(istart - base) + srcSize);
+ const U32 lowLimit = ZSTD_getLowestMatchIndex(ms, endIndex, cParams->windowLog);
+ const U32 dictStartIndex = lowLimit;
+ const U32 dictLimit = ms->window.dictLimit;
+ const U32 prefixStartIndex = (dictLimit > lowLimit) ? dictLimit : lowLimit;
+ const BYTE* const prefixStart = base + prefixStartIndex;
+ const BYTE* const dictBase = ms->window.dictBase;
+ const BYTE* const dictStart = dictBase + dictStartIndex;
+ const BYTE* const dictEnd = dictBase + prefixStartIndex;
+ U32 offset_1=rep[0], offset_2=rep[1];
+
+ DEBUGLOG(5, "ZSTD_compressBlock_doubleFast_extDict_generic (srcSize=%zu)", srcSize);
+
+ /* if extDict is invalidated due to maxDistance, switch to "regular" variant */
+ if (prefixStartIndex == dictStartIndex)
+ return ZSTD_compressBlock_doubleFast(ms, seqStore, rep, src, srcSize);
+
+ /* Search Loop */
+ while (ip < ilimit) { /* < instead of <=, because (ip+1) */
+ const size_t hSmall = ZSTD_hashPtr(ip, hBitsS, mls);
+ const U32 matchIndex = hashSmall[hSmall];
+ const BYTE* const matchBase = matchIndex < prefixStartIndex ? dictBase : base;
+ const BYTE* match = matchBase + matchIndex;
+
+ const size_t hLong = ZSTD_hashPtr(ip, hBitsL, 8);
+ const U32 matchLongIndex = hashLong[hLong];
+ const BYTE* const matchLongBase = matchLongIndex < prefixStartIndex ? dictBase : base;
+ const BYTE* matchLong = matchLongBase + matchLongIndex;
+
+ const U32 curr = (U32)(ip-base);
+ const U32 repIndex = curr + 1 - offset_1; /* offset_1 expected <= curr +1 */
+ const BYTE* const repBase = repIndex < prefixStartIndex ? dictBase : base;
+ const BYTE* const repMatch = repBase + repIndex;
+ size_t mLength;
+ hashSmall[hSmall] = hashLong[hLong] = curr; /* update hash table */
+
+ if ((((U32)((prefixStartIndex-1) - repIndex) >= 3) /* intentional underflow : ensure repIndex doesn't overlap dict + prefix */
+ & (offset_1 <= curr+1 - dictStartIndex)) /* note: we are searching at curr+1 */
+ && (MEM_read32(repMatch) == MEM_read32(ip+1)) ) {
+ const BYTE* repMatchEnd = repIndex < prefixStartIndex ? dictEnd : iend;
+ mLength = ZSTD_count_2segments(ip+1+4, repMatch+4, iend, repMatchEnd, prefixStart) + 4;
+ ip++;
+ ZSTD_storeSeq(seqStore, (size_t)(ip-anchor), anchor, iend, STORE_REPCODE_1, mLength);
+ } else {
+ if ((matchLongIndex > dictStartIndex) && (MEM_read64(matchLong) == MEM_read64(ip))) {
+ const BYTE* const matchEnd = matchLongIndex < prefixStartIndex ? dictEnd : iend;
+ const BYTE* const lowMatchPtr = matchLongIndex < prefixStartIndex ? dictStart : prefixStart;
+ U32 offset;
+ mLength = ZSTD_count_2segments(ip+8, matchLong+8, iend, matchEnd, prefixStart) + 8;
+ offset = curr - matchLongIndex;
+ while (((ip>anchor) & (matchLong>lowMatchPtr)) && (ip[-1] == matchLong[-1])) { ip--; matchLong--; mLength++; } /* catch up */
+ offset_2 = offset_1;
+ offset_1 = offset;
+ ZSTD_storeSeq(seqStore, (size_t)(ip-anchor), anchor, iend, STORE_OFFSET(offset), mLength);
+
+ } else if ((matchIndex > dictStartIndex) && (MEM_read32(match) == MEM_read32(ip))) {
+ size_t const h3 = ZSTD_hashPtr(ip+1, hBitsL, 8);
+ U32 const matchIndex3 = hashLong[h3];
+ const BYTE* const match3Base = matchIndex3 < prefixStartIndex ? dictBase : base;
+ const BYTE* match3 = match3Base + matchIndex3;
+ U32 offset;
+ hashLong[h3] = curr + 1;
+ if ( (matchIndex3 > dictStartIndex) && (MEM_read64(match3) == MEM_read64(ip+1)) ) {
+ const BYTE* const matchEnd = matchIndex3 < prefixStartIndex ? dictEnd : iend;
+ const BYTE* const lowMatchPtr = matchIndex3 < prefixStartIndex ? dictStart : prefixStart;
+ mLength = ZSTD_count_2segments(ip+9, match3+8, iend, matchEnd, prefixStart) + 8;
+ ip++;
+ offset = curr+1 - matchIndex3;
+ while (((ip>anchor) & (match3>lowMatchPtr)) && (ip[-1] == match3[-1])) { ip--; match3--; mLength++; } /* catch up */
+ } else {
+ const BYTE* const matchEnd = matchIndex < prefixStartIndex ? dictEnd : iend;
+ const BYTE* const lowMatchPtr = matchIndex < prefixStartIndex ? dictStart : prefixStart;
+ mLength = ZSTD_count_2segments(ip+4, match+4, iend, matchEnd, prefixStart) + 4;
+ offset = curr - matchIndex;
+ while (((ip>anchor) & (match>lowMatchPtr)) && (ip[-1] == match[-1])) { ip--; match--; mLength++; } /* catch up */
+ }
+ offset_2 = offset_1;
+ offset_1 = offset;
+ ZSTD_storeSeq(seqStore, (size_t)(ip-anchor), anchor, iend, STORE_OFFSET(offset), mLength);
+
+ } else {
+ ip += ((ip-anchor) >> kSearchStrength) + 1;
+ continue;
+ } }
+
+ /* move to next sequence start */
+ ip += mLength;
+ anchor = ip;
+
+ if (ip <= ilimit) {
+ /* Complementary insertion */
+ /* done after iLimit test, as candidates could be > iend-8 */
+ { U32 const indexToInsert = curr+2;
+ hashLong[ZSTD_hashPtr(base+indexToInsert, hBitsL, 8)] = indexToInsert;
+ hashLong[ZSTD_hashPtr(ip-2, hBitsL, 8)] = (U32)(ip-2-base);
+ hashSmall[ZSTD_hashPtr(base+indexToInsert, hBitsS, mls)] = indexToInsert;
+ hashSmall[ZSTD_hashPtr(ip-1, hBitsS, mls)] = (U32)(ip-1-base);
+ }
+
+ /* check immediate repcode */
+ while (ip <= ilimit) {
+ U32 const current2 = (U32)(ip-base);
+ U32 const repIndex2 = current2 - offset_2;
+ const BYTE* repMatch2 = repIndex2 < prefixStartIndex ? dictBase + repIndex2 : base + repIndex2;
+ if ( (((U32)((prefixStartIndex-1) - repIndex2) >= 3) /* intentional overflow : ensure repIndex2 doesn't overlap dict + prefix */
+ & (offset_2 <= current2 - dictStartIndex))
+ && (MEM_read32(repMatch2) == MEM_read32(ip)) ) {
+ const BYTE* const repEnd2 = repIndex2 < prefixStartIndex ? dictEnd : iend;
+ size_t const repLength2 = ZSTD_count_2segments(ip+4, repMatch2+4, iend, repEnd2, prefixStart) + 4;
+ U32 const tmpOffset = offset_2; offset_2 = offset_1; offset_1 = tmpOffset; /* swap offset_2 <=> offset_1 */
+ ZSTD_storeSeq(seqStore, 0, anchor, iend, STORE_REPCODE_1, repLength2);
+ hashSmall[ZSTD_hashPtr(ip, hBitsS, mls)] = current2;
+ hashLong[ZSTD_hashPtr(ip, hBitsL, 8)] = current2;
+ ip += repLength2;
+ anchor = ip;
+ continue;
+ }
+ break;
+ } } }
+
+ /* save reps for next block */
+ rep[0] = offset_1;
+ rep[1] = offset_2;
+
+ /* Return the last literals size */
+ return (size_t)(iend - anchor);
+}
+
+ZSTD_GEN_DFAST_FN(extDict, 4)
+ZSTD_GEN_DFAST_FN(extDict, 5)
+ZSTD_GEN_DFAST_FN(extDict, 6)
+ZSTD_GEN_DFAST_FN(extDict, 7)
+
+size_t ZSTD_compressBlock_doubleFast_extDict(
+ ZSTD_matchState_t* ms, seqStore_t* seqStore, U32 rep[ZSTD_REP_NUM],
+ void const* src, size_t srcSize)
+{
+ U32 const mls = ms->cParams.minMatch;
+ switch(mls)
+ {
+ default: /* includes case 3 */
+ case 4 :
+ return ZSTD_compressBlock_doubleFast_extDict_4(ms, seqStore, rep, src, srcSize);
+ case 5 :
+ return ZSTD_compressBlock_doubleFast_extDict_5(ms, seqStore, rep, src, srcSize);
+ case 6 :
+ return ZSTD_compressBlock_doubleFast_extDict_6(ms, seqStore, rep, src, srcSize);
+ case 7 :
+ return ZSTD_compressBlock_doubleFast_extDict_7(ms, seqStore, rep, src, srcSize);
+ }
+}
diff --git a/lib/zstd/compress/zstd_double_fast.h b/lib/zstd/compress/zstd_double_fast.h
new file mode 100644
index 0000000000..6822bde65a
--- /dev/null
+++ b/lib/zstd/compress/zstd_double_fast.h
@@ -0,0 +1,32 @@
+/*
+ * Copyright (c) Yann Collet, Facebook, Inc.
+ * All rights reserved.
+ *
+ * This source code is licensed under both the BSD-style license (found in the
+ * LICENSE file in the root directory of this source tree) and the GPLv2 (found
+ * in the COPYING file in the root directory of this source tree).
+ * You may select, at your option, one of the above-listed licenses.
+ */
+
+#ifndef ZSTD_DOUBLE_FAST_H
+#define ZSTD_DOUBLE_FAST_H
+
+
+#include "../common/mem.h" /* U32 */
+#include "zstd_compress_internal.h" /* ZSTD_CCtx, size_t */
+
+void ZSTD_fillDoubleHashTable(ZSTD_matchState_t* ms,
+ void const* end, ZSTD_dictTableLoadMethod_e dtlm);
+size_t ZSTD_compressBlock_doubleFast(
+ ZSTD_matchState_t* ms, seqStore_t* seqStore, U32 rep[ZSTD_REP_NUM],
+ void const* src, size_t srcSize);
+size_t ZSTD_compressBlock_doubleFast_dictMatchState(
+ ZSTD_matchState_t* ms, seqStore_t* seqStore, U32 rep[ZSTD_REP_NUM],
+ void const* src, size_t srcSize);
+size_t ZSTD_compressBlock_doubleFast_extDict(
+ ZSTD_matchState_t* ms, seqStore_t* seqStore, U32 rep[ZSTD_REP_NUM],
+ void const* src, size_t srcSize);
+
+
+
+#endif /* ZSTD_DOUBLE_FAST_H */
diff --git a/lib/zstd/compress/zstd_fast.c b/lib/zstd/compress/zstd_fast.c
new file mode 100644
index 0000000000..a752e6beab
--- /dev/null
+++ b/lib/zstd/compress/zstd_fast.c
@@ -0,0 +1,675 @@
+/*
+ * Copyright (c) Yann Collet, Facebook, Inc.
+ * All rights reserved.
+ *
+ * This source code is licensed under both the BSD-style license (found in the
+ * LICENSE file in the root directory of this source tree) and the GPLv2 (found
+ * in the COPYING file in the root directory of this source tree).
+ * You may select, at your option, one of the above-listed licenses.
+ */
+
+#include "zstd_compress_internal.h" /* ZSTD_hashPtr, ZSTD_count, ZSTD_storeSeq */
+#include "zstd_fast.h"
+
+
+void ZSTD_fillHashTable(ZSTD_matchState_t* ms,
+ const void* const end,
+ ZSTD_dictTableLoadMethod_e dtlm)
+{
+ const ZSTD_compressionParameters* const cParams = &ms->cParams;
+ U32* const hashTable = ms->hashTable;
+ U32 const hBits = cParams->hashLog;
+ U32 const mls = cParams->minMatch;
+ const BYTE* const base = ms->window.base;
+ const BYTE* ip = base + ms->nextToUpdate;
+ const BYTE* const iend = ((const BYTE*)end) - HASH_READ_SIZE;
+ const U32 fastHashFillStep = 3;
+
+ /* Always insert every fastHashFillStep position into the hash table.
+ * Insert the other positions if their hash entry is empty.
+ */
+ for ( ; ip + fastHashFillStep < iend + 2; ip += fastHashFillStep) {
+ U32 const curr = (U32)(ip - base);
+ size_t const hash0 = ZSTD_hashPtr(ip, hBits, mls);
+ hashTable[hash0] = curr;
+ if (dtlm == ZSTD_dtlm_fast) continue;
+ /* Only load extra positions for ZSTD_dtlm_full */
+ { U32 p;
+ for (p = 1; p < fastHashFillStep; ++p) {
+ size_t const hash = ZSTD_hashPtr(ip + p, hBits, mls);
+ if (hashTable[hash] == 0) { /* not yet filled */
+ hashTable[hash] = curr + p;
+ } } } }
+}
+
+
+/*
+ * If you squint hard enough (and ignore repcodes), the search operation at any
+ * given position is broken into 4 stages:
+ *
+ * 1. Hash (map position to hash value via input read)
+ * 2. Lookup (map hash val to index via hashtable read)
+ * 3. Load (map index to value at that position via input read)
+ * 4. Compare
+ *
+ * Each of these steps involves a memory read at an address which is computed
+ * from the previous step. This means these steps must be sequenced and their
+ * latencies are cumulative.
+ *
+ * Rather than do 1->2->3->4 sequentially for a single position before moving
+ * onto the next, this implementation interleaves these operations across the
+ * next few positions:
+ *
+ * R = Repcode Read & Compare
+ * H = Hash
+ * T = Table Lookup
+ * M = Match Read & Compare
+ *
+ * Pos | Time -->
+ * ----+-------------------
+ * N | ... M
+ * N+1 | ... TM
+ * N+2 | R H T M
+ * N+3 | H TM
+ * N+4 | R H T M
+ * N+5 | H ...
+ * N+6 | R ...
+ *
+ * This is very much analogous to the pipelining of execution in a CPU. And just
+ * like a CPU, we have to dump the pipeline when we find a match (i.e., take a
+ * branch).
+ *
+ * When this happens, we throw away our current state, and do the following prep
+ * to re-enter the loop:
+ *
+ * Pos | Time -->
+ * ----+-------------------
+ * N | H T
+ * N+1 | H
+ *
+ * This is also the work we do at the beginning to enter the loop initially.
+ */
+FORCE_INLINE_TEMPLATE size_t
+ZSTD_compressBlock_fast_noDict_generic(
+ ZSTD_matchState_t* ms, seqStore_t* seqStore, U32 rep[ZSTD_REP_NUM],
+ void const* src, size_t srcSize,
+ U32 const mls, U32 const hasStep)
+{
+ const ZSTD_compressionParameters* const cParams = &ms->cParams;
+ U32* const hashTable = ms->hashTable;
+ U32 const hlog = cParams->hashLog;
+ /* support stepSize of 0 */
+ size_t const stepSize = hasStep ? (cParams->targetLength + !(cParams->targetLength) + 1) : 2;
+ const BYTE* const base = ms->window.base;
+ const BYTE* const istart = (const BYTE*)src;
+ const U32 endIndex = (U32)((size_t)(istart - base) + srcSize);
+ const U32 prefixStartIndex = ZSTD_getLowestPrefixIndex(ms, endIndex, cParams->windowLog);
+ const BYTE* const prefixStart = base + prefixStartIndex;
+ const BYTE* const iend = istart + srcSize;
+ const BYTE* const ilimit = iend - HASH_READ_SIZE;
+
+ const BYTE* anchor = istart;
+ const BYTE* ip0 = istart;
+ const BYTE* ip1;
+ const BYTE* ip2;
+ const BYTE* ip3;
+ U32 current0;
+
+ U32 rep_offset1 = rep[0];
+ U32 rep_offset2 = rep[1];
+ U32 offsetSaved = 0;
+
+ size_t hash0; /* hash for ip0 */
+ size_t hash1; /* hash for ip1 */
+ U32 idx; /* match idx for ip0 */
+ U32 mval; /* src value at match idx */
+
+ U32 offcode;
+ const BYTE* match0;
+ size_t mLength;
+
+ /* ip0 and ip1 are always adjacent. The targetLength skipping and
+ * uncompressibility acceleration is applied to every other position,
+ * matching the behavior of #1562. step therefore represents the gap
+ * between pairs of positions, from ip0 to ip2 or ip1 to ip3. */
+ size_t step;
+ const BYTE* nextStep;
+ const size_t kStepIncr = (1 << (kSearchStrength - 1));
+
+ DEBUGLOG(5, "ZSTD_compressBlock_fast_generic");
+ ip0 += (ip0 == prefixStart);
+ { U32 const curr = (U32)(ip0 - base);
+ U32 const windowLow = ZSTD_getLowestPrefixIndex(ms, curr, cParams->windowLog);
+ U32 const maxRep = curr - windowLow;
+ if (rep_offset2 > maxRep) offsetSaved = rep_offset2, rep_offset2 = 0;
+ if (rep_offset1 > maxRep) offsetSaved = rep_offset1, rep_offset1 = 0;
+ }
+
+ /* start each op */
+_start: /* Requires: ip0 */
+
+ step = stepSize;
+ nextStep = ip0 + kStepIncr;
+
+ /* calculate positions, ip0 - anchor == 0, so we skip step calc */
+ ip1 = ip0 + 1;
+ ip2 = ip0 + step;
+ ip3 = ip2 + 1;
+
+ if (ip3 >= ilimit) {
+ goto _cleanup;
+ }
+
+ hash0 = ZSTD_hashPtr(ip0, hlog, mls);
+ hash1 = ZSTD_hashPtr(ip1, hlog, mls);
+
+ idx = hashTable[hash0];
+
+ do {
+ /* load repcode match for ip[2]*/
+ const U32 rval = MEM_read32(ip2 - rep_offset1);
+
+ /* write back hash table entry */
+ current0 = (U32)(ip0 - base);
+ hashTable[hash0] = current0;
+
+ /* check repcode at ip[2] */
+ if ((MEM_read32(ip2) == rval) & (rep_offset1 > 0)) {
+ ip0 = ip2;
+ match0 = ip0 - rep_offset1;
+ mLength = ip0[-1] == match0[-1];
+ ip0 -= mLength;
+ match0 -= mLength;
+ offcode = STORE_REPCODE_1;
+ mLength += 4;
+ goto _match;
+ }
+
+ /* load match for ip[0] */
+ if (idx >= prefixStartIndex) {
+ mval = MEM_read32(base + idx);
+ } else {
+ mval = MEM_read32(ip0) ^ 1; /* guaranteed to not match. */
+ }
+
+ /* check match at ip[0] */
+ if (MEM_read32(ip0) == mval) {
+ /* found a match! */
+ goto _offset;
+ }
+
+ /* lookup ip[1] */
+ idx = hashTable[hash1];
+
+ /* hash ip[2] */
+ hash0 = hash1;
+ hash1 = ZSTD_hashPtr(ip2, hlog, mls);
+
+ /* advance to next positions */
+ ip0 = ip1;
+ ip1 = ip2;
+ ip2 = ip3;
+
+ /* write back hash table entry */
+ current0 = (U32)(ip0 - base);
+ hashTable[hash0] = current0;
+
+ /* load match for ip[0] */
+ if (idx >= prefixStartIndex) {
+ mval = MEM_read32(base + idx);
+ } else {
+ mval = MEM_read32(ip0) ^ 1; /* guaranteed to not match. */
+ }
+
+ /* check match at ip[0] */
+ if (MEM_read32(ip0) == mval) {
+ /* found a match! */
+ goto _offset;
+ }
+
+ /* lookup ip[1] */
+ idx = hashTable[hash1];
+
+ /* hash ip[2] */
+ hash0 = hash1;
+ hash1 = ZSTD_hashPtr(ip2, hlog, mls);
+
+ /* advance to next positions */
+ ip0 = ip1;
+ ip1 = ip2;
+ ip2 = ip0 + step;
+ ip3 = ip1 + step;
+
+ /* calculate step */
+ if (ip2 >= nextStep) {
+ step++;
+ PREFETCH_L1(ip1 + 64);
+ PREFETCH_L1(ip1 + 128);
+ nextStep += kStepIncr;
+ }
+ } while (ip3 < ilimit);
+
+_cleanup:
+ /* Note that there are probably still a couple positions we could search.
+ * However, it seems to be a meaningful performance hit to try to search
+ * them. So let's not. */
+
+ /* save reps for next block */
+ rep[0] = rep_offset1 ? rep_offset1 : offsetSaved;
+ rep[1] = rep_offset2 ? rep_offset2 : offsetSaved;
+
+ /* Return the last literals size */
+ return (size_t)(iend - anchor);
+
+_offset: /* Requires: ip0, idx */
+
+ /* Compute the offset code. */
+ match0 = base + idx;
+ rep_offset2 = rep_offset1;
+ rep_offset1 = (U32)(ip0-match0);
+ offcode = STORE_OFFSET(rep_offset1);
+ mLength = 4;
+
+ /* Count the backwards match length. */
+ while (((ip0>anchor) & (match0>prefixStart)) && (ip0[-1] == match0[-1])) {
+ ip0--;
+ match0--;
+ mLength++;
+ }
+
+_match: /* Requires: ip0, match0, offcode */
+
+ /* Count the forward length. */
+ mLength += ZSTD_count(ip0 + mLength, match0 + mLength, iend);
+
+ ZSTD_storeSeq(seqStore, (size_t)(ip0 - anchor), anchor, iend, offcode, mLength);
+
+ ip0 += mLength;
+ anchor = ip0;
+
+ /* write next hash table entry */
+ if (ip1 < ip0) {
+ hashTable[hash1] = (U32)(ip1 - base);
+ }
+
+ /* Fill table and check for immediate repcode. */
+ if (ip0 <= ilimit) {
+ /* Fill Table */
+ assert(base+current0+2 > istart); /* check base overflow */
+ hashTable[ZSTD_hashPtr(base+current0+2, hlog, mls)] = current0+2; /* here because current+2 could be > iend-8 */
+ hashTable[ZSTD_hashPtr(ip0-2, hlog, mls)] = (U32)(ip0-2-base);
+
+ if (rep_offset2 > 0) { /* rep_offset2==0 means rep_offset2 is invalidated */
+ while ( (ip0 <= ilimit) && (MEM_read32(ip0) == MEM_read32(ip0 - rep_offset2)) ) {
+ /* store sequence */
+ size_t const rLength = ZSTD_count(ip0+4, ip0+4-rep_offset2, iend) + 4;
+ { U32 const tmpOff = rep_offset2; rep_offset2 = rep_offset1; rep_offset1 = tmpOff; } /* swap rep_offset2 <=> rep_offset1 */
+ hashTable[ZSTD_hashPtr(ip0, hlog, mls)] = (U32)(ip0-base);
+ ip0 += rLength;
+ ZSTD_storeSeq(seqStore, 0 /*litLen*/, anchor, iend, STORE_REPCODE_1, rLength);
+ anchor = ip0;
+ continue; /* faster when present (confirmed on gcc-8) ... (?) */
+ } } }
+
+ goto _start;
+}
+
+#define ZSTD_GEN_FAST_FN(dictMode, mls, step) \
+ static size_t ZSTD_compressBlock_fast_##dictMode##_##mls##_##step( \
+ ZSTD_matchState_t* ms, seqStore_t* seqStore, U32 rep[ZSTD_REP_NUM], \
+ void const* src, size_t srcSize) \
+ { \
+ return ZSTD_compressBlock_fast_##dictMode##_generic(ms, seqStore, rep, src, srcSize, mls, step); \
+ }
+
+ZSTD_GEN_FAST_FN(noDict, 4, 1)
+ZSTD_GEN_FAST_FN(noDict, 5, 1)
+ZSTD_GEN_FAST_FN(noDict, 6, 1)
+ZSTD_GEN_FAST_FN(noDict, 7, 1)
+
+ZSTD_GEN_FAST_FN(noDict, 4, 0)
+ZSTD_GEN_FAST_FN(noDict, 5, 0)
+ZSTD_GEN_FAST_FN(noDict, 6, 0)
+ZSTD_GEN_FAST_FN(noDict, 7, 0)
+
+size_t ZSTD_compressBlock_fast(
+ ZSTD_matchState_t* ms, seqStore_t* seqStore, U32 rep[ZSTD_REP_NUM],
+ void const* src, size_t srcSize)
+{
+ U32 const mls = ms->cParams.minMatch;
+ assert(ms->dictMatchState == NULL);
+ if (ms->cParams.targetLength > 1) {
+ switch(mls)
+ {
+ default: /* includes case 3 */
+ case 4 :
+ return ZSTD_compressBlock_fast_noDict_4_1(ms, seqStore, rep, src, srcSize);
+ case 5 :
+ return ZSTD_compressBlock_fast_noDict_5_1(ms, seqStore, rep, src, srcSize);
+ case 6 :
+ return ZSTD_compressBlock_fast_noDict_6_1(ms, seqStore, rep, src, srcSize);
+ case 7 :
+ return ZSTD_compressBlock_fast_noDict_7_1(ms, seqStore, rep, src, srcSize);
+ }
+ } else {
+ switch(mls)
+ {
+ default: /* includes case 3 */
+ case 4 :
+ return ZSTD_compressBlock_fast_noDict_4_0(ms, seqStore, rep, src, srcSize);
+ case 5 :
+ return ZSTD_compressBlock_fast_noDict_5_0(ms, seqStore, rep, src, srcSize);
+ case 6 :
+ return ZSTD_compressBlock_fast_noDict_6_0(ms, seqStore, rep, src, srcSize);
+ case 7 :
+ return ZSTD_compressBlock_fast_noDict_7_0(ms, seqStore, rep, src, srcSize);
+ }
+
+ }
+}
+
+FORCE_INLINE_TEMPLATE
+size_t ZSTD_compressBlock_fast_dictMatchState_generic(
+ ZSTD_matchState_t* ms, seqStore_t* seqStore, U32 rep[ZSTD_REP_NUM],
+ void const* src, size_t srcSize, U32 const mls, U32 const hasStep)
+{
+ const ZSTD_compressionParameters* const cParams = &ms->cParams;
+ U32* const hashTable = ms->hashTable;
+ U32 const hlog = cParams->hashLog;
+ /* support stepSize of 0 */
+ U32 const stepSize = cParams->targetLength + !(cParams->targetLength);
+ const BYTE* const base = ms->window.base;
+ const BYTE* const istart = (const BYTE*)src;
+ const BYTE* ip = istart;
+ const BYTE* anchor = istart;
+ const U32 prefixStartIndex = ms->window.dictLimit;
+ const BYTE* const prefixStart = base + prefixStartIndex;
+ const BYTE* const iend = istart + srcSize;
+ const BYTE* const ilimit = iend - HASH_READ_SIZE;
+ U32 offset_1=rep[0], offset_2=rep[1];
+ U32 offsetSaved = 0;
+
+ const ZSTD_matchState_t* const dms = ms->dictMatchState;
+ const ZSTD_compressionParameters* const dictCParams = &dms->cParams ;
+ const U32* const dictHashTable = dms->hashTable;
+ const U32 dictStartIndex = dms->window.dictLimit;
+ const BYTE* const dictBase = dms->window.base;
+ const BYTE* const dictStart = dictBase + dictStartIndex;
+ const BYTE* const dictEnd = dms->window.nextSrc;
+ const U32 dictIndexDelta = prefixStartIndex - (U32)(dictEnd - dictBase);
+ const U32 dictAndPrefixLength = (U32)(ip - prefixStart + dictEnd - dictStart);
+ const U32 dictHLog = dictCParams->hashLog;
+
+ /* if a dictionary is still attached, it necessarily means that
+ * it is within window size. So we just check it. */
+ const U32 maxDistance = 1U << cParams->windowLog;
+ const U32 endIndex = (U32)((size_t)(ip - base) + srcSize);
+ assert(endIndex - prefixStartIndex <= maxDistance);
+ (void)maxDistance; (void)endIndex; /* these variables are not used when assert() is disabled */
+
+ (void)hasStep; /* not currently specialized on whether it's accelerated */
+
+ /* ensure there will be no underflow
+ * when translating a dict index into a local index */
+ assert(prefixStartIndex >= (U32)(dictEnd - dictBase));
+
+ /* init */
+ DEBUGLOG(5, "ZSTD_compressBlock_fast_dictMatchState_generic");
+ ip += (dictAndPrefixLength == 0);
+ /* dictMatchState repCode checks don't currently handle repCode == 0
+ * disabling. */
+ assert(offset_1 <= dictAndPrefixLength);
+ assert(offset_2 <= dictAndPrefixLength);
+
+ /* Main Search Loop */
+ while (ip < ilimit) { /* < instead of <=, because repcode check at (ip+1) */
+ size_t mLength;
+ size_t const h = ZSTD_hashPtr(ip, hlog, mls);
+ U32 const curr = (U32)(ip-base);
+ U32 const matchIndex = hashTable[h];
+ const BYTE* match = base + matchIndex;
+ const U32 repIndex = curr + 1 - offset_1;
+ const BYTE* repMatch = (repIndex < prefixStartIndex) ?
+ dictBase + (repIndex - dictIndexDelta) :
+ base + repIndex;
+ hashTable[h] = curr; /* update hash table */
+
+ if ( ((U32)((prefixStartIndex-1) - repIndex) >= 3) /* intentional underflow : ensure repIndex isn't overlapping dict + prefix */
+ && (MEM_read32(repMatch) == MEM_read32(ip+1)) ) {
+ const BYTE* const repMatchEnd = repIndex < prefixStartIndex ? dictEnd : iend;
+ mLength = ZSTD_count_2segments(ip+1+4, repMatch+4, iend, repMatchEnd, prefixStart) + 4;
+ ip++;
+ ZSTD_storeSeq(seqStore, (size_t)(ip-anchor), anchor, iend, STORE_REPCODE_1, mLength);
+ } else if ( (matchIndex <= prefixStartIndex) ) {
+ size_t const dictHash = ZSTD_hashPtr(ip, dictHLog, mls);
+ U32 const dictMatchIndex = dictHashTable[dictHash];
+ const BYTE* dictMatch = dictBase + dictMatchIndex;
+ if (dictMatchIndex <= dictStartIndex ||
+ MEM_read32(dictMatch) != MEM_read32(ip)) {
+ assert(stepSize >= 1);
+ ip += ((ip-anchor) >> kSearchStrength) + stepSize;
+ continue;
+ } else {
+ /* found a dict match */
+ U32 const offset = (U32)(curr-dictMatchIndex-dictIndexDelta);
+ mLength = ZSTD_count_2segments(ip+4, dictMatch+4, iend, dictEnd, prefixStart) + 4;
+ while (((ip>anchor) & (dictMatch>dictStart))
+ && (ip[-1] == dictMatch[-1])) {
+ ip--; dictMatch--; mLength++;
+ } /* catch up */
+ offset_2 = offset_1;
+ offset_1 = offset;
+ ZSTD_storeSeq(seqStore, (size_t)(ip-anchor), anchor, iend, STORE_OFFSET(offset), mLength);
+ }
+ } else if (MEM_read32(match) != MEM_read32(ip)) {
+ /* it's not a match, and we're not going to check the dictionary */
+ assert(stepSize >= 1);
+ ip += ((ip-anchor) >> kSearchStrength) + stepSize;
+ continue;
+ } else {
+ /* found a regular match */
+ U32 const offset = (U32)(ip-match);
+ mLength = ZSTD_count(ip+4, match+4, iend) + 4;
+ while (((ip>anchor) & (match>prefixStart))
+ && (ip[-1] == match[-1])) { ip--; match--; mLength++; } /* catch up */
+ offset_2 = offset_1;
+ offset_1 = offset;
+ ZSTD_storeSeq(seqStore, (size_t)(ip-anchor), anchor, iend, STORE_OFFSET(offset), mLength);
+ }
+
+ /* match found */
+ ip += mLength;
+ anchor = ip;
+
+ if (ip <= ilimit) {
+ /* Fill Table */
+ assert(base+curr+2 > istart); /* check base overflow */
+ hashTable[ZSTD_hashPtr(base+curr+2, hlog, mls)] = curr+2; /* here because curr+2 could be > iend-8 */
+ hashTable[ZSTD_hashPtr(ip-2, hlog, mls)] = (U32)(ip-2-base);
+
+ /* check immediate repcode */
+ while (ip <= ilimit) {
+ U32 const current2 = (U32)(ip-base);
+ U32 const repIndex2 = current2 - offset_2;
+ const BYTE* repMatch2 = repIndex2 < prefixStartIndex ?
+ dictBase - dictIndexDelta + repIndex2 :
+ base + repIndex2;
+ if ( ((U32)((prefixStartIndex-1) - (U32)repIndex2) >= 3 /* intentional overflow */)
+ && (MEM_read32(repMatch2) == MEM_read32(ip)) ) {
+ const BYTE* const repEnd2 = repIndex2 < prefixStartIndex ? dictEnd : iend;
+ size_t const repLength2 = ZSTD_count_2segments(ip+4, repMatch2+4, iend, repEnd2, prefixStart) + 4;
+ U32 tmpOffset = offset_2; offset_2 = offset_1; offset_1 = tmpOffset; /* swap offset_2 <=> offset_1 */
+ ZSTD_storeSeq(seqStore, 0, anchor, iend, STORE_REPCODE_1, repLength2);
+ hashTable[ZSTD_hashPtr(ip, hlog, mls)] = current2;
+ ip += repLength2;
+ anchor = ip;
+ continue;
+ }
+ break;
+ }
+ }
+ }
+
+ /* save reps for next block */
+ rep[0] = offset_1 ? offset_1 : offsetSaved;
+ rep[1] = offset_2 ? offset_2 : offsetSaved;
+
+ /* Return the last literals size */
+ return (size_t)(iend - anchor);
+}
+
+
+ZSTD_GEN_FAST_FN(dictMatchState, 4, 0)
+ZSTD_GEN_FAST_FN(dictMatchState, 5, 0)
+ZSTD_GEN_FAST_FN(dictMatchState, 6, 0)
+ZSTD_GEN_FAST_FN(dictMatchState, 7, 0)
+
+size_t ZSTD_compressBlock_fast_dictMatchState(
+ ZSTD_matchState_t* ms, seqStore_t* seqStore, U32 rep[ZSTD_REP_NUM],
+ void const* src, size_t srcSize)
+{
+ U32 const mls = ms->cParams.minMatch;
+ assert(ms->dictMatchState != NULL);
+ switch(mls)
+ {
+ default: /* includes case 3 */
+ case 4 :
+ return ZSTD_compressBlock_fast_dictMatchState_4_0(ms, seqStore, rep, src, srcSize);
+ case 5 :
+ return ZSTD_compressBlock_fast_dictMatchState_5_0(ms, seqStore, rep, src, srcSize);
+ case 6 :
+ return ZSTD_compressBlock_fast_dictMatchState_6_0(ms, seqStore, rep, src, srcSize);
+ case 7 :
+ return ZSTD_compressBlock_fast_dictMatchState_7_0(ms, seqStore, rep, src, srcSize);
+ }
+}
+
+
+static size_t ZSTD_compressBlock_fast_extDict_generic(
+ ZSTD_matchState_t* ms, seqStore_t* seqStore, U32 rep[ZSTD_REP_NUM],
+ void const* src, size_t srcSize, U32 const mls, U32 const hasStep)
+{
+ const ZSTD_compressionParameters* const cParams = &ms->cParams;
+ U32* const hashTable = ms->hashTable;
+ U32 const hlog = cParams->hashLog;
+ /* support stepSize of 0 */
+ U32 const stepSize = cParams->targetLength + !(cParams->targetLength);
+ const BYTE* const base = ms->window.base;
+ const BYTE* const dictBase = ms->window.dictBase;
+ const BYTE* const istart = (const BYTE*)src;
+ const BYTE* ip = istart;
+ const BYTE* anchor = istart;
+ const U32 endIndex = (U32)((size_t)(istart - base) + srcSize);
+ const U32 lowLimit = ZSTD_getLowestMatchIndex(ms, endIndex, cParams->windowLog);
+ const U32 dictStartIndex = lowLimit;
+ const BYTE* const dictStart = dictBase + dictStartIndex;
+ const U32 dictLimit = ms->window.dictLimit;
+ const U32 prefixStartIndex = dictLimit < lowLimit ? lowLimit : dictLimit;
+ const BYTE* const prefixStart = base + prefixStartIndex;
+ const BYTE* const dictEnd = dictBase + prefixStartIndex;
+ const BYTE* const iend = istart + srcSize;
+ const BYTE* const ilimit = iend - 8;
+ U32 offset_1=rep[0], offset_2=rep[1];
+
+ (void)hasStep; /* not currently specialized on whether it's accelerated */
+
+ DEBUGLOG(5, "ZSTD_compressBlock_fast_extDict_generic (offset_1=%u)", offset_1);
+
+ /* switch to "regular" variant if extDict is invalidated due to maxDistance */
+ if (prefixStartIndex == dictStartIndex)
+ return ZSTD_compressBlock_fast(ms, seqStore, rep, src, srcSize);
+
+ /* Search Loop */
+ while (ip < ilimit) { /* < instead of <=, because (ip+1) */
+ const size_t h = ZSTD_hashPtr(ip, hlog, mls);
+ const U32 matchIndex = hashTable[h];
+ const BYTE* const matchBase = matchIndex < prefixStartIndex ? dictBase : base;
+ const BYTE* match = matchBase + matchIndex;
+ const U32 curr = (U32)(ip-base);
+ const U32 repIndex = curr + 1 - offset_1;
+ const BYTE* const repBase = repIndex < prefixStartIndex ? dictBase : base;
+ const BYTE* const repMatch = repBase + repIndex;
+ hashTable[h] = curr; /* update hash table */
+ DEBUGLOG(7, "offset_1 = %u , curr = %u", offset_1, curr);
+
+ if ( ( ((U32)((prefixStartIndex-1) - repIndex) >= 3) /* intentional underflow */
+ & (offset_1 <= curr+1 - dictStartIndex) ) /* note: we are searching at curr+1 */
+ && (MEM_read32(repMatch) == MEM_read32(ip+1)) ) {
+ const BYTE* const repMatchEnd = repIndex < prefixStartIndex ? dictEnd : iend;
+ size_t const rLength = ZSTD_count_2segments(ip+1 +4, repMatch +4, iend, repMatchEnd, prefixStart) + 4;
+ ip++;
+ ZSTD_storeSeq(seqStore, (size_t)(ip-anchor), anchor, iend, STORE_REPCODE_1, rLength);
+ ip += rLength;
+ anchor = ip;
+ } else {
+ if ( (matchIndex < dictStartIndex) ||
+ (MEM_read32(match) != MEM_read32(ip)) ) {
+ assert(stepSize >= 1);
+ ip += ((ip-anchor) >> kSearchStrength) + stepSize;
+ continue;
+ }
+ { const BYTE* const matchEnd = matchIndex < prefixStartIndex ? dictEnd : iend;
+ const BYTE* const lowMatchPtr = matchIndex < prefixStartIndex ? dictStart : prefixStart;
+ U32 const offset = curr - matchIndex;
+ size_t mLength = ZSTD_count_2segments(ip+4, match+4, iend, matchEnd, prefixStart) + 4;
+ while (((ip>anchor) & (match>lowMatchPtr)) && (ip[-1] == match[-1])) { ip--; match--; mLength++; } /* catch up */
+ offset_2 = offset_1; offset_1 = offset; /* update offset history */
+ ZSTD_storeSeq(seqStore, (size_t)(ip-anchor), anchor, iend, STORE_OFFSET(offset), mLength);
+ ip += mLength;
+ anchor = ip;
+ } }
+
+ if (ip <= ilimit) {
+ /* Fill Table */
+ hashTable[ZSTD_hashPtr(base+curr+2, hlog, mls)] = curr+2;
+ hashTable[ZSTD_hashPtr(ip-2, hlog, mls)] = (U32)(ip-2-base);
+ /* check immediate repcode */
+ while (ip <= ilimit) {
+ U32 const current2 = (U32)(ip-base);
+ U32 const repIndex2 = current2 - offset_2;
+ const BYTE* const repMatch2 = repIndex2 < prefixStartIndex ? dictBase + repIndex2 : base + repIndex2;
+ if ( (((U32)((prefixStartIndex-1) - repIndex2) >= 3) & (offset_2 <= curr - dictStartIndex)) /* intentional overflow */
+ && (MEM_read32(repMatch2) == MEM_read32(ip)) ) {
+ const BYTE* const repEnd2 = repIndex2 < prefixStartIndex ? dictEnd : iend;
+ size_t const repLength2 = ZSTD_count_2segments(ip+4, repMatch2+4, iend, repEnd2, prefixStart) + 4;
+ { U32 const tmpOffset = offset_2; offset_2 = offset_1; offset_1 = tmpOffset; } /* swap offset_2 <=> offset_1 */
+ ZSTD_storeSeq(seqStore, 0 /*litlen*/, anchor, iend, STORE_REPCODE_1, repLength2);
+ hashTable[ZSTD_hashPtr(ip, hlog, mls)] = current2;
+ ip += repLength2;
+ anchor = ip;
+ continue;
+ }
+ break;
+ } } }
+
+ /* save reps for next block */
+ rep[0] = offset_1;
+ rep[1] = offset_2;
+
+ /* Return the last literals size */
+ return (size_t)(iend - anchor);
+}
+
+ZSTD_GEN_FAST_FN(extDict, 4, 0)
+ZSTD_GEN_FAST_FN(extDict, 5, 0)
+ZSTD_GEN_FAST_FN(extDict, 6, 0)
+ZSTD_GEN_FAST_FN(extDict, 7, 0)
+
+size_t ZSTD_compressBlock_fast_extDict(
+ ZSTD_matchState_t* ms, seqStore_t* seqStore, U32 rep[ZSTD_REP_NUM],
+ void const* src, size_t srcSize)
+{
+ U32 const mls = ms->cParams.minMatch;
+ switch(mls)
+ {
+ default: /* includes case 3 */
+ case 4 :
+ return ZSTD_compressBlock_fast_extDict_4_0(ms, seqStore, rep, src, srcSize);
+ case 5 :
+ return ZSTD_compressBlock_fast_extDict_5_0(ms, seqStore, rep, src, srcSize);
+ case 6 :
+ return ZSTD_compressBlock_fast_extDict_6_0(ms, seqStore, rep, src, srcSize);
+ case 7 :
+ return ZSTD_compressBlock_fast_extDict_7_0(ms, seqStore, rep, src, srcSize);
+ }
+}
diff --git a/lib/zstd/compress/zstd_fast.h b/lib/zstd/compress/zstd_fast.h
new file mode 100644
index 0000000000..fddc2f532d
--- /dev/null
+++ b/lib/zstd/compress/zstd_fast.h
@@ -0,0 +1,31 @@
+/*
+ * Copyright (c) Yann Collet, Facebook, Inc.
+ * All rights reserved.
+ *
+ * This source code is licensed under both the BSD-style license (found in the
+ * LICENSE file in the root directory of this source tree) and the GPLv2 (found
+ * in the COPYING file in the root directory of this source tree).
+ * You may select, at your option, one of the above-listed licenses.
+ */
+
+#ifndef ZSTD_FAST_H
+#define ZSTD_FAST_H
+
+
+#include "../common/mem.h" /* U32 */
+#include "zstd_compress_internal.h"
+
+void ZSTD_fillHashTable(ZSTD_matchState_t* ms,
+ void const* end, ZSTD_dictTableLoadMethod_e dtlm);
+size_t ZSTD_compressBlock_fast(
+ ZSTD_matchState_t* ms, seqStore_t* seqStore, U32 rep[ZSTD_REP_NUM],
+ void const* src, size_t srcSize);
+size_t ZSTD_compressBlock_fast_dictMatchState(
+ ZSTD_matchState_t* ms, seqStore_t* seqStore, U32 rep[ZSTD_REP_NUM],
+ void const* src, size_t srcSize);
+size_t ZSTD_compressBlock_fast_extDict(
+ ZSTD_matchState_t* ms, seqStore_t* seqStore, U32 rep[ZSTD_REP_NUM],
+ void const* src, size_t srcSize);
+
+
+#endif /* ZSTD_FAST_H */
diff --git a/lib/zstd/compress/zstd_lazy.c b/lib/zstd/compress/zstd_lazy.c
new file mode 100644
index 0000000000..0298a01a75
--- /dev/null
+++ b/lib/zstd/compress/zstd_lazy.c
@@ -0,0 +1,2102 @@
+/*
+ * Copyright (c) Yann Collet, Facebook, Inc.
+ * All rights reserved.
+ *
+ * This source code is licensed under both the BSD-style license (found in the
+ * LICENSE file in the root directory of this source tree) and the GPLv2 (found
+ * in the COPYING file in the root directory of this source tree).
+ * You may select, at your option, one of the above-listed licenses.
+ */
+
+#include "zstd_compress_internal.h"
+#include "zstd_lazy.h"
+
+
+/*-*************************************
+* Binary Tree search
+***************************************/
+
+static void
+ZSTD_updateDUBT(ZSTD_matchState_t* ms,
+ const BYTE* ip, const BYTE* iend,
+ U32 mls)
+{
+ const ZSTD_compressionParameters* const cParams = &ms->cParams;
+ U32* const hashTable = ms->hashTable;
+ U32 const hashLog = cParams->hashLog;
+
+ U32* const bt = ms->chainTable;
+ U32 const btLog = cParams->chainLog - 1;
+ U32 const btMask = (1 << btLog) - 1;
+
+ const BYTE* const base = ms->window.base;
+ U32 const target = (U32)(ip - base);
+ U32 idx = ms->nextToUpdate;
+
+ if (idx != target)
+ DEBUGLOG(7, "ZSTD_updateDUBT, from %u to %u (dictLimit:%u)",
+ idx, target, ms->window.dictLimit);
+ assert(ip + 8 <= iend); /* condition for ZSTD_hashPtr */
+ (void)iend;
+
+ assert(idx >= ms->window.dictLimit); /* condition for valid base+idx */
+ for ( ; idx < target ; idx++) {
+ size_t const h = ZSTD_hashPtr(base + idx, hashLog, mls); /* assumption : ip + 8 <= iend */
+ U32 const matchIndex = hashTable[h];
+
+ U32* const nextCandidatePtr = bt + 2*(idx&btMask);
+ U32* const sortMarkPtr = nextCandidatePtr + 1;
+
+ DEBUGLOG(8, "ZSTD_updateDUBT: insert %u", idx);
+ hashTable[h] = idx; /* Update Hash Table */
+ *nextCandidatePtr = matchIndex; /* update BT like a chain */
+ *sortMarkPtr = ZSTD_DUBT_UNSORTED_MARK;
+ }
+ ms->nextToUpdate = target;
+}
+
+
+/* ZSTD_insertDUBT1() :
+ * sort one already inserted but unsorted position
+ * assumption : curr >= btlow == (curr - btmask)
+ * doesn't fail */
+static void
+ZSTD_insertDUBT1(const ZSTD_matchState_t* ms,
+ U32 curr, const BYTE* inputEnd,
+ U32 nbCompares, U32 btLow,
+ const ZSTD_dictMode_e dictMode)
+{
+ const ZSTD_compressionParameters* const cParams = &ms->cParams;
+ U32* const bt = ms->chainTable;
+ U32 const btLog = cParams->chainLog - 1;
+ U32 const btMask = (1 << btLog) - 1;
+ size_t commonLengthSmaller=0, commonLengthLarger=0;
+ const BYTE* const base = ms->window.base;
+ const BYTE* const dictBase = ms->window.dictBase;
+ const U32 dictLimit = ms->window.dictLimit;
+ const BYTE* const ip = (curr>=dictLimit) ? base + curr : dictBase + curr;
+ const BYTE* const iend = (curr>=dictLimit) ? inputEnd : dictBase + dictLimit;
+ const BYTE* const dictEnd = dictBase + dictLimit;
+ const BYTE* const prefixStart = base + dictLimit;
+ const BYTE* match;
+ U32* smallerPtr = bt + 2*(curr&btMask);
+ U32* largerPtr = smallerPtr + 1;
+ U32 matchIndex = *smallerPtr; /* this candidate is unsorted : next sorted candidate is reached through *smallerPtr, while *largerPtr contains previous unsorted candidate (which is already saved and can be overwritten) */
+ U32 dummy32; /* to be nullified at the end */
+ U32 const windowValid = ms->window.lowLimit;
+ U32 const maxDistance = 1U << cParams->windowLog;
+ U32 const windowLow = (curr - windowValid > maxDistance) ? curr - maxDistance : windowValid;
+
+
+ DEBUGLOG(8, "ZSTD_insertDUBT1(%u) (dictLimit=%u, lowLimit=%u)",
+ curr, dictLimit, windowLow);
+ assert(curr >= btLow);
+ assert(ip < iend); /* condition for ZSTD_count */
+
+ for (; nbCompares && (matchIndex > windowLow); --nbCompares) {
+ U32* const nextPtr = bt + 2*(matchIndex & btMask);
+ size_t matchLength = MIN(commonLengthSmaller, commonLengthLarger); /* guaranteed minimum nb of common bytes */
+ assert(matchIndex < curr);
+ /* note : all candidates are now supposed sorted,
+ * but it's still possible to have nextPtr[1] == ZSTD_DUBT_UNSORTED_MARK
+ * when a real index has the same value as ZSTD_DUBT_UNSORTED_MARK */
+
+ if ( (dictMode != ZSTD_extDict)
+ || (matchIndex+matchLength >= dictLimit) /* both in current segment*/
+ || (curr < dictLimit) /* both in extDict */) {
+ const BYTE* const mBase = ( (dictMode != ZSTD_extDict)
+ || (matchIndex+matchLength >= dictLimit)) ?
+ base : dictBase;
+ assert( (matchIndex+matchLength >= dictLimit) /* might be wrong if extDict is incorrectly set to 0 */
+ || (curr < dictLimit) );
+ match = mBase + matchIndex;
+ matchLength += ZSTD_count(ip+matchLength, match+matchLength, iend);
+ } else {
+ match = dictBase + matchIndex;
+ matchLength += ZSTD_count_2segments(ip+matchLength, match+matchLength, iend, dictEnd, prefixStart);
+ if (matchIndex+matchLength >= dictLimit)
+ match = base + matchIndex; /* preparation for next read of match[matchLength] */
+ }
+
+ DEBUGLOG(8, "ZSTD_insertDUBT1: comparing %u with %u : found %u common bytes ",
+ curr, matchIndex, (U32)matchLength);
+
+ if (ip+matchLength == iend) { /* equal : no way to know if inf or sup */
+ break; /* drop , to guarantee consistency ; miss a bit of compression, but other solutions can corrupt tree */
+ }
+
+ if (match[matchLength] < ip[matchLength]) { /* necessarily within buffer */
+ /* match is smaller than current */
+ *smallerPtr = matchIndex; /* update smaller idx */
+ commonLengthSmaller = matchLength; /* all smaller will now have at least this guaranteed common length */
+ if (matchIndex <= btLow) { smallerPtr=&dummy32; break; } /* beyond tree size, stop searching */
+ DEBUGLOG(8, "ZSTD_insertDUBT1: %u (>btLow=%u) is smaller : next => %u",
+ matchIndex, btLow, nextPtr[1]);
+ smallerPtr = nextPtr+1; /* new "candidate" => larger than match, which was smaller than target */
+ matchIndex = nextPtr[1]; /* new matchIndex, larger than previous and closer to current */
+ } else {
+ /* match is larger than current */
+ *largerPtr = matchIndex;
+ commonLengthLarger = matchLength;
+ if (matchIndex <= btLow) { largerPtr=&dummy32; break; } /* beyond tree size, stop searching */
+ DEBUGLOG(8, "ZSTD_insertDUBT1: %u (>btLow=%u) is larger => %u",
+ matchIndex, btLow, nextPtr[0]);
+ largerPtr = nextPtr;
+ matchIndex = nextPtr[0];
+ } }
+
+ *smallerPtr = *largerPtr = 0;
+}
+
+
+static size_t
+ZSTD_DUBT_findBetterDictMatch (
+ const ZSTD_matchState_t* ms,
+ const BYTE* const ip, const BYTE* const iend,
+ size_t* offsetPtr,
+ size_t bestLength,
+ U32 nbCompares,
+ U32 const mls,
+ const ZSTD_dictMode_e dictMode)
+{
+ const ZSTD_matchState_t * const dms = ms->dictMatchState;
+ const ZSTD_compressionParameters* const dmsCParams = &dms->cParams;
+ const U32 * const dictHashTable = dms->hashTable;
+ U32 const hashLog = dmsCParams->hashLog;
+ size_t const h = ZSTD_hashPtr(ip, hashLog, mls);
+ U32 dictMatchIndex = dictHashTable[h];
+
+ const BYTE* const base = ms->window.base;
+ const BYTE* const prefixStart = base + ms->window.dictLimit;
+ U32 const curr = (U32)(ip-base);
+ const BYTE* const dictBase = dms->window.base;
+ const BYTE* const dictEnd = dms->window.nextSrc;
+ U32 const dictHighLimit = (U32)(dms->window.nextSrc - dms->window.base);
+ U32 const dictLowLimit = dms->window.lowLimit;
+ U32 const dictIndexDelta = ms->window.lowLimit - dictHighLimit;
+
+ U32* const dictBt = dms->chainTable;
+ U32 const btLog = dmsCParams->chainLog - 1;
+ U32 const btMask = (1 << btLog) - 1;
+ U32 const btLow = (btMask >= dictHighLimit - dictLowLimit) ? dictLowLimit : dictHighLimit - btMask;
+
+ size_t commonLengthSmaller=0, commonLengthLarger=0;
+
+ (void)dictMode;
+ assert(dictMode == ZSTD_dictMatchState);
+
+ for (; nbCompares && (dictMatchIndex > dictLowLimit); --nbCompares) {
+ U32* const nextPtr = dictBt + 2*(dictMatchIndex & btMask);
+ size_t matchLength = MIN(commonLengthSmaller, commonLengthLarger); /* guaranteed minimum nb of common bytes */
+ const BYTE* match = dictBase + dictMatchIndex;
+ matchLength += ZSTD_count_2segments(ip+matchLength, match+matchLength, iend, dictEnd, prefixStart);
+ if (dictMatchIndex+matchLength >= dictHighLimit)
+ match = base + dictMatchIndex + dictIndexDelta; /* to prepare for next usage of match[matchLength] */
+
+ if (matchLength > bestLength) {
+ U32 matchIndex = dictMatchIndex + dictIndexDelta;
+ if ( (4*(int)(matchLength-bestLength)) > (int)(ZSTD_highbit32(curr-matchIndex+1) - ZSTD_highbit32((U32)offsetPtr[0]+1)) ) {
+ DEBUGLOG(9, "ZSTD_DUBT_findBetterDictMatch(%u) : found better match length %u -> %u and offsetCode %u -> %u (dictMatchIndex %u, matchIndex %u)",
+ curr, (U32)bestLength, (U32)matchLength, (U32)*offsetPtr, STORE_OFFSET(curr - matchIndex), dictMatchIndex, matchIndex);
+ bestLength = matchLength, *offsetPtr = STORE_OFFSET(curr - matchIndex);
+ }
+ if (ip+matchLength == iend) { /* reached end of input : ip[matchLength] is not valid, no way to know if it's larger or smaller than match */
+ break; /* drop, to guarantee consistency (miss a little bit of compression) */
+ }
+ }
+
+ if (match[matchLength] < ip[matchLength]) {
+ if (dictMatchIndex <= btLow) { break; } /* beyond tree size, stop the search */
+ commonLengthSmaller = matchLength; /* all smaller will now have at least this guaranteed common length */
+ dictMatchIndex = nextPtr[1]; /* new matchIndex larger than previous (closer to current) */
+ } else {
+ /* match is larger than current */
+ if (dictMatchIndex <= btLow) { break; } /* beyond tree size, stop the search */
+ commonLengthLarger = matchLength;
+ dictMatchIndex = nextPtr[0];
+ }
+ }
+
+ if (bestLength >= MINMATCH) {
+ U32 const mIndex = curr - (U32)STORED_OFFSET(*offsetPtr); (void)mIndex;
+ DEBUGLOG(8, "ZSTD_DUBT_findBetterDictMatch(%u) : found match of length %u and offsetCode %u (pos %u)",
+ curr, (U32)bestLength, (U32)*offsetPtr, mIndex);
+ }
+ return bestLength;
+
+}
+
+
+static size_t
+ZSTD_DUBT_findBestMatch(ZSTD_matchState_t* ms,
+ const BYTE* const ip, const BYTE* const iend,
+ size_t* offsetPtr,
+ U32 const mls,
+ const ZSTD_dictMode_e dictMode)
+{
+ const ZSTD_compressionParameters* const cParams = &ms->cParams;
+ U32* const hashTable = ms->hashTable;
+ U32 const hashLog = cParams->hashLog;
+ size_t const h = ZSTD_hashPtr(ip, hashLog, mls);
+ U32 matchIndex = hashTable[h];
+
+ const BYTE* const base = ms->window.base;
+ U32 const curr = (U32)(ip-base);
+ U32 const windowLow = ZSTD_getLowestMatchIndex(ms, curr, cParams->windowLog);
+
+ U32* const bt = ms->chainTable;
+ U32 const btLog = cParams->chainLog - 1;
+ U32 const btMask = (1 << btLog) - 1;
+ U32 const btLow = (btMask >= curr) ? 0 : curr - btMask;
+ U32 const unsortLimit = MAX(btLow, windowLow);
+
+ U32* nextCandidate = bt + 2*(matchIndex&btMask);
+ U32* unsortedMark = bt + 2*(matchIndex&btMask) + 1;
+ U32 nbCompares = 1U << cParams->searchLog;
+ U32 nbCandidates = nbCompares;
+ U32 previousCandidate = 0;
+
+ DEBUGLOG(7, "ZSTD_DUBT_findBestMatch (%u) ", curr);
+ assert(ip <= iend-8); /* required for h calculation */
+ assert(dictMode != ZSTD_dedicatedDictSearch);
+
+ /* reach end of unsorted candidates list */
+ while ( (matchIndex > unsortLimit)
+ && (*unsortedMark == ZSTD_DUBT_UNSORTED_MARK)
+ && (nbCandidates > 1) ) {
+ DEBUGLOG(8, "ZSTD_DUBT_findBestMatch: candidate %u is unsorted",
+ matchIndex);
+ *unsortedMark = previousCandidate; /* the unsortedMark becomes a reversed chain, to move up back to original position */
+ previousCandidate = matchIndex;
+ matchIndex = *nextCandidate;
+ nextCandidate = bt + 2*(matchIndex&btMask);
+ unsortedMark = bt + 2*(matchIndex&btMask) + 1;
+ nbCandidates --;
+ }
+
+ /* nullify last candidate if it's still unsorted
+ * simplification, detrimental to compression ratio, beneficial for speed */
+ if ( (matchIndex > unsortLimit)
+ && (*unsortedMark==ZSTD_DUBT_UNSORTED_MARK) ) {
+ DEBUGLOG(7, "ZSTD_DUBT_findBestMatch: nullify last unsorted candidate %u",
+ matchIndex);
+ *nextCandidate = *unsortedMark = 0;
+ }
+
+ /* batch sort stacked candidates */
+ matchIndex = previousCandidate;
+ while (matchIndex) { /* will end on matchIndex == 0 */
+ U32* const nextCandidateIdxPtr = bt + 2*(matchIndex&btMask) + 1;
+ U32 const nextCandidateIdx = *nextCandidateIdxPtr;
+ ZSTD_insertDUBT1(ms, matchIndex, iend,
+ nbCandidates, unsortLimit, dictMode);
+ matchIndex = nextCandidateIdx;
+ nbCandidates++;
+ }
+
+ /* find longest match */
+ { size_t commonLengthSmaller = 0, commonLengthLarger = 0;
+ const BYTE* const dictBase = ms->window.dictBase;
+ const U32 dictLimit = ms->window.dictLimit;
+ const BYTE* const dictEnd = dictBase + dictLimit;
+ const BYTE* const prefixStart = base + dictLimit;
+ U32* smallerPtr = bt + 2*(curr&btMask);
+ U32* largerPtr = bt + 2*(curr&btMask) + 1;
+ U32 matchEndIdx = curr + 8 + 1;
+ U32 dummy32; /* to be nullified at the end */
+ size_t bestLength = 0;
+
+ matchIndex = hashTable[h];
+ hashTable[h] = curr; /* Update Hash Table */
+
+ for (; nbCompares && (matchIndex > windowLow); --nbCompares) {
+ U32* const nextPtr = bt + 2*(matchIndex & btMask);
+ size_t matchLength = MIN(commonLengthSmaller, commonLengthLarger); /* guaranteed minimum nb of common bytes */
+ const BYTE* match;
+
+ if ((dictMode != ZSTD_extDict) || (matchIndex+matchLength >= dictLimit)) {
+ match = base + matchIndex;
+ matchLength += ZSTD_count(ip+matchLength, match+matchLength, iend);
+ } else {
+ match = dictBase + matchIndex;
+ matchLength += ZSTD_count_2segments(ip+matchLength, match+matchLength, iend, dictEnd, prefixStart);
+ if (matchIndex+matchLength >= dictLimit)
+ match = base + matchIndex; /* to prepare for next usage of match[matchLength] */
+ }
+
+ if (matchLength > bestLength) {
+ if (matchLength > matchEndIdx - matchIndex)
+ matchEndIdx = matchIndex + (U32)matchLength;
+ if ( (4*(int)(matchLength-bestLength)) > (int)(ZSTD_highbit32(curr-matchIndex+1) - ZSTD_highbit32((U32)offsetPtr[0]+1)) )
+ bestLength = matchLength, *offsetPtr = STORE_OFFSET(curr - matchIndex);
+ if (ip+matchLength == iend) { /* equal : no way to know if inf or sup */
+ if (dictMode == ZSTD_dictMatchState) {
+ nbCompares = 0; /* in addition to avoiding checking any
+ * further in this loop, make sure we
+ * skip checking in the dictionary. */
+ }
+ break; /* drop, to guarantee consistency (miss a little bit of compression) */
+ }
+ }
+
+ if (match[matchLength] < ip[matchLength]) {
+ /* match is smaller than current */
+ *smallerPtr = matchIndex; /* update smaller idx */
+ commonLengthSmaller = matchLength; /* all smaller will now have at least this guaranteed common length */
+ if (matchIndex <= btLow) { smallerPtr=&dummy32; break; } /* beyond tree size, stop the search */
+ smallerPtr = nextPtr+1; /* new "smaller" => larger of match */
+ matchIndex = nextPtr[1]; /* new matchIndex larger than previous (closer to current) */
+ } else {
+ /* match is larger than current */
+ *largerPtr = matchIndex;
+ commonLengthLarger = matchLength;
+ if (matchIndex <= btLow) { largerPtr=&dummy32; break; } /* beyond tree size, stop the search */
+ largerPtr = nextPtr;
+ matchIndex = nextPtr[0];
+ } }
+
+ *smallerPtr = *largerPtr = 0;
+
+ assert(nbCompares <= (1U << ZSTD_SEARCHLOG_MAX)); /* Check we haven't underflowed. */
+ if (dictMode == ZSTD_dictMatchState && nbCompares) {
+ bestLength = ZSTD_DUBT_findBetterDictMatch(
+ ms, ip, iend,
+ offsetPtr, bestLength, nbCompares,
+ mls, dictMode);
+ }
+
+ assert(matchEndIdx > curr+8); /* ensure nextToUpdate is increased */
+ ms->nextToUpdate = matchEndIdx - 8; /* skip repetitive patterns */
+ if (bestLength >= MINMATCH) {
+ U32 const mIndex = curr - (U32)STORED_OFFSET(*offsetPtr); (void)mIndex;
+ DEBUGLOG(8, "ZSTD_DUBT_findBestMatch(%u) : found match of length %u and offsetCode %u (pos %u)",
+ curr, (U32)bestLength, (U32)*offsetPtr, mIndex);
+ }
+ return bestLength;
+ }
+}
+
+
+/* ZSTD_BtFindBestMatch() : Tree updater, providing best match */
+FORCE_INLINE_TEMPLATE size_t
+ZSTD_BtFindBestMatch( ZSTD_matchState_t* ms,
+ const BYTE* const ip, const BYTE* const iLimit,
+ size_t* offsetPtr,
+ const U32 mls /* template */,
+ const ZSTD_dictMode_e dictMode)
+{
+ DEBUGLOG(7, "ZSTD_BtFindBestMatch");
+ if (ip < ms->window.base + ms->nextToUpdate) return 0; /* skipped area */
+ ZSTD_updateDUBT(ms, ip, iLimit, mls);
+ return ZSTD_DUBT_findBestMatch(ms, ip, iLimit, offsetPtr, mls, dictMode);
+}
+
+/* *********************************
+* Dedicated dict search
+***********************************/
+
+void ZSTD_dedicatedDictSearch_lazy_loadDictionary(ZSTD_matchState_t* ms, const BYTE* const ip)
+{
+ const BYTE* const base = ms->window.base;
+ U32 const target = (U32)(ip - base);
+ U32* const hashTable = ms->hashTable;
+ U32* const chainTable = ms->chainTable;
+ U32 const chainSize = 1 << ms->cParams.chainLog;
+ U32 idx = ms->nextToUpdate;
+ U32 const minChain = chainSize < target - idx ? target - chainSize : idx;
+ U32 const bucketSize = 1 << ZSTD_LAZY_DDSS_BUCKET_LOG;
+ U32 const cacheSize = bucketSize - 1;
+ U32 const chainAttempts = (1 << ms->cParams.searchLog) - cacheSize;
+ U32 const chainLimit = chainAttempts > 255 ? 255 : chainAttempts;
+
+ /* We know the hashtable is oversized by a factor of `bucketSize`.
+ * We are going to temporarily pretend `bucketSize == 1`, keeping only a
+ * single entry. We will use the rest of the space to construct a temporary
+ * chaintable.
+ */
+ U32 const hashLog = ms->cParams.hashLog - ZSTD_LAZY_DDSS_BUCKET_LOG;
+ U32* const tmpHashTable = hashTable;
+ U32* const tmpChainTable = hashTable + ((size_t)1 << hashLog);
+ U32 const tmpChainSize = (U32)((1 << ZSTD_LAZY_DDSS_BUCKET_LOG) - 1) << hashLog;
+ U32 const tmpMinChain = tmpChainSize < target ? target - tmpChainSize : idx;
+ U32 hashIdx;
+
+ assert(ms->cParams.chainLog <= 24);
+ assert(ms->cParams.hashLog > ms->cParams.chainLog);
+ assert(idx != 0);
+ assert(tmpMinChain <= minChain);
+
+ /* fill conventional hash table and conventional chain table */
+ for ( ; idx < target; idx++) {
+ U32 const h = (U32)ZSTD_hashPtr(base + idx, hashLog, ms->cParams.minMatch);
+ if (idx >= tmpMinChain) {
+ tmpChainTable[idx - tmpMinChain] = hashTable[h];
+ }
+ tmpHashTable[h] = idx;
+ }
+
+ /* sort chains into ddss chain table */
+ {
+ U32 chainPos = 0;
+ for (hashIdx = 0; hashIdx < (1U << hashLog); hashIdx++) {
+ U32 count;
+ U32 countBeyondMinChain = 0;
+ U32 i = tmpHashTable[hashIdx];
+ for (count = 0; i >= tmpMinChain && count < cacheSize; count++) {
+ /* skip through the chain to the first position that won't be
+ * in the hash cache bucket */
+ if (i < minChain) {
+ countBeyondMinChain++;
+ }
+ i = tmpChainTable[i - tmpMinChain];
+ }
+ if (count == cacheSize) {
+ for (count = 0; count < chainLimit;) {
+ if (i < minChain) {
+ if (!i || ++countBeyondMinChain > cacheSize) {
+ /* only allow pulling `cacheSize` number of entries
+ * into the cache or chainTable beyond `minChain`,
+ * to replace the entries pulled out of the
+ * chainTable into the cache. This lets us reach
+ * back further without increasing the total number
+ * of entries in the chainTable, guaranteeing the
+ * DDSS chain table will fit into the space
+ * allocated for the regular one. */
+ break;
+ }
+ }
+ chainTable[chainPos++] = i;
+ count++;
+ if (i < tmpMinChain) {
+ break;
+ }
+ i = tmpChainTable[i - tmpMinChain];
+ }
+ } else {
+ count = 0;
+ }
+ if (count) {
+ tmpHashTable[hashIdx] = ((chainPos - count) << 8) + count;
+ } else {
+ tmpHashTable[hashIdx] = 0;
+ }
+ }
+ assert(chainPos <= chainSize); /* I believe this is guaranteed... */
+ }
+
+ /* move chain pointers into the last entry of each hash bucket */
+ for (hashIdx = (1 << hashLog); hashIdx; ) {
+ U32 const bucketIdx = --hashIdx << ZSTD_LAZY_DDSS_BUCKET_LOG;
+ U32 const chainPackedPointer = tmpHashTable[hashIdx];
+ U32 i;
+ for (i = 0; i < cacheSize; i++) {
+ hashTable[bucketIdx + i] = 0;
+ }
+ hashTable[bucketIdx + bucketSize - 1] = chainPackedPointer;
+ }
+
+ /* fill the buckets of the hash table */
+ for (idx = ms->nextToUpdate; idx < target; idx++) {
+ U32 const h = (U32)ZSTD_hashPtr(base + idx, hashLog, ms->cParams.minMatch)
+ << ZSTD_LAZY_DDSS_BUCKET_LOG;
+ U32 i;
+ /* Shift hash cache down 1. */
+ for (i = cacheSize - 1; i; i--)
+ hashTable[h + i] = hashTable[h + i - 1];
+ hashTable[h] = idx;
+ }
+
+ ms->nextToUpdate = target;
+}
+
+/* Returns the longest match length found in the dedicated dict search structure.
+ * If none are longer than the argument ml, then ml will be returned.
+ */
+FORCE_INLINE_TEMPLATE
+size_t ZSTD_dedicatedDictSearch_lazy_search(size_t* offsetPtr, size_t ml, U32 nbAttempts,
+ const ZSTD_matchState_t* const dms,
+ const BYTE* const ip, const BYTE* const iLimit,
+ const BYTE* const prefixStart, const U32 curr,
+ const U32 dictLimit, const size_t ddsIdx) {
+ const U32 ddsLowestIndex = dms->window.dictLimit;
+ const BYTE* const ddsBase = dms->window.base;
+ const BYTE* const ddsEnd = dms->window.nextSrc;
+ const U32 ddsSize = (U32)(ddsEnd - ddsBase);
+ const U32 ddsIndexDelta = dictLimit - ddsSize;
+ const U32 bucketSize = (1 << ZSTD_LAZY_DDSS_BUCKET_LOG);
+ const U32 bucketLimit = nbAttempts < bucketSize - 1 ? nbAttempts : bucketSize - 1;
+ U32 ddsAttempt;
+ U32 matchIndex;
+
+ for (ddsAttempt = 0; ddsAttempt < bucketSize - 1; ddsAttempt++) {
+ PREFETCH_L1(ddsBase + dms->hashTable[ddsIdx + ddsAttempt]);
+ }
+
+ {
+ U32 const chainPackedPointer = dms->hashTable[ddsIdx + bucketSize - 1];
+ U32 const chainIndex = chainPackedPointer >> 8;
+
+ PREFETCH_L1(&dms->chainTable[chainIndex]);
+ }
+
+ for (ddsAttempt = 0; ddsAttempt < bucketLimit; ddsAttempt++) {
+ size_t currentMl=0;
+ const BYTE* match;
+ matchIndex = dms->hashTable[ddsIdx + ddsAttempt];
+ match = ddsBase + matchIndex;
+
+ if (!matchIndex) {
+ return ml;
+ }
+
+ /* guaranteed by table construction */
+ (void)ddsLowestIndex;
+ assert(matchIndex >= ddsLowestIndex);
+ assert(match+4 <= ddsEnd);
+ if (MEM_read32(match) == MEM_read32(ip)) {
+ /* assumption : matchIndex <= dictLimit-4 (by table construction) */
+ currentMl = ZSTD_count_2segments(ip+4, match+4, iLimit, ddsEnd, prefixStart) + 4;
+ }
+
+ /* save best solution */
+ if (currentMl > ml) {
+ ml = currentMl;
+ *offsetPtr = STORE_OFFSET(curr - (matchIndex + ddsIndexDelta));
+ if (ip+currentMl == iLimit) {
+ /* best possible, avoids read overflow on next attempt */
+ return ml;
+ }
+ }
+ }
+
+ {
+ U32 const chainPackedPointer = dms->hashTable[ddsIdx + bucketSize - 1];
+ U32 chainIndex = chainPackedPointer >> 8;
+ U32 const chainLength = chainPackedPointer & 0xFF;
+ U32 const chainAttempts = nbAttempts - ddsAttempt;
+ U32 const chainLimit = chainAttempts > chainLength ? chainLength : chainAttempts;
+ U32 chainAttempt;
+
+ for (chainAttempt = 0 ; chainAttempt < chainLimit; chainAttempt++) {
+ PREFETCH_L1(ddsBase + dms->chainTable[chainIndex + chainAttempt]);
+ }
+
+ for (chainAttempt = 0 ; chainAttempt < chainLimit; chainAttempt++, chainIndex++) {
+ size_t currentMl=0;
+ const BYTE* match;
+ matchIndex = dms->chainTable[chainIndex];
+ match = ddsBase + matchIndex;
+
+ /* guaranteed by table construction */
+ assert(matchIndex >= ddsLowestIndex);
+ assert(match+4 <= ddsEnd);
+ if (MEM_read32(match) == MEM_read32(ip)) {
+ /* assumption : matchIndex <= dictLimit-4 (by table construction) */
+ currentMl = ZSTD_count_2segments(ip+4, match+4, iLimit, ddsEnd, prefixStart) + 4;
+ }
+
+ /* save best solution */
+ if (currentMl > ml) {
+ ml = currentMl;
+ *offsetPtr = STORE_OFFSET(curr - (matchIndex + ddsIndexDelta));
+ if (ip+currentMl == iLimit) break; /* best possible, avoids read overflow on next attempt */
+ }
+ }
+ }
+ return ml;
+}
+
+
+/* *********************************
+* Hash Chain
+***********************************/
+#define NEXT_IN_CHAIN(d, mask) chainTable[(d) & (mask)]
+
+/* Update chains up to ip (excluded)
+ Assumption : always within prefix (i.e. not within extDict) */
+FORCE_INLINE_TEMPLATE U32 ZSTD_insertAndFindFirstIndex_internal(
+ ZSTD_matchState_t* ms,
+ const ZSTD_compressionParameters* const cParams,
+ const BYTE* ip, U32 const mls)
+{
+ U32* const hashTable = ms->hashTable;
+ const U32 hashLog = cParams->hashLog;
+ U32* const chainTable = ms->chainTable;
+ const U32 chainMask = (1 << cParams->chainLog) - 1;
+ const BYTE* const base = ms->window.base;
+ const U32 target = (U32)(ip - base);
+ U32 idx = ms->nextToUpdate;
+
+ while(idx < target) { /* catch up */
+ size_t const h = ZSTD_hashPtr(base+idx, hashLog, mls);
+ NEXT_IN_CHAIN(idx, chainMask) = hashTable[h];
+ hashTable[h] = idx;
+ idx++;
+ }
+
+ ms->nextToUpdate = target;
+ return hashTable[ZSTD_hashPtr(ip, hashLog, mls)];
+}
+
+U32 ZSTD_insertAndFindFirstIndex(ZSTD_matchState_t* ms, const BYTE* ip) {
+ const ZSTD_compressionParameters* const cParams = &ms->cParams;
+ return ZSTD_insertAndFindFirstIndex_internal(ms, cParams, ip, ms->cParams.minMatch);
+}
+
+/* inlining is important to hardwire a hot branch (template emulation) */
+FORCE_INLINE_TEMPLATE
+size_t ZSTD_HcFindBestMatch(
+ ZSTD_matchState_t* ms,
+ const BYTE* const ip, const BYTE* const iLimit,
+ size_t* offsetPtr,
+ const U32 mls, const ZSTD_dictMode_e dictMode)
+{
+ const ZSTD_compressionParameters* const cParams = &ms->cParams;
+ U32* const chainTable = ms->chainTable;
+ const U32 chainSize = (1 << cParams->chainLog);
+ const U32 chainMask = chainSize-1;
+ const BYTE* const base = ms->window.base;
+ const BYTE* const dictBase = ms->window.dictBase;
+ const U32 dictLimit = ms->window.dictLimit;
+ const BYTE* const prefixStart = base + dictLimit;
+ const BYTE* const dictEnd = dictBase + dictLimit;
+ const U32 curr = (U32)(ip-base);
+ const U32 maxDistance = 1U << cParams->windowLog;
+ const U32 lowestValid = ms->window.lowLimit;
+ const U32 withinMaxDistance = (curr - lowestValid > maxDistance) ? curr - maxDistance : lowestValid;
+ const U32 isDictionary = (ms->loadedDictEnd != 0);
+ const U32 lowLimit = isDictionary ? lowestValid : withinMaxDistance;
+ const U32 minChain = curr > chainSize ? curr - chainSize : 0;
+ U32 nbAttempts = 1U << cParams->searchLog;
+ size_t ml=4-1;
+
+ const ZSTD_matchState_t* const dms = ms->dictMatchState;
+ const U32 ddsHashLog = dictMode == ZSTD_dedicatedDictSearch
+ ? dms->cParams.hashLog - ZSTD_LAZY_DDSS_BUCKET_LOG : 0;
+ const size_t ddsIdx = dictMode == ZSTD_dedicatedDictSearch
+ ? ZSTD_hashPtr(ip, ddsHashLog, mls) << ZSTD_LAZY_DDSS_BUCKET_LOG : 0;
+
+ U32 matchIndex;
+
+ if (dictMode == ZSTD_dedicatedDictSearch) {
+ const U32* entry = &dms->hashTable[ddsIdx];
+ PREFETCH_L1(entry);
+ }
+
+ /* HC4 match finder */
+ matchIndex = ZSTD_insertAndFindFirstIndex_internal(ms, cParams, ip, mls);
+
+ for ( ; (matchIndex>=lowLimit) & (nbAttempts>0) ; nbAttempts--) {
+ size_t currentMl=0;
+ if ((dictMode != ZSTD_extDict) || matchIndex >= dictLimit) {
+ const BYTE* const match = base + matchIndex;
+ assert(matchIndex >= dictLimit); /* ensures this is true if dictMode != ZSTD_extDict */
+ if (match[ml] == ip[ml]) /* potentially better */
+ currentMl = ZSTD_count(ip, match, iLimit);
+ } else {
+ const BYTE* const match = dictBase + matchIndex;
+ assert(match+4 <= dictEnd);
+ if (MEM_read32(match) == MEM_read32(ip)) /* assumption : matchIndex <= dictLimit-4 (by table construction) */
+ currentMl = ZSTD_count_2segments(ip+4, match+4, iLimit, dictEnd, prefixStart) + 4;
+ }
+
+ /* save best solution */
+ if (currentMl > ml) {
+ ml = currentMl;
+ *offsetPtr = STORE_OFFSET(curr - matchIndex);
+ if (ip+currentMl == iLimit) break; /* best possible, avoids read overflow on next attempt */
+ }
+
+ if (matchIndex <= minChain) break;
+ matchIndex = NEXT_IN_CHAIN(matchIndex, chainMask);
+ }
+
+ assert(nbAttempts <= (1U << ZSTD_SEARCHLOG_MAX)); /* Check we haven't underflowed. */
+ if (dictMode == ZSTD_dedicatedDictSearch) {
+ ml = ZSTD_dedicatedDictSearch_lazy_search(offsetPtr, ml, nbAttempts, dms,
+ ip, iLimit, prefixStart, curr, dictLimit, ddsIdx);
+ } else if (dictMode == ZSTD_dictMatchState) {
+ const U32* const dmsChainTable = dms->chainTable;
+ const U32 dmsChainSize = (1 << dms->cParams.chainLog);
+ const U32 dmsChainMask = dmsChainSize - 1;
+ const U32 dmsLowestIndex = dms->window.dictLimit;
+ const BYTE* const dmsBase = dms->window.base;
+ const BYTE* const dmsEnd = dms->window.nextSrc;
+ const U32 dmsSize = (U32)(dmsEnd - dmsBase);
+ const U32 dmsIndexDelta = dictLimit - dmsSize;
+ const U32 dmsMinChain = dmsSize > dmsChainSize ? dmsSize - dmsChainSize : 0;
+
+ matchIndex = dms->hashTable[ZSTD_hashPtr(ip, dms->cParams.hashLog, mls)];
+
+ for ( ; (matchIndex>=dmsLowestIndex) & (nbAttempts>0) ; nbAttempts--) {
+ size_t currentMl=0;
+ const BYTE* const match = dmsBase + matchIndex;
+ assert(match+4 <= dmsEnd);
+ if (MEM_read32(match) == MEM_read32(ip)) /* assumption : matchIndex <= dictLimit-4 (by table construction) */
+ currentMl = ZSTD_count_2segments(ip+4, match+4, iLimit, dmsEnd, prefixStart) + 4;
+
+ /* save best solution */
+ if (currentMl > ml) {
+ ml = currentMl;
+ assert(curr > matchIndex + dmsIndexDelta);
+ *offsetPtr = STORE_OFFSET(curr - (matchIndex + dmsIndexDelta));
+ if (ip+currentMl == iLimit) break; /* best possible, avoids read overflow on next attempt */
+ }
+
+ if (matchIndex <= dmsMinChain) break;
+
+ matchIndex = dmsChainTable[matchIndex & dmsChainMask];
+ }
+ }
+
+ return ml;
+}
+
+/* *********************************
+* (SIMD) Row-based matchfinder
+***********************************/
+/* Constants for row-based hash */
+#define ZSTD_ROW_HASH_TAG_OFFSET 16 /* byte offset of hashes in the match state's tagTable from the beginning of a row */
+#define ZSTD_ROW_HASH_TAG_BITS 8 /* nb bits to use for the tag */
+#define ZSTD_ROW_HASH_TAG_MASK ((1u << ZSTD_ROW_HASH_TAG_BITS) - 1)
+#define ZSTD_ROW_HASH_MAX_ENTRIES 64 /* absolute maximum number of entries per row, for all configurations */
+
+#define ZSTD_ROW_HASH_CACHE_MASK (ZSTD_ROW_HASH_CACHE_SIZE - 1)
+
+typedef U64 ZSTD_VecMask; /* Clarifies when we are interacting with a U64 representing a mask of matches */
+
+/* ZSTD_VecMask_next():
+ * Starting from the LSB, returns the idx of the next non-zero bit.
+ * Basically counting the nb of trailing zeroes.
+ */
+static U32 ZSTD_VecMask_next(ZSTD_VecMask val) {
+ assert(val != 0);
+# if (defined(__GNUC__) && ((__GNUC__ > 3) || ((__GNUC__ == 3) && (__GNUC_MINOR__ >= 4))))
+ if (sizeof(size_t) == 4) {
+ U32 mostSignificantWord = (U32)(val >> 32);
+ U32 leastSignificantWord = (U32)val;
+ if (leastSignificantWord == 0) {
+ return 32 + (U32)__builtin_ctz(mostSignificantWord);
+ } else {
+ return (U32)__builtin_ctz(leastSignificantWord);
+ }
+ } else {
+ return (U32)__builtin_ctzll(val);
+ }
+# else
+ /* Software ctz version: http://aggregate.org/MAGIC/#Trailing%20Zero%20Count
+ * and: https://stackoverflow.com/questions/2709430/count-number-of-bits-in-a-64-bit-long-big-integer
+ */
+ val = ~val & (val - 1ULL); /* Lowest set bit mask */
+ val = val - ((val >> 1) & 0x5555555555555555);
+ val = (val & 0x3333333333333333ULL) + ((val >> 2) & 0x3333333333333333ULL);
+ return (U32)((((val + (val >> 4)) & 0xF0F0F0F0F0F0F0FULL) * 0x101010101010101ULL) >> 56);
+# endif
+}
+
+/* ZSTD_rotateRight_*():
+ * Rotates a bitfield to the right by "count" bits.
+ * https://en.wikipedia.org/w/index.php?title=Circular_shift&oldid=991635599#Implementing_circular_shifts
+ */
+FORCE_INLINE_TEMPLATE
+U64 ZSTD_rotateRight_U64(U64 const value, U32 count) {
+ assert(count < 64);
+ count &= 0x3F; /* for fickle pattern recognition */
+ return (value >> count) | (U64)(value << ((0U - count) & 0x3F));
+}
+
+FORCE_INLINE_TEMPLATE
+U32 ZSTD_rotateRight_U32(U32 const value, U32 count) {
+ assert(count < 32);
+ count &= 0x1F; /* for fickle pattern recognition */
+ return (value >> count) | (U32)(value << ((0U - count) & 0x1F));
+}
+
+FORCE_INLINE_TEMPLATE
+U16 ZSTD_rotateRight_U16(U16 const value, U32 count) {
+ assert(count < 16);
+ count &= 0x0F; /* for fickle pattern recognition */
+ return (value >> count) | (U16)(value << ((0U - count) & 0x0F));
+}
+
+/* ZSTD_row_nextIndex():
+ * Returns the next index to insert at within a tagTable row, and updates the "head"
+ * value to reflect the update. Essentially cycles backwards from [0, {entries per row})
+ */
+FORCE_INLINE_TEMPLATE U32 ZSTD_row_nextIndex(BYTE* const tagRow, U32 const rowMask) {
+ U32 const next = (*tagRow - 1) & rowMask;
+ *tagRow = (BYTE)next;
+ return next;
+}
+
+/* ZSTD_isAligned():
+ * Checks that a pointer is aligned to "align" bytes which must be a power of 2.
+ */
+MEM_STATIC int ZSTD_isAligned(void const* ptr, size_t align) {
+ assert((align & (align - 1)) == 0);
+ return (((size_t)ptr) & (align - 1)) == 0;
+}
+
+/* ZSTD_row_prefetch():
+ * Performs prefetching for the hashTable and tagTable at a given row.
+ */
+FORCE_INLINE_TEMPLATE void ZSTD_row_prefetch(U32 const* hashTable, U16 const* tagTable, U32 const relRow, U32 const rowLog) {
+ PREFETCH_L1(hashTable + relRow);
+ if (rowLog >= 5) {
+ PREFETCH_L1(hashTable + relRow + 16);
+ /* Note: prefetching more of the hash table does not appear to be beneficial for 128-entry rows */
+ }
+ PREFETCH_L1(tagTable + relRow);
+ if (rowLog == 6) {
+ PREFETCH_L1(tagTable + relRow + 32);
+ }
+ assert(rowLog == 4 || rowLog == 5 || rowLog == 6);
+ assert(ZSTD_isAligned(hashTable + relRow, 64)); /* prefetched hash row always 64-byte aligned */
+ assert(ZSTD_isAligned(tagTable + relRow, (size_t)1 << rowLog)); /* prefetched tagRow sits on correct multiple of bytes (32,64,128) */
+}
+
+/* ZSTD_row_fillHashCache():
+ * Fill up the hash cache starting at idx, prefetching up to ZSTD_ROW_HASH_CACHE_SIZE entries,
+ * but not beyond iLimit.
+ */
+FORCE_INLINE_TEMPLATE void ZSTD_row_fillHashCache(ZSTD_matchState_t* ms, const BYTE* base,
+ U32 const rowLog, U32 const mls,
+ U32 idx, const BYTE* const iLimit)
+{
+ U32 const* const hashTable = ms->hashTable;
+ U16 const* const tagTable = ms->tagTable;
+ U32 const hashLog = ms->rowHashLog;
+ U32 const maxElemsToPrefetch = (base + idx) > iLimit ? 0 : (U32)(iLimit - (base + idx) + 1);
+ U32 const lim = idx + MIN(ZSTD_ROW_HASH_CACHE_SIZE, maxElemsToPrefetch);
+
+ for (; idx < lim; ++idx) {
+ U32 const hash = (U32)ZSTD_hashPtr(base + idx, hashLog + ZSTD_ROW_HASH_TAG_BITS, mls);
+ U32 const row = (hash >> ZSTD_ROW_HASH_TAG_BITS) << rowLog;
+ ZSTD_row_prefetch(hashTable, tagTable, row, rowLog);
+ ms->hashCache[idx & ZSTD_ROW_HASH_CACHE_MASK] = hash;
+ }
+
+ DEBUGLOG(6, "ZSTD_row_fillHashCache(): [%u %u %u %u %u %u %u %u]", ms->hashCache[0], ms->hashCache[1],
+ ms->hashCache[2], ms->hashCache[3], ms->hashCache[4],
+ ms->hashCache[5], ms->hashCache[6], ms->hashCache[7]);
+}
+
+/* ZSTD_row_nextCachedHash():
+ * Returns the hash of base + idx, and replaces the hash in the hash cache with the byte at
+ * base + idx + ZSTD_ROW_HASH_CACHE_SIZE. Also prefetches the appropriate rows from hashTable and tagTable.
+ */
+FORCE_INLINE_TEMPLATE U32 ZSTD_row_nextCachedHash(U32* cache, U32 const* hashTable,
+ U16 const* tagTable, BYTE const* base,
+ U32 idx, U32 const hashLog,
+ U32 const rowLog, U32 const mls)
+{
+ U32 const newHash = (U32)ZSTD_hashPtr(base+idx+ZSTD_ROW_HASH_CACHE_SIZE, hashLog + ZSTD_ROW_HASH_TAG_BITS, mls);
+ U32 const row = (newHash >> ZSTD_ROW_HASH_TAG_BITS) << rowLog;
+ ZSTD_row_prefetch(hashTable, tagTable, row, rowLog);
+ { U32 const hash = cache[idx & ZSTD_ROW_HASH_CACHE_MASK];
+ cache[idx & ZSTD_ROW_HASH_CACHE_MASK] = newHash;
+ return hash;
+ }
+}
+
+/* ZSTD_row_update_internalImpl():
+ * Updates the hash table with positions starting from updateStartIdx until updateEndIdx.
+ */
+FORCE_INLINE_TEMPLATE void ZSTD_row_update_internalImpl(ZSTD_matchState_t* ms,
+ U32 updateStartIdx, U32 const updateEndIdx,
+ U32 const mls, U32 const rowLog,
+ U32 const rowMask, U32 const useCache)
+{
+ U32* const hashTable = ms->hashTable;
+ U16* const tagTable = ms->tagTable;
+ U32 const hashLog = ms->rowHashLog;
+ const BYTE* const base = ms->window.base;
+
+ DEBUGLOG(6, "ZSTD_row_update_internalImpl(): updateStartIdx=%u, updateEndIdx=%u", updateStartIdx, updateEndIdx);
+ for (; updateStartIdx < updateEndIdx; ++updateStartIdx) {
+ U32 const hash = useCache ? ZSTD_row_nextCachedHash(ms->hashCache, hashTable, tagTable, base, updateStartIdx, hashLog, rowLog, mls)
+ : (U32)ZSTD_hashPtr(base + updateStartIdx, hashLog + ZSTD_ROW_HASH_TAG_BITS, mls);
+ U32 const relRow = (hash >> ZSTD_ROW_HASH_TAG_BITS) << rowLog;
+ U32* const row = hashTable + relRow;
+ BYTE* tagRow = (BYTE*)(tagTable + relRow); /* Though tagTable is laid out as a table of U16, each tag is only 1 byte.
+ Explicit cast allows us to get exact desired position within each row */
+ U32 const pos = ZSTD_row_nextIndex(tagRow, rowMask);
+
+ assert(hash == ZSTD_hashPtr(base + updateStartIdx, hashLog + ZSTD_ROW_HASH_TAG_BITS, mls));
+ ((BYTE*)tagRow)[pos + ZSTD_ROW_HASH_TAG_OFFSET] = hash & ZSTD_ROW_HASH_TAG_MASK;
+ row[pos] = updateStartIdx;
+ }
+}
+
+/* ZSTD_row_update_internal():
+ * Inserts the byte at ip into the appropriate position in the hash table, and updates ms->nextToUpdate.
+ * Skips sections of long matches as is necessary.
+ */
+FORCE_INLINE_TEMPLATE void ZSTD_row_update_internal(ZSTD_matchState_t* ms, const BYTE* ip,
+ U32 const mls, U32 const rowLog,
+ U32 const rowMask, U32 const useCache)
+{
+ U32 idx = ms->nextToUpdate;
+ const BYTE* const base = ms->window.base;
+ const U32 target = (U32)(ip - base);
+ const U32 kSkipThreshold = 384;
+ const U32 kMaxMatchStartPositionsToUpdate = 96;
+ const U32 kMaxMatchEndPositionsToUpdate = 32;
+
+ if (useCache) {
+ /* Only skip positions when using hash cache, i.e.
+ * if we are loading a dict, don't skip anything.
+ * If we decide to skip, then we only update a set number
+ * of positions at the beginning and end of the match.
+ */
+ if (UNLIKELY(target - idx > kSkipThreshold)) {
+ U32 const bound = idx + kMaxMatchStartPositionsToUpdate;
+ ZSTD_row_update_internalImpl(ms, idx, bound, mls, rowLog, rowMask, useCache);
+ idx = target - kMaxMatchEndPositionsToUpdate;
+ ZSTD_row_fillHashCache(ms, base, rowLog, mls, idx, ip+1);
+ }
+ }
+ assert(target >= idx);
+ ZSTD_row_update_internalImpl(ms, idx, target, mls, rowLog, rowMask, useCache);
+ ms->nextToUpdate = target;
+}
+
+/* ZSTD_row_update():
+ * External wrapper for ZSTD_row_update_internal(). Used for filling the hashtable during dictionary
+ * processing.
+ */
+void ZSTD_row_update(ZSTD_matchState_t* const ms, const BYTE* ip) {
+ const U32 rowLog = BOUNDED(4, ms->cParams.searchLog, 6);
+ const U32 rowMask = (1u << rowLog) - 1;
+ const U32 mls = MIN(ms->cParams.minMatch, 6 /* mls caps out at 6 */);
+
+ DEBUGLOG(5, "ZSTD_row_update(), rowLog=%u", rowLog);
+ ZSTD_row_update_internal(ms, ip, mls, rowLog, rowMask, 0 /* dont use cache */);
+}
+
+#if defined(ZSTD_ARCH_X86_SSE2)
+FORCE_INLINE_TEMPLATE ZSTD_VecMask
+ZSTD_row_getSSEMask(int nbChunks, const BYTE* const src, const BYTE tag, const U32 head)
+{
+ const __m128i comparisonMask = _mm_set1_epi8((char)tag);
+ int matches[4] = {0};
+ int i;
+ assert(nbChunks == 1 || nbChunks == 2 || nbChunks == 4);
+ for (i=0; i<nbChunks; i++) {
+ const __m128i chunk = _mm_loadu_si128((const __m128i*)(const void*)(src + 16*i));
+ const __m128i equalMask = _mm_cmpeq_epi8(chunk, comparisonMask);
+ matches[i] = _mm_movemask_epi8(equalMask);
+ }
+ if (nbChunks == 1) return ZSTD_rotateRight_U16((U16)matches[0], head);
+ if (nbChunks == 2) return ZSTD_rotateRight_U32((U32)matches[1] << 16 | (U32)matches[0], head);
+ assert(nbChunks == 4);
+ return ZSTD_rotateRight_U64((U64)matches[3] << 48 | (U64)matches[2] << 32 | (U64)matches[1] << 16 | (U64)matches[0], head);
+}
+#endif
+
+/* Returns a ZSTD_VecMask (U32) that has the nth bit set to 1 if the newly-computed "tag" matches
+ * the hash at the nth position in a row of the tagTable.
+ * Each row is a circular buffer beginning at the value of "head". So we must rotate the "matches" bitfield
+ * to match up with the actual layout of the entries within the hashTable */
+FORCE_INLINE_TEMPLATE ZSTD_VecMask
+ZSTD_row_getMatchMask(const BYTE* const tagRow, const BYTE tag, const U32 head, const U32 rowEntries)
+{
+ const BYTE* const src = tagRow + ZSTD_ROW_HASH_TAG_OFFSET;
+ assert((rowEntries == 16) || (rowEntries == 32) || rowEntries == 64);
+ assert(rowEntries <= ZSTD_ROW_HASH_MAX_ENTRIES);
+
+#if defined(ZSTD_ARCH_X86_SSE2)
+
+ return ZSTD_row_getSSEMask(rowEntries / 16, src, tag, head);
+
+#else /* SW or NEON-LE */
+
+# if defined(ZSTD_ARCH_ARM_NEON)
+ /* This NEON path only works for little endian - otherwise use SWAR below */
+ if (MEM_isLittleEndian()) {
+ if (rowEntries == 16) {
+ const uint8x16_t chunk = vld1q_u8(src);
+ const uint16x8_t equalMask = vreinterpretq_u16_u8(vceqq_u8(chunk, vdupq_n_u8(tag)));
+ const uint16x8_t t0 = vshlq_n_u16(equalMask, 7);
+ const uint32x4_t t1 = vreinterpretq_u32_u16(vsriq_n_u16(t0, t0, 14));
+ const uint64x2_t t2 = vreinterpretq_u64_u32(vshrq_n_u32(t1, 14));
+ const uint8x16_t t3 = vreinterpretq_u8_u64(vsraq_n_u64(t2, t2, 28));
+ const U16 hi = (U16)vgetq_lane_u8(t3, 8);
+ const U16 lo = (U16)vgetq_lane_u8(t3, 0);
+ return ZSTD_rotateRight_U16((hi << 8) | lo, head);
+ } else if (rowEntries == 32) {
+ const uint16x8x2_t chunk = vld2q_u16((const U16*)(const void*)src);
+ const uint8x16_t chunk0 = vreinterpretq_u8_u16(chunk.val[0]);
+ const uint8x16_t chunk1 = vreinterpretq_u8_u16(chunk.val[1]);
+ const uint8x16_t equalMask0 = vceqq_u8(chunk0, vdupq_n_u8(tag));
+ const uint8x16_t equalMask1 = vceqq_u8(chunk1, vdupq_n_u8(tag));
+ const int8x8_t pack0 = vqmovn_s16(vreinterpretq_s16_u8(equalMask0));
+ const int8x8_t pack1 = vqmovn_s16(vreinterpretq_s16_u8(equalMask1));
+ const uint8x8_t t0 = vreinterpret_u8_s8(pack0);
+ const uint8x8_t t1 = vreinterpret_u8_s8(pack1);
+ const uint8x8_t t2 = vsri_n_u8(t1, t0, 2);
+ const uint8x8x2_t t3 = vuzp_u8(t2, t0);
+ const uint8x8_t t4 = vsri_n_u8(t3.val[1], t3.val[0], 4);
+ const U32 matches = vget_lane_u32(vreinterpret_u32_u8(t4), 0);
+ return ZSTD_rotateRight_U32(matches, head);
+ } else { /* rowEntries == 64 */
+ const uint8x16x4_t chunk = vld4q_u8(src);
+ const uint8x16_t dup = vdupq_n_u8(tag);
+ const uint8x16_t cmp0 = vceqq_u8(chunk.val[0], dup);
+ const uint8x16_t cmp1 = vceqq_u8(chunk.val[1], dup);
+ const uint8x16_t cmp2 = vceqq_u8(chunk.val[2], dup);
+ const uint8x16_t cmp3 = vceqq_u8(chunk.val[3], dup);
+
+ const uint8x16_t t0 = vsriq_n_u8(cmp1, cmp0, 1);
+ const uint8x16_t t1 = vsriq_n_u8(cmp3, cmp2, 1);
+ const uint8x16_t t2 = vsriq_n_u8(t1, t0, 2);
+ const uint8x16_t t3 = vsriq_n_u8(t2, t2, 4);
+ const uint8x8_t t4 = vshrn_n_u16(vreinterpretq_u16_u8(t3), 4);
+ const U64 matches = vget_lane_u64(vreinterpret_u64_u8(t4), 0);
+ return ZSTD_rotateRight_U64(matches, head);
+ }
+ }
+# endif /* ZSTD_ARCH_ARM_NEON */
+ /* SWAR */
+ { const size_t chunkSize = sizeof(size_t);
+ const size_t shiftAmount = ((chunkSize * 8) - chunkSize);
+ const size_t xFF = ~((size_t)0);
+ const size_t x01 = xFF / 0xFF;
+ const size_t x80 = x01 << 7;
+ const size_t splatChar = tag * x01;
+ ZSTD_VecMask matches = 0;
+ int i = rowEntries - chunkSize;
+ assert((sizeof(size_t) == 4) || (sizeof(size_t) == 8));
+ if (MEM_isLittleEndian()) { /* runtime check so have two loops */
+ const size_t extractMagic = (xFF / 0x7F) >> chunkSize;
+ do {
+ size_t chunk = MEM_readST(&src[i]);
+ chunk ^= splatChar;
+ chunk = (((chunk | x80) - x01) | chunk) & x80;
+ matches <<= chunkSize;
+ matches |= (chunk * extractMagic) >> shiftAmount;
+ i -= chunkSize;
+ } while (i >= 0);
+ } else { /* big endian: reverse bits during extraction */
+ const size_t msb = xFF ^ (xFF >> 1);
+ const size_t extractMagic = (msb / 0x1FF) | msb;
+ do {
+ size_t chunk = MEM_readST(&src[i]);
+ chunk ^= splatChar;
+ chunk = (((chunk | x80) - x01) | chunk) & x80;
+ matches <<= chunkSize;
+ matches |= ((chunk >> 7) * extractMagic) >> shiftAmount;
+ i -= chunkSize;
+ } while (i >= 0);
+ }
+ matches = ~matches;
+ if (rowEntries == 16) {
+ return ZSTD_rotateRight_U16((U16)matches, head);
+ } else if (rowEntries == 32) {
+ return ZSTD_rotateRight_U32((U32)matches, head);
+ } else {
+ return ZSTD_rotateRight_U64((U64)matches, head);
+ }
+ }
+#endif
+}
+
+/* The high-level approach of the SIMD row based match finder is as follows:
+ * - Figure out where to insert the new entry:
+ * - Generate a hash from a byte along with an additional 1-byte "short hash". The additional byte is our "tag"
+ * - The hashTable is effectively split into groups or "rows" of 16 or 32 entries of U32, and the hash determines
+ * which row to insert into.
+ * - Determine the correct position within the row to insert the entry into. Each row of 16 or 32 can
+ * be considered as a circular buffer with a "head" index that resides in the tagTable.
+ * - Also insert the "tag" into the equivalent row and position in the tagTable.
+ * - Note: The tagTable has 17 or 33 1-byte entries per row, due to 16 or 32 tags, and 1 "head" entry.
+ * The 17 or 33 entry rows are spaced out to occur every 32 or 64 bytes, respectively,
+ * for alignment/performance reasons, leaving some bytes unused.
+ * - Use SIMD to efficiently compare the tags in the tagTable to the 1-byte "short hash" and
+ * generate a bitfield that we can cycle through to check the collisions in the hash table.
+ * - Pick the longest match.
+ */
+FORCE_INLINE_TEMPLATE
+size_t ZSTD_RowFindBestMatch(
+ ZSTD_matchState_t* ms,
+ const BYTE* const ip, const BYTE* const iLimit,
+ size_t* offsetPtr,
+ const U32 mls, const ZSTD_dictMode_e dictMode,
+ const U32 rowLog)
+{
+ U32* const hashTable = ms->hashTable;
+ U16* const tagTable = ms->tagTable;
+ U32* const hashCache = ms->hashCache;
+ const U32 hashLog = ms->rowHashLog;
+ const ZSTD_compressionParameters* const cParams = &ms->cParams;
+ const BYTE* const base = ms->window.base;
+ const BYTE* const dictBase = ms->window.dictBase;
+ const U32 dictLimit = ms->window.dictLimit;
+ const BYTE* const prefixStart = base + dictLimit;
+ const BYTE* const dictEnd = dictBase + dictLimit;
+ const U32 curr = (U32)(ip-base);
+ const U32 maxDistance = 1U << cParams->windowLog;
+ const U32 lowestValid = ms->window.lowLimit;
+ const U32 withinMaxDistance = (curr - lowestValid > maxDistance) ? curr - maxDistance : lowestValid;
+ const U32 isDictionary = (ms->loadedDictEnd != 0);
+ const U32 lowLimit = isDictionary ? lowestValid : withinMaxDistance;
+ const U32 rowEntries = (1U << rowLog);
+ const U32 rowMask = rowEntries - 1;
+ const U32 cappedSearchLog = MIN(cParams->searchLog, rowLog); /* nb of searches is capped at nb entries per row */
+ U32 nbAttempts = 1U << cappedSearchLog;
+ size_t ml=4-1;
+
+ /* DMS/DDS variables that may be referenced laster */
+ const ZSTD_matchState_t* const dms = ms->dictMatchState;
+
+ /* Initialize the following variables to satisfy static analyzer */
+ size_t ddsIdx = 0;
+ U32 ddsExtraAttempts = 0; /* cctx hash tables are limited in searches, but allow extra searches into DDS */
+ U32 dmsTag = 0;
+ U32* dmsRow = NULL;
+ BYTE* dmsTagRow = NULL;
+
+ if (dictMode == ZSTD_dedicatedDictSearch) {
+ const U32 ddsHashLog = dms->cParams.hashLog - ZSTD_LAZY_DDSS_BUCKET_LOG;
+ { /* Prefetch DDS hashtable entry */
+ ddsIdx = ZSTD_hashPtr(ip, ddsHashLog, mls) << ZSTD_LAZY_DDSS_BUCKET_LOG;
+ PREFETCH_L1(&dms->hashTable[ddsIdx]);
+ }
+ ddsExtraAttempts = cParams->searchLog > rowLog ? 1U << (cParams->searchLog - rowLog) : 0;
+ }
+
+ if (dictMode == ZSTD_dictMatchState) {
+ /* Prefetch DMS rows */
+ U32* const dmsHashTable = dms->hashTable;
+ U16* const dmsTagTable = dms->tagTable;
+ U32 const dmsHash = (U32)ZSTD_hashPtr(ip, dms->rowHashLog + ZSTD_ROW_HASH_TAG_BITS, mls);
+ U32 const dmsRelRow = (dmsHash >> ZSTD_ROW_HASH_TAG_BITS) << rowLog;
+ dmsTag = dmsHash & ZSTD_ROW_HASH_TAG_MASK;
+ dmsTagRow = (BYTE*)(dmsTagTable + dmsRelRow);
+ dmsRow = dmsHashTable + dmsRelRow;
+ ZSTD_row_prefetch(dmsHashTable, dmsTagTable, dmsRelRow, rowLog);
+ }
+
+ /* Update the hashTable and tagTable up to (but not including) ip */
+ ZSTD_row_update_internal(ms, ip, mls, rowLog, rowMask, 1 /* useCache */);
+ { /* Get the hash for ip, compute the appropriate row */
+ U32 const hash = ZSTD_row_nextCachedHash(hashCache, hashTable, tagTable, base, curr, hashLog, rowLog, mls);
+ U32 const relRow = (hash >> ZSTD_ROW_HASH_TAG_BITS) << rowLog;
+ U32 const tag = hash & ZSTD_ROW_HASH_TAG_MASK;
+ U32* const row = hashTable + relRow;
+ BYTE* tagRow = (BYTE*)(tagTable + relRow);
+ U32 const head = *tagRow & rowMask;
+ U32 matchBuffer[ZSTD_ROW_HASH_MAX_ENTRIES];
+ size_t numMatches = 0;
+ size_t currMatch = 0;
+ ZSTD_VecMask matches = ZSTD_row_getMatchMask(tagRow, (BYTE)tag, head, rowEntries);
+
+ /* Cycle through the matches and prefetch */
+ for (; (matches > 0) && (nbAttempts > 0); --nbAttempts, matches &= (matches - 1)) {
+ U32 const matchPos = (head + ZSTD_VecMask_next(matches)) & rowMask;
+ U32 const matchIndex = row[matchPos];
+ assert(numMatches < rowEntries);
+ if (matchIndex < lowLimit)
+ break;
+ if ((dictMode != ZSTD_extDict) || matchIndex >= dictLimit) {
+ PREFETCH_L1(base + matchIndex);
+ } else {
+ PREFETCH_L1(dictBase + matchIndex);
+ }
+ matchBuffer[numMatches++] = matchIndex;
+ }
+
+ /* Speed opt: insert current byte into hashtable too. This allows us to avoid one iteration of the loop
+ in ZSTD_row_update_internal() at the next search. */
+ {
+ U32 const pos = ZSTD_row_nextIndex(tagRow, rowMask);
+ tagRow[pos + ZSTD_ROW_HASH_TAG_OFFSET] = (BYTE)tag;
+ row[pos] = ms->nextToUpdate++;
+ }
+
+ /* Return the longest match */
+ for (; currMatch < numMatches; ++currMatch) {
+ U32 const matchIndex = matchBuffer[currMatch];
+ size_t currentMl=0;
+ assert(matchIndex < curr);
+ assert(matchIndex >= lowLimit);
+
+ if ((dictMode != ZSTD_extDict) || matchIndex >= dictLimit) {
+ const BYTE* const match = base + matchIndex;
+ assert(matchIndex >= dictLimit); /* ensures this is true if dictMode != ZSTD_extDict */
+ if (match[ml] == ip[ml]) /* potentially better */
+ currentMl = ZSTD_count(ip, match, iLimit);
+ } else {
+ const BYTE* const match = dictBase + matchIndex;
+ assert(match+4 <= dictEnd);
+ if (MEM_read32(match) == MEM_read32(ip)) /* assumption : matchIndex <= dictLimit-4 (by table construction) */
+ currentMl = ZSTD_count_2segments(ip+4, match+4, iLimit, dictEnd, prefixStart) + 4;
+ }
+
+ /* Save best solution */
+ if (currentMl > ml) {
+ ml = currentMl;
+ *offsetPtr = STORE_OFFSET(curr - matchIndex);
+ if (ip+currentMl == iLimit) break; /* best possible, avoids read overflow on next attempt */
+ }
+ }
+ }
+
+ assert(nbAttempts <= (1U << ZSTD_SEARCHLOG_MAX)); /* Check we haven't underflowed. */
+ if (dictMode == ZSTD_dedicatedDictSearch) {
+ ml = ZSTD_dedicatedDictSearch_lazy_search(offsetPtr, ml, nbAttempts + ddsExtraAttempts, dms,
+ ip, iLimit, prefixStart, curr, dictLimit, ddsIdx);
+ } else if (dictMode == ZSTD_dictMatchState) {
+ /* TODO: Measure and potentially add prefetching to DMS */
+ const U32 dmsLowestIndex = dms->window.dictLimit;
+ const BYTE* const dmsBase = dms->window.base;
+ const BYTE* const dmsEnd = dms->window.nextSrc;
+ const U32 dmsSize = (U32)(dmsEnd - dmsBase);
+ const U32 dmsIndexDelta = dictLimit - dmsSize;
+
+ { U32 const head = *dmsTagRow & rowMask;
+ U32 matchBuffer[ZSTD_ROW_HASH_MAX_ENTRIES];
+ size_t numMatches = 0;
+ size_t currMatch = 0;
+ ZSTD_VecMask matches = ZSTD_row_getMatchMask(dmsTagRow, (BYTE)dmsTag, head, rowEntries);
+
+ for (; (matches > 0) && (nbAttempts > 0); --nbAttempts, matches &= (matches - 1)) {
+ U32 const matchPos = (head + ZSTD_VecMask_next(matches)) & rowMask;
+ U32 const matchIndex = dmsRow[matchPos];
+ if (matchIndex < dmsLowestIndex)
+ break;
+ PREFETCH_L1(dmsBase + matchIndex);
+ matchBuffer[numMatches++] = matchIndex;
+ }
+
+ /* Return the longest match */
+ for (; currMatch < numMatches; ++currMatch) {
+ U32 const matchIndex = matchBuffer[currMatch];
+ size_t currentMl=0;
+ assert(matchIndex >= dmsLowestIndex);
+ assert(matchIndex < curr);
+
+ { const BYTE* const match = dmsBase + matchIndex;
+ assert(match+4 <= dmsEnd);
+ if (MEM_read32(match) == MEM_read32(ip))
+ currentMl = ZSTD_count_2segments(ip+4, match+4, iLimit, dmsEnd, prefixStart) + 4;
+ }
+
+ if (currentMl > ml) {
+ ml = currentMl;
+ assert(curr > matchIndex + dmsIndexDelta);
+ *offsetPtr = STORE_OFFSET(curr - (matchIndex + dmsIndexDelta));
+ if (ip+currentMl == iLimit) break;
+ }
+ }
+ }
+ }
+ return ml;
+}
+
+
+/*
+ * Generate search functions templated on (dictMode, mls, rowLog).
+ * These functions are outlined for code size & compilation time.
+ * ZSTD_searchMax() dispatches to the correct implementation function.
+ *
+ * TODO: The start of the search function involves loading and calculating a
+ * bunch of constants from the ZSTD_matchState_t. These computations could be
+ * done in an initialization function, and saved somewhere in the match state.
+ * Then we could pass a pointer to the saved state instead of the match state,
+ * and avoid duplicate computations.
+ *
+ * TODO: Move the match re-winding into searchMax. This improves compression
+ * ratio, and unlocks further simplifications with the next TODO.
+ *
+ * TODO: Try moving the repcode search into searchMax. After the re-winding
+ * and repcode search are in searchMax, there is no more logic in the match
+ * finder loop that requires knowledge about the dictMode. So we should be
+ * able to avoid force inlining it, and we can join the extDict loop with
+ * the single segment loop. It should go in searchMax instead of its own
+ * function to avoid having multiple virtual function calls per search.
+ */
+
+#define ZSTD_BT_SEARCH_FN(dictMode, mls) ZSTD_BtFindBestMatch_##dictMode##_##mls
+#define ZSTD_HC_SEARCH_FN(dictMode, mls) ZSTD_HcFindBestMatch_##dictMode##_##mls
+#define ZSTD_ROW_SEARCH_FN(dictMode, mls, rowLog) ZSTD_RowFindBestMatch_##dictMode##_##mls##_##rowLog
+
+#define ZSTD_SEARCH_FN_ATTRS FORCE_NOINLINE
+
+#define GEN_ZSTD_BT_SEARCH_FN(dictMode, mls) \
+ ZSTD_SEARCH_FN_ATTRS size_t ZSTD_BT_SEARCH_FN(dictMode, mls)( \
+ ZSTD_matchState_t* ms, \
+ const BYTE* ip, const BYTE* const iLimit, \
+ size_t* offBasePtr) \
+ { \
+ assert(MAX(4, MIN(6, ms->cParams.minMatch)) == mls); \
+ return ZSTD_BtFindBestMatch(ms, ip, iLimit, offBasePtr, mls, ZSTD_##dictMode); \
+ } \
+
+#define GEN_ZSTD_HC_SEARCH_FN(dictMode, mls) \
+ ZSTD_SEARCH_FN_ATTRS size_t ZSTD_HC_SEARCH_FN(dictMode, mls)( \
+ ZSTD_matchState_t* ms, \
+ const BYTE* ip, const BYTE* const iLimit, \
+ size_t* offsetPtr) \
+ { \
+ assert(MAX(4, MIN(6, ms->cParams.minMatch)) == mls); \
+ return ZSTD_HcFindBestMatch(ms, ip, iLimit, offsetPtr, mls, ZSTD_##dictMode); \
+ } \
+
+#define GEN_ZSTD_ROW_SEARCH_FN(dictMode, mls, rowLog) \
+ ZSTD_SEARCH_FN_ATTRS size_t ZSTD_ROW_SEARCH_FN(dictMode, mls, rowLog)( \
+ ZSTD_matchState_t* ms, \
+ const BYTE* ip, const BYTE* const iLimit, \
+ size_t* offsetPtr) \
+ { \
+ assert(MAX(4, MIN(6, ms->cParams.minMatch)) == mls); \
+ assert(MAX(4, MIN(6, ms->cParams.searchLog)) == rowLog); \
+ return ZSTD_RowFindBestMatch(ms, ip, iLimit, offsetPtr, mls, ZSTD_##dictMode, rowLog); \
+ } \
+
+#define ZSTD_FOR_EACH_ROWLOG(X, dictMode, mls) \
+ X(dictMode, mls, 4) \
+ X(dictMode, mls, 5) \
+ X(dictMode, mls, 6)
+
+#define ZSTD_FOR_EACH_MLS_ROWLOG(X, dictMode) \
+ ZSTD_FOR_EACH_ROWLOG(X, dictMode, 4) \
+ ZSTD_FOR_EACH_ROWLOG(X, dictMode, 5) \
+ ZSTD_FOR_EACH_ROWLOG(X, dictMode, 6)
+
+#define ZSTD_FOR_EACH_MLS(X, dictMode) \
+ X(dictMode, 4) \
+ X(dictMode, 5) \
+ X(dictMode, 6)
+
+#define ZSTD_FOR_EACH_DICT_MODE(X, ...) \
+ X(__VA_ARGS__, noDict) \
+ X(__VA_ARGS__, extDict) \
+ X(__VA_ARGS__, dictMatchState) \
+ X(__VA_ARGS__, dedicatedDictSearch)
+
+/* Generate row search fns for each combination of (dictMode, mls, rowLog) */
+ZSTD_FOR_EACH_DICT_MODE(ZSTD_FOR_EACH_MLS_ROWLOG, GEN_ZSTD_ROW_SEARCH_FN)
+/* Generate binary Tree search fns for each combination of (dictMode, mls) */
+ZSTD_FOR_EACH_DICT_MODE(ZSTD_FOR_EACH_MLS, GEN_ZSTD_BT_SEARCH_FN)
+/* Generate hash chain search fns for each combination of (dictMode, mls) */
+ZSTD_FOR_EACH_DICT_MODE(ZSTD_FOR_EACH_MLS, GEN_ZSTD_HC_SEARCH_FN)
+
+typedef enum { search_hashChain=0, search_binaryTree=1, search_rowHash=2 } searchMethod_e;
+
+#define GEN_ZSTD_CALL_BT_SEARCH_FN(dictMode, mls) \
+ case mls: \
+ return ZSTD_BT_SEARCH_FN(dictMode, mls)(ms, ip, iend, offsetPtr);
+#define GEN_ZSTD_CALL_HC_SEARCH_FN(dictMode, mls) \
+ case mls: \
+ return ZSTD_HC_SEARCH_FN(dictMode, mls)(ms, ip, iend, offsetPtr);
+#define GEN_ZSTD_CALL_ROW_SEARCH_FN(dictMode, mls, rowLog) \
+ case rowLog: \
+ return ZSTD_ROW_SEARCH_FN(dictMode, mls, rowLog)(ms, ip, iend, offsetPtr);
+
+#define ZSTD_SWITCH_MLS(X, dictMode) \
+ switch (mls) { \
+ ZSTD_FOR_EACH_MLS(X, dictMode) \
+ }
+
+#define ZSTD_SWITCH_ROWLOG(dictMode, mls) \
+ case mls: \
+ switch (rowLog) { \
+ ZSTD_FOR_EACH_ROWLOG(GEN_ZSTD_CALL_ROW_SEARCH_FN, dictMode, mls) \
+ } \
+ ZSTD_UNREACHABLE; \
+ break;
+
+#define ZSTD_SWITCH_SEARCH_METHOD(dictMode) \
+ switch (searchMethod) { \
+ case search_hashChain: \
+ ZSTD_SWITCH_MLS(GEN_ZSTD_CALL_HC_SEARCH_FN, dictMode) \
+ break; \
+ case search_binaryTree: \
+ ZSTD_SWITCH_MLS(GEN_ZSTD_CALL_BT_SEARCH_FN, dictMode) \
+ break; \
+ case search_rowHash: \
+ ZSTD_SWITCH_MLS(ZSTD_SWITCH_ROWLOG, dictMode) \
+ break; \
+ } \
+ ZSTD_UNREACHABLE;
+
+/*
+ * Searches for the longest match at @p ip.
+ * Dispatches to the correct implementation function based on the
+ * (searchMethod, dictMode, mls, rowLog). We use switch statements
+ * here instead of using an indirect function call through a function
+ * pointer because after Spectre and Meltdown mitigations, indirect
+ * function calls can be very costly, especially in the kernel.
+ *
+ * NOTE: dictMode and searchMethod should be templated, so those switch
+ * statements should be optimized out. Only the mls & rowLog switches
+ * should be left.
+ *
+ * @param ms The match state.
+ * @param ip The position to search at.
+ * @param iend The end of the input data.
+ * @param[out] offsetPtr Stores the match offset into this pointer.
+ * @param mls The minimum search length, in the range [4, 6].
+ * @param rowLog The row log (if applicable), in the range [4, 6].
+ * @param searchMethod The search method to use (templated).
+ * @param dictMode The dictMode (templated).
+ *
+ * @returns The length of the longest match found, or < mls if no match is found.
+ * If a match is found its offset is stored in @p offsetPtr.
+ */
+FORCE_INLINE_TEMPLATE size_t ZSTD_searchMax(
+ ZSTD_matchState_t* ms,
+ const BYTE* ip,
+ const BYTE* iend,
+ size_t* offsetPtr,
+ U32 const mls,
+ U32 const rowLog,
+ searchMethod_e const searchMethod,
+ ZSTD_dictMode_e const dictMode)
+{
+ if (dictMode == ZSTD_noDict) {
+ ZSTD_SWITCH_SEARCH_METHOD(noDict)
+ } else if (dictMode == ZSTD_extDict) {
+ ZSTD_SWITCH_SEARCH_METHOD(extDict)
+ } else if (dictMode == ZSTD_dictMatchState) {
+ ZSTD_SWITCH_SEARCH_METHOD(dictMatchState)
+ } else if (dictMode == ZSTD_dedicatedDictSearch) {
+ ZSTD_SWITCH_SEARCH_METHOD(dedicatedDictSearch)
+ }
+ ZSTD_UNREACHABLE;
+ return 0;
+}
+
+/* *******************************
+* Common parser - lazy strategy
+*********************************/
+
+FORCE_INLINE_TEMPLATE size_t
+ZSTD_compressBlock_lazy_generic(
+ ZSTD_matchState_t* ms, seqStore_t* seqStore,
+ U32 rep[ZSTD_REP_NUM],
+ const void* src, size_t srcSize,
+ const searchMethod_e searchMethod, const U32 depth,
+ ZSTD_dictMode_e const dictMode)
+{
+ const BYTE* const istart = (const BYTE*)src;
+ const BYTE* ip = istart;
+ const BYTE* anchor = istart;
+ const BYTE* const iend = istart + srcSize;
+ const BYTE* const ilimit = (searchMethod == search_rowHash) ? iend - 8 - ZSTD_ROW_HASH_CACHE_SIZE : iend - 8;
+ const BYTE* const base = ms->window.base;
+ const U32 prefixLowestIndex = ms->window.dictLimit;
+ const BYTE* const prefixLowest = base + prefixLowestIndex;
+ const U32 mls = BOUNDED(4, ms->cParams.minMatch, 6);
+ const U32 rowLog = BOUNDED(4, ms->cParams.searchLog, 6);
+
+ U32 offset_1 = rep[0], offset_2 = rep[1], savedOffset=0;
+
+ const int isDMS = dictMode == ZSTD_dictMatchState;
+ const int isDDS = dictMode == ZSTD_dedicatedDictSearch;
+ const int isDxS = isDMS || isDDS;
+ const ZSTD_matchState_t* const dms = ms->dictMatchState;
+ const U32 dictLowestIndex = isDxS ? dms->window.dictLimit : 0;
+ const BYTE* const dictBase = isDxS ? dms->window.base : NULL;
+ const BYTE* const dictLowest = isDxS ? dictBase + dictLowestIndex : NULL;
+ const BYTE* const dictEnd = isDxS ? dms->window.nextSrc : NULL;
+ const U32 dictIndexDelta = isDxS ?
+ prefixLowestIndex - (U32)(dictEnd - dictBase) :
+ 0;
+ const U32 dictAndPrefixLength = (U32)((ip - prefixLowest) + (dictEnd - dictLowest));
+
+ DEBUGLOG(5, "ZSTD_compressBlock_lazy_generic (dictMode=%u) (searchFunc=%u)", (U32)dictMode, (U32)searchMethod);
+ ip += (dictAndPrefixLength == 0);
+ if (dictMode == ZSTD_noDict) {
+ U32 const curr = (U32)(ip - base);
+ U32 const windowLow = ZSTD_getLowestPrefixIndex(ms, curr, ms->cParams.windowLog);
+ U32 const maxRep = curr - windowLow;
+ if (offset_2 > maxRep) savedOffset = offset_2, offset_2 = 0;
+ if (offset_1 > maxRep) savedOffset = offset_1, offset_1 = 0;
+ }
+ if (isDxS) {
+ /* dictMatchState repCode checks don't currently handle repCode == 0
+ * disabling. */
+ assert(offset_1 <= dictAndPrefixLength);
+ assert(offset_2 <= dictAndPrefixLength);
+ }
+
+ if (searchMethod == search_rowHash) {
+ ZSTD_row_fillHashCache(ms, base, rowLog,
+ MIN(ms->cParams.minMatch, 6 /* mls caps out at 6 */),
+ ms->nextToUpdate, ilimit);
+ }
+
+ /* Match Loop */
+#if defined(__x86_64__)
+ /* I've measured random a 5% speed loss on levels 5 & 6 (greedy) when the
+ * code alignment is perturbed. To fix the instability align the loop on 32-bytes.
+ */
+ __asm__(".p2align 5");
+#endif
+ while (ip < ilimit) {
+ size_t matchLength=0;
+ size_t offcode=STORE_REPCODE_1;
+ const BYTE* start=ip+1;
+ DEBUGLOG(7, "search baseline (depth 0)");
+
+ /* check repCode */
+ if (isDxS) {
+ const U32 repIndex = (U32)(ip - base) + 1 - offset_1;
+ const BYTE* repMatch = ((dictMode == ZSTD_dictMatchState || dictMode == ZSTD_dedicatedDictSearch)
+ && repIndex < prefixLowestIndex) ?
+ dictBase + (repIndex - dictIndexDelta) :
+ base + repIndex;
+ if (((U32)((prefixLowestIndex-1) - repIndex) >= 3 /* intentional underflow */)
+ && (MEM_read32(repMatch) == MEM_read32(ip+1)) ) {
+ const BYTE* repMatchEnd = repIndex < prefixLowestIndex ? dictEnd : iend;
+ matchLength = ZSTD_count_2segments(ip+1+4, repMatch+4, iend, repMatchEnd, prefixLowest) + 4;
+ if (depth==0) goto _storeSequence;
+ }
+ }
+ if ( dictMode == ZSTD_noDict
+ && ((offset_1 > 0) & (MEM_read32(ip+1-offset_1) == MEM_read32(ip+1)))) {
+ matchLength = ZSTD_count(ip+1+4, ip+1+4-offset_1, iend) + 4;
+ if (depth==0) goto _storeSequence;
+ }
+
+ /* first search (depth 0) */
+ { size_t offsetFound = 999999999;
+ size_t const ml2 = ZSTD_searchMax(ms, ip, iend, &offsetFound, mls, rowLog, searchMethod, dictMode);
+ if (ml2 > matchLength)
+ matchLength = ml2, start = ip, offcode=offsetFound;
+ }
+
+ if (matchLength < 4) {
+ ip += ((ip-anchor) >> kSearchStrength) + 1; /* jump faster over incompressible sections */
+ continue;
+ }
+
+ /* let's try to find a better solution */
+ if (depth>=1)
+ while (ip<ilimit) {
+ DEBUGLOG(7, "search depth 1");
+ ip ++;
+ if ( (dictMode == ZSTD_noDict)
+ && (offcode) && ((offset_1>0) & (MEM_read32(ip) == MEM_read32(ip - offset_1)))) {
+ size_t const mlRep = ZSTD_count(ip+4, ip+4-offset_1, iend) + 4;
+ int const gain2 = (int)(mlRep * 3);
+ int const gain1 = (int)(matchLength*3 - ZSTD_highbit32((U32)STORED_TO_OFFBASE(offcode)) + 1);
+ if ((mlRep >= 4) && (gain2 > gain1))
+ matchLength = mlRep, offcode = STORE_REPCODE_1, start = ip;
+ }
+ if (isDxS) {
+ const U32 repIndex = (U32)(ip - base) - offset_1;
+ const BYTE* repMatch = repIndex < prefixLowestIndex ?
+ dictBase + (repIndex - dictIndexDelta) :
+ base + repIndex;
+ if (((U32)((prefixLowestIndex-1) - repIndex) >= 3 /* intentional underflow */)
+ && (MEM_read32(repMatch) == MEM_read32(ip)) ) {
+ const BYTE* repMatchEnd = repIndex < prefixLowestIndex ? dictEnd : iend;
+ size_t const mlRep = ZSTD_count_2segments(ip+4, repMatch+4, iend, repMatchEnd, prefixLowest) + 4;
+ int const gain2 = (int)(mlRep * 3);
+ int const gain1 = (int)(matchLength*3 - ZSTD_highbit32((U32)STORED_TO_OFFBASE(offcode)) + 1);
+ if ((mlRep >= 4) && (gain2 > gain1))
+ matchLength = mlRep, offcode = STORE_REPCODE_1, start = ip;
+ }
+ }
+ { size_t offset2=999999999;
+ size_t const ml2 = ZSTD_searchMax(ms, ip, iend, &offset2, mls, rowLog, searchMethod, dictMode);
+ int const gain2 = (int)(ml2*4 - ZSTD_highbit32((U32)STORED_TO_OFFBASE(offset2))); /* raw approx */
+ int const gain1 = (int)(matchLength*4 - ZSTD_highbit32((U32)STORED_TO_OFFBASE(offcode)) + 4);
+ if ((ml2 >= 4) && (gain2 > gain1)) {
+ matchLength = ml2, offcode = offset2, start = ip;
+ continue; /* search a better one */
+ } }
+
+ /* let's find an even better one */
+ if ((depth==2) && (ip<ilimit)) {
+ DEBUGLOG(7, "search depth 2");
+ ip ++;
+ if ( (dictMode == ZSTD_noDict)
+ && (offcode) && ((offset_1>0) & (MEM_read32(ip) == MEM_read32(ip - offset_1)))) {
+ size_t const mlRep = ZSTD_count(ip+4, ip+4-offset_1, iend) + 4;
+ int const gain2 = (int)(mlRep * 4);
+ int const gain1 = (int)(matchLength*4 - ZSTD_highbit32((U32)STORED_TO_OFFBASE(offcode)) + 1);
+ if ((mlRep >= 4) && (gain2 > gain1))
+ matchLength = mlRep, offcode = STORE_REPCODE_1, start = ip;
+ }
+ if (isDxS) {
+ const U32 repIndex = (U32)(ip - base) - offset_1;
+ const BYTE* repMatch = repIndex < prefixLowestIndex ?
+ dictBase + (repIndex - dictIndexDelta) :
+ base + repIndex;
+ if (((U32)((prefixLowestIndex-1) - repIndex) >= 3 /* intentional underflow */)
+ && (MEM_read32(repMatch) == MEM_read32(ip)) ) {
+ const BYTE* repMatchEnd = repIndex < prefixLowestIndex ? dictEnd : iend;
+ size_t const mlRep = ZSTD_count_2segments(ip+4, repMatch+4, iend, repMatchEnd, prefixLowest) + 4;
+ int const gain2 = (int)(mlRep * 4);
+ int const gain1 = (int)(matchLength*4 - ZSTD_highbit32((U32)STORED_TO_OFFBASE(offcode)) + 1);
+ if ((mlRep >= 4) && (gain2 > gain1))
+ matchLength = mlRep, offcode = STORE_REPCODE_1, start = ip;
+ }
+ }
+ { size_t offset2=999999999;
+ size_t const ml2 = ZSTD_searchMax(ms, ip, iend, &offset2, mls, rowLog, searchMethod, dictMode);
+ int const gain2 = (int)(ml2*4 - ZSTD_highbit32((U32)STORED_TO_OFFBASE(offset2))); /* raw approx */
+ int const gain1 = (int)(matchLength*4 - ZSTD_highbit32((U32)STORED_TO_OFFBASE(offcode)) + 7);
+ if ((ml2 >= 4) && (gain2 > gain1)) {
+ matchLength = ml2, offcode = offset2, start = ip;
+ continue;
+ } } }
+ break; /* nothing found : store previous solution */
+ }
+
+ /* NOTE:
+ * Pay attention that `start[-value]` can lead to strange undefined behavior
+ * notably if `value` is unsigned, resulting in a large positive `-value`.
+ */
+ /* catch up */
+ if (STORED_IS_OFFSET(offcode)) {
+ if (dictMode == ZSTD_noDict) {
+ while ( ((start > anchor) & (start - STORED_OFFSET(offcode) > prefixLowest))
+ && (start[-1] == (start-STORED_OFFSET(offcode))[-1]) ) /* only search for offset within prefix */
+ { start--; matchLength++; }
+ }
+ if (isDxS) {
+ U32 const matchIndex = (U32)((size_t)(start-base) - STORED_OFFSET(offcode));
+ const BYTE* match = (matchIndex < prefixLowestIndex) ? dictBase + matchIndex - dictIndexDelta : base + matchIndex;
+ const BYTE* const mStart = (matchIndex < prefixLowestIndex) ? dictLowest : prefixLowest;
+ while ((start>anchor) && (match>mStart) && (start[-1] == match[-1])) { start--; match--; matchLength++; } /* catch up */
+ }
+ offset_2 = offset_1; offset_1 = (U32)STORED_OFFSET(offcode);
+ }
+ /* store sequence */
+_storeSequence:
+ { size_t const litLength = (size_t)(start - anchor);
+ ZSTD_storeSeq(seqStore, litLength, anchor, iend, (U32)offcode, matchLength);
+ anchor = ip = start + matchLength;
+ }
+
+ /* check immediate repcode */
+ if (isDxS) {
+ while (ip <= ilimit) {
+ U32 const current2 = (U32)(ip-base);
+ U32 const repIndex = current2 - offset_2;
+ const BYTE* repMatch = repIndex < prefixLowestIndex ?
+ dictBase - dictIndexDelta + repIndex :
+ base + repIndex;
+ if ( ((U32)((prefixLowestIndex-1) - (U32)repIndex) >= 3 /* intentional overflow */)
+ && (MEM_read32(repMatch) == MEM_read32(ip)) ) {
+ const BYTE* const repEnd2 = repIndex < prefixLowestIndex ? dictEnd : iend;
+ matchLength = ZSTD_count_2segments(ip+4, repMatch+4, iend, repEnd2, prefixLowest) + 4;
+ offcode = offset_2; offset_2 = offset_1; offset_1 = (U32)offcode; /* swap offset_2 <=> offset_1 */
+ ZSTD_storeSeq(seqStore, 0, anchor, iend, STORE_REPCODE_1, matchLength);
+ ip += matchLength;
+ anchor = ip;
+ continue;
+ }
+ break;
+ }
+ }
+
+ if (dictMode == ZSTD_noDict) {
+ while ( ((ip <= ilimit) & (offset_2>0))
+ && (MEM_read32(ip) == MEM_read32(ip - offset_2)) ) {
+ /* store sequence */
+ matchLength = ZSTD_count(ip+4, ip+4-offset_2, iend) + 4;
+ offcode = offset_2; offset_2 = offset_1; offset_1 = (U32)offcode; /* swap repcodes */
+ ZSTD_storeSeq(seqStore, 0, anchor, iend, STORE_REPCODE_1, matchLength);
+ ip += matchLength;
+ anchor = ip;
+ continue; /* faster when present ... (?) */
+ } } }
+
+ /* Save reps for next block */
+ rep[0] = offset_1 ? offset_1 : savedOffset;
+ rep[1] = offset_2 ? offset_2 : savedOffset;
+
+ /* Return the last literals size */
+ return (size_t)(iend - anchor);
+}
+
+
+size_t ZSTD_compressBlock_btlazy2(
+ ZSTD_matchState_t* ms, seqStore_t* seqStore, U32 rep[ZSTD_REP_NUM],
+ void const* src, size_t srcSize)
+{
+ return ZSTD_compressBlock_lazy_generic(ms, seqStore, rep, src, srcSize, search_binaryTree, 2, ZSTD_noDict);
+}
+
+size_t ZSTD_compressBlock_lazy2(
+ ZSTD_matchState_t* ms, seqStore_t* seqStore, U32 rep[ZSTD_REP_NUM],
+ void const* src, size_t srcSize)
+{
+ return ZSTD_compressBlock_lazy_generic(ms, seqStore, rep, src, srcSize, search_hashChain, 2, ZSTD_noDict);
+}
+
+size_t ZSTD_compressBlock_lazy(
+ ZSTD_matchState_t* ms, seqStore_t* seqStore, U32 rep[ZSTD_REP_NUM],
+ void const* src, size_t srcSize)
+{
+ return ZSTD_compressBlock_lazy_generic(ms, seqStore, rep, src, srcSize, search_hashChain, 1, ZSTD_noDict);
+}
+
+size_t ZSTD_compressBlock_greedy(
+ ZSTD_matchState_t* ms, seqStore_t* seqStore, U32 rep[ZSTD_REP_NUM],
+ void const* src, size_t srcSize)
+{
+ return ZSTD_compressBlock_lazy_generic(ms, seqStore, rep, src, srcSize, search_hashChain, 0, ZSTD_noDict);
+}
+
+size_t ZSTD_compressBlock_btlazy2_dictMatchState(
+ ZSTD_matchState_t* ms, seqStore_t* seqStore, U32 rep[ZSTD_REP_NUM],
+ void const* src, size_t srcSize)
+{
+ return ZSTD_compressBlock_lazy_generic(ms, seqStore, rep, src, srcSize, search_binaryTree, 2, ZSTD_dictMatchState);
+}
+
+size_t ZSTD_compressBlock_lazy2_dictMatchState(
+ ZSTD_matchState_t* ms, seqStore_t* seqStore, U32 rep[ZSTD_REP_NUM],
+ void const* src, size_t srcSize)
+{
+ return ZSTD_compressBlock_lazy_generic(ms, seqStore, rep, src, srcSize, search_hashChain, 2, ZSTD_dictMatchState);
+}
+
+size_t ZSTD_compressBlock_lazy_dictMatchState(
+ ZSTD_matchState_t* ms, seqStore_t* seqStore, U32 rep[ZSTD_REP_NUM],
+ void const* src, size_t srcSize)
+{
+ return ZSTD_compressBlock_lazy_generic(ms, seqStore, rep, src, srcSize, search_hashChain, 1, ZSTD_dictMatchState);
+}
+
+size_t ZSTD_compressBlock_greedy_dictMatchState(
+ ZSTD_matchState_t* ms, seqStore_t* seqStore, U32 rep[ZSTD_REP_NUM],
+ void const* src, size_t srcSize)
+{
+ return ZSTD_compressBlock_lazy_generic(ms, seqStore, rep, src, srcSize, search_hashChain, 0, ZSTD_dictMatchState);
+}
+
+
+size_t ZSTD_compressBlock_lazy2_dedicatedDictSearch(
+ ZSTD_matchState_t* ms, seqStore_t* seqStore, U32 rep[ZSTD_REP_NUM],
+ void const* src, size_t srcSize)
+{
+ return ZSTD_compressBlock_lazy_generic(ms, seqStore, rep, src, srcSize, search_hashChain, 2, ZSTD_dedicatedDictSearch);
+}
+
+size_t ZSTD_compressBlock_lazy_dedicatedDictSearch(
+ ZSTD_matchState_t* ms, seqStore_t* seqStore, U32 rep[ZSTD_REP_NUM],
+ void const* src, size_t srcSize)
+{
+ return ZSTD_compressBlock_lazy_generic(ms, seqStore, rep, src, srcSize, search_hashChain, 1, ZSTD_dedicatedDictSearch);
+}
+
+size_t ZSTD_compressBlock_greedy_dedicatedDictSearch(
+ ZSTD_matchState_t* ms, seqStore_t* seqStore, U32 rep[ZSTD_REP_NUM],
+ void const* src, size_t srcSize)
+{
+ return ZSTD_compressBlock_lazy_generic(ms, seqStore, rep, src, srcSize, search_hashChain, 0, ZSTD_dedicatedDictSearch);
+}
+
+/* Row-based matchfinder */
+size_t ZSTD_compressBlock_lazy2_row(
+ ZSTD_matchState_t* ms, seqStore_t* seqStore, U32 rep[ZSTD_REP_NUM],
+ void const* src, size_t srcSize)
+{
+ return ZSTD_compressBlock_lazy_generic(ms, seqStore, rep, src, srcSize, search_rowHash, 2, ZSTD_noDict);
+}
+
+size_t ZSTD_compressBlock_lazy_row(
+ ZSTD_matchState_t* ms, seqStore_t* seqStore, U32 rep[ZSTD_REP_NUM],
+ void const* src, size_t srcSize)
+{
+ return ZSTD_compressBlock_lazy_generic(ms, seqStore, rep, src, srcSize, search_rowHash, 1, ZSTD_noDict);
+}
+
+size_t ZSTD_compressBlock_greedy_row(
+ ZSTD_matchState_t* ms, seqStore_t* seqStore, U32 rep[ZSTD_REP_NUM],
+ void const* src, size_t srcSize)
+{
+ return ZSTD_compressBlock_lazy_generic(ms, seqStore, rep, src, srcSize, search_rowHash, 0, ZSTD_noDict);
+}
+
+size_t ZSTD_compressBlock_lazy2_dictMatchState_row(
+ ZSTD_matchState_t* ms, seqStore_t* seqStore, U32 rep[ZSTD_REP_NUM],
+ void const* src, size_t srcSize)
+{
+ return ZSTD_compressBlock_lazy_generic(ms, seqStore, rep, src, srcSize, search_rowHash, 2, ZSTD_dictMatchState);
+}
+
+size_t ZSTD_compressBlock_lazy_dictMatchState_row(
+ ZSTD_matchState_t* ms, seqStore_t* seqStore, U32 rep[ZSTD_REP_NUM],
+ void const* src, size_t srcSize)
+{
+ return ZSTD_compressBlock_lazy_generic(ms, seqStore, rep, src, srcSize, search_rowHash, 1, ZSTD_dictMatchState);
+}
+
+size_t ZSTD_compressBlock_greedy_dictMatchState_row(
+ ZSTD_matchState_t* ms, seqStore_t* seqStore, U32 rep[ZSTD_REP_NUM],
+ void const* src, size_t srcSize)
+{
+ return ZSTD_compressBlock_lazy_generic(ms, seqStore, rep, src, srcSize, search_rowHash, 0, ZSTD_dictMatchState);
+}
+
+
+size_t ZSTD_compressBlock_lazy2_dedicatedDictSearch_row(
+ ZSTD_matchState_t* ms, seqStore_t* seqStore, U32 rep[ZSTD_REP_NUM],
+ void const* src, size_t srcSize)
+{
+ return ZSTD_compressBlock_lazy_generic(ms, seqStore, rep, src, srcSize, search_rowHash, 2, ZSTD_dedicatedDictSearch);
+}
+
+size_t ZSTD_compressBlock_lazy_dedicatedDictSearch_row(
+ ZSTD_matchState_t* ms, seqStore_t* seqStore, U32 rep[ZSTD_REP_NUM],
+ void const* src, size_t srcSize)
+{
+ return ZSTD_compressBlock_lazy_generic(ms, seqStore, rep, src, srcSize, search_rowHash, 1, ZSTD_dedicatedDictSearch);
+}
+
+size_t ZSTD_compressBlock_greedy_dedicatedDictSearch_row(
+ ZSTD_matchState_t* ms, seqStore_t* seqStore, U32 rep[ZSTD_REP_NUM],
+ void const* src, size_t srcSize)
+{
+ return ZSTD_compressBlock_lazy_generic(ms, seqStore, rep, src, srcSize, search_rowHash, 0, ZSTD_dedicatedDictSearch);
+}
+
+FORCE_INLINE_TEMPLATE
+size_t ZSTD_compressBlock_lazy_extDict_generic(
+ ZSTD_matchState_t* ms, seqStore_t* seqStore,
+ U32 rep[ZSTD_REP_NUM],
+ const void* src, size_t srcSize,
+ const searchMethod_e searchMethod, const U32 depth)
+{
+ const BYTE* const istart = (const BYTE*)src;
+ const BYTE* ip = istart;
+ const BYTE* anchor = istart;
+ const BYTE* const iend = istart + srcSize;
+ const BYTE* const ilimit = searchMethod == search_rowHash ? iend - 8 - ZSTD_ROW_HASH_CACHE_SIZE : iend - 8;
+ const BYTE* const base = ms->window.base;
+ const U32 dictLimit = ms->window.dictLimit;
+ const BYTE* const prefixStart = base + dictLimit;
+ const BYTE* const dictBase = ms->window.dictBase;
+ const BYTE* const dictEnd = dictBase + dictLimit;
+ const BYTE* const dictStart = dictBase + ms->window.lowLimit;
+ const U32 windowLog = ms->cParams.windowLog;
+ const U32 mls = BOUNDED(4, ms->cParams.minMatch, 6);
+ const U32 rowLog = BOUNDED(4, ms->cParams.searchLog, 6);
+
+ U32 offset_1 = rep[0], offset_2 = rep[1];
+
+ DEBUGLOG(5, "ZSTD_compressBlock_lazy_extDict_generic (searchFunc=%u)", (U32)searchMethod);
+
+ /* init */
+ ip += (ip == prefixStart);
+ if (searchMethod == search_rowHash) {
+ ZSTD_row_fillHashCache(ms, base, rowLog,
+ MIN(ms->cParams.minMatch, 6 /* mls caps out at 6 */),
+ ms->nextToUpdate, ilimit);
+ }
+
+ /* Match Loop */
+#if defined(__x86_64__)
+ /* I've measured random a 5% speed loss on levels 5 & 6 (greedy) when the
+ * code alignment is perturbed. To fix the instability align the loop on 32-bytes.
+ */
+ __asm__(".p2align 5");
+#endif
+ while (ip < ilimit) {
+ size_t matchLength=0;
+ size_t offcode=STORE_REPCODE_1;
+ const BYTE* start=ip+1;
+ U32 curr = (U32)(ip-base);
+
+ /* check repCode */
+ { const U32 windowLow = ZSTD_getLowestMatchIndex(ms, curr+1, windowLog);
+ const U32 repIndex = (U32)(curr+1 - offset_1);
+ const BYTE* const repBase = repIndex < dictLimit ? dictBase : base;
+ const BYTE* const repMatch = repBase + repIndex;
+ if ( ((U32)((dictLimit-1) - repIndex) >= 3) /* intentional overflow */
+ & (offset_1 <= curr+1 - windowLow) ) /* note: we are searching at curr+1 */
+ if (MEM_read32(ip+1) == MEM_read32(repMatch)) {
+ /* repcode detected we should take it */
+ const BYTE* const repEnd = repIndex < dictLimit ? dictEnd : iend;
+ matchLength = ZSTD_count_2segments(ip+1+4, repMatch+4, iend, repEnd, prefixStart) + 4;
+ if (depth==0) goto _storeSequence;
+ } }
+
+ /* first search (depth 0) */
+ { size_t offsetFound = 999999999;
+ size_t const ml2 = ZSTD_searchMax(ms, ip, iend, &offsetFound, mls, rowLog, searchMethod, ZSTD_extDict);
+ if (ml2 > matchLength)
+ matchLength = ml2, start = ip, offcode=offsetFound;
+ }
+
+ if (matchLength < 4) {
+ ip += ((ip-anchor) >> kSearchStrength) + 1; /* jump faster over incompressible sections */
+ continue;
+ }
+
+ /* let's try to find a better solution */
+ if (depth>=1)
+ while (ip<ilimit) {
+ ip ++;
+ curr++;
+ /* check repCode */
+ if (offcode) {
+ const U32 windowLow = ZSTD_getLowestMatchIndex(ms, curr, windowLog);
+ const U32 repIndex = (U32)(curr - offset_1);
+ const BYTE* const repBase = repIndex < dictLimit ? dictBase : base;
+ const BYTE* const repMatch = repBase + repIndex;
+ if ( ((U32)((dictLimit-1) - repIndex) >= 3) /* intentional overflow : do not test positions overlapping 2 memory segments */
+ & (offset_1 <= curr - windowLow) ) /* equivalent to `curr > repIndex >= windowLow` */
+ if (MEM_read32(ip) == MEM_read32(repMatch)) {
+ /* repcode detected */
+ const BYTE* const repEnd = repIndex < dictLimit ? dictEnd : iend;
+ size_t const repLength = ZSTD_count_2segments(ip+4, repMatch+4, iend, repEnd, prefixStart) + 4;
+ int const gain2 = (int)(repLength * 3);
+ int const gain1 = (int)(matchLength*3 - ZSTD_highbit32((U32)STORED_TO_OFFBASE(offcode)) + 1);
+ if ((repLength >= 4) && (gain2 > gain1))
+ matchLength = repLength, offcode = STORE_REPCODE_1, start = ip;
+ } }
+
+ /* search match, depth 1 */
+ { size_t offset2=999999999;
+ size_t const ml2 = ZSTD_searchMax(ms, ip, iend, &offset2, mls, rowLog, searchMethod, ZSTD_extDict);
+ int const gain2 = (int)(ml2*4 - ZSTD_highbit32((U32)STORED_TO_OFFBASE(offset2))); /* raw approx */
+ int const gain1 = (int)(matchLength*4 - ZSTD_highbit32((U32)STORED_TO_OFFBASE(offcode)) + 4);
+ if ((ml2 >= 4) && (gain2 > gain1)) {
+ matchLength = ml2, offcode = offset2, start = ip;
+ continue; /* search a better one */
+ } }
+
+ /* let's find an even better one */
+ if ((depth==2) && (ip<ilimit)) {
+ ip ++;
+ curr++;
+ /* check repCode */
+ if (offcode) {
+ const U32 windowLow = ZSTD_getLowestMatchIndex(ms, curr, windowLog);
+ const U32 repIndex = (U32)(curr - offset_1);
+ const BYTE* const repBase = repIndex < dictLimit ? dictBase : base;
+ const BYTE* const repMatch = repBase + repIndex;
+ if ( ((U32)((dictLimit-1) - repIndex) >= 3) /* intentional overflow : do not test positions overlapping 2 memory segments */
+ & (offset_1 <= curr - windowLow) ) /* equivalent to `curr > repIndex >= windowLow` */
+ if (MEM_read32(ip) == MEM_read32(repMatch)) {
+ /* repcode detected */
+ const BYTE* const repEnd = repIndex < dictLimit ? dictEnd : iend;
+ size_t const repLength = ZSTD_count_2segments(ip+4, repMatch+4, iend, repEnd, prefixStart) + 4;
+ int const gain2 = (int)(repLength * 4);
+ int const gain1 = (int)(matchLength*4 - ZSTD_highbit32((U32)STORED_TO_OFFBASE(offcode)) + 1);
+ if ((repLength >= 4) && (gain2 > gain1))
+ matchLength = repLength, offcode = STORE_REPCODE_1, start = ip;
+ } }
+
+ /* search match, depth 2 */
+ { size_t offset2=999999999;
+ size_t const ml2 = ZSTD_searchMax(ms, ip, iend, &offset2, mls, rowLog, searchMethod, ZSTD_extDict);
+ int const gain2 = (int)(ml2*4 - ZSTD_highbit32((U32)STORED_TO_OFFBASE(offset2))); /* raw approx */
+ int const gain1 = (int)(matchLength*4 - ZSTD_highbit32((U32)STORED_TO_OFFBASE(offcode)) + 7);
+ if ((ml2 >= 4) && (gain2 > gain1)) {
+ matchLength = ml2, offcode = offset2, start = ip;
+ continue;
+ } } }
+ break; /* nothing found : store previous solution */
+ }
+
+ /* catch up */
+ if (STORED_IS_OFFSET(offcode)) {
+ U32 const matchIndex = (U32)((size_t)(start-base) - STORED_OFFSET(offcode));
+ const BYTE* match = (matchIndex < dictLimit) ? dictBase + matchIndex : base + matchIndex;
+ const BYTE* const mStart = (matchIndex < dictLimit) ? dictStart : prefixStart;
+ while ((start>anchor) && (match>mStart) && (start[-1] == match[-1])) { start--; match--; matchLength++; } /* catch up */
+ offset_2 = offset_1; offset_1 = (U32)STORED_OFFSET(offcode);
+ }
+
+ /* store sequence */
+_storeSequence:
+ { size_t const litLength = (size_t)(start - anchor);
+ ZSTD_storeSeq(seqStore, litLength, anchor, iend, (U32)offcode, matchLength);
+ anchor = ip = start + matchLength;
+ }
+
+ /* check immediate repcode */
+ while (ip <= ilimit) {
+ const U32 repCurrent = (U32)(ip-base);
+ const U32 windowLow = ZSTD_getLowestMatchIndex(ms, repCurrent, windowLog);
+ const U32 repIndex = repCurrent - offset_2;
+ const BYTE* const repBase = repIndex < dictLimit ? dictBase : base;
+ const BYTE* const repMatch = repBase + repIndex;
+ if ( ((U32)((dictLimit-1) - repIndex) >= 3) /* intentional overflow : do not test positions overlapping 2 memory segments */
+ & (offset_2 <= repCurrent - windowLow) ) /* equivalent to `curr > repIndex >= windowLow` */
+ if (MEM_read32(ip) == MEM_read32(repMatch)) {
+ /* repcode detected we should take it */
+ const BYTE* const repEnd = repIndex < dictLimit ? dictEnd : iend;
+ matchLength = ZSTD_count_2segments(ip+4, repMatch+4, iend, repEnd, prefixStart) + 4;
+ offcode = offset_2; offset_2 = offset_1; offset_1 = (U32)offcode; /* swap offset history */
+ ZSTD_storeSeq(seqStore, 0, anchor, iend, STORE_REPCODE_1, matchLength);
+ ip += matchLength;
+ anchor = ip;
+ continue; /* faster when present ... (?) */
+ }
+ break;
+ } }
+
+ /* Save reps for next block */
+ rep[0] = offset_1;
+ rep[1] = offset_2;
+
+ /* Return the last literals size */
+ return (size_t)(iend - anchor);
+}
+
+
+size_t ZSTD_compressBlock_greedy_extDict(
+ ZSTD_matchState_t* ms, seqStore_t* seqStore, U32 rep[ZSTD_REP_NUM],
+ void const* src, size_t srcSize)
+{
+ return ZSTD_compressBlock_lazy_extDict_generic(ms, seqStore, rep, src, srcSize, search_hashChain, 0);
+}
+
+size_t ZSTD_compressBlock_lazy_extDict(
+ ZSTD_matchState_t* ms, seqStore_t* seqStore, U32 rep[ZSTD_REP_NUM],
+ void const* src, size_t srcSize)
+
+{
+ return ZSTD_compressBlock_lazy_extDict_generic(ms, seqStore, rep, src, srcSize, search_hashChain, 1);
+}
+
+size_t ZSTD_compressBlock_lazy2_extDict(
+ ZSTD_matchState_t* ms, seqStore_t* seqStore, U32 rep[ZSTD_REP_NUM],
+ void const* src, size_t srcSize)
+
+{
+ return ZSTD_compressBlock_lazy_extDict_generic(ms, seqStore, rep, src, srcSize, search_hashChain, 2);
+}
+
+size_t ZSTD_compressBlock_btlazy2_extDict(
+ ZSTD_matchState_t* ms, seqStore_t* seqStore, U32 rep[ZSTD_REP_NUM],
+ void const* src, size_t srcSize)
+
+{
+ return ZSTD_compressBlock_lazy_extDict_generic(ms, seqStore, rep, src, srcSize, search_binaryTree, 2);
+}
+
+size_t ZSTD_compressBlock_greedy_extDict_row(
+ ZSTD_matchState_t* ms, seqStore_t* seqStore, U32 rep[ZSTD_REP_NUM],
+ void const* src, size_t srcSize)
+{
+ return ZSTD_compressBlock_lazy_extDict_generic(ms, seqStore, rep, src, srcSize, search_rowHash, 0);
+}
+
+size_t ZSTD_compressBlock_lazy_extDict_row(
+ ZSTD_matchState_t* ms, seqStore_t* seqStore, U32 rep[ZSTD_REP_NUM],
+ void const* src, size_t srcSize)
+
+{
+ return ZSTD_compressBlock_lazy_extDict_generic(ms, seqStore, rep, src, srcSize, search_rowHash, 1);
+}
+
+size_t ZSTD_compressBlock_lazy2_extDict_row(
+ ZSTD_matchState_t* ms, seqStore_t* seqStore, U32 rep[ZSTD_REP_NUM],
+ void const* src, size_t srcSize)
+
+{
+ return ZSTD_compressBlock_lazy_extDict_generic(ms, seqStore, rep, src, srcSize, search_rowHash, 2);
+}
diff --git a/lib/zstd/compress/zstd_lazy.h b/lib/zstd/compress/zstd_lazy.h
new file mode 100644
index 0000000000..e5bdf4df8d
--- /dev/null
+++ b/lib/zstd/compress/zstd_lazy.h
@@ -0,0 +1,119 @@
+/*
+ * Copyright (c) Yann Collet, Facebook, Inc.
+ * All rights reserved.
+ *
+ * This source code is licensed under both the BSD-style license (found in the
+ * LICENSE file in the root directory of this source tree) and the GPLv2 (found
+ * in the COPYING file in the root directory of this source tree).
+ * You may select, at your option, one of the above-listed licenses.
+ */
+
+#ifndef ZSTD_LAZY_H
+#define ZSTD_LAZY_H
+
+
+#include "zstd_compress_internal.h"
+
+/*
+ * Dedicated Dictionary Search Structure bucket log. In the
+ * ZSTD_dedicatedDictSearch mode, the hashTable has
+ * 2 ** ZSTD_LAZY_DDSS_BUCKET_LOG entries in each bucket, rather than just
+ * one.
+ */
+#define ZSTD_LAZY_DDSS_BUCKET_LOG 2
+
+U32 ZSTD_insertAndFindFirstIndex(ZSTD_matchState_t* ms, const BYTE* ip);
+void ZSTD_row_update(ZSTD_matchState_t* const ms, const BYTE* ip);
+
+void ZSTD_dedicatedDictSearch_lazy_loadDictionary(ZSTD_matchState_t* ms, const BYTE* const ip);
+
+void ZSTD_preserveUnsortedMark (U32* const table, U32 const size, U32 const reducerValue); /*! used in ZSTD_reduceIndex(). preemptively increase value of ZSTD_DUBT_UNSORTED_MARK */
+
+size_t ZSTD_compressBlock_btlazy2(
+ ZSTD_matchState_t* ms, seqStore_t* seqStore, U32 rep[ZSTD_REP_NUM],
+ void const* src, size_t srcSize);
+size_t ZSTD_compressBlock_lazy2(
+ ZSTD_matchState_t* ms, seqStore_t* seqStore, U32 rep[ZSTD_REP_NUM],
+ void const* src, size_t srcSize);
+size_t ZSTD_compressBlock_lazy(
+ ZSTD_matchState_t* ms, seqStore_t* seqStore, U32 rep[ZSTD_REP_NUM],
+ void const* src, size_t srcSize);
+size_t ZSTD_compressBlock_greedy(
+ ZSTD_matchState_t* ms, seqStore_t* seqStore, U32 rep[ZSTD_REP_NUM],
+ void const* src, size_t srcSize);
+size_t ZSTD_compressBlock_lazy2_row(
+ ZSTD_matchState_t* ms, seqStore_t* seqStore, U32 rep[ZSTD_REP_NUM],
+ void const* src, size_t srcSize);
+size_t ZSTD_compressBlock_lazy_row(
+ ZSTD_matchState_t* ms, seqStore_t* seqStore, U32 rep[ZSTD_REP_NUM],
+ void const* src, size_t srcSize);
+size_t ZSTD_compressBlock_greedy_row(
+ ZSTD_matchState_t* ms, seqStore_t* seqStore, U32 rep[ZSTD_REP_NUM],
+ void const* src, size_t srcSize);
+
+size_t ZSTD_compressBlock_btlazy2_dictMatchState(
+ ZSTD_matchState_t* ms, seqStore_t* seqStore, U32 rep[ZSTD_REP_NUM],
+ void const* src, size_t srcSize);
+size_t ZSTD_compressBlock_lazy2_dictMatchState(
+ ZSTD_matchState_t* ms, seqStore_t* seqStore, U32 rep[ZSTD_REP_NUM],
+ void const* src, size_t srcSize);
+size_t ZSTD_compressBlock_lazy_dictMatchState(
+ ZSTD_matchState_t* ms, seqStore_t* seqStore, U32 rep[ZSTD_REP_NUM],
+ void const* src, size_t srcSize);
+size_t ZSTD_compressBlock_greedy_dictMatchState(
+ ZSTD_matchState_t* ms, seqStore_t* seqStore, U32 rep[ZSTD_REP_NUM],
+ void const* src, size_t srcSize);
+size_t ZSTD_compressBlock_lazy2_dictMatchState_row(
+ ZSTD_matchState_t* ms, seqStore_t* seqStore, U32 rep[ZSTD_REP_NUM],
+ void const* src, size_t srcSize);
+size_t ZSTD_compressBlock_lazy_dictMatchState_row(
+ ZSTD_matchState_t* ms, seqStore_t* seqStore, U32 rep[ZSTD_REP_NUM],
+ void const* src, size_t srcSize);
+size_t ZSTD_compressBlock_greedy_dictMatchState_row(
+ ZSTD_matchState_t* ms, seqStore_t* seqStore, U32 rep[ZSTD_REP_NUM],
+ void const* src, size_t srcSize);
+
+size_t ZSTD_compressBlock_lazy2_dedicatedDictSearch(
+ ZSTD_matchState_t* ms, seqStore_t* seqStore, U32 rep[ZSTD_REP_NUM],
+ void const* src, size_t srcSize);
+size_t ZSTD_compressBlock_lazy_dedicatedDictSearch(
+ ZSTD_matchState_t* ms, seqStore_t* seqStore, U32 rep[ZSTD_REP_NUM],
+ void const* src, size_t srcSize);
+size_t ZSTD_compressBlock_greedy_dedicatedDictSearch(
+ ZSTD_matchState_t* ms, seqStore_t* seqStore, U32 rep[ZSTD_REP_NUM],
+ void const* src, size_t srcSize);
+size_t ZSTD_compressBlock_lazy2_dedicatedDictSearch_row(
+ ZSTD_matchState_t* ms, seqStore_t* seqStore, U32 rep[ZSTD_REP_NUM],
+ void const* src, size_t srcSize);
+size_t ZSTD_compressBlock_lazy_dedicatedDictSearch_row(
+ ZSTD_matchState_t* ms, seqStore_t* seqStore, U32 rep[ZSTD_REP_NUM],
+ void const* src, size_t srcSize);
+size_t ZSTD_compressBlock_greedy_dedicatedDictSearch_row(
+ ZSTD_matchState_t* ms, seqStore_t* seqStore, U32 rep[ZSTD_REP_NUM],
+ void const* src, size_t srcSize);
+
+size_t ZSTD_compressBlock_greedy_extDict(
+ ZSTD_matchState_t* ms, seqStore_t* seqStore, U32 rep[ZSTD_REP_NUM],
+ void const* src, size_t srcSize);
+size_t ZSTD_compressBlock_lazy_extDict(
+ ZSTD_matchState_t* ms, seqStore_t* seqStore, U32 rep[ZSTD_REP_NUM],
+ void const* src, size_t srcSize);
+size_t ZSTD_compressBlock_lazy2_extDict(
+ ZSTD_matchState_t* ms, seqStore_t* seqStore, U32 rep[ZSTD_REP_NUM],
+ void const* src, size_t srcSize);
+size_t ZSTD_compressBlock_greedy_extDict_row(
+ ZSTD_matchState_t* ms, seqStore_t* seqStore, U32 rep[ZSTD_REP_NUM],
+ void const* src, size_t srcSize);
+size_t ZSTD_compressBlock_lazy_extDict_row(
+ ZSTD_matchState_t* ms, seqStore_t* seqStore, U32 rep[ZSTD_REP_NUM],
+ void const* src, size_t srcSize);
+size_t ZSTD_compressBlock_lazy2_extDict_row(
+ ZSTD_matchState_t* ms, seqStore_t* seqStore, U32 rep[ZSTD_REP_NUM],
+ void const* src, size_t srcSize);
+size_t ZSTD_compressBlock_btlazy2_extDict(
+ ZSTD_matchState_t* ms, seqStore_t* seqStore, U32 rep[ZSTD_REP_NUM],
+ void const* src, size_t srcSize);
+
+
+
+#endif /* ZSTD_LAZY_H */
diff --git a/lib/zstd/compress/zstd_ldm.c b/lib/zstd/compress/zstd_ldm.c
new file mode 100644
index 0000000000..dd86fc83e7
--- /dev/null
+++ b/lib/zstd/compress/zstd_ldm.c
@@ -0,0 +1,724 @@
+/*
+ * Copyright (c) Yann Collet, Facebook, Inc.
+ * All rights reserved.
+ *
+ * This source code is licensed under both the BSD-style license (found in the
+ * LICENSE file in the root directory of this source tree) and the GPLv2 (found
+ * in the COPYING file in the root directory of this source tree).
+ * You may select, at your option, one of the above-listed licenses.
+ */
+
+#include "zstd_ldm.h"
+
+#include "../common/debug.h"
+#include <linux/xxhash.h>
+#include "zstd_fast.h" /* ZSTD_fillHashTable() */
+#include "zstd_double_fast.h" /* ZSTD_fillDoubleHashTable() */
+#include "zstd_ldm_geartab.h"
+
+#define LDM_BUCKET_SIZE_LOG 3
+#define LDM_MIN_MATCH_LENGTH 64
+#define LDM_HASH_RLOG 7
+
+typedef struct {
+ U64 rolling;
+ U64 stopMask;
+} ldmRollingHashState_t;
+
+/* ZSTD_ldm_gear_init():
+ *
+ * Initializes the rolling hash state such that it will honor the
+ * settings in params. */
+static void ZSTD_ldm_gear_init(ldmRollingHashState_t* state, ldmParams_t const* params)
+{
+ unsigned maxBitsInMask = MIN(params->minMatchLength, 64);
+ unsigned hashRateLog = params->hashRateLog;
+
+ state->rolling = ~(U32)0;
+
+ /* The choice of the splitting criterion is subject to two conditions:
+ * 1. it has to trigger on average every 2^(hashRateLog) bytes;
+ * 2. ideally, it has to depend on a window of minMatchLength bytes.
+ *
+ * In the gear hash algorithm, bit n depends on the last n bytes;
+ * so in order to obtain a good quality splitting criterion it is
+ * preferable to use bits with high weight.
+ *
+ * To match condition 1 we use a mask with hashRateLog bits set
+ * and, because of the previous remark, we make sure these bits
+ * have the highest possible weight while still respecting
+ * condition 2.
+ */
+ if (hashRateLog > 0 && hashRateLog <= maxBitsInMask) {
+ state->stopMask = (((U64)1 << hashRateLog) - 1) << (maxBitsInMask - hashRateLog);
+ } else {
+ /* In this degenerate case we simply honor the hash rate. */
+ state->stopMask = ((U64)1 << hashRateLog) - 1;
+ }
+}
+
+/* ZSTD_ldm_gear_reset()
+ * Feeds [data, data + minMatchLength) into the hash without registering any
+ * splits. This effectively resets the hash state. This is used when skipping
+ * over data, either at the beginning of a block, or skipping sections.
+ */
+static void ZSTD_ldm_gear_reset(ldmRollingHashState_t* state,
+ BYTE const* data, size_t minMatchLength)
+{
+ U64 hash = state->rolling;
+ size_t n = 0;
+
+#define GEAR_ITER_ONCE() do { \
+ hash = (hash << 1) + ZSTD_ldm_gearTab[data[n] & 0xff]; \
+ n += 1; \
+ } while (0)
+ while (n + 3 < minMatchLength) {
+ GEAR_ITER_ONCE();
+ GEAR_ITER_ONCE();
+ GEAR_ITER_ONCE();
+ GEAR_ITER_ONCE();
+ }
+ while (n < minMatchLength) {
+ GEAR_ITER_ONCE();
+ }
+#undef GEAR_ITER_ONCE
+}
+
+/* ZSTD_ldm_gear_feed():
+ *
+ * Registers in the splits array all the split points found in the first
+ * size bytes following the data pointer. This function terminates when
+ * either all the data has been processed or LDM_BATCH_SIZE splits are
+ * present in the splits array.
+ *
+ * Precondition: The splits array must not be full.
+ * Returns: The number of bytes processed. */
+static size_t ZSTD_ldm_gear_feed(ldmRollingHashState_t* state,
+ BYTE const* data, size_t size,
+ size_t* splits, unsigned* numSplits)
+{
+ size_t n;
+ U64 hash, mask;
+
+ hash = state->rolling;
+ mask = state->stopMask;
+ n = 0;
+
+#define GEAR_ITER_ONCE() do { \
+ hash = (hash << 1) + ZSTD_ldm_gearTab[data[n] & 0xff]; \
+ n += 1; \
+ if (UNLIKELY((hash & mask) == 0)) { \
+ splits[*numSplits] = n; \
+ *numSplits += 1; \
+ if (*numSplits == LDM_BATCH_SIZE) \
+ goto done; \
+ } \
+ } while (0)
+
+ while (n + 3 < size) {
+ GEAR_ITER_ONCE();
+ GEAR_ITER_ONCE();
+ GEAR_ITER_ONCE();
+ GEAR_ITER_ONCE();
+ }
+ while (n < size) {
+ GEAR_ITER_ONCE();
+ }
+
+#undef GEAR_ITER_ONCE
+
+done:
+ state->rolling = hash;
+ return n;
+}
+
+void ZSTD_ldm_adjustParameters(ldmParams_t* params,
+ ZSTD_compressionParameters const* cParams)
+{
+ params->windowLog = cParams->windowLog;
+ ZSTD_STATIC_ASSERT(LDM_BUCKET_SIZE_LOG <= ZSTD_LDM_BUCKETSIZELOG_MAX);
+ DEBUGLOG(4, "ZSTD_ldm_adjustParameters");
+ if (!params->bucketSizeLog) params->bucketSizeLog = LDM_BUCKET_SIZE_LOG;
+ if (!params->minMatchLength) params->minMatchLength = LDM_MIN_MATCH_LENGTH;
+ if (params->hashLog == 0) {
+ params->hashLog = MAX(ZSTD_HASHLOG_MIN, params->windowLog - LDM_HASH_RLOG);
+ assert(params->hashLog <= ZSTD_HASHLOG_MAX);
+ }
+ if (params->hashRateLog == 0) {
+ params->hashRateLog = params->windowLog < params->hashLog
+ ? 0
+ : params->windowLog - params->hashLog;
+ }
+ params->bucketSizeLog = MIN(params->bucketSizeLog, params->hashLog);
+}
+
+size_t ZSTD_ldm_getTableSize(ldmParams_t params)
+{
+ size_t const ldmHSize = ((size_t)1) << params.hashLog;
+ size_t const ldmBucketSizeLog = MIN(params.bucketSizeLog, params.hashLog);
+ size_t const ldmBucketSize = ((size_t)1) << (params.hashLog - ldmBucketSizeLog);
+ size_t const totalSize = ZSTD_cwksp_alloc_size(ldmBucketSize)
+ + ZSTD_cwksp_alloc_size(ldmHSize * sizeof(ldmEntry_t));
+ return params.enableLdm == ZSTD_ps_enable ? totalSize : 0;
+}
+
+size_t ZSTD_ldm_getMaxNbSeq(ldmParams_t params, size_t maxChunkSize)
+{
+ return params.enableLdm == ZSTD_ps_enable ? (maxChunkSize / params.minMatchLength) : 0;
+}
+
+/* ZSTD_ldm_getBucket() :
+ * Returns a pointer to the start of the bucket associated with hash. */
+static ldmEntry_t* ZSTD_ldm_getBucket(
+ ldmState_t* ldmState, size_t hash, ldmParams_t const ldmParams)
+{
+ return ldmState->hashTable + (hash << ldmParams.bucketSizeLog);
+}
+
+/* ZSTD_ldm_insertEntry() :
+ * Insert the entry with corresponding hash into the hash table */
+static void ZSTD_ldm_insertEntry(ldmState_t* ldmState,
+ size_t const hash, const ldmEntry_t entry,
+ ldmParams_t const ldmParams)
+{
+ BYTE* const pOffset = ldmState->bucketOffsets + hash;
+ unsigned const offset = *pOffset;
+
+ *(ZSTD_ldm_getBucket(ldmState, hash, ldmParams) + offset) = entry;
+ *pOffset = (BYTE)((offset + 1) & ((1u << ldmParams.bucketSizeLog) - 1));
+
+}
+
+/* ZSTD_ldm_countBackwardsMatch() :
+ * Returns the number of bytes that match backwards before pIn and pMatch.
+ *
+ * We count only bytes where pMatch >= pBase and pIn >= pAnchor. */
+static size_t ZSTD_ldm_countBackwardsMatch(
+ const BYTE* pIn, const BYTE* pAnchor,
+ const BYTE* pMatch, const BYTE* pMatchBase)
+{
+ size_t matchLength = 0;
+ while (pIn > pAnchor && pMatch > pMatchBase && pIn[-1] == pMatch[-1]) {
+ pIn--;
+ pMatch--;
+ matchLength++;
+ }
+ return matchLength;
+}
+
+/* ZSTD_ldm_countBackwardsMatch_2segments() :
+ * Returns the number of bytes that match backwards from pMatch,
+ * even with the backwards match spanning 2 different segments.
+ *
+ * On reaching `pMatchBase`, start counting from mEnd */
+static size_t ZSTD_ldm_countBackwardsMatch_2segments(
+ const BYTE* pIn, const BYTE* pAnchor,
+ const BYTE* pMatch, const BYTE* pMatchBase,
+ const BYTE* pExtDictStart, const BYTE* pExtDictEnd)
+{
+ size_t matchLength = ZSTD_ldm_countBackwardsMatch(pIn, pAnchor, pMatch, pMatchBase);
+ if (pMatch - matchLength != pMatchBase || pMatchBase == pExtDictStart) {
+ /* If backwards match is entirely in the extDict or prefix, immediately return */
+ return matchLength;
+ }
+ DEBUGLOG(7, "ZSTD_ldm_countBackwardsMatch_2segments: found 2-parts backwards match (length in prefix==%zu)", matchLength);
+ matchLength += ZSTD_ldm_countBackwardsMatch(pIn - matchLength, pAnchor, pExtDictEnd, pExtDictStart);
+ DEBUGLOG(7, "final backwards match length = %zu", matchLength);
+ return matchLength;
+}
+
+/* ZSTD_ldm_fillFastTables() :
+ *
+ * Fills the relevant tables for the ZSTD_fast and ZSTD_dfast strategies.
+ * This is similar to ZSTD_loadDictionaryContent.
+ *
+ * The tables for the other strategies are filled within their
+ * block compressors. */
+static size_t ZSTD_ldm_fillFastTables(ZSTD_matchState_t* ms,
+ void const* end)
+{
+ const BYTE* const iend = (const BYTE*)end;
+
+ switch(ms->cParams.strategy)
+ {
+ case ZSTD_fast:
+ ZSTD_fillHashTable(ms, iend, ZSTD_dtlm_fast);
+ break;
+
+ case ZSTD_dfast:
+ ZSTD_fillDoubleHashTable(ms, iend, ZSTD_dtlm_fast);
+ break;
+
+ case ZSTD_greedy:
+ case ZSTD_lazy:
+ case ZSTD_lazy2:
+ case ZSTD_btlazy2:
+ case ZSTD_btopt:
+ case ZSTD_btultra:
+ case ZSTD_btultra2:
+ break;
+ default:
+ assert(0); /* not possible : not a valid strategy id */
+ }
+
+ return 0;
+}
+
+void ZSTD_ldm_fillHashTable(
+ ldmState_t* ldmState, const BYTE* ip,
+ const BYTE* iend, ldmParams_t const* params)
+{
+ U32 const minMatchLength = params->minMatchLength;
+ U32 const hBits = params->hashLog - params->bucketSizeLog;
+ BYTE const* const base = ldmState->window.base;
+ BYTE const* const istart = ip;
+ ldmRollingHashState_t hashState;
+ size_t* const splits = ldmState->splitIndices;
+ unsigned numSplits;
+
+ DEBUGLOG(5, "ZSTD_ldm_fillHashTable");
+
+ ZSTD_ldm_gear_init(&hashState, params);
+ while (ip < iend) {
+ size_t hashed;
+ unsigned n;
+
+ numSplits = 0;
+ hashed = ZSTD_ldm_gear_feed(&hashState, ip, iend - ip, splits, &numSplits);
+
+ for (n = 0; n < numSplits; n++) {
+ if (ip + splits[n] >= istart + minMatchLength) {
+ BYTE const* const split = ip + splits[n] - minMatchLength;
+ U64 const xxhash = xxh64(split, minMatchLength, 0);
+ U32 const hash = (U32)(xxhash & (((U32)1 << hBits) - 1));
+ ldmEntry_t entry;
+
+ entry.offset = (U32)(split - base);
+ entry.checksum = (U32)(xxhash >> 32);
+ ZSTD_ldm_insertEntry(ldmState, hash, entry, *params);
+ }
+ }
+
+ ip += hashed;
+ }
+}
+
+
+/* ZSTD_ldm_limitTableUpdate() :
+ *
+ * Sets cctx->nextToUpdate to a position corresponding closer to anchor
+ * if it is far way
+ * (after a long match, only update tables a limited amount). */
+static void ZSTD_ldm_limitTableUpdate(ZSTD_matchState_t* ms, const BYTE* anchor)
+{
+ U32 const curr = (U32)(anchor - ms->window.base);
+ if (curr > ms->nextToUpdate + 1024) {
+ ms->nextToUpdate =
+ curr - MIN(512, curr - ms->nextToUpdate - 1024);
+ }
+}
+
+static size_t ZSTD_ldm_generateSequences_internal(
+ ldmState_t* ldmState, rawSeqStore_t* rawSeqStore,
+ ldmParams_t const* params, void const* src, size_t srcSize)
+{
+ /* LDM parameters */
+ int const extDict = ZSTD_window_hasExtDict(ldmState->window);
+ U32 const minMatchLength = params->minMatchLength;
+ U32 const entsPerBucket = 1U << params->bucketSizeLog;
+ U32 const hBits = params->hashLog - params->bucketSizeLog;
+ /* Prefix and extDict parameters */
+ U32 const dictLimit = ldmState->window.dictLimit;
+ U32 const lowestIndex = extDict ? ldmState->window.lowLimit : dictLimit;
+ BYTE const* const base = ldmState->window.base;
+ BYTE const* const dictBase = extDict ? ldmState->window.dictBase : NULL;
+ BYTE const* const dictStart = extDict ? dictBase + lowestIndex : NULL;
+ BYTE const* const dictEnd = extDict ? dictBase + dictLimit : NULL;
+ BYTE const* const lowPrefixPtr = base + dictLimit;
+ /* Input bounds */
+ BYTE const* const istart = (BYTE const*)src;
+ BYTE const* const iend = istart + srcSize;
+ BYTE const* const ilimit = iend - HASH_READ_SIZE;
+ /* Input positions */
+ BYTE const* anchor = istart;
+ BYTE const* ip = istart;
+ /* Rolling hash state */
+ ldmRollingHashState_t hashState;
+ /* Arrays for staged-processing */
+ size_t* const splits = ldmState->splitIndices;
+ ldmMatchCandidate_t* const candidates = ldmState->matchCandidates;
+ unsigned numSplits;
+
+ if (srcSize < minMatchLength)
+ return iend - anchor;
+
+ /* Initialize the rolling hash state with the first minMatchLength bytes */
+ ZSTD_ldm_gear_init(&hashState, params);
+ ZSTD_ldm_gear_reset(&hashState, ip, minMatchLength);
+ ip += minMatchLength;
+
+ while (ip < ilimit) {
+ size_t hashed;
+ unsigned n;
+
+ numSplits = 0;
+ hashed = ZSTD_ldm_gear_feed(&hashState, ip, ilimit - ip,
+ splits, &numSplits);
+
+ for (n = 0; n < numSplits; n++) {
+ BYTE const* const split = ip + splits[n] - minMatchLength;
+ U64 const xxhash = xxh64(split, minMatchLength, 0);
+ U32 const hash = (U32)(xxhash & (((U32)1 << hBits) - 1));
+
+ candidates[n].split = split;
+ candidates[n].hash = hash;
+ candidates[n].checksum = (U32)(xxhash >> 32);
+ candidates[n].bucket = ZSTD_ldm_getBucket(ldmState, hash, *params);
+ PREFETCH_L1(candidates[n].bucket);
+ }
+
+ for (n = 0; n < numSplits; n++) {
+ size_t forwardMatchLength = 0, backwardMatchLength = 0,
+ bestMatchLength = 0, mLength;
+ U32 offset;
+ BYTE const* const split = candidates[n].split;
+ U32 const checksum = candidates[n].checksum;
+ U32 const hash = candidates[n].hash;
+ ldmEntry_t* const bucket = candidates[n].bucket;
+ ldmEntry_t const* cur;
+ ldmEntry_t const* bestEntry = NULL;
+ ldmEntry_t newEntry;
+
+ newEntry.offset = (U32)(split - base);
+ newEntry.checksum = checksum;
+
+ /* If a split point would generate a sequence overlapping with
+ * the previous one, we merely register it in the hash table and
+ * move on */
+ if (split < anchor) {
+ ZSTD_ldm_insertEntry(ldmState, hash, newEntry, *params);
+ continue;
+ }
+
+ for (cur = bucket; cur < bucket + entsPerBucket; cur++) {
+ size_t curForwardMatchLength, curBackwardMatchLength,
+ curTotalMatchLength;
+ if (cur->checksum != checksum || cur->offset <= lowestIndex) {
+ continue;
+ }
+ if (extDict) {
+ BYTE const* const curMatchBase =
+ cur->offset < dictLimit ? dictBase : base;
+ BYTE const* const pMatch = curMatchBase + cur->offset;
+ BYTE const* const matchEnd =
+ cur->offset < dictLimit ? dictEnd : iend;
+ BYTE const* const lowMatchPtr =
+ cur->offset < dictLimit ? dictStart : lowPrefixPtr;
+ curForwardMatchLength =
+ ZSTD_count_2segments(split, pMatch, iend, matchEnd, lowPrefixPtr);
+ if (curForwardMatchLength < minMatchLength) {
+ continue;
+ }
+ curBackwardMatchLength = ZSTD_ldm_countBackwardsMatch_2segments(
+ split, anchor, pMatch, lowMatchPtr, dictStart, dictEnd);
+ } else { /* !extDict */
+ BYTE const* const pMatch = base + cur->offset;
+ curForwardMatchLength = ZSTD_count(split, pMatch, iend);
+ if (curForwardMatchLength < minMatchLength) {
+ continue;
+ }
+ curBackwardMatchLength =
+ ZSTD_ldm_countBackwardsMatch(split, anchor, pMatch, lowPrefixPtr);
+ }
+ curTotalMatchLength = curForwardMatchLength + curBackwardMatchLength;
+
+ if (curTotalMatchLength > bestMatchLength) {
+ bestMatchLength = curTotalMatchLength;
+ forwardMatchLength = curForwardMatchLength;
+ backwardMatchLength = curBackwardMatchLength;
+ bestEntry = cur;
+ }
+ }
+
+ /* No match found -- insert an entry into the hash table
+ * and process the next candidate match */
+ if (bestEntry == NULL) {
+ ZSTD_ldm_insertEntry(ldmState, hash, newEntry, *params);
+ continue;
+ }
+
+ /* Match found */
+ offset = (U32)(split - base) - bestEntry->offset;
+ mLength = forwardMatchLength + backwardMatchLength;
+ {
+ rawSeq* const seq = rawSeqStore->seq + rawSeqStore->size;
+
+ /* Out of sequence storage */
+ if (rawSeqStore->size == rawSeqStore->capacity)
+ return ERROR(dstSize_tooSmall);
+ seq->litLength = (U32)(split - backwardMatchLength - anchor);
+ seq->matchLength = (U32)mLength;
+ seq->offset = offset;
+ rawSeqStore->size++;
+ }
+
+ /* Insert the current entry into the hash table --- it must be
+ * done after the previous block to avoid clobbering bestEntry */
+ ZSTD_ldm_insertEntry(ldmState, hash, newEntry, *params);
+
+ anchor = split + forwardMatchLength;
+
+ /* If we find a match that ends after the data that we've hashed
+ * then we have a repeating, overlapping, pattern. E.g. all zeros.
+ * If one repetition of the pattern matches our `stopMask` then all
+ * repetitions will. We don't need to insert them all into out table,
+ * only the first one. So skip over overlapping matches.
+ * This is a major speed boost (20x) for compressing a single byte
+ * repeated, when that byte ends up in the table.
+ */
+ if (anchor > ip + hashed) {
+ ZSTD_ldm_gear_reset(&hashState, anchor - minMatchLength, minMatchLength);
+ /* Continue the outer loop at anchor (ip + hashed == anchor). */
+ ip = anchor - hashed;
+ break;
+ }
+ }
+
+ ip += hashed;
+ }
+
+ return iend - anchor;
+}
+
+/*! ZSTD_ldm_reduceTable() :
+ * reduce table indexes by `reducerValue` */
+static void ZSTD_ldm_reduceTable(ldmEntry_t* const table, U32 const size,
+ U32 const reducerValue)
+{
+ U32 u;
+ for (u = 0; u < size; u++) {
+ if (table[u].offset < reducerValue) table[u].offset = 0;
+ else table[u].offset -= reducerValue;
+ }
+}
+
+size_t ZSTD_ldm_generateSequences(
+ ldmState_t* ldmState, rawSeqStore_t* sequences,
+ ldmParams_t const* params, void const* src, size_t srcSize)
+{
+ U32 const maxDist = 1U << params->windowLog;
+ BYTE const* const istart = (BYTE const*)src;
+ BYTE const* const iend = istart + srcSize;
+ size_t const kMaxChunkSize = 1 << 20;
+ size_t const nbChunks = (srcSize / kMaxChunkSize) + ((srcSize % kMaxChunkSize) != 0);
+ size_t chunk;
+ size_t leftoverSize = 0;
+
+ assert(ZSTD_CHUNKSIZE_MAX >= kMaxChunkSize);
+ /* Check that ZSTD_window_update() has been called for this chunk prior
+ * to passing it to this function.
+ */
+ assert(ldmState->window.nextSrc >= (BYTE const*)src + srcSize);
+ /* The input could be very large (in zstdmt), so it must be broken up into
+ * chunks to enforce the maximum distance and handle overflow correction.
+ */
+ assert(sequences->pos <= sequences->size);
+ assert(sequences->size <= sequences->capacity);
+ for (chunk = 0; chunk < nbChunks && sequences->size < sequences->capacity; ++chunk) {
+ BYTE const* const chunkStart = istart + chunk * kMaxChunkSize;
+ size_t const remaining = (size_t)(iend - chunkStart);
+ BYTE const *const chunkEnd =
+ (remaining < kMaxChunkSize) ? iend : chunkStart + kMaxChunkSize;
+ size_t const chunkSize = chunkEnd - chunkStart;
+ size_t newLeftoverSize;
+ size_t const prevSize = sequences->size;
+
+ assert(chunkStart < iend);
+ /* 1. Perform overflow correction if necessary. */
+ if (ZSTD_window_needOverflowCorrection(ldmState->window, 0, maxDist, ldmState->loadedDictEnd, chunkStart, chunkEnd)) {
+ U32 const ldmHSize = 1U << params->hashLog;
+ U32 const correction = ZSTD_window_correctOverflow(
+ &ldmState->window, /* cycleLog */ 0, maxDist, chunkStart);
+ ZSTD_ldm_reduceTable(ldmState->hashTable, ldmHSize, correction);
+ /* invalidate dictionaries on overflow correction */
+ ldmState->loadedDictEnd = 0;
+ }
+ /* 2. We enforce the maximum offset allowed.
+ *
+ * kMaxChunkSize should be small enough that we don't lose too much of
+ * the window through early invalidation.
+ * TODO: * Test the chunk size.
+ * * Try invalidation after the sequence generation and test the
+ * the offset against maxDist directly.
+ *
+ * NOTE: Because of dictionaries + sequence splitting we MUST make sure
+ * that any offset used is valid at the END of the sequence, since it may
+ * be split into two sequences. This condition holds when using
+ * ZSTD_window_enforceMaxDist(), but if we move to checking offsets
+ * against maxDist directly, we'll have to carefully handle that case.
+ */
+ ZSTD_window_enforceMaxDist(&ldmState->window, chunkEnd, maxDist, &ldmState->loadedDictEnd, NULL);
+ /* 3. Generate the sequences for the chunk, and get newLeftoverSize. */
+ newLeftoverSize = ZSTD_ldm_generateSequences_internal(
+ ldmState, sequences, params, chunkStart, chunkSize);
+ if (ZSTD_isError(newLeftoverSize))
+ return newLeftoverSize;
+ /* 4. We add the leftover literals from previous iterations to the first
+ * newly generated sequence, or add the `newLeftoverSize` if none are
+ * generated.
+ */
+ /* Prepend the leftover literals from the last call */
+ if (prevSize < sequences->size) {
+ sequences->seq[prevSize].litLength += (U32)leftoverSize;
+ leftoverSize = newLeftoverSize;
+ } else {
+ assert(newLeftoverSize == chunkSize);
+ leftoverSize += chunkSize;
+ }
+ }
+ return 0;
+}
+
+void
+ZSTD_ldm_skipSequences(rawSeqStore_t* rawSeqStore, size_t srcSize, U32 const minMatch)
+{
+ while (srcSize > 0 && rawSeqStore->pos < rawSeqStore->size) {
+ rawSeq* seq = rawSeqStore->seq + rawSeqStore->pos;
+ if (srcSize <= seq->litLength) {
+ /* Skip past srcSize literals */
+ seq->litLength -= (U32)srcSize;
+ return;
+ }
+ srcSize -= seq->litLength;
+ seq->litLength = 0;
+ if (srcSize < seq->matchLength) {
+ /* Skip past the first srcSize of the match */
+ seq->matchLength -= (U32)srcSize;
+ if (seq->matchLength < minMatch) {
+ /* The match is too short, omit it */
+ if (rawSeqStore->pos + 1 < rawSeqStore->size) {
+ seq[1].litLength += seq[0].matchLength;
+ }
+ rawSeqStore->pos++;
+ }
+ return;
+ }
+ srcSize -= seq->matchLength;
+ seq->matchLength = 0;
+ rawSeqStore->pos++;
+ }
+}
+
+/*
+ * If the sequence length is longer than remaining then the sequence is split
+ * between this block and the next.
+ *
+ * Returns the current sequence to handle, or if the rest of the block should
+ * be literals, it returns a sequence with offset == 0.
+ */
+static rawSeq maybeSplitSequence(rawSeqStore_t* rawSeqStore,
+ U32 const remaining, U32 const minMatch)
+{
+ rawSeq sequence = rawSeqStore->seq[rawSeqStore->pos];
+ assert(sequence.offset > 0);
+ /* Likely: No partial sequence */
+ if (remaining >= sequence.litLength + sequence.matchLength) {
+ rawSeqStore->pos++;
+ return sequence;
+ }
+ /* Cut the sequence short (offset == 0 ==> rest is literals). */
+ if (remaining <= sequence.litLength) {
+ sequence.offset = 0;
+ } else if (remaining < sequence.litLength + sequence.matchLength) {
+ sequence.matchLength = remaining - sequence.litLength;
+ if (sequence.matchLength < minMatch) {
+ sequence.offset = 0;
+ }
+ }
+ /* Skip past `remaining` bytes for the future sequences. */
+ ZSTD_ldm_skipSequences(rawSeqStore, remaining, minMatch);
+ return sequence;
+}
+
+void ZSTD_ldm_skipRawSeqStoreBytes(rawSeqStore_t* rawSeqStore, size_t nbBytes) {
+ U32 currPos = (U32)(rawSeqStore->posInSequence + nbBytes);
+ while (currPos && rawSeqStore->pos < rawSeqStore->size) {
+ rawSeq currSeq = rawSeqStore->seq[rawSeqStore->pos];
+ if (currPos >= currSeq.litLength + currSeq.matchLength) {
+ currPos -= currSeq.litLength + currSeq.matchLength;
+ rawSeqStore->pos++;
+ } else {
+ rawSeqStore->posInSequence = currPos;
+ break;
+ }
+ }
+ if (currPos == 0 || rawSeqStore->pos == rawSeqStore->size) {
+ rawSeqStore->posInSequence = 0;
+ }
+}
+
+size_t ZSTD_ldm_blockCompress(rawSeqStore_t* rawSeqStore,
+ ZSTD_matchState_t* ms, seqStore_t* seqStore, U32 rep[ZSTD_REP_NUM],
+ ZSTD_paramSwitch_e useRowMatchFinder,
+ void const* src, size_t srcSize)
+{
+ const ZSTD_compressionParameters* const cParams = &ms->cParams;
+ unsigned const minMatch = cParams->minMatch;
+ ZSTD_blockCompressor const blockCompressor =
+ ZSTD_selectBlockCompressor(cParams->strategy, useRowMatchFinder, ZSTD_matchState_dictMode(ms));
+ /* Input bounds */
+ BYTE const* const istart = (BYTE const*)src;
+ BYTE const* const iend = istart + srcSize;
+ /* Input positions */
+ BYTE const* ip = istart;
+
+ DEBUGLOG(5, "ZSTD_ldm_blockCompress: srcSize=%zu", srcSize);
+ /* If using opt parser, use LDMs only as candidates rather than always accepting them */
+ if (cParams->strategy >= ZSTD_btopt) {
+ size_t lastLLSize;
+ ms->ldmSeqStore = rawSeqStore;
+ lastLLSize = blockCompressor(ms, seqStore, rep, src, srcSize);
+ ZSTD_ldm_skipRawSeqStoreBytes(rawSeqStore, srcSize);
+ return lastLLSize;
+ }
+
+ assert(rawSeqStore->pos <= rawSeqStore->size);
+ assert(rawSeqStore->size <= rawSeqStore->capacity);
+ /* Loop through each sequence and apply the block compressor to the literals */
+ while (rawSeqStore->pos < rawSeqStore->size && ip < iend) {
+ /* maybeSplitSequence updates rawSeqStore->pos */
+ rawSeq const sequence = maybeSplitSequence(rawSeqStore,
+ (U32)(iend - ip), minMatch);
+ int i;
+ /* End signal */
+ if (sequence.offset == 0)
+ break;
+
+ assert(ip + sequence.litLength + sequence.matchLength <= iend);
+
+ /* Fill tables for block compressor */
+ ZSTD_ldm_limitTableUpdate(ms, ip);
+ ZSTD_ldm_fillFastTables(ms, ip);
+ /* Run the block compressor */
+ DEBUGLOG(5, "pos %u : calling block compressor on segment of size %u", (unsigned)(ip-istart), sequence.litLength);
+ {
+ size_t const newLitLength =
+ blockCompressor(ms, seqStore, rep, ip, sequence.litLength);
+ ip += sequence.litLength;
+ /* Update the repcodes */
+ for (i = ZSTD_REP_NUM - 1; i > 0; i--)
+ rep[i] = rep[i-1];
+ rep[0] = sequence.offset;
+ /* Store the sequence */
+ ZSTD_storeSeq(seqStore, newLitLength, ip - newLitLength, iend,
+ STORE_OFFSET(sequence.offset),
+ sequence.matchLength);
+ ip += sequence.matchLength;
+ }
+ }
+ /* Fill the tables for the block compressor */
+ ZSTD_ldm_limitTableUpdate(ms, ip);
+ ZSTD_ldm_fillFastTables(ms, ip);
+ /* Compress the last literals */
+ return blockCompressor(ms, seqStore, rep, ip, iend - ip);
+}
diff --git a/lib/zstd/compress/zstd_ldm.h b/lib/zstd/compress/zstd_ldm.h
new file mode 100644
index 0000000000..fbc6a5e88f
--- /dev/null
+++ b/lib/zstd/compress/zstd_ldm.h
@@ -0,0 +1,111 @@
+/*
+ * Copyright (c) Yann Collet, Facebook, Inc.
+ * All rights reserved.
+ *
+ * This source code is licensed under both the BSD-style license (found in the
+ * LICENSE file in the root directory of this source tree) and the GPLv2 (found
+ * in the COPYING file in the root directory of this source tree).
+ * You may select, at your option, one of the above-listed licenses.
+ */
+
+#ifndef ZSTD_LDM_H
+#define ZSTD_LDM_H
+
+
+#include "zstd_compress_internal.h" /* ldmParams_t, U32 */
+#include <linux/zstd.h> /* ZSTD_CCtx, size_t */
+
+/*-*************************************
+* Long distance matching
+***************************************/
+
+#define ZSTD_LDM_DEFAULT_WINDOW_LOG ZSTD_WINDOWLOG_LIMIT_DEFAULT
+
+void ZSTD_ldm_fillHashTable(
+ ldmState_t* state, const BYTE* ip,
+ const BYTE* iend, ldmParams_t const* params);
+
+/*
+ * ZSTD_ldm_generateSequences():
+ *
+ * Generates the sequences using the long distance match finder.
+ * Generates long range matching sequences in `sequences`, which parse a prefix
+ * of the source. `sequences` must be large enough to store every sequence,
+ * which can be checked with `ZSTD_ldm_getMaxNbSeq()`.
+ * @returns 0 or an error code.
+ *
+ * NOTE: The user must have called ZSTD_window_update() for all of the input
+ * they have, even if they pass it to ZSTD_ldm_generateSequences() in chunks.
+ * NOTE: This function returns an error if it runs out of space to store
+ * sequences.
+ */
+size_t ZSTD_ldm_generateSequences(
+ ldmState_t* ldms, rawSeqStore_t* sequences,
+ ldmParams_t const* params, void const* src, size_t srcSize);
+
+/*
+ * ZSTD_ldm_blockCompress():
+ *
+ * Compresses a block using the predefined sequences, along with a secondary
+ * block compressor. The literals section of every sequence is passed to the
+ * secondary block compressor, and those sequences are interspersed with the
+ * predefined sequences. Returns the length of the last literals.
+ * Updates `rawSeqStore.pos` to indicate how many sequences have been consumed.
+ * `rawSeqStore.seq` may also be updated to split the last sequence between two
+ * blocks.
+ * @return The length of the last literals.
+ *
+ * NOTE: The source must be at most the maximum block size, but the predefined
+ * sequences can be any size, and may be longer than the block. In the case that
+ * they are longer than the block, the last sequences may need to be split into
+ * two. We handle that case correctly, and update `rawSeqStore` appropriately.
+ * NOTE: This function does not return any errors.
+ */
+size_t ZSTD_ldm_blockCompress(rawSeqStore_t* rawSeqStore,
+ ZSTD_matchState_t* ms, seqStore_t* seqStore, U32 rep[ZSTD_REP_NUM],
+ ZSTD_paramSwitch_e useRowMatchFinder,
+ void const* src, size_t srcSize);
+
+/*
+ * ZSTD_ldm_skipSequences():
+ *
+ * Skip past `srcSize` bytes worth of sequences in `rawSeqStore`.
+ * Avoids emitting matches less than `minMatch` bytes.
+ * Must be called for data that is not passed to ZSTD_ldm_blockCompress().
+ */
+void ZSTD_ldm_skipSequences(rawSeqStore_t* rawSeqStore, size_t srcSize,
+ U32 const minMatch);
+
+/* ZSTD_ldm_skipRawSeqStoreBytes():
+ * Moves forward in rawSeqStore by nbBytes, updating fields 'pos' and 'posInSequence'.
+ * Not to be used in conjunction with ZSTD_ldm_skipSequences().
+ * Must be called for data with is not passed to ZSTD_ldm_blockCompress().
+ */
+void ZSTD_ldm_skipRawSeqStoreBytes(rawSeqStore_t* rawSeqStore, size_t nbBytes);
+
+/* ZSTD_ldm_getTableSize() :
+ * Estimate the space needed for long distance matching tables or 0 if LDM is
+ * disabled.
+ */
+size_t ZSTD_ldm_getTableSize(ldmParams_t params);
+
+/* ZSTD_ldm_getSeqSpace() :
+ * Return an upper bound on the number of sequences that can be produced by
+ * the long distance matcher, or 0 if LDM is disabled.
+ */
+size_t ZSTD_ldm_getMaxNbSeq(ldmParams_t params, size_t maxChunkSize);
+
+/* ZSTD_ldm_adjustParameters() :
+ * If the params->hashRateLog is not set, set it to its default value based on
+ * windowLog and params->hashLog.
+ *
+ * Ensures that params->bucketSizeLog is <= params->hashLog (setting it to
+ * params->hashLog if it is not).
+ *
+ * Ensures that the minMatchLength >= targetLength during optimal parsing.
+ */
+void ZSTD_ldm_adjustParameters(ldmParams_t* params,
+ ZSTD_compressionParameters const* cParams);
+
+
+#endif /* ZSTD_FAST_H */
diff --git a/lib/zstd/compress/zstd_ldm_geartab.h b/lib/zstd/compress/zstd_ldm_geartab.h
new file mode 100644
index 0000000000..647f865be2
--- /dev/null
+++ b/lib/zstd/compress/zstd_ldm_geartab.h
@@ -0,0 +1,106 @@
+/*
+ * Copyright (c) Yann Collet, Facebook, Inc.
+ * All rights reserved.
+ *
+ * This source code is licensed under both the BSD-style license (found in the
+ * LICENSE file in the root directory of this source tree) and the GPLv2 (found
+ * in the COPYING file in the root directory of this source tree).
+ * You may select, at your option, one of the above-listed licenses.
+ */
+
+#ifndef ZSTD_LDM_GEARTAB_H
+#define ZSTD_LDM_GEARTAB_H
+
+#include "../common/compiler.h" /* UNUSED_ATTR */
+#include "../common/mem.h" /* U64 */
+
+static UNUSED_ATTR const U64 ZSTD_ldm_gearTab[256] = {
+ 0xf5b8f72c5f77775c, 0x84935f266b7ac412, 0xb647ada9ca730ccc,
+ 0xb065bb4b114fb1de, 0x34584e7e8c3a9fd0, 0x4e97e17c6ae26b05,
+ 0x3a03d743bc99a604, 0xcecd042422c4044f, 0x76de76c58524259e,
+ 0x9c8528f65badeaca, 0x86563706e2097529, 0x2902475fa375d889,
+ 0xafb32a9739a5ebe6, 0xce2714da3883e639, 0x21eaf821722e69e,
+ 0x37b628620b628, 0x49a8d455d88caf5, 0x8556d711e6958140,
+ 0x4f7ae74fc605c1f, 0x829f0c3468bd3a20, 0x4ffdc885c625179e,
+ 0x8473de048a3daf1b, 0x51008822b05646b2, 0x69d75d12b2d1cc5f,
+ 0x8c9d4a19159154bc, 0xc3cc10f4abbd4003, 0xd06ddc1cecb97391,
+ 0xbe48e6e7ed80302e, 0x3481db31cee03547, 0xacc3f67cdaa1d210,
+ 0x65cb771d8c7f96cc, 0x8eb27177055723dd, 0xc789950d44cd94be,
+ 0x934feadc3700b12b, 0x5e485f11edbdf182, 0x1e2e2a46fd64767a,
+ 0x2969ca71d82efa7c, 0x9d46e9935ebbba2e, 0xe056b67e05e6822b,
+ 0x94d73f55739d03a0, 0xcd7010bdb69b5a03, 0x455ef9fcd79b82f4,
+ 0x869cb54a8749c161, 0x38d1a4fa6185d225, 0xb475166f94bbe9bb,
+ 0xa4143548720959f1, 0x7aed4780ba6b26ba, 0xd0ce264439e02312,
+ 0x84366d746078d508, 0xa8ce973c72ed17be, 0x21c323a29a430b01,
+ 0x9962d617e3af80ee, 0xab0ce91d9c8cf75b, 0x530e8ee6d19a4dbc,
+ 0x2ef68c0cf53f5d72, 0xc03a681640a85506, 0x496e4e9f9c310967,
+ 0x78580472b59b14a0, 0x273824c23b388577, 0x66bf923ad45cb553,
+ 0x47ae1a5a2492ba86, 0x35e304569e229659, 0x4765182a46870b6f,
+ 0x6cbab625e9099412, 0xddac9a2e598522c1, 0x7172086e666624f2,
+ 0xdf5003ca503b7837, 0x88c0c1db78563d09, 0x58d51865acfc289d,
+ 0x177671aec65224f1, 0xfb79d8a241e967d7, 0x2be1e101cad9a49a,
+ 0x6625682f6e29186b, 0x399553457ac06e50, 0x35dffb4c23abb74,
+ 0x429db2591f54aade, 0xc52802a8037d1009, 0x6acb27381f0b25f3,
+ 0xf45e2551ee4f823b, 0x8b0ea2d99580c2f7, 0x3bed519cbcb4e1e1,
+ 0xff452823dbb010a, 0x9d42ed614f3dd267, 0x5b9313c06257c57b,
+ 0xa114b8008b5e1442, 0xc1fe311c11c13d4b, 0x66e8763ea34c5568,
+ 0x8b982af1c262f05d, 0xee8876faaa75fbb7, 0x8a62a4d0d172bb2a,
+ 0xc13d94a3b7449a97, 0x6dbbba9dc15d037c, 0xc786101f1d92e0f1,
+ 0xd78681a907a0b79b, 0xf61aaf2962c9abb9, 0x2cfd16fcd3cb7ad9,
+ 0x868c5b6744624d21, 0x25e650899c74ddd7, 0xba042af4a7c37463,
+ 0x4eb1a539465a3eca, 0xbe09dbf03b05d5ca, 0x774e5a362b5472ba,
+ 0x47a1221229d183cd, 0x504b0ca18ef5a2df, 0xdffbdfbde2456eb9,
+ 0x46cd2b2fbee34634, 0xf2aef8fe819d98c3, 0x357f5276d4599d61,
+ 0x24a5483879c453e3, 0x88026889192b4b9, 0x28da96671782dbec,
+ 0x4ef37c40588e9aaa, 0x8837b90651bc9fb3, 0xc164f741d3f0e5d6,
+ 0xbc135a0a704b70ba, 0x69cd868f7622ada, 0xbc37ba89e0b9c0ab,
+ 0x47c14a01323552f6, 0x4f00794bacee98bb, 0x7107de7d637a69d5,
+ 0x88af793bb6f2255e, 0xf3c6466b8799b598, 0xc288c616aa7f3b59,
+ 0x81ca63cf42fca3fd, 0x88d85ace36a2674b, 0xd056bd3792389e7,
+ 0xe55c396c4e9dd32d, 0xbefb504571e6c0a6, 0x96ab32115e91e8cc,
+ 0xbf8acb18de8f38d1, 0x66dae58801672606, 0x833b6017872317fb,
+ 0xb87c16f2d1c92864, 0xdb766a74e58b669c, 0x89659f85c61417be,
+ 0xc8daad856011ea0c, 0x76a4b565b6fe7eae, 0xa469d085f6237312,
+ 0xaaf0365683a3e96c, 0x4dbb746f8424f7b8, 0x638755af4e4acc1,
+ 0x3d7807f5bde64486, 0x17be6d8f5bbb7639, 0x903f0cd44dc35dc,
+ 0x67b672eafdf1196c, 0xa676ff93ed4c82f1, 0x521d1004c5053d9d,
+ 0x37ba9ad09ccc9202, 0x84e54d297aacfb51, 0xa0b4b776a143445,
+ 0x820d471e20b348e, 0x1874383cb83d46dc, 0x97edeec7a1efe11c,
+ 0xb330e50b1bdc42aa, 0x1dd91955ce70e032, 0xa514cdb88f2939d5,
+ 0x2791233fd90db9d3, 0x7b670a4cc50f7a9b, 0x77c07d2a05c6dfa5,
+ 0xe3778b6646d0a6fa, 0xb39c8eda47b56749, 0x933ed448addbef28,
+ 0xaf846af6ab7d0bf4, 0xe5af208eb666e49, 0x5e6622f73534cd6a,
+ 0x297daeca42ef5b6e, 0x862daef3d35539a6, 0xe68722498f8e1ea9,
+ 0x981c53093dc0d572, 0xfa09b0bfbf86fbf5, 0x30b1e96166219f15,
+ 0x70e7d466bdc4fb83, 0x5a66736e35f2a8e9, 0xcddb59d2b7c1baef,
+ 0xd6c7d247d26d8996, 0xea4e39eac8de1ba3, 0x539c8bb19fa3aff2,
+ 0x9f90e4c5fd508d8, 0xa34e5956fbaf3385, 0x2e2f8e151d3ef375,
+ 0x173691e9b83faec1, 0xb85a8d56bf016379, 0x8382381267408ae3,
+ 0xb90f901bbdc0096d, 0x7c6ad32933bcec65, 0x76bb5e2f2c8ad595,
+ 0x390f851a6cf46d28, 0xc3e6064da1c2da72, 0xc52a0c101cfa5389,
+ 0xd78eaf84a3fbc530, 0x3781b9e2288b997e, 0x73c2f6dea83d05c4,
+ 0x4228e364c5b5ed7, 0x9d7a3edf0da43911, 0x8edcfeda24686756,
+ 0x5e7667a7b7a9b3a1, 0x4c4f389fa143791d, 0xb08bc1023da7cddc,
+ 0x7ab4be3ae529b1cc, 0x754e6132dbe74ff9, 0x71635442a839df45,
+ 0x2f6fb1643fbe52de, 0x961e0a42cf7a8177, 0xf3b45d83d89ef2ea,
+ 0xee3de4cf4a6e3e9b, 0xcd6848542c3295e7, 0xe4cee1664c78662f,
+ 0x9947548b474c68c4, 0x25d73777a5ed8b0b, 0xc915b1d636b7fc,
+ 0x21c2ba75d9b0d2da, 0x5f6b5dcf608a64a1, 0xdcf333255ff9570c,
+ 0x633b922418ced4ee, 0xc136dde0b004b34a, 0x58cc83b05d4b2f5a,
+ 0x5eb424dda28e42d2, 0x62df47369739cd98, 0xb4e0b42485e4ce17,
+ 0x16e1f0c1f9a8d1e7, 0x8ec3916707560ebf, 0x62ba6e2df2cc9db3,
+ 0xcbf9f4ff77d83a16, 0x78d9d7d07d2bbcc4, 0xef554ce1e02c41f4,
+ 0x8d7581127eccf94d, 0xa9b53336cb3c8a05, 0x38c42c0bf45c4f91,
+ 0x640893cdf4488863, 0x80ec34bc575ea568, 0x39f324f5b48eaa40,
+ 0xe9d9ed1f8eff527f, 0x9224fc058cc5a214, 0xbaba00b04cfe7741,
+ 0x309a9f120fcf52af, 0xa558f3ec65626212, 0x424bec8b7adabe2f,
+ 0x41622513a6aea433, 0xb88da2d5324ca798, 0xd287733b245528a4,
+ 0x9a44697e6d68aec3, 0x7b1093be2f49bb28, 0x50bbec632e3d8aad,
+ 0x6cd90723e1ea8283, 0x897b9e7431b02bf3, 0x219efdcb338a7047,
+ 0x3b0311f0a27c0656, 0xdb17bf91c0db96e7, 0x8cd4fd6b4e85a5b2,
+ 0xfab071054ba6409d, 0x40d6fe831fa9dfd9, 0xaf358debad7d791e,
+ 0xeb8d0e25a65e3e58, 0xbbcbd3df14e08580, 0xcf751f27ecdab2b,
+ 0x2b4da14f2613d8f4
+};
+
+#endif /* ZSTD_LDM_GEARTAB_H */
diff --git a/lib/zstd/compress/zstd_opt.c b/lib/zstd/compress/zstd_opt.c
new file mode 100644
index 0000000000..fd82acfda6
--- /dev/null
+++ b/lib/zstd/compress/zstd_opt.c
@@ -0,0 +1,1446 @@
+/*
+ * Copyright (c) Przemyslaw Skibinski, Yann Collet, Facebook, Inc.
+ * All rights reserved.
+ *
+ * This source code is licensed under both the BSD-style license (found in the
+ * LICENSE file in the root directory of this source tree) and the GPLv2 (found
+ * in the COPYING file in the root directory of this source tree).
+ * You may select, at your option, one of the above-listed licenses.
+ */
+
+#include "zstd_compress_internal.h"
+#include "hist.h"
+#include "zstd_opt.h"
+
+
+#define ZSTD_LITFREQ_ADD 2 /* scaling factor for litFreq, so that frequencies adapt faster to new stats */
+#define ZSTD_MAX_PRICE (1<<30)
+
+#define ZSTD_PREDEF_THRESHOLD 1024 /* if srcSize < ZSTD_PREDEF_THRESHOLD, symbols' cost is assumed static, directly determined by pre-defined distributions */
+
+
+/*-*************************************
+* Price functions for optimal parser
+***************************************/
+
+#if 0 /* approximation at bit level (for tests) */
+# define BITCOST_ACCURACY 0
+# define BITCOST_MULTIPLIER (1 << BITCOST_ACCURACY)
+# define WEIGHT(stat, opt) ((void)opt, ZSTD_bitWeight(stat))
+#elif 0 /* fractional bit accuracy (for tests) */
+# define BITCOST_ACCURACY 8
+# define BITCOST_MULTIPLIER (1 << BITCOST_ACCURACY)
+# define WEIGHT(stat,opt) ((void)opt, ZSTD_fracWeight(stat))
+#else /* opt==approx, ultra==accurate */
+# define BITCOST_ACCURACY 8
+# define BITCOST_MULTIPLIER (1 << BITCOST_ACCURACY)
+# define WEIGHT(stat,opt) (opt ? ZSTD_fracWeight(stat) : ZSTD_bitWeight(stat))
+#endif
+
+MEM_STATIC U32 ZSTD_bitWeight(U32 stat)
+{
+ return (ZSTD_highbit32(stat+1) * BITCOST_MULTIPLIER);
+}
+
+MEM_STATIC U32 ZSTD_fracWeight(U32 rawStat)
+{
+ U32 const stat = rawStat + 1;
+ U32 const hb = ZSTD_highbit32(stat);
+ U32 const BWeight = hb * BITCOST_MULTIPLIER;
+ U32 const FWeight = (stat << BITCOST_ACCURACY) >> hb;
+ U32 const weight = BWeight + FWeight;
+ assert(hb + BITCOST_ACCURACY < 31);
+ return weight;
+}
+
+#if (DEBUGLEVEL>=2)
+/* debugging function,
+ * @return price in bytes as fractional value
+ * for debug messages only */
+MEM_STATIC double ZSTD_fCost(U32 price)
+{
+ return (double)price / (BITCOST_MULTIPLIER*8);
+}
+#endif
+
+static int ZSTD_compressedLiterals(optState_t const* const optPtr)
+{
+ return optPtr->literalCompressionMode != ZSTD_ps_disable;
+}
+
+static void ZSTD_setBasePrices(optState_t* optPtr, int optLevel)
+{
+ if (ZSTD_compressedLiterals(optPtr))
+ optPtr->litSumBasePrice = WEIGHT(optPtr->litSum, optLevel);
+ optPtr->litLengthSumBasePrice = WEIGHT(optPtr->litLengthSum, optLevel);
+ optPtr->matchLengthSumBasePrice = WEIGHT(optPtr->matchLengthSum, optLevel);
+ optPtr->offCodeSumBasePrice = WEIGHT(optPtr->offCodeSum, optLevel);
+}
+
+
+static U32 sum_u32(const unsigned table[], size_t nbElts)
+{
+ size_t n;
+ U32 total = 0;
+ for (n=0; n<nbElts; n++) {
+ total += table[n];
+ }
+ return total;
+}
+
+static U32 ZSTD_downscaleStats(unsigned* table, U32 lastEltIndex, U32 shift)
+{
+ U32 s, sum=0;
+ DEBUGLOG(5, "ZSTD_downscaleStats (nbElts=%u, shift=%u)", (unsigned)lastEltIndex+1, (unsigned)shift);
+ assert(shift < 30);
+ for (s=0; s<lastEltIndex+1; s++) {
+ table[s] = 1 + (table[s] >> shift);
+ sum += table[s];
+ }
+ return sum;
+}
+
+/* ZSTD_scaleStats() :
+ * reduce all elements in table is sum too large
+ * return the resulting sum of elements */
+static U32 ZSTD_scaleStats(unsigned* table, U32 lastEltIndex, U32 logTarget)
+{
+ U32 const prevsum = sum_u32(table, lastEltIndex+1);
+ U32 const factor = prevsum >> logTarget;
+ DEBUGLOG(5, "ZSTD_scaleStats (nbElts=%u, target=%u)", (unsigned)lastEltIndex+1, (unsigned)logTarget);
+ assert(logTarget < 30);
+ if (factor <= 1) return prevsum;
+ return ZSTD_downscaleStats(table, lastEltIndex, ZSTD_highbit32(factor));
+}
+
+/* ZSTD_rescaleFreqs() :
+ * if first block (detected by optPtr->litLengthSum == 0) : init statistics
+ * take hints from dictionary if there is one
+ * and init from zero if there is none,
+ * using src for literals stats, and baseline stats for sequence symbols
+ * otherwise downscale existing stats, to be used as seed for next block.
+ */
+static void
+ZSTD_rescaleFreqs(optState_t* const optPtr,
+ const BYTE* const src, size_t const srcSize,
+ int const optLevel)
+{
+ int const compressedLiterals = ZSTD_compressedLiterals(optPtr);
+ DEBUGLOG(5, "ZSTD_rescaleFreqs (srcSize=%u)", (unsigned)srcSize);
+ optPtr->priceType = zop_dynamic;
+
+ if (optPtr->litLengthSum == 0) { /* first block : init */
+ if (srcSize <= ZSTD_PREDEF_THRESHOLD) { /* heuristic */
+ DEBUGLOG(5, "(srcSize <= ZSTD_PREDEF_THRESHOLD) => zop_predef");
+ optPtr->priceType = zop_predef;
+ }
+
+ assert(optPtr->symbolCosts != NULL);
+ if (optPtr->symbolCosts->huf.repeatMode == HUF_repeat_valid) {
+ /* huffman table presumed generated by dictionary */
+ optPtr->priceType = zop_dynamic;
+
+ if (compressedLiterals) {
+ unsigned lit;
+ assert(optPtr->litFreq != NULL);
+ optPtr->litSum = 0;
+ for (lit=0; lit<=MaxLit; lit++) {
+ U32 const scaleLog = 11; /* scale to 2K */
+ U32 const bitCost = HUF_getNbBitsFromCTable(optPtr->symbolCosts->huf.CTable, lit);
+ assert(bitCost <= scaleLog);
+ optPtr->litFreq[lit] = bitCost ? 1 << (scaleLog-bitCost) : 1 /*minimum to calculate cost*/;
+ optPtr->litSum += optPtr->litFreq[lit];
+ } }
+
+ { unsigned ll;
+ FSE_CState_t llstate;
+ FSE_initCState(&llstate, optPtr->symbolCosts->fse.litlengthCTable);
+ optPtr->litLengthSum = 0;
+ for (ll=0; ll<=MaxLL; ll++) {
+ U32 const scaleLog = 10; /* scale to 1K */
+ U32 const bitCost = FSE_getMaxNbBits(llstate.symbolTT, ll);
+ assert(bitCost < scaleLog);
+ optPtr->litLengthFreq[ll] = bitCost ? 1 << (scaleLog-bitCost) : 1 /*minimum to calculate cost*/;
+ optPtr->litLengthSum += optPtr->litLengthFreq[ll];
+ } }
+
+ { unsigned ml;
+ FSE_CState_t mlstate;
+ FSE_initCState(&mlstate, optPtr->symbolCosts->fse.matchlengthCTable);
+ optPtr->matchLengthSum = 0;
+ for (ml=0; ml<=MaxML; ml++) {
+ U32 const scaleLog = 10;
+ U32 const bitCost = FSE_getMaxNbBits(mlstate.symbolTT, ml);
+ assert(bitCost < scaleLog);
+ optPtr->matchLengthFreq[ml] = bitCost ? 1 << (scaleLog-bitCost) : 1 /*minimum to calculate cost*/;
+ optPtr->matchLengthSum += optPtr->matchLengthFreq[ml];
+ } }
+
+ { unsigned of;
+ FSE_CState_t ofstate;
+ FSE_initCState(&ofstate, optPtr->symbolCosts->fse.offcodeCTable);
+ optPtr->offCodeSum = 0;
+ for (of=0; of<=MaxOff; of++) {
+ U32 const scaleLog = 10;
+ U32 const bitCost = FSE_getMaxNbBits(ofstate.symbolTT, of);
+ assert(bitCost < scaleLog);
+ optPtr->offCodeFreq[of] = bitCost ? 1 << (scaleLog-bitCost) : 1 /*minimum to calculate cost*/;
+ optPtr->offCodeSum += optPtr->offCodeFreq[of];
+ } }
+
+ } else { /* not a dictionary */
+
+ assert(optPtr->litFreq != NULL);
+ if (compressedLiterals) {
+ unsigned lit = MaxLit;
+ HIST_count_simple(optPtr->litFreq, &lit, src, srcSize); /* use raw first block to init statistics */
+ optPtr->litSum = ZSTD_downscaleStats(optPtr->litFreq, MaxLit, 8);
+ }
+
+ { unsigned const baseLLfreqs[MaxLL+1] = {
+ 4, 2, 1, 1, 1, 1, 1, 1,
+ 1, 1, 1, 1, 1, 1, 1, 1,
+ 1, 1, 1, 1, 1, 1, 1, 1,
+ 1, 1, 1, 1, 1, 1, 1, 1,
+ 1, 1, 1, 1
+ };
+ ZSTD_memcpy(optPtr->litLengthFreq, baseLLfreqs, sizeof(baseLLfreqs));
+ optPtr->litLengthSum = sum_u32(baseLLfreqs, MaxLL+1);
+ }
+
+ { unsigned ml;
+ for (ml=0; ml<=MaxML; ml++)
+ optPtr->matchLengthFreq[ml] = 1;
+ }
+ optPtr->matchLengthSum = MaxML+1;
+
+ { unsigned const baseOFCfreqs[MaxOff+1] = {
+ 6, 2, 1, 1, 2, 3, 4, 4,
+ 4, 3, 2, 1, 1, 1, 1, 1,
+ 1, 1, 1, 1, 1, 1, 1, 1,
+ 1, 1, 1, 1, 1, 1, 1, 1
+ };
+ ZSTD_memcpy(optPtr->offCodeFreq, baseOFCfreqs, sizeof(baseOFCfreqs));
+ optPtr->offCodeSum = sum_u32(baseOFCfreqs, MaxOff+1);
+ }
+
+
+ }
+
+ } else { /* new block : re-use previous statistics, scaled down */
+
+ if (compressedLiterals)
+ optPtr->litSum = ZSTD_scaleStats(optPtr->litFreq, MaxLit, 12);
+ optPtr->litLengthSum = ZSTD_scaleStats(optPtr->litLengthFreq, MaxLL, 11);
+ optPtr->matchLengthSum = ZSTD_scaleStats(optPtr->matchLengthFreq, MaxML, 11);
+ optPtr->offCodeSum = ZSTD_scaleStats(optPtr->offCodeFreq, MaxOff, 11);
+ }
+
+ ZSTD_setBasePrices(optPtr, optLevel);
+}
+
+/* ZSTD_rawLiteralsCost() :
+ * price of literals (only) in specified segment (which length can be 0).
+ * does not include price of literalLength symbol */
+static U32 ZSTD_rawLiteralsCost(const BYTE* const literals, U32 const litLength,
+ const optState_t* const optPtr,
+ int optLevel)
+{
+ if (litLength == 0) return 0;
+
+ if (!ZSTD_compressedLiterals(optPtr))
+ return (litLength << 3) * BITCOST_MULTIPLIER; /* Uncompressed - 8 bytes per literal. */
+
+ if (optPtr->priceType == zop_predef)
+ return (litLength*6) * BITCOST_MULTIPLIER; /* 6 bit per literal - no statistic used */
+
+ /* dynamic statistics */
+ { U32 price = litLength * optPtr->litSumBasePrice;
+ U32 u;
+ for (u=0; u < litLength; u++) {
+ assert(WEIGHT(optPtr->litFreq[literals[u]], optLevel) <= optPtr->litSumBasePrice); /* literal cost should never be negative */
+ price -= WEIGHT(optPtr->litFreq[literals[u]], optLevel);
+ }
+ return price;
+ }
+}
+
+/* ZSTD_litLengthPrice() :
+ * cost of literalLength symbol */
+static U32 ZSTD_litLengthPrice(U32 const litLength, const optState_t* const optPtr, int optLevel)
+{
+ assert(litLength <= ZSTD_BLOCKSIZE_MAX);
+ if (optPtr->priceType == zop_predef)
+ return WEIGHT(litLength, optLevel);
+ /* We can't compute the litLength price for sizes >= ZSTD_BLOCKSIZE_MAX
+ * because it isn't representable in the zstd format. So instead just
+ * call it 1 bit more than ZSTD_BLOCKSIZE_MAX - 1. In this case the block
+ * would be all literals.
+ */
+ if (litLength == ZSTD_BLOCKSIZE_MAX)
+ return BITCOST_MULTIPLIER + ZSTD_litLengthPrice(ZSTD_BLOCKSIZE_MAX - 1, optPtr, optLevel);
+
+ /* dynamic statistics */
+ { U32 const llCode = ZSTD_LLcode(litLength);
+ return (LL_bits[llCode] * BITCOST_MULTIPLIER)
+ + optPtr->litLengthSumBasePrice
+ - WEIGHT(optPtr->litLengthFreq[llCode], optLevel);
+ }
+}
+
+/* ZSTD_getMatchPrice() :
+ * Provides the cost of the match part (offset + matchLength) of a sequence
+ * Must be combined with ZSTD_fullLiteralsCost() to get the full cost of a sequence.
+ * @offcode : expects a scale where 0,1,2 are repcodes 1-3, and 3+ are real_offsets+2
+ * @optLevel: when <2, favors small offset for decompression speed (improved cache efficiency)
+ */
+FORCE_INLINE_TEMPLATE U32
+ZSTD_getMatchPrice(U32 const offcode,
+ U32 const matchLength,
+ const optState_t* const optPtr,
+ int const optLevel)
+{
+ U32 price;
+ U32 const offCode = ZSTD_highbit32(STORED_TO_OFFBASE(offcode));
+ U32 const mlBase = matchLength - MINMATCH;
+ assert(matchLength >= MINMATCH);
+
+ if (optPtr->priceType == zop_predef) /* fixed scheme, do not use statistics */
+ return WEIGHT(mlBase, optLevel) + ((16 + offCode) * BITCOST_MULTIPLIER);
+
+ /* dynamic statistics */
+ price = (offCode * BITCOST_MULTIPLIER) + (optPtr->offCodeSumBasePrice - WEIGHT(optPtr->offCodeFreq[offCode], optLevel));
+ if ((optLevel<2) /*static*/ && offCode >= 20)
+ price += (offCode-19)*2 * BITCOST_MULTIPLIER; /* handicap for long distance offsets, favor decompression speed */
+
+ /* match Length */
+ { U32 const mlCode = ZSTD_MLcode(mlBase);
+ price += (ML_bits[mlCode] * BITCOST_MULTIPLIER) + (optPtr->matchLengthSumBasePrice - WEIGHT(optPtr->matchLengthFreq[mlCode], optLevel));
+ }
+
+ price += BITCOST_MULTIPLIER / 5; /* heuristic : make matches a bit more costly to favor less sequences -> faster decompression speed */
+
+ DEBUGLOG(8, "ZSTD_getMatchPrice(ml:%u) = %u", matchLength, price);
+ return price;
+}
+
+/* ZSTD_updateStats() :
+ * assumption : literals + litLengtn <= iend */
+static void ZSTD_updateStats(optState_t* const optPtr,
+ U32 litLength, const BYTE* literals,
+ U32 offsetCode, U32 matchLength)
+{
+ /* literals */
+ if (ZSTD_compressedLiterals(optPtr)) {
+ U32 u;
+ for (u=0; u < litLength; u++)
+ optPtr->litFreq[literals[u]] += ZSTD_LITFREQ_ADD;
+ optPtr->litSum += litLength*ZSTD_LITFREQ_ADD;
+ }
+
+ /* literal Length */
+ { U32 const llCode = ZSTD_LLcode(litLength);
+ optPtr->litLengthFreq[llCode]++;
+ optPtr->litLengthSum++;
+ }
+
+ /* offset code : expected to follow storeSeq() numeric representation */
+ { U32 const offCode = ZSTD_highbit32(STORED_TO_OFFBASE(offsetCode));
+ assert(offCode <= MaxOff);
+ optPtr->offCodeFreq[offCode]++;
+ optPtr->offCodeSum++;
+ }
+
+ /* match Length */
+ { U32 const mlBase = matchLength - MINMATCH;
+ U32 const mlCode = ZSTD_MLcode(mlBase);
+ optPtr->matchLengthFreq[mlCode]++;
+ optPtr->matchLengthSum++;
+ }
+}
+
+
+/* ZSTD_readMINMATCH() :
+ * function safe only for comparisons
+ * assumption : memPtr must be at least 4 bytes before end of buffer */
+MEM_STATIC U32 ZSTD_readMINMATCH(const void* memPtr, U32 length)
+{
+ switch (length)
+ {
+ default :
+ case 4 : return MEM_read32(memPtr);
+ case 3 : if (MEM_isLittleEndian())
+ return MEM_read32(memPtr)<<8;
+ else
+ return MEM_read32(memPtr)>>8;
+ }
+}
+
+
+/* Update hashTable3 up to ip (excluded)
+ Assumption : always within prefix (i.e. not within extDict) */
+static U32 ZSTD_insertAndFindFirstIndexHash3 (const ZSTD_matchState_t* ms,
+ U32* nextToUpdate3,
+ const BYTE* const ip)
+{
+ U32* const hashTable3 = ms->hashTable3;
+ U32 const hashLog3 = ms->hashLog3;
+ const BYTE* const base = ms->window.base;
+ U32 idx = *nextToUpdate3;
+ U32 const target = (U32)(ip - base);
+ size_t const hash3 = ZSTD_hash3Ptr(ip, hashLog3);
+ assert(hashLog3 > 0);
+
+ while(idx < target) {
+ hashTable3[ZSTD_hash3Ptr(base+idx, hashLog3)] = idx;
+ idx++;
+ }
+
+ *nextToUpdate3 = target;
+ return hashTable3[hash3];
+}
+
+
+/*-*************************************
+* Binary Tree search
+***************************************/
+/* ZSTD_insertBt1() : add one or multiple positions to tree.
+ * @param ip assumed <= iend-8 .
+ * @param target The target of ZSTD_updateTree_internal() - we are filling to this position
+ * @return : nb of positions added */
+static U32 ZSTD_insertBt1(
+ const ZSTD_matchState_t* ms,
+ const BYTE* const ip, const BYTE* const iend,
+ U32 const target,
+ U32 const mls, const int extDict)
+{
+ const ZSTD_compressionParameters* const cParams = &ms->cParams;
+ U32* const hashTable = ms->hashTable;
+ U32 const hashLog = cParams->hashLog;
+ size_t const h = ZSTD_hashPtr(ip, hashLog, mls);
+ U32* const bt = ms->chainTable;
+ U32 const btLog = cParams->chainLog - 1;
+ U32 const btMask = (1 << btLog) - 1;
+ U32 matchIndex = hashTable[h];
+ size_t commonLengthSmaller=0, commonLengthLarger=0;
+ const BYTE* const base = ms->window.base;
+ const BYTE* const dictBase = ms->window.dictBase;
+ const U32 dictLimit = ms->window.dictLimit;
+ const BYTE* const dictEnd = dictBase + dictLimit;
+ const BYTE* const prefixStart = base + dictLimit;
+ const BYTE* match;
+ const U32 curr = (U32)(ip-base);
+ const U32 btLow = btMask >= curr ? 0 : curr - btMask;
+ U32* smallerPtr = bt + 2*(curr&btMask);
+ U32* largerPtr = smallerPtr + 1;
+ U32 dummy32; /* to be nullified at the end */
+ /* windowLow is based on target because
+ * we only need positions that will be in the window at the end of the tree update.
+ */
+ U32 const windowLow = ZSTD_getLowestMatchIndex(ms, target, cParams->windowLog);
+ U32 matchEndIdx = curr+8+1;
+ size_t bestLength = 8;
+ U32 nbCompares = 1U << cParams->searchLog;
+#ifdef ZSTD_C_PREDICT
+ U32 predictedSmall = *(bt + 2*((curr-1)&btMask) + 0);
+ U32 predictedLarge = *(bt + 2*((curr-1)&btMask) + 1);
+ predictedSmall += (predictedSmall>0);
+ predictedLarge += (predictedLarge>0);
+#endif /* ZSTD_C_PREDICT */
+
+ DEBUGLOG(8, "ZSTD_insertBt1 (%u)", curr);
+
+ assert(curr <= target);
+ assert(ip <= iend-8); /* required for h calculation */
+ hashTable[h] = curr; /* Update Hash Table */
+
+ assert(windowLow > 0);
+ for (; nbCompares && (matchIndex >= windowLow); --nbCompares) {
+ U32* const nextPtr = bt + 2*(matchIndex & btMask);
+ size_t matchLength = MIN(commonLengthSmaller, commonLengthLarger); /* guaranteed minimum nb of common bytes */
+ assert(matchIndex < curr);
+
+#ifdef ZSTD_C_PREDICT /* note : can create issues when hlog small <= 11 */
+ const U32* predictPtr = bt + 2*((matchIndex-1) & btMask); /* written this way, as bt is a roll buffer */
+ if (matchIndex == predictedSmall) {
+ /* no need to check length, result known */
+ *smallerPtr = matchIndex;
+ if (matchIndex <= btLow) { smallerPtr=&dummy32; break; } /* beyond tree size, stop the search */
+ smallerPtr = nextPtr+1; /* new "smaller" => larger of match */
+ matchIndex = nextPtr[1]; /* new matchIndex larger than previous (closer to current) */
+ predictedSmall = predictPtr[1] + (predictPtr[1]>0);
+ continue;
+ }
+ if (matchIndex == predictedLarge) {
+ *largerPtr = matchIndex;
+ if (matchIndex <= btLow) { largerPtr=&dummy32; break; } /* beyond tree size, stop the search */
+ largerPtr = nextPtr;
+ matchIndex = nextPtr[0];
+ predictedLarge = predictPtr[0] + (predictPtr[0]>0);
+ continue;
+ }
+#endif
+
+ if (!extDict || (matchIndex+matchLength >= dictLimit)) {
+ assert(matchIndex+matchLength >= dictLimit); /* might be wrong if actually extDict */
+ match = base + matchIndex;
+ matchLength += ZSTD_count(ip+matchLength, match+matchLength, iend);
+ } else {
+ match = dictBase + matchIndex;
+ matchLength += ZSTD_count_2segments(ip+matchLength, match+matchLength, iend, dictEnd, prefixStart);
+ if (matchIndex+matchLength >= dictLimit)
+ match = base + matchIndex; /* to prepare for next usage of match[matchLength] */
+ }
+
+ if (matchLength > bestLength) {
+ bestLength = matchLength;
+ if (matchLength > matchEndIdx - matchIndex)
+ matchEndIdx = matchIndex + (U32)matchLength;
+ }
+
+ if (ip+matchLength == iend) { /* equal : no way to know if inf or sup */
+ break; /* drop , to guarantee consistency ; miss a bit of compression, but other solutions can corrupt tree */
+ }
+
+ if (match[matchLength] < ip[matchLength]) { /* necessarily within buffer */
+ /* match is smaller than current */
+ *smallerPtr = matchIndex; /* update smaller idx */
+ commonLengthSmaller = matchLength; /* all smaller will now have at least this guaranteed common length */
+ if (matchIndex <= btLow) { smallerPtr=&dummy32; break; } /* beyond tree size, stop searching */
+ smallerPtr = nextPtr+1; /* new "candidate" => larger than match, which was smaller than target */
+ matchIndex = nextPtr[1]; /* new matchIndex, larger than previous and closer to current */
+ } else {
+ /* match is larger than current */
+ *largerPtr = matchIndex;
+ commonLengthLarger = matchLength;
+ if (matchIndex <= btLow) { largerPtr=&dummy32; break; } /* beyond tree size, stop searching */
+ largerPtr = nextPtr;
+ matchIndex = nextPtr[0];
+ } }
+
+ *smallerPtr = *largerPtr = 0;
+ { U32 positions = 0;
+ if (bestLength > 384) positions = MIN(192, (U32)(bestLength - 384)); /* speed optimization */
+ assert(matchEndIdx > curr + 8);
+ return MAX(positions, matchEndIdx - (curr + 8));
+ }
+}
+
+FORCE_INLINE_TEMPLATE
+void ZSTD_updateTree_internal(
+ ZSTD_matchState_t* ms,
+ const BYTE* const ip, const BYTE* const iend,
+ const U32 mls, const ZSTD_dictMode_e dictMode)
+{
+ const BYTE* const base = ms->window.base;
+ U32 const target = (U32)(ip - base);
+ U32 idx = ms->nextToUpdate;
+ DEBUGLOG(6, "ZSTD_updateTree_internal, from %u to %u (dictMode:%u)",
+ idx, target, dictMode);
+
+ while(idx < target) {
+ U32 const forward = ZSTD_insertBt1(ms, base+idx, iend, target, mls, dictMode == ZSTD_extDict);
+ assert(idx < (U32)(idx + forward));
+ idx += forward;
+ }
+ assert((size_t)(ip - base) <= (size_t)(U32)(-1));
+ assert((size_t)(iend - base) <= (size_t)(U32)(-1));
+ ms->nextToUpdate = target;
+}
+
+void ZSTD_updateTree(ZSTD_matchState_t* ms, const BYTE* ip, const BYTE* iend) {
+ ZSTD_updateTree_internal(ms, ip, iend, ms->cParams.minMatch, ZSTD_noDict);
+}
+
+FORCE_INLINE_TEMPLATE
+U32 ZSTD_insertBtAndGetAllMatches (
+ ZSTD_match_t* matches, /* store result (found matches) in this table (presumed large enough) */
+ ZSTD_matchState_t* ms,
+ U32* nextToUpdate3,
+ const BYTE* const ip, const BYTE* const iLimit, const ZSTD_dictMode_e dictMode,
+ const U32 rep[ZSTD_REP_NUM],
+ U32 const ll0, /* tells if associated literal length is 0 or not. This value must be 0 or 1 */
+ const U32 lengthToBeat,
+ U32 const mls /* template */)
+{
+ const ZSTD_compressionParameters* const cParams = &ms->cParams;
+ U32 const sufficient_len = MIN(cParams->targetLength, ZSTD_OPT_NUM -1);
+ const BYTE* const base = ms->window.base;
+ U32 const curr = (U32)(ip-base);
+ U32 const hashLog = cParams->hashLog;
+ U32 const minMatch = (mls==3) ? 3 : 4;
+ U32* const hashTable = ms->hashTable;
+ size_t const h = ZSTD_hashPtr(ip, hashLog, mls);
+ U32 matchIndex = hashTable[h];
+ U32* const bt = ms->chainTable;
+ U32 const btLog = cParams->chainLog - 1;
+ U32 const btMask= (1U << btLog) - 1;
+ size_t commonLengthSmaller=0, commonLengthLarger=0;
+ const BYTE* const dictBase = ms->window.dictBase;
+ U32 const dictLimit = ms->window.dictLimit;
+ const BYTE* const dictEnd = dictBase + dictLimit;
+ const BYTE* const prefixStart = base + dictLimit;
+ U32 const btLow = (btMask >= curr) ? 0 : curr - btMask;
+ U32 const windowLow = ZSTD_getLowestMatchIndex(ms, curr, cParams->windowLog);
+ U32 const matchLow = windowLow ? windowLow : 1;
+ U32* smallerPtr = bt + 2*(curr&btMask);
+ U32* largerPtr = bt + 2*(curr&btMask) + 1;
+ U32 matchEndIdx = curr+8+1; /* farthest referenced position of any match => detects repetitive patterns */
+ U32 dummy32; /* to be nullified at the end */
+ U32 mnum = 0;
+ U32 nbCompares = 1U << cParams->searchLog;
+
+ const ZSTD_matchState_t* dms = dictMode == ZSTD_dictMatchState ? ms->dictMatchState : NULL;
+ const ZSTD_compressionParameters* const dmsCParams =
+ dictMode == ZSTD_dictMatchState ? &dms->cParams : NULL;
+ const BYTE* const dmsBase = dictMode == ZSTD_dictMatchState ? dms->window.base : NULL;
+ const BYTE* const dmsEnd = dictMode == ZSTD_dictMatchState ? dms->window.nextSrc : NULL;
+ U32 const dmsHighLimit = dictMode == ZSTD_dictMatchState ? (U32)(dmsEnd - dmsBase) : 0;
+ U32 const dmsLowLimit = dictMode == ZSTD_dictMatchState ? dms->window.lowLimit : 0;
+ U32 const dmsIndexDelta = dictMode == ZSTD_dictMatchState ? windowLow - dmsHighLimit : 0;
+ U32 const dmsHashLog = dictMode == ZSTD_dictMatchState ? dmsCParams->hashLog : hashLog;
+ U32 const dmsBtLog = dictMode == ZSTD_dictMatchState ? dmsCParams->chainLog - 1 : btLog;
+ U32 const dmsBtMask = dictMode == ZSTD_dictMatchState ? (1U << dmsBtLog) - 1 : 0;
+ U32 const dmsBtLow = dictMode == ZSTD_dictMatchState && dmsBtMask < dmsHighLimit - dmsLowLimit ? dmsHighLimit - dmsBtMask : dmsLowLimit;
+
+ size_t bestLength = lengthToBeat-1;
+ DEBUGLOG(8, "ZSTD_insertBtAndGetAllMatches: current=%u", curr);
+
+ /* check repCode */
+ assert(ll0 <= 1); /* necessarily 1 or 0 */
+ { U32 const lastR = ZSTD_REP_NUM + ll0;
+ U32 repCode;
+ for (repCode = ll0; repCode < lastR; repCode++) {
+ U32 const repOffset = (repCode==ZSTD_REP_NUM) ? (rep[0] - 1) : rep[repCode];
+ U32 const repIndex = curr - repOffset;
+ U32 repLen = 0;
+ assert(curr >= dictLimit);
+ if (repOffset-1 /* intentional overflow, discards 0 and -1 */ < curr-dictLimit) { /* equivalent to `curr > repIndex >= dictLimit` */
+ /* We must validate the repcode offset because when we're using a dictionary the
+ * valid offset range shrinks when the dictionary goes out of bounds.
+ */
+ if ((repIndex >= windowLow) & (ZSTD_readMINMATCH(ip, minMatch) == ZSTD_readMINMATCH(ip - repOffset, minMatch))) {
+ repLen = (U32)ZSTD_count(ip+minMatch, ip+minMatch-repOffset, iLimit) + minMatch;
+ }
+ } else { /* repIndex < dictLimit || repIndex >= curr */
+ const BYTE* const repMatch = dictMode == ZSTD_dictMatchState ?
+ dmsBase + repIndex - dmsIndexDelta :
+ dictBase + repIndex;
+ assert(curr >= windowLow);
+ if ( dictMode == ZSTD_extDict
+ && ( ((repOffset-1) /*intentional overflow*/ < curr - windowLow) /* equivalent to `curr > repIndex >= windowLow` */
+ & (((U32)((dictLimit-1) - repIndex) >= 3) ) /* intentional overflow : do not test positions overlapping 2 memory segments */)
+ && (ZSTD_readMINMATCH(ip, minMatch) == ZSTD_readMINMATCH(repMatch, minMatch)) ) {
+ repLen = (U32)ZSTD_count_2segments(ip+minMatch, repMatch+minMatch, iLimit, dictEnd, prefixStart) + minMatch;
+ }
+ if (dictMode == ZSTD_dictMatchState
+ && ( ((repOffset-1) /*intentional overflow*/ < curr - (dmsLowLimit + dmsIndexDelta)) /* equivalent to `curr > repIndex >= dmsLowLimit` */
+ & ((U32)((dictLimit-1) - repIndex) >= 3) ) /* intentional overflow : do not test positions overlapping 2 memory segments */
+ && (ZSTD_readMINMATCH(ip, minMatch) == ZSTD_readMINMATCH(repMatch, minMatch)) ) {
+ repLen = (U32)ZSTD_count_2segments(ip+minMatch, repMatch+minMatch, iLimit, dmsEnd, prefixStart) + minMatch;
+ } }
+ /* save longer solution */
+ if (repLen > bestLength) {
+ DEBUGLOG(8, "found repCode %u (ll0:%u, offset:%u) of length %u",
+ repCode, ll0, repOffset, repLen);
+ bestLength = repLen;
+ matches[mnum].off = STORE_REPCODE(repCode - ll0 + 1); /* expect value between 1 and 3 */
+ matches[mnum].len = (U32)repLen;
+ mnum++;
+ if ( (repLen > sufficient_len)
+ | (ip+repLen == iLimit) ) { /* best possible */
+ return mnum;
+ } } } }
+
+ /* HC3 match finder */
+ if ((mls == 3) /*static*/ && (bestLength < mls)) {
+ U32 const matchIndex3 = ZSTD_insertAndFindFirstIndexHash3(ms, nextToUpdate3, ip);
+ if ((matchIndex3 >= matchLow)
+ & (curr - matchIndex3 < (1<<18)) /*heuristic : longer distance likely too expensive*/ ) {
+ size_t mlen;
+ if ((dictMode == ZSTD_noDict) /*static*/ || (dictMode == ZSTD_dictMatchState) /*static*/ || (matchIndex3 >= dictLimit)) {
+ const BYTE* const match = base + matchIndex3;
+ mlen = ZSTD_count(ip, match, iLimit);
+ } else {
+ const BYTE* const match = dictBase + matchIndex3;
+ mlen = ZSTD_count_2segments(ip, match, iLimit, dictEnd, prefixStart);
+ }
+
+ /* save best solution */
+ if (mlen >= mls /* == 3 > bestLength */) {
+ DEBUGLOG(8, "found small match with hlog3, of length %u",
+ (U32)mlen);
+ bestLength = mlen;
+ assert(curr > matchIndex3);
+ assert(mnum==0); /* no prior solution */
+ matches[0].off = STORE_OFFSET(curr - matchIndex3);
+ matches[0].len = (U32)mlen;
+ mnum = 1;
+ if ( (mlen > sufficient_len) |
+ (ip+mlen == iLimit) ) { /* best possible length */
+ ms->nextToUpdate = curr+1; /* skip insertion */
+ return 1;
+ } } }
+ /* no dictMatchState lookup: dicts don't have a populated HC3 table */
+ } /* if (mls == 3) */
+
+ hashTable[h] = curr; /* Update Hash Table */
+
+ for (; nbCompares && (matchIndex >= matchLow); --nbCompares) {
+ U32* const nextPtr = bt + 2*(matchIndex & btMask);
+ const BYTE* match;
+ size_t matchLength = MIN(commonLengthSmaller, commonLengthLarger); /* guaranteed minimum nb of common bytes */
+ assert(curr > matchIndex);
+
+ if ((dictMode == ZSTD_noDict) || (dictMode == ZSTD_dictMatchState) || (matchIndex+matchLength >= dictLimit)) {
+ assert(matchIndex+matchLength >= dictLimit); /* ensure the condition is correct when !extDict */
+ match = base + matchIndex;
+ if (matchIndex >= dictLimit) assert(memcmp(match, ip, matchLength) == 0); /* ensure early section of match is equal as expected */
+ matchLength += ZSTD_count(ip+matchLength, match+matchLength, iLimit);
+ } else {
+ match = dictBase + matchIndex;
+ assert(memcmp(match, ip, matchLength) == 0); /* ensure early section of match is equal as expected */
+ matchLength += ZSTD_count_2segments(ip+matchLength, match+matchLength, iLimit, dictEnd, prefixStart);
+ if (matchIndex+matchLength >= dictLimit)
+ match = base + matchIndex; /* prepare for match[matchLength] read */
+ }
+
+ if (matchLength > bestLength) {
+ DEBUGLOG(8, "found match of length %u at distance %u (offCode=%u)",
+ (U32)matchLength, curr - matchIndex, STORE_OFFSET(curr - matchIndex));
+ assert(matchEndIdx > matchIndex);
+ if (matchLength > matchEndIdx - matchIndex)
+ matchEndIdx = matchIndex + (U32)matchLength;
+ bestLength = matchLength;
+ matches[mnum].off = STORE_OFFSET(curr - matchIndex);
+ matches[mnum].len = (U32)matchLength;
+ mnum++;
+ if ( (matchLength > ZSTD_OPT_NUM)
+ | (ip+matchLength == iLimit) /* equal : no way to know if inf or sup */) {
+ if (dictMode == ZSTD_dictMatchState) nbCompares = 0; /* break should also skip searching dms */
+ break; /* drop, to preserve bt consistency (miss a little bit of compression) */
+ } }
+
+ if (match[matchLength] < ip[matchLength]) {
+ /* match smaller than current */
+ *smallerPtr = matchIndex; /* update smaller idx */
+ commonLengthSmaller = matchLength; /* all smaller will now have at least this guaranteed common length */
+ if (matchIndex <= btLow) { smallerPtr=&dummy32; break; } /* beyond tree size, stop the search */
+ smallerPtr = nextPtr+1; /* new candidate => larger than match, which was smaller than current */
+ matchIndex = nextPtr[1]; /* new matchIndex, larger than previous, closer to current */
+ } else {
+ *largerPtr = matchIndex;
+ commonLengthLarger = matchLength;
+ if (matchIndex <= btLow) { largerPtr=&dummy32; break; } /* beyond tree size, stop the search */
+ largerPtr = nextPtr;
+ matchIndex = nextPtr[0];
+ } }
+
+ *smallerPtr = *largerPtr = 0;
+
+ assert(nbCompares <= (1U << ZSTD_SEARCHLOG_MAX)); /* Check we haven't underflowed. */
+ if (dictMode == ZSTD_dictMatchState && nbCompares) {
+ size_t const dmsH = ZSTD_hashPtr(ip, dmsHashLog, mls);
+ U32 dictMatchIndex = dms->hashTable[dmsH];
+ const U32* const dmsBt = dms->chainTable;
+ commonLengthSmaller = commonLengthLarger = 0;
+ for (; nbCompares && (dictMatchIndex > dmsLowLimit); --nbCompares) {
+ const U32* const nextPtr = dmsBt + 2*(dictMatchIndex & dmsBtMask);
+ size_t matchLength = MIN(commonLengthSmaller, commonLengthLarger); /* guaranteed minimum nb of common bytes */
+ const BYTE* match = dmsBase + dictMatchIndex;
+ matchLength += ZSTD_count_2segments(ip+matchLength, match+matchLength, iLimit, dmsEnd, prefixStart);
+ if (dictMatchIndex+matchLength >= dmsHighLimit)
+ match = base + dictMatchIndex + dmsIndexDelta; /* to prepare for next usage of match[matchLength] */
+
+ if (matchLength > bestLength) {
+ matchIndex = dictMatchIndex + dmsIndexDelta;
+ DEBUGLOG(8, "found dms match of length %u at distance %u (offCode=%u)",
+ (U32)matchLength, curr - matchIndex, STORE_OFFSET(curr - matchIndex));
+ if (matchLength > matchEndIdx - matchIndex)
+ matchEndIdx = matchIndex + (U32)matchLength;
+ bestLength = matchLength;
+ matches[mnum].off = STORE_OFFSET(curr - matchIndex);
+ matches[mnum].len = (U32)matchLength;
+ mnum++;
+ if ( (matchLength > ZSTD_OPT_NUM)
+ | (ip+matchLength == iLimit) /* equal : no way to know if inf or sup */) {
+ break; /* drop, to guarantee consistency (miss a little bit of compression) */
+ } }
+
+ if (dictMatchIndex <= dmsBtLow) { break; } /* beyond tree size, stop the search */
+ if (match[matchLength] < ip[matchLength]) {
+ commonLengthSmaller = matchLength; /* all smaller will now have at least this guaranteed common length */
+ dictMatchIndex = nextPtr[1]; /* new matchIndex larger than previous (closer to current) */
+ } else {
+ /* match is larger than current */
+ commonLengthLarger = matchLength;
+ dictMatchIndex = nextPtr[0];
+ } } } /* if (dictMode == ZSTD_dictMatchState) */
+
+ assert(matchEndIdx > curr+8);
+ ms->nextToUpdate = matchEndIdx - 8; /* skip repetitive patterns */
+ return mnum;
+}
+
+typedef U32 (*ZSTD_getAllMatchesFn)(
+ ZSTD_match_t*,
+ ZSTD_matchState_t*,
+ U32*,
+ const BYTE*,
+ const BYTE*,
+ const U32 rep[ZSTD_REP_NUM],
+ U32 const ll0,
+ U32 const lengthToBeat);
+
+FORCE_INLINE_TEMPLATE U32 ZSTD_btGetAllMatches_internal(
+ ZSTD_match_t* matches,
+ ZSTD_matchState_t* ms,
+ U32* nextToUpdate3,
+ const BYTE* ip,
+ const BYTE* const iHighLimit,
+ const U32 rep[ZSTD_REP_NUM],
+ U32 const ll0,
+ U32 const lengthToBeat,
+ const ZSTD_dictMode_e dictMode,
+ const U32 mls)
+{
+ assert(BOUNDED(3, ms->cParams.minMatch, 6) == mls);
+ DEBUGLOG(8, "ZSTD_BtGetAllMatches(dictMode=%d, mls=%u)", (int)dictMode, mls);
+ if (ip < ms->window.base + ms->nextToUpdate)
+ return 0; /* skipped area */
+ ZSTD_updateTree_internal(ms, ip, iHighLimit, mls, dictMode);
+ return ZSTD_insertBtAndGetAllMatches(matches, ms, nextToUpdate3, ip, iHighLimit, dictMode, rep, ll0, lengthToBeat, mls);
+}
+
+#define ZSTD_BT_GET_ALL_MATCHES_FN(dictMode, mls) ZSTD_btGetAllMatches_##dictMode##_##mls
+
+#define GEN_ZSTD_BT_GET_ALL_MATCHES_(dictMode, mls) \
+ static U32 ZSTD_BT_GET_ALL_MATCHES_FN(dictMode, mls)( \
+ ZSTD_match_t* matches, \
+ ZSTD_matchState_t* ms, \
+ U32* nextToUpdate3, \
+ const BYTE* ip, \
+ const BYTE* const iHighLimit, \
+ const U32 rep[ZSTD_REP_NUM], \
+ U32 const ll0, \
+ U32 const lengthToBeat) \
+ { \
+ return ZSTD_btGetAllMatches_internal( \
+ matches, ms, nextToUpdate3, ip, iHighLimit, \
+ rep, ll0, lengthToBeat, ZSTD_##dictMode, mls); \
+ }
+
+#define GEN_ZSTD_BT_GET_ALL_MATCHES(dictMode) \
+ GEN_ZSTD_BT_GET_ALL_MATCHES_(dictMode, 3) \
+ GEN_ZSTD_BT_GET_ALL_MATCHES_(dictMode, 4) \
+ GEN_ZSTD_BT_GET_ALL_MATCHES_(dictMode, 5) \
+ GEN_ZSTD_BT_GET_ALL_MATCHES_(dictMode, 6)
+
+GEN_ZSTD_BT_GET_ALL_MATCHES(noDict)
+GEN_ZSTD_BT_GET_ALL_MATCHES(extDict)
+GEN_ZSTD_BT_GET_ALL_MATCHES(dictMatchState)
+
+#define ZSTD_BT_GET_ALL_MATCHES_ARRAY(dictMode) \
+ { \
+ ZSTD_BT_GET_ALL_MATCHES_FN(dictMode, 3), \
+ ZSTD_BT_GET_ALL_MATCHES_FN(dictMode, 4), \
+ ZSTD_BT_GET_ALL_MATCHES_FN(dictMode, 5), \
+ ZSTD_BT_GET_ALL_MATCHES_FN(dictMode, 6) \
+ }
+
+static ZSTD_getAllMatchesFn
+ZSTD_selectBtGetAllMatches(ZSTD_matchState_t const* ms, ZSTD_dictMode_e const dictMode)
+{
+ ZSTD_getAllMatchesFn const getAllMatchesFns[3][4] = {
+ ZSTD_BT_GET_ALL_MATCHES_ARRAY(noDict),
+ ZSTD_BT_GET_ALL_MATCHES_ARRAY(extDict),
+ ZSTD_BT_GET_ALL_MATCHES_ARRAY(dictMatchState)
+ };
+ U32 const mls = BOUNDED(3, ms->cParams.minMatch, 6);
+ assert((U32)dictMode < 3);
+ assert(mls - 3 < 4);
+ return getAllMatchesFns[(int)dictMode][mls - 3];
+}
+
+/* ***********************
+* LDM helper functions *
+*************************/
+
+/* Struct containing info needed to make decision about ldm inclusion */
+typedef struct {
+ rawSeqStore_t seqStore; /* External match candidates store for this block */
+ U32 startPosInBlock; /* Start position of the current match candidate */
+ U32 endPosInBlock; /* End position of the current match candidate */
+ U32 offset; /* Offset of the match candidate */
+} ZSTD_optLdm_t;
+
+/* ZSTD_optLdm_skipRawSeqStoreBytes():
+ * Moves forward in @rawSeqStore by @nbBytes,
+ * which will update the fields 'pos' and 'posInSequence'.
+ */
+static void ZSTD_optLdm_skipRawSeqStoreBytes(rawSeqStore_t* rawSeqStore, size_t nbBytes)
+{
+ U32 currPos = (U32)(rawSeqStore->posInSequence + nbBytes);
+ while (currPos && rawSeqStore->pos < rawSeqStore->size) {
+ rawSeq currSeq = rawSeqStore->seq[rawSeqStore->pos];
+ if (currPos >= currSeq.litLength + currSeq.matchLength) {
+ currPos -= currSeq.litLength + currSeq.matchLength;
+ rawSeqStore->pos++;
+ } else {
+ rawSeqStore->posInSequence = currPos;
+ break;
+ }
+ }
+ if (currPos == 0 || rawSeqStore->pos == rawSeqStore->size) {
+ rawSeqStore->posInSequence = 0;
+ }
+}
+
+/* ZSTD_opt_getNextMatchAndUpdateSeqStore():
+ * Calculates the beginning and end of the next match in the current block.
+ * Updates 'pos' and 'posInSequence' of the ldmSeqStore.
+ */
+static void
+ZSTD_opt_getNextMatchAndUpdateSeqStore(ZSTD_optLdm_t* optLdm, U32 currPosInBlock,
+ U32 blockBytesRemaining)
+{
+ rawSeq currSeq;
+ U32 currBlockEndPos;
+ U32 literalsBytesRemaining;
+ U32 matchBytesRemaining;
+
+ /* Setting match end position to MAX to ensure we never use an LDM during this block */
+ if (optLdm->seqStore.size == 0 || optLdm->seqStore.pos >= optLdm->seqStore.size) {
+ optLdm->startPosInBlock = UINT_MAX;
+ optLdm->endPosInBlock = UINT_MAX;
+ return;
+ }
+ /* Calculate appropriate bytes left in matchLength and litLength
+ * after adjusting based on ldmSeqStore->posInSequence */
+ currSeq = optLdm->seqStore.seq[optLdm->seqStore.pos];
+ assert(optLdm->seqStore.posInSequence <= currSeq.litLength + currSeq.matchLength);
+ currBlockEndPos = currPosInBlock + blockBytesRemaining;
+ literalsBytesRemaining = (optLdm->seqStore.posInSequence < currSeq.litLength) ?
+ currSeq.litLength - (U32)optLdm->seqStore.posInSequence :
+ 0;
+ matchBytesRemaining = (literalsBytesRemaining == 0) ?
+ currSeq.matchLength - ((U32)optLdm->seqStore.posInSequence - currSeq.litLength) :
+ currSeq.matchLength;
+
+ /* If there are more literal bytes than bytes remaining in block, no ldm is possible */
+ if (literalsBytesRemaining >= blockBytesRemaining) {
+ optLdm->startPosInBlock = UINT_MAX;
+ optLdm->endPosInBlock = UINT_MAX;
+ ZSTD_optLdm_skipRawSeqStoreBytes(&optLdm->seqStore, blockBytesRemaining);
+ return;
+ }
+
+ /* Matches may be < MINMATCH by this process. In that case, we will reject them
+ when we are deciding whether or not to add the ldm */
+ optLdm->startPosInBlock = currPosInBlock + literalsBytesRemaining;
+ optLdm->endPosInBlock = optLdm->startPosInBlock + matchBytesRemaining;
+ optLdm->offset = currSeq.offset;
+
+ if (optLdm->endPosInBlock > currBlockEndPos) {
+ /* Match ends after the block ends, we can't use the whole match */
+ optLdm->endPosInBlock = currBlockEndPos;
+ ZSTD_optLdm_skipRawSeqStoreBytes(&optLdm->seqStore, currBlockEndPos - currPosInBlock);
+ } else {
+ /* Consume nb of bytes equal to size of sequence left */
+ ZSTD_optLdm_skipRawSeqStoreBytes(&optLdm->seqStore, literalsBytesRemaining + matchBytesRemaining);
+ }
+}
+
+/* ZSTD_optLdm_maybeAddMatch():
+ * Adds a match if it's long enough,
+ * based on it's 'matchStartPosInBlock' and 'matchEndPosInBlock',
+ * into 'matches'. Maintains the correct ordering of 'matches'.
+ */
+static void ZSTD_optLdm_maybeAddMatch(ZSTD_match_t* matches, U32* nbMatches,
+ const ZSTD_optLdm_t* optLdm, U32 currPosInBlock)
+{
+ U32 const posDiff = currPosInBlock - optLdm->startPosInBlock;
+ /* Note: ZSTD_match_t actually contains offCode and matchLength (before subtracting MINMATCH) */
+ U32 const candidateMatchLength = optLdm->endPosInBlock - optLdm->startPosInBlock - posDiff;
+
+ /* Ensure that current block position is not outside of the match */
+ if (currPosInBlock < optLdm->startPosInBlock
+ || currPosInBlock >= optLdm->endPosInBlock
+ || candidateMatchLength < MINMATCH) {
+ return;
+ }
+
+ if (*nbMatches == 0 || ((candidateMatchLength > matches[*nbMatches-1].len) && *nbMatches < ZSTD_OPT_NUM)) {
+ U32 const candidateOffCode = STORE_OFFSET(optLdm->offset);
+ DEBUGLOG(6, "ZSTD_optLdm_maybeAddMatch(): Adding ldm candidate match (offCode: %u matchLength %u) at block position=%u",
+ candidateOffCode, candidateMatchLength, currPosInBlock);
+ matches[*nbMatches].len = candidateMatchLength;
+ matches[*nbMatches].off = candidateOffCode;
+ (*nbMatches)++;
+ }
+}
+
+/* ZSTD_optLdm_processMatchCandidate():
+ * Wrapper function to update ldm seq store and call ldm functions as necessary.
+ */
+static void
+ZSTD_optLdm_processMatchCandidate(ZSTD_optLdm_t* optLdm,
+ ZSTD_match_t* matches, U32* nbMatches,
+ U32 currPosInBlock, U32 remainingBytes)
+{
+ if (optLdm->seqStore.size == 0 || optLdm->seqStore.pos >= optLdm->seqStore.size) {
+ return;
+ }
+
+ if (currPosInBlock >= optLdm->endPosInBlock) {
+ if (currPosInBlock > optLdm->endPosInBlock) {
+ /* The position at which ZSTD_optLdm_processMatchCandidate() is called is not necessarily
+ * at the end of a match from the ldm seq store, and will often be some bytes
+ * over beyond matchEndPosInBlock. As such, we need to correct for these "overshoots"
+ */
+ U32 const posOvershoot = currPosInBlock - optLdm->endPosInBlock;
+ ZSTD_optLdm_skipRawSeqStoreBytes(&optLdm->seqStore, posOvershoot);
+ }
+ ZSTD_opt_getNextMatchAndUpdateSeqStore(optLdm, currPosInBlock, remainingBytes);
+ }
+ ZSTD_optLdm_maybeAddMatch(matches, nbMatches, optLdm, currPosInBlock);
+}
+
+
+/*-*******************************
+* Optimal parser
+*********************************/
+
+static U32 ZSTD_totalLen(ZSTD_optimal_t sol)
+{
+ return sol.litlen + sol.mlen;
+}
+
+#if 0 /* debug */
+
+static void
+listStats(const U32* table, int lastEltID)
+{
+ int const nbElts = lastEltID + 1;
+ int enb;
+ for (enb=0; enb < nbElts; enb++) {
+ (void)table;
+ /* RAWLOG(2, "%3i:%3i, ", enb, table[enb]); */
+ RAWLOG(2, "%4i,", table[enb]);
+ }
+ RAWLOG(2, " \n");
+}
+
+#endif
+
+FORCE_INLINE_TEMPLATE size_t
+ZSTD_compressBlock_opt_generic(ZSTD_matchState_t* ms,
+ seqStore_t* seqStore,
+ U32 rep[ZSTD_REP_NUM],
+ const void* src, size_t srcSize,
+ const int optLevel,
+ const ZSTD_dictMode_e dictMode)
+{
+ optState_t* const optStatePtr = &ms->opt;
+ const BYTE* const istart = (const BYTE*)src;
+ const BYTE* ip = istart;
+ const BYTE* anchor = istart;
+ const BYTE* const iend = istart + srcSize;
+ const BYTE* const ilimit = iend - 8;
+ const BYTE* const base = ms->window.base;
+ const BYTE* const prefixStart = base + ms->window.dictLimit;
+ const ZSTD_compressionParameters* const cParams = &ms->cParams;
+
+ ZSTD_getAllMatchesFn getAllMatches = ZSTD_selectBtGetAllMatches(ms, dictMode);
+
+ U32 const sufficient_len = MIN(cParams->targetLength, ZSTD_OPT_NUM -1);
+ U32 const minMatch = (cParams->minMatch == 3) ? 3 : 4;
+ U32 nextToUpdate3 = ms->nextToUpdate;
+
+ ZSTD_optimal_t* const opt = optStatePtr->priceTable;
+ ZSTD_match_t* const matches = optStatePtr->matchTable;
+ ZSTD_optimal_t lastSequence;
+ ZSTD_optLdm_t optLdm;
+
+ optLdm.seqStore = ms->ldmSeqStore ? *ms->ldmSeqStore : kNullRawSeqStore;
+ optLdm.endPosInBlock = optLdm.startPosInBlock = optLdm.offset = 0;
+ ZSTD_opt_getNextMatchAndUpdateSeqStore(&optLdm, (U32)(ip-istart), (U32)(iend-ip));
+
+ /* init */
+ DEBUGLOG(5, "ZSTD_compressBlock_opt_generic: current=%u, prefix=%u, nextToUpdate=%u",
+ (U32)(ip - base), ms->window.dictLimit, ms->nextToUpdate);
+ assert(optLevel <= 2);
+ ZSTD_rescaleFreqs(optStatePtr, (const BYTE*)src, srcSize, optLevel);
+ ip += (ip==prefixStart);
+
+ /* Match Loop */
+ while (ip < ilimit) {
+ U32 cur, last_pos = 0;
+
+ /* find first match */
+ { U32 const litlen = (U32)(ip - anchor);
+ U32 const ll0 = !litlen;
+ U32 nbMatches = getAllMatches(matches, ms, &nextToUpdate3, ip, iend, rep, ll0, minMatch);
+ ZSTD_optLdm_processMatchCandidate(&optLdm, matches, &nbMatches,
+ (U32)(ip-istart), (U32)(iend - ip));
+ if (!nbMatches) { ip++; continue; }
+
+ /* initialize opt[0] */
+ { U32 i ; for (i=0; i<ZSTD_REP_NUM; i++) opt[0].rep[i] = rep[i]; }
+ opt[0].mlen = 0; /* means is_a_literal */
+ opt[0].litlen = litlen;
+ /* We don't need to include the actual price of the literals because
+ * it is static for the duration of the forward pass, and is included
+ * in every price. We include the literal length to avoid negative
+ * prices when we subtract the previous literal length.
+ */
+ opt[0].price = (int)ZSTD_litLengthPrice(litlen, optStatePtr, optLevel);
+
+ /* large match -> immediate encoding */
+ { U32 const maxML = matches[nbMatches-1].len;
+ U32 const maxOffcode = matches[nbMatches-1].off;
+ DEBUGLOG(6, "found %u matches of maxLength=%u and maxOffCode=%u at cPos=%u => start new series",
+ nbMatches, maxML, maxOffcode, (U32)(ip-prefixStart));
+
+ if (maxML > sufficient_len) {
+ lastSequence.litlen = litlen;
+ lastSequence.mlen = maxML;
+ lastSequence.off = maxOffcode;
+ DEBUGLOG(6, "large match (%u>%u), immediate encoding",
+ maxML, sufficient_len);
+ cur = 0;
+ last_pos = ZSTD_totalLen(lastSequence);
+ goto _shortestPath;
+ } }
+
+ /* set prices for first matches starting position == 0 */
+ assert(opt[0].price >= 0);
+ { U32 const literalsPrice = (U32)opt[0].price + ZSTD_litLengthPrice(0, optStatePtr, optLevel);
+ U32 pos;
+ U32 matchNb;
+ for (pos = 1; pos < minMatch; pos++) {
+ opt[pos].price = ZSTD_MAX_PRICE; /* mlen, litlen and price will be fixed during forward scanning */
+ }
+ for (matchNb = 0; matchNb < nbMatches; matchNb++) {
+ U32 const offcode = matches[matchNb].off;
+ U32 const end = matches[matchNb].len;
+ for ( ; pos <= end ; pos++ ) {
+ U32 const matchPrice = ZSTD_getMatchPrice(offcode, pos, optStatePtr, optLevel);
+ U32 const sequencePrice = literalsPrice + matchPrice;
+ DEBUGLOG(7, "rPos:%u => set initial price : %.2f",
+ pos, ZSTD_fCost(sequencePrice));
+ opt[pos].mlen = pos;
+ opt[pos].off = offcode;
+ opt[pos].litlen = litlen;
+ opt[pos].price = (int)sequencePrice;
+ } }
+ last_pos = pos-1;
+ }
+ }
+
+ /* check further positions */
+ for (cur = 1; cur <= last_pos; cur++) {
+ const BYTE* const inr = ip + cur;
+ assert(cur < ZSTD_OPT_NUM);
+ DEBUGLOG(7, "cPos:%zi==rPos:%u", inr-istart, cur)
+
+ /* Fix current position with one literal if cheaper */
+ { U32 const litlen = (opt[cur-1].mlen == 0) ? opt[cur-1].litlen + 1 : 1;
+ int const price = opt[cur-1].price
+ + (int)ZSTD_rawLiteralsCost(ip+cur-1, 1, optStatePtr, optLevel)
+ + (int)ZSTD_litLengthPrice(litlen, optStatePtr, optLevel)
+ - (int)ZSTD_litLengthPrice(litlen-1, optStatePtr, optLevel);
+ assert(price < 1000000000); /* overflow check */
+ if (price <= opt[cur].price) {
+ DEBUGLOG(7, "cPos:%zi==rPos:%u : better price (%.2f<=%.2f) using literal (ll==%u) (hist:%u,%u,%u)",
+ inr-istart, cur, ZSTD_fCost(price), ZSTD_fCost(opt[cur].price), litlen,
+ opt[cur-1].rep[0], opt[cur-1].rep[1], opt[cur-1].rep[2]);
+ opt[cur].mlen = 0;
+ opt[cur].off = 0;
+ opt[cur].litlen = litlen;
+ opt[cur].price = price;
+ } else {
+ DEBUGLOG(7, "cPos:%zi==rPos:%u : literal would cost more (%.2f>%.2f) (hist:%u,%u,%u)",
+ inr-istart, cur, ZSTD_fCost(price), ZSTD_fCost(opt[cur].price),
+ opt[cur].rep[0], opt[cur].rep[1], opt[cur].rep[2]);
+ }
+ }
+
+ /* Set the repcodes of the current position. We must do it here
+ * because we rely on the repcodes of the 2nd to last sequence being
+ * correct to set the next chunks repcodes during the backward
+ * traversal.
+ */
+ ZSTD_STATIC_ASSERT(sizeof(opt[cur].rep) == sizeof(repcodes_t));
+ assert(cur >= opt[cur].mlen);
+ if (opt[cur].mlen != 0) {
+ U32 const prev = cur - opt[cur].mlen;
+ repcodes_t const newReps = ZSTD_newRep(opt[prev].rep, opt[cur].off, opt[cur].litlen==0);
+ ZSTD_memcpy(opt[cur].rep, &newReps, sizeof(repcodes_t));
+ } else {
+ ZSTD_memcpy(opt[cur].rep, opt[cur - 1].rep, sizeof(repcodes_t));
+ }
+
+ /* last match must start at a minimum distance of 8 from oend */
+ if (inr > ilimit) continue;
+
+ if (cur == last_pos) break;
+
+ if ( (optLevel==0) /*static_test*/
+ && (opt[cur+1].price <= opt[cur].price + (BITCOST_MULTIPLIER/2)) ) {
+ DEBUGLOG(7, "move to next rPos:%u : price is <=", cur+1);
+ continue; /* skip unpromising positions; about ~+6% speed, -0.01 ratio */
+ }
+
+ assert(opt[cur].price >= 0);
+ { U32 const ll0 = (opt[cur].mlen != 0);
+ U32 const litlen = (opt[cur].mlen == 0) ? opt[cur].litlen : 0;
+ U32 const previousPrice = (U32)opt[cur].price;
+ U32 const basePrice = previousPrice + ZSTD_litLengthPrice(0, optStatePtr, optLevel);
+ U32 nbMatches = getAllMatches(matches, ms, &nextToUpdate3, inr, iend, opt[cur].rep, ll0, minMatch);
+ U32 matchNb;
+
+ ZSTD_optLdm_processMatchCandidate(&optLdm, matches, &nbMatches,
+ (U32)(inr-istart), (U32)(iend-inr));
+
+ if (!nbMatches) {
+ DEBUGLOG(7, "rPos:%u : no match found", cur);
+ continue;
+ }
+
+ { U32 const maxML = matches[nbMatches-1].len;
+ DEBUGLOG(7, "cPos:%zi==rPos:%u, found %u matches, of maxLength=%u",
+ inr-istart, cur, nbMatches, maxML);
+
+ if ( (maxML > sufficient_len)
+ || (cur + maxML >= ZSTD_OPT_NUM) ) {
+ lastSequence.mlen = maxML;
+ lastSequence.off = matches[nbMatches-1].off;
+ lastSequence.litlen = litlen;
+ cur -= (opt[cur].mlen==0) ? opt[cur].litlen : 0; /* last sequence is actually only literals, fix cur to last match - note : may underflow, in which case, it's first sequence, and it's okay */
+ last_pos = cur + ZSTD_totalLen(lastSequence);
+ if (cur > ZSTD_OPT_NUM) cur = 0; /* underflow => first match */
+ goto _shortestPath;
+ } }
+
+ /* set prices using matches found at position == cur */
+ for (matchNb = 0; matchNb < nbMatches; matchNb++) {
+ U32 const offset = matches[matchNb].off;
+ U32 const lastML = matches[matchNb].len;
+ U32 const startML = (matchNb>0) ? matches[matchNb-1].len+1 : minMatch;
+ U32 mlen;
+
+ DEBUGLOG(7, "testing match %u => offCode=%4u, mlen=%2u, llen=%2u",
+ matchNb, matches[matchNb].off, lastML, litlen);
+
+ for (mlen = lastML; mlen >= startML; mlen--) { /* scan downward */
+ U32 const pos = cur + mlen;
+ int const price = (int)basePrice + (int)ZSTD_getMatchPrice(offset, mlen, optStatePtr, optLevel);
+
+ if ((pos > last_pos) || (price < opt[pos].price)) {
+ DEBUGLOG(7, "rPos:%u (ml=%2u) => new better price (%.2f<%.2f)",
+ pos, mlen, ZSTD_fCost(price), ZSTD_fCost(opt[pos].price));
+ while (last_pos < pos) { opt[last_pos+1].price = ZSTD_MAX_PRICE; last_pos++; } /* fill empty positions */
+ opt[pos].mlen = mlen;
+ opt[pos].off = offset;
+ opt[pos].litlen = litlen;
+ opt[pos].price = price;
+ } else {
+ DEBUGLOG(7, "rPos:%u (ml=%2u) => new price is worse (%.2f>=%.2f)",
+ pos, mlen, ZSTD_fCost(price), ZSTD_fCost(opt[pos].price));
+ if (optLevel==0) break; /* early update abort; gets ~+10% speed for about -0.01 ratio loss */
+ }
+ } } }
+ } /* for (cur = 1; cur <= last_pos; cur++) */
+
+ lastSequence = opt[last_pos];
+ cur = last_pos > ZSTD_totalLen(lastSequence) ? last_pos - ZSTD_totalLen(lastSequence) : 0; /* single sequence, and it starts before `ip` */
+ assert(cur < ZSTD_OPT_NUM); /* control overflow*/
+
+_shortestPath: /* cur, last_pos, best_mlen, best_off have to be set */
+ assert(opt[0].mlen == 0);
+
+ /* Set the next chunk's repcodes based on the repcodes of the beginning
+ * of the last match, and the last sequence. This avoids us having to
+ * update them while traversing the sequences.
+ */
+ if (lastSequence.mlen != 0) {
+ repcodes_t const reps = ZSTD_newRep(opt[cur].rep, lastSequence.off, lastSequence.litlen==0);
+ ZSTD_memcpy(rep, &reps, sizeof(reps));
+ } else {
+ ZSTD_memcpy(rep, opt[cur].rep, sizeof(repcodes_t));
+ }
+
+ { U32 const storeEnd = cur + 1;
+ U32 storeStart = storeEnd;
+ U32 seqPos = cur;
+
+ DEBUGLOG(6, "start reverse traversal (last_pos:%u, cur:%u)",
+ last_pos, cur); (void)last_pos;
+ assert(storeEnd < ZSTD_OPT_NUM);
+ DEBUGLOG(6, "last sequence copied into pos=%u (llen=%u,mlen=%u,ofc=%u)",
+ storeEnd, lastSequence.litlen, lastSequence.mlen, lastSequence.off);
+ opt[storeEnd] = lastSequence;
+ while (seqPos > 0) {
+ U32 const backDist = ZSTD_totalLen(opt[seqPos]);
+ storeStart--;
+ DEBUGLOG(6, "sequence from rPos=%u copied into pos=%u (llen=%u,mlen=%u,ofc=%u)",
+ seqPos, storeStart, opt[seqPos].litlen, opt[seqPos].mlen, opt[seqPos].off);
+ opt[storeStart] = opt[seqPos];
+ seqPos = (seqPos > backDist) ? seqPos - backDist : 0;
+ }
+
+ /* save sequences */
+ DEBUGLOG(6, "sending selected sequences into seqStore")
+ { U32 storePos;
+ for (storePos=storeStart; storePos <= storeEnd; storePos++) {
+ U32 const llen = opt[storePos].litlen;
+ U32 const mlen = opt[storePos].mlen;
+ U32 const offCode = opt[storePos].off;
+ U32 const advance = llen + mlen;
+ DEBUGLOG(6, "considering seq starting at %zi, llen=%u, mlen=%u",
+ anchor - istart, (unsigned)llen, (unsigned)mlen);
+
+ if (mlen==0) { /* only literals => must be last "sequence", actually starting a new stream of sequences */
+ assert(storePos == storeEnd); /* must be last sequence */
+ ip = anchor + llen; /* last "sequence" is a bunch of literals => don't progress anchor */
+ continue; /* will finish */
+ }
+
+ assert(anchor + llen <= iend);
+ ZSTD_updateStats(optStatePtr, llen, anchor, offCode, mlen);
+ ZSTD_storeSeq(seqStore, llen, anchor, iend, offCode, mlen);
+ anchor += advance;
+ ip = anchor;
+ } }
+ ZSTD_setBasePrices(optStatePtr, optLevel);
+ }
+ } /* while (ip < ilimit) */
+
+ /* Return the last literals size */
+ return (size_t)(iend - anchor);
+}
+
+static size_t ZSTD_compressBlock_opt0(
+ ZSTD_matchState_t* ms, seqStore_t* seqStore, U32 rep[ZSTD_REP_NUM],
+ const void* src, size_t srcSize, const ZSTD_dictMode_e dictMode)
+{
+ return ZSTD_compressBlock_opt_generic(ms, seqStore, rep, src, srcSize, 0 /* optLevel */, dictMode);
+}
+
+static size_t ZSTD_compressBlock_opt2(
+ ZSTD_matchState_t* ms, seqStore_t* seqStore, U32 rep[ZSTD_REP_NUM],
+ const void* src, size_t srcSize, const ZSTD_dictMode_e dictMode)
+{
+ return ZSTD_compressBlock_opt_generic(ms, seqStore, rep, src, srcSize, 2 /* optLevel */, dictMode);
+}
+
+size_t ZSTD_compressBlock_btopt(
+ ZSTD_matchState_t* ms, seqStore_t* seqStore, U32 rep[ZSTD_REP_NUM],
+ const void* src, size_t srcSize)
+{
+ DEBUGLOG(5, "ZSTD_compressBlock_btopt");
+ return ZSTD_compressBlock_opt0(ms, seqStore, rep, src, srcSize, ZSTD_noDict);
+}
+
+
+
+
+/* ZSTD_initStats_ultra():
+ * make a first compression pass, just to seed stats with more accurate starting values.
+ * only works on first block, with no dictionary and no ldm.
+ * this function cannot error, hence its contract must be respected.
+ */
+static void
+ZSTD_initStats_ultra(ZSTD_matchState_t* ms,
+ seqStore_t* seqStore,
+ U32 rep[ZSTD_REP_NUM],
+ const void* src, size_t srcSize)
+{
+ U32 tmpRep[ZSTD_REP_NUM]; /* updated rep codes will sink here */
+ ZSTD_memcpy(tmpRep, rep, sizeof(tmpRep));
+
+ DEBUGLOG(4, "ZSTD_initStats_ultra (srcSize=%zu)", srcSize);
+ assert(ms->opt.litLengthSum == 0); /* first block */
+ assert(seqStore->sequences == seqStore->sequencesStart); /* no ldm */
+ assert(ms->window.dictLimit == ms->window.lowLimit); /* no dictionary */
+ assert(ms->window.dictLimit - ms->nextToUpdate <= 1); /* no prefix (note: intentional overflow, defined as 2-complement) */
+
+ ZSTD_compressBlock_opt2(ms, seqStore, tmpRep, src, srcSize, ZSTD_noDict); /* generate stats into ms->opt*/
+
+ /* invalidate first scan from history */
+ ZSTD_resetSeqStore(seqStore);
+ ms->window.base -= srcSize;
+ ms->window.dictLimit += (U32)srcSize;
+ ms->window.lowLimit = ms->window.dictLimit;
+ ms->nextToUpdate = ms->window.dictLimit;
+
+}
+
+size_t ZSTD_compressBlock_btultra(
+ ZSTD_matchState_t* ms, seqStore_t* seqStore, U32 rep[ZSTD_REP_NUM],
+ const void* src, size_t srcSize)
+{
+ DEBUGLOG(5, "ZSTD_compressBlock_btultra (srcSize=%zu)", srcSize);
+ return ZSTD_compressBlock_opt2(ms, seqStore, rep, src, srcSize, ZSTD_noDict);
+}
+
+size_t ZSTD_compressBlock_btultra2(
+ ZSTD_matchState_t* ms, seqStore_t* seqStore, U32 rep[ZSTD_REP_NUM],
+ const void* src, size_t srcSize)
+{
+ U32 const curr = (U32)((const BYTE*)src - ms->window.base);
+ DEBUGLOG(5, "ZSTD_compressBlock_btultra2 (srcSize=%zu)", srcSize);
+
+ /* 2-pass strategy:
+ * this strategy makes a first pass over first block to collect statistics
+ * and seed next round's statistics with it.
+ * After 1st pass, function forgets everything, and starts a new block.
+ * Consequently, this can only work if no data has been previously loaded in tables,
+ * aka, no dictionary, no prefix, no ldm preprocessing.
+ * The compression ratio gain is generally small (~0.5% on first block),
+ * the cost is 2x cpu time on first block. */
+ assert(srcSize <= ZSTD_BLOCKSIZE_MAX);
+ if ( (ms->opt.litLengthSum==0) /* first block */
+ && (seqStore->sequences == seqStore->sequencesStart) /* no ldm */
+ && (ms->window.dictLimit == ms->window.lowLimit) /* no dictionary */
+ && (curr == ms->window.dictLimit) /* start of frame, nothing already loaded nor skipped */
+ && (srcSize > ZSTD_PREDEF_THRESHOLD)
+ ) {
+ ZSTD_initStats_ultra(ms, seqStore, rep, src, srcSize);
+ }
+
+ return ZSTD_compressBlock_opt2(ms, seqStore, rep, src, srcSize, ZSTD_noDict);
+}
+
+size_t ZSTD_compressBlock_btopt_dictMatchState(
+ ZSTD_matchState_t* ms, seqStore_t* seqStore, U32 rep[ZSTD_REP_NUM],
+ const void* src, size_t srcSize)
+{
+ return ZSTD_compressBlock_opt0(ms, seqStore, rep, src, srcSize, ZSTD_dictMatchState);
+}
+
+size_t ZSTD_compressBlock_btultra_dictMatchState(
+ ZSTD_matchState_t* ms, seqStore_t* seqStore, U32 rep[ZSTD_REP_NUM],
+ const void* src, size_t srcSize)
+{
+ return ZSTD_compressBlock_opt2(ms, seqStore, rep, src, srcSize, ZSTD_dictMatchState);
+}
+
+size_t ZSTD_compressBlock_btopt_extDict(
+ ZSTD_matchState_t* ms, seqStore_t* seqStore, U32 rep[ZSTD_REP_NUM],
+ const void* src, size_t srcSize)
+{
+ return ZSTD_compressBlock_opt0(ms, seqStore, rep, src, srcSize, ZSTD_extDict);
+}
+
+size_t ZSTD_compressBlock_btultra_extDict(
+ ZSTD_matchState_t* ms, seqStore_t* seqStore, U32 rep[ZSTD_REP_NUM],
+ const void* src, size_t srcSize)
+{
+ return ZSTD_compressBlock_opt2(ms, seqStore, rep, src, srcSize, ZSTD_extDict);
+}
+
+/* note : no btultra2 variant for extDict nor dictMatchState,
+ * because btultra2 is not meant to work with dictionaries
+ * and is only specific for the first block (no prefix) */
diff --git a/lib/zstd/compress/zstd_opt.h b/lib/zstd/compress/zstd_opt.h
new file mode 100644
index 0000000000..22b862858b
--- /dev/null
+++ b/lib/zstd/compress/zstd_opt.h
@@ -0,0 +1,50 @@
+/*
+ * Copyright (c) Yann Collet, Facebook, Inc.
+ * All rights reserved.
+ *
+ * This source code is licensed under both the BSD-style license (found in the
+ * LICENSE file in the root directory of this source tree) and the GPLv2 (found
+ * in the COPYING file in the root directory of this source tree).
+ * You may select, at your option, one of the above-listed licenses.
+ */
+
+#ifndef ZSTD_OPT_H
+#define ZSTD_OPT_H
+
+
+#include "zstd_compress_internal.h"
+
+/* used in ZSTD_loadDictionaryContent() */
+void ZSTD_updateTree(ZSTD_matchState_t* ms, const BYTE* ip, const BYTE* iend);
+
+size_t ZSTD_compressBlock_btopt(
+ ZSTD_matchState_t* ms, seqStore_t* seqStore, U32 rep[ZSTD_REP_NUM],
+ void const* src, size_t srcSize);
+size_t ZSTD_compressBlock_btultra(
+ ZSTD_matchState_t* ms, seqStore_t* seqStore, U32 rep[ZSTD_REP_NUM],
+ void const* src, size_t srcSize);
+size_t ZSTD_compressBlock_btultra2(
+ ZSTD_matchState_t* ms, seqStore_t* seqStore, U32 rep[ZSTD_REP_NUM],
+ void const* src, size_t srcSize);
+
+
+size_t ZSTD_compressBlock_btopt_dictMatchState(
+ ZSTD_matchState_t* ms, seqStore_t* seqStore, U32 rep[ZSTD_REP_NUM],
+ void const* src, size_t srcSize);
+size_t ZSTD_compressBlock_btultra_dictMatchState(
+ ZSTD_matchState_t* ms, seqStore_t* seqStore, U32 rep[ZSTD_REP_NUM],
+ void const* src, size_t srcSize);
+
+size_t ZSTD_compressBlock_btopt_extDict(
+ ZSTD_matchState_t* ms, seqStore_t* seqStore, U32 rep[ZSTD_REP_NUM],
+ void const* src, size_t srcSize);
+size_t ZSTD_compressBlock_btultra_extDict(
+ ZSTD_matchState_t* ms, seqStore_t* seqStore, U32 rep[ZSTD_REP_NUM],
+ void const* src, size_t srcSize);
+
+ /* note : no btultra2 variant for extDict nor dictMatchState,
+ * because btultra2 is not meant to work with dictionaries
+ * and is only specific for the first block (no prefix) */
+
+
+#endif /* ZSTD_OPT_H */