summaryrefslogtreecommitdiffstats
path: root/src/zstd/lib/compress
diff options
context:
space:
mode:
Diffstat (limited to 'src/zstd/lib/compress')
-rw-r--r--src/zstd/lib/compress/fse_compress.c698
-rw-r--r--src/zstd/lib/compress/hist.c183
-rw-r--r--src/zstd/lib/compress/hist.h75
-rw-r--r--src/zstd/lib/compress/huf_compress.c798
-rw-r--r--src/zstd/lib/compress/zstd_compress.c4278
-rw-r--r--src/zstd/lib/compress/zstd_compress_internal.h1125
-rw-r--r--src/zstd/lib/compress/zstd_compress_literals.c158
-rw-r--r--src/zstd/lib/compress/zstd_compress_literals.h29
-rw-r--r--src/zstd/lib/compress/zstd_compress_sequences.c419
-rw-r--r--src/zstd/lib/compress/zstd_compress_sequences.h54
-rw-r--r--src/zstd/lib/compress/zstd_compress_superblock.c845
-rw-r--r--src/zstd/lib/compress/zstd_compress_superblock.h32
-rw-r--r--src/zstd/lib/compress/zstd_cwksp.h525
-rw-r--r--src/zstd/lib/compress/zstd_double_fast.c521
-rw-r--r--src/zstd/lib/compress/zstd_double_fast.h38
-rw-r--r--src/zstd/lib/compress/zstd_fast.c496
-rw-r--r--src/zstd/lib/compress/zstd_fast.h37
-rw-r--r--src/zstd/lib/compress/zstd_lazy.c1138
-rw-r--r--src/zstd/lib/compress/zstd_lazy.h67
-rw-r--r--src/zstd/lib/compress/zstd_ldm.c619
-rw-r--r--src/zstd/lib/compress/zstd_ldm.h110
-rw-r--r--src/zstd/lib/compress/zstd_opt.c1200
-rw-r--r--src/zstd/lib/compress/zstd_opt.h56
-rw-r--r--src/zstd/lib/compress/zstdmt_compress.c2143
-rw-r--r--src/zstd/lib/compress/zstdmt_compress.h192
25 files changed, 15836 insertions, 0 deletions
diff --git a/src/zstd/lib/compress/fse_compress.c b/src/zstd/lib/compress/fse_compress.c
new file mode 100644
index 000000000..a42759814
--- /dev/null
+++ b/src/zstd/lib/compress/fse_compress.c
@@ -0,0 +1,698 @@
+/* ******************************************************************
+ * FSE : Finite State Entropy encoder
+ * Copyright (c) 2013-2020, 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 <stdlib.h> /* malloc, free, qsort */
+#include <string.h> /* memcpy, memset */
+#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"
+
+
+/* **************************************************************
+* 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 cumul[FSE_MAX_SYMBOL_VALUE+2];
+
+ FSE_FUNCTION_TYPE* const tableSymbol = (FSE_FUNCTION_TYPE*)workSpace;
+ U32 highThreshold = tableSize-1;
+
+ /* CTable header */
+ if (((size_t)1 << tableLog) * sizeof(FSE_FUNCTION_TYPE) > wkspSize) return ERROR(tableLog_tooLarge);
+ 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__
+ 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 <= maxSymbolValue+1; u++) {
+ if (normalizedCounter[u-1]==-1) { /* Low proba symbol */
+ cumul[u] = cumul[u-1] + 1;
+ tableSymbol[highThreshold--] = (FSE_FUNCTION_TYPE)(u-1);
+ } else {
+ cumul[u] = cumul[u-1] + normalizedCounter[u-1];
+ } }
+ cumul[maxSymbolValue+1] = tableSize+1;
+ }
+
+ /* Spread symbols */
+ { U32 position = 0;
+ U32 symbol;
+ for (symbol=0; symbol<=maxSymbolValue; 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);
+ symbolTT[s].deltaFindState = total - 1;
+ total ++;
+ break;
+ default :
+ {
+ U32 const maxBitsOut = tableLog - BIT_highbit32 (normalizedCounter[s]-1);
+ U32 const minStatePlus = normalizedCounter[s] << maxBitsOut;
+ symbolTT[s].deltaNbBits = (maxBitsOut << 16) - minStatePlus;
+ symbolTT[s].deltaFindState = total - normalizedCounter[s];
+ total += 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;
+}
+
+
+size_t FSE_buildCTable(FSE_CTable* ct, const short* normalizedCounter, unsigned maxSymbolValue, unsigned tableLog)
+{
+ FSE_FUNCTION_TYPE tableSymbol[FSE_MAX_TABLESIZE]; /* memset() is not necessary, even if static analyzer complain about it */
+ return FSE_buildCTable_wksp(ct, normalizedCounter, maxSymbolValue, tableLog, tableSymbol, sizeof(tableSymbol));
+}
+
+
+
+#ifndef FSE_COMMONDEFS_ONLY
+
+
+/*-**************************************************************
+* FSE NCount encoding
+****************************************************************/
+size_t FSE_NCountWriteBound(unsigned maxSymbolValue, unsigned tableLog)
+{
+ size_t const maxHeaderSize = (((maxSymbolValue+1) * tableLog) >> 3) + 3;
+ 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*)malloc(size);
+}
+
+void FSE_freeCTable (FSE_CTable* ct) { 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 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] = -1;
+ 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 = ((((U64)1<<vStepLog) * ToDistribute) + mid) / 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)
+{
+ /* 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 };
+ U64 const scale = 62 - tableLog;
+ U64 const step = ((U64)1<<62) / 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] = -1;
+ 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);
+ 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); }
+
+/* FSE_compress_wksp() :
+ * Same as FSE_compress2(), but using an externally allocated scratch buffer (`workSpace`).
+ * `wkspSize` size must be `(1<<tableLog)`.
+ */
+size_t FSE_compress_wksp (void* dst, size_t dstSize, const void* src, size_t srcSize, unsigned maxSymbolValue, unsigned tableLog, void* workSpace, size_t wkspSize)
+{
+ BYTE* const ostart = (BYTE*) dst;
+ BYTE* op = ostart;
+ BYTE* const oend = ostart + dstSize;
+
+ unsigned count[FSE_MAX_SYMBOL_VALUE+1];
+ S16 norm[FSE_MAX_SYMBOL_VALUE+1];
+ FSE_CTable* CTable = (FSE_CTable*)workSpace;
+ size_t const CTableSize = FSE_CTABLE_SIZE_U32(tableLog, maxSymbolValue);
+ void* scratchBuffer = (void*)(CTable + CTableSize);
+ size_t const scratchBufferSize = wkspSize - (CTableSize * sizeof(FSE_CTable));
+
+ /* init conditions */
+ if (wkspSize < FSE_WKSP_SIZE_U32(tableLog, maxSymbolValue)) return ERROR(tableLog_tooLarge);
+ if (srcSize <= 1) return 0; /* Not compressible */
+ if (!maxSymbolValue) maxSymbolValue = FSE_MAX_SYMBOL_VALUE;
+ if (!tableLog) tableLog = FSE_DEFAULT_TABLELOG;
+
+ /* Scan input and build symbol stats */
+ { CHECK_V_F(maxCount, HIST_count_wksp(count, &maxSymbolValue, src, srcSize, scratchBuffer, scratchBufferSize) );
+ if (maxCount == srcSize) return 1; /* only a single symbol in src : rle */
+ if (maxCount == 1) return 0; /* each symbol present maximum once => not compressible */
+ if (maxCount < (srcSize >> 7)) return 0; /* Heuristic : not compressible enough */
+ }
+
+ tableLog = FSE_optimalTableLog(tableLog, srcSize, maxSymbolValue);
+ CHECK_F( FSE_normalizeCount(norm, tableLog, count, srcSize, maxSymbolValue) );
+
+ /* Write table description header */
+ { CHECK_V_F(nc_err, FSE_writeNCount(op, oend-op, norm, maxSymbolValue, tableLog) );
+ op += nc_err;
+ }
+
+ /* Compress */
+ CHECK_F( FSE_buildCTable_wksp(CTable, norm, maxSymbolValue, tableLog, scratchBuffer, scratchBufferSize) );
+ { CHECK_V_F(cSize, FSE_compress_usingCTable(op, oend - op, src, srcSize, CTable) );
+ if (cSize == 0) return 0; /* not enough space for compressed data */
+ op += cSize;
+ }
+
+ /* check compressibility */
+ if ( (size_t)(op-ostart) >= srcSize-1 ) return 0;
+
+ return op-ostart;
+}
+
+typedef struct {
+ FSE_CTable CTable_max[FSE_CTABLE_SIZE_U32(FSE_MAX_TABLELOG, FSE_MAX_SYMBOL_VALUE)];
+ BYTE scratchBuffer[1 << FSE_MAX_TABLELOG];
+} fseWkspMax_t;
+
+size_t FSE_compress2 (void* dst, size_t dstCapacity, const void* src, size_t srcSize, unsigned maxSymbolValue, unsigned tableLog)
+{
+ fseWkspMax_t scratchBuffer;
+ DEBUG_STATIC_ASSERT(sizeof(scratchBuffer) >= FSE_WKSP_SIZE_U32(FSE_MAX_TABLELOG, FSE_MAX_SYMBOL_VALUE)); /* compilation failures here means scratchBuffer is not large enough */
+ if (tableLog > FSE_MAX_TABLELOG) return ERROR(tableLog_tooLarge);
+ return FSE_compress_wksp(dst, dstCapacity, src, srcSize, maxSymbolValue, tableLog, &scratchBuffer, sizeof(scratchBuffer));
+}
+
+size_t FSE_compress (void* dst, size_t dstCapacity, const void* src, size_t srcSize)
+{
+ return FSE_compress2(dst, dstCapacity, src, srcSize, FSE_MAX_SYMBOL_VALUE, FSE_DEFAULT_TABLELOG);
+}
+
+
+#endif /* FSE_COMMONDEFS_ONLY */
diff --git a/src/zstd/lib/compress/hist.c b/src/zstd/lib/compress/hist.c
new file mode 100644
index 000000000..61e08c796
--- /dev/null
+++ b/src/zstd/lib/compress/hist.c
@@ -0,0 +1,183 @@
+/* ******************************************************************
+ * hist : Histogram functions
+ * part of Finite State Entropy project
+ * Copyright (c) 2013-2020, 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;
+
+ 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` size must be a table of size >= HIST_WKSP_SIZE_U32.
+ * @return : largest histogram frequency,
+ * or an error code (notably when histogram would be 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;
+ unsigned maxSymbolValue = *maxSymbolValuePtr;
+ unsigned max=0;
+ U32* const Counting1 = workSpace;
+ U32* const Counting2 = Counting1 + 256;
+ U32* const Counting3 = Counting2 + 256;
+ U32* const Counting4 = Counting3 + 256;
+
+ memset(workSpace, 0, 4*256*sizeof(unsigned));
+
+ /* safety checks */
+ if (!sourceSize) {
+ memset(count, 0, maxSymbolValue + 1);
+ *maxSymbolValuePtr = 0;
+ return 0;
+ }
+ if (!maxSymbolValue) maxSymbolValue = 255; /* 0 == default */
+
+ /* 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++]++;
+
+ if (check) { /* verify stats will fit into destination table */
+ U32 s; for (s=255; s>maxSymbolValue; s--) {
+ Counting1[s] += Counting2[s] + Counting3[s] + Counting4[s];
+ if (Counting1[s]) return ERROR(maxSymbolValue_tooSmall);
+ } }
+
+ { U32 s;
+ if (maxSymbolValue > 255) maxSymbolValue = 255;
+ for (s=0; s<=maxSymbolValue; s++) {
+ count[s] = Counting1[s] + Counting2[s] + Counting3[s] + Counting4[s];
+ if (count[s] > max) max = count[s];
+ } }
+
+ while (!count[maxSymbolValue]) maxSymbolValue--;
+ *maxSymbolValuePtr = maxSymbolValue;
+ 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);
+}
+
+/* fast variant (unsafe : won't check if src contains values beyond count[] limit) */
+size_t HIST_countFast(unsigned* count, unsigned* maxSymbolValuePtr,
+ const void* source, size_t sourceSize)
+{
+ unsigned tmpCounters[HIST_WKSP_SIZE_U32];
+ return HIST_countFast_wksp(count, maxSymbolValuePtr, source, sourceSize, tmpCounters, sizeof(tmpCounters));
+}
+
+/* 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);
+}
+
+size_t HIST_count(unsigned* count, unsigned* maxSymbolValuePtr,
+ const void* src, size_t srcSize)
+{
+ unsigned tmpCounters[HIST_WKSP_SIZE_U32];
+ return HIST_count_wksp(count, maxSymbolValuePtr, src, srcSize, tmpCounters, sizeof(tmpCounters));
+}
diff --git a/src/zstd/lib/compress/hist.h b/src/zstd/lib/compress/hist.h
new file mode 100644
index 000000000..77e3ec4fb
--- /dev/null
+++ b/src/zstd/lib/compress/hist.h
@@ -0,0 +1,75 @@
+/* ******************************************************************
+ * hist : Histogram functions
+ * part of Finite State Entropy project
+ * Copyright (c) 2013-2020, 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 <stddef.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/src/zstd/lib/compress/huf_compress.c b/src/zstd/lib/compress/huf_compress.c
new file mode 100644
index 000000000..546879868
--- /dev/null
+++ b/src/zstd/lib/compress/huf_compress.c
@@ -0,0 +1,798 @@
+/* ******************************************************************
+ * Huffman encoder, part of New Generation Entropy library
+ * Copyright (c) 2013-2020, 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
+****************************************************************/
+#ifdef _MSC_VER /* Visual Studio */
+# pragma warning(disable : 4127) /* disable: C4127: conditional expression is constant */
+#endif
+
+
+/* **************************************************************
+* Includes
+****************************************************************/
+#include <string.h> /* memcpy, memset */
+#include <stdio.h> /* printf (debug) */
+#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
+*********************************************************/
+/* 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
+static size_t HUF_compressWeights (void* dst, size_t dstSize, const void* weightTable, size_t wtSize)
+{
+ 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;
+
+ FSE_CTable CTable[FSE_CTABLE_SIZE_U32(MAX_FSE_TABLELOG_FOR_HUFF_HEADER, HUF_TABLELOG_MAX)];
+ BYTE scratchBuffer[1<<MAX_FSE_TABLELOG_FOR_HUFF_HEADER];
+
+ unsigned count[HUF_TABLELOG_MAX+1];
+ S16 norm[HUF_TABLELOG_MAX+1];
+
+ /* init conditions */
+ if (wtSize <= 1) return 0; /* Not compressible */
+
+ /* Scan input and build symbol stats */
+ { unsigned const maxCount = HIST_count_simple(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(norm, tableLog, count, wtSize, maxSymbolValue) );
+
+ /* Write table description header */
+ { CHECK_V_F(hSize, FSE_writeNCount(op, (size_t)(oend-op), norm, maxSymbolValue, tableLog) );
+ op += hSize;
+ }
+
+ /* Compress */
+ CHECK_F( FSE_buildCTable_wksp(CTable, norm, maxSymbolValue, tableLog, scratchBuffer, sizeof(scratchBuffer)) );
+ { CHECK_V_F(cSize, FSE_compress_usingCTable(op, (size_t)(oend - op), weightTable, wtSize, CTable) );
+ if (cSize == 0) return 0; /* not enough space for compressed data */
+ op += cSize;
+ }
+
+ return (size_t)(op-ostart);
+}
+
+
+struct HUF_CElt_s {
+ U16 val;
+ BYTE nbBits;
+}; /* typedef'd to HUF_CElt within "huf.h" */
+
+/*! 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)
+{
+ BYTE bitsToWeight[HUF_TABLELOG_MAX + 1]; /* precomputed conversion table */
+ BYTE huffWeight[HUF_SYMBOLVALUE_MAX];
+ BYTE* op = (BYTE*)dst;
+ U32 n;
+
+ /* check conditions */
+ if (maxSymbolValue > HUF_SYMBOLVALUE_MAX) return ERROR(maxSymbolValue_tooLarge);
+
+ /* convert to weight */
+ bitsToWeight[0] = 0;
+ for (n=1; n<huffLog+1; n++)
+ bitsToWeight[n] = (BYTE)(huffLog + 1 - n);
+ for (n=0; n<maxSymbolValue; n++)
+ huffWeight[n] = bitsToWeight[CTable[n].nbBits];
+
+ /* attempt weights compression by FSE */
+ { CHECK_V_F(hSize, HUF_compressWeights(op+1, maxDstSize-1, huffWeight, maxSymbolValue) );
+ 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));
+ 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)((huffWeight[n] << 4) + huffWeight[n+1]);
+ return ((maxSymbolValue+1)/2) + 1;
+}
+
+
+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;
+
+ /* get symbol weights */
+ CHECK_V_F(readSize, HUF_readStats(huffWeight, HUF_SYMBOLVALUE_MAX+1, rankVal, &nbSymbols, &tableLog, src, srcSize));
+
+ /* check result */
+ if (tableLog > HUF_TABLELOG_MAX) return ERROR(tableLog_tooLarge);
+ if (nbSymbols > *maxSymbolValuePtr+1) return ERROR(maxSymbolValue_tooSmall);
+
+ /* Prepare base value per rank */
+ { U32 n, nextRankStart = 0;
+ for (n=1; n<=tableLog; n++) {
+ U32 current = nextRankStart;
+ nextRankStart += (rankVal[n] << (n-1));
+ rankVal[n] = current;
+ } }
+
+ /* fill nbBits */
+ *hasZeroWeights = 0;
+ { U32 n; for (n=0; n<nbSymbols; n++) {
+ const U32 w = huffWeight[n];
+ *hasZeroWeights |= (w == 0);
+ CTable[n].nbBits = (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[CTable[n].nbBits]++; }
+ /* 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++) CTable[n].val = valPerRank[CTable[n].nbBits]++; }
+ }
+
+ *maxSymbolValuePtr = nbSymbols - 1;
+ return readSize;
+}
+
+U32 HUF_getNbBits(const void* symbolTable, U32 symbolValue)
+{
+ const HUF_CElt* table = (const HUF_CElt*)symbolTable;
+ assert(symbolValue <= HUF_SYMBOLVALUE_MAX);
+ return table[symbolValue].nbBits;
+}
+
+
+typedef struct nodeElt_s {
+ U32 count;
+ U16 parent;
+ BYTE byte;
+ BYTE nbBits;
+} nodeElt;
+
+static U32 HUF_setMaxHeight(nodeElt* huffNode, U32 lastNonNull, U32 maxNbBits)
+{
+ const U32 largestBits = huffNode[lastNonNull].nbBits;
+ if (largestBits <= maxNbBits) return largestBits; /* early exit : no elt > maxNbBits */
+
+ /* there are several too large elements (at least >= 2) */
+ { int totalCost = 0;
+ const U32 baseCost = 1 << (largestBits - maxNbBits);
+ int n = (int)lastNonNull;
+
+ 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 */
+ while (huffNode[n].nbBits == maxNbBits) n--; /* n end at index of smallest symbol using < maxNbBits */
+
+ /* renorm totalCost */
+ totalCost >>= (largestBits - maxNbBits); /* note : totalCost is necessarily a multiple of baseCost */
+
+ /* repay normalized cost */
+ { U32 const noSymbol = 0xF0F0F0F0;
+ U32 rankLast[HUF_TABLELOG_MAX+2];
+
+ /* Get pos of last (smallest) symbol per rank */
+ 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) {
+ 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;
+ 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 !) */
+ /* HUF_MAX_TABLELOG test just to please gcc 5+; but it should not be necessary */
+ while ((nBitsToDecrease<=HUF_TABLELOG_MAX) && (rankLast[nBitsToDecrease] == noSymbol))
+ nBitsToDecrease ++;
+ totalCost -= 1 << (nBitsToDecrease-1);
+ if (rankLast[nBitsToDecrease-1] == noSymbol)
+ rankLast[nBitsToDecrease-1] = rankLast[nBitsToDecrease]; /* this rank is no longer empty */
+ huffNode[rankLast[nBitsToDecrease]].nbBits ++;
+ 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) */
+
+ while (totalCost < 0) { /* Sometimes, cost correction overshoot */
+ if (rankLast[1] == noSymbol) { /* special case : no rank 1 symbol (using maxNbBits-1); let's create one from largest rank 0 (using maxNbBits) */
+ 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 ++;
+ } } } /* there are several too large elements (at least >= 2) */
+
+ return maxNbBits;
+}
+
+typedef struct {
+ U32 base;
+ U32 current;
+} rankPos;
+
+typedef nodeElt huffNodeTable[HUF_CTABLE_WORKSPACE_SIZE_U32];
+
+#define RANK_POSITION_TABLE_SIZE 32
+
+typedef struct {
+ huffNodeTable huffNodeTbl;
+ rankPos rankPosition[RANK_POSITION_TABLE_SIZE];
+} HUF_buildCTable_wksp_tables;
+
+static void HUF_sort(nodeElt* huffNode, const unsigned* count, U32 maxSymbolValue, rankPos* rankPosition)
+{
+ U32 n;
+
+ memset(rankPosition, 0, sizeof(*rankPosition) * RANK_POSITION_TABLE_SIZE);
+ for (n=0; n<=maxSymbolValue; n++) {
+ U32 r = BIT_highbit32(count[n] + 1);
+ rankPosition[r].base ++;
+ }
+ for (n=30; n>0; n--) rankPosition[n-1].base += rankPosition[n].base;
+ for (n=0; n<32; n++) rankPosition[n].current = rankPosition[n].base;
+ for (n=0; n<=maxSymbolValue; n++) {
+ U32 const c = count[n];
+ U32 const r = BIT_highbit32(c+1) + 1;
+ U32 pos = rankPosition[r].current++;
+ while ((pos > rankPosition[r].base) && (c > huffNode[pos-1].count)) {
+ huffNode[pos] = huffNode[pos-1];
+ pos--;
+ }
+ huffNode[pos].count = c;
+ huffNode[pos].byte = (BYTE)n;
+ }
+}
+
+
+/** 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)
+
+size_t HUF_buildCTable_wksp (HUF_CElt* tree, const unsigned* count, U32 maxSymbolValue, U32 maxNbBits, void* workSpace, size_t wkspSize)
+{
+ HUF_buildCTable_wksp_tables* const wksp_tables = (HUF_buildCTable_wksp_tables*)workSpace;
+ nodeElt* const huffNode0 = wksp_tables->huffNodeTbl;
+ nodeElt* const huffNode = huffNode0+1;
+ int nonNullRank;
+ int lowS, lowN;
+ int nodeNb = STARTNODE;
+ int n, nodeRoot;
+
+ /* safety checks */
+ if (((size_t)workSpace & 3) != 0) return ERROR(GENERIC); /* must be aligned on 4-bytes boundaries */
+ 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);
+ memset(huffNode0, 0, sizeof(huffNodeTable));
+
+ /* sort, decreasing order */
+ HUF_sort(huffNode, count, maxSymbolValue, wksp_tables->rankPosition);
+
+ /* 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;
+
+ /* enforce maxTableLog */
+ maxNbBits = HUF_setMaxHeight(huffNode, (U32)nonNullRank, maxNbBits);
+
+ /* fill result into tree (val, nbBits) */
+ { U16 nbPerRank[HUF_TABLELOG_MAX+1] = {0};
+ U16 valPerRank[HUF_TABLELOG_MAX+1] = {0};
+ int const alphabetSize = (int)(maxSymbolValue + 1);
+ if (maxNbBits > HUF_TABLELOG_MAX) return ERROR(GENERIC); /* check fit into table */
+ for (n=0; n<=nonNullRank; n++)
+ nbPerRank[huffNode[n].nbBits]++;
+ /* determine stating 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++)
+ tree[huffNode[n].byte].nbBits = huffNode[n].nbBits; /* push nbBits per symbol, symbol order */
+ for (n=0; n<alphabetSize; n++)
+ tree[n].val = valPerRank[tree[n].nbBits]++; /* assign value within rank, symbol order */
+ }
+
+ return maxNbBits;
+}
+
+/** HUF_buildCTable() :
+ * @return : maxNbBits
+ * Note : count is used before tree is written, so they can safely overlap
+ */
+size_t HUF_buildCTable (HUF_CElt* tree, const unsigned* count, unsigned maxSymbolValue, unsigned maxNbBits)
+{
+ HUF_buildCTable_wksp_tables workspace;
+ return HUF_buildCTable_wksp(tree, count, maxSymbolValue, maxNbBits, &workspace, sizeof(workspace));
+}
+
+size_t HUF_estimateCompressedSize(const HUF_CElt* CTable, const unsigned* count, unsigned maxSymbolValue)
+{
+ size_t nbBits = 0;
+ int s;
+ for (s = 0; s <= (int)maxSymbolValue; ++s) {
+ nbBits += CTable[s].nbBits * count[s];
+ }
+ return nbBits >> 3;
+}
+
+int HUF_validateCTable(const HUF_CElt* CTable, const unsigned* count, unsigned maxSymbolValue) {
+ int bad = 0;
+ int s;
+ for (s = 0; s <= (int)maxSymbolValue; ++s) {
+ bad |= (count[s] != 0) & (CTable[s].nbBits == 0);
+ }
+ return !bad;
+}
+
+size_t HUF_compressBound(size_t size) { return HUF_COMPRESSBOUND(size); }
+
+FORCE_INLINE_TEMPLATE void
+HUF_encodeSymbol(BIT_CStream_t* bitCPtr, U32 symbol, const HUF_CElt* CTable)
+{
+ BIT_addBitsFast(bitCPtr, CTable[symbol].val, CTable[symbol].nbBits);
+}
+
+#define HUF_FLUSHBITS(s) BIT_flushBits(s)
+
+#define HUF_FLUSHBITS_1(stream) \
+ if (sizeof((stream)->bitContainer)*8 < HUF_TABLELOG_MAX*2+7) HUF_FLUSHBITS(stream)
+
+#define HUF_FLUSHBITS_2(stream) \
+ if (sizeof((stream)->bitContainer)*8 < HUF_TABLELOG_MAX*4+7) HUF_FLUSHBITS(stream)
+
+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)
+{
+ const BYTE* ip = (const BYTE*) src;
+ BYTE* const ostart = (BYTE*)dst;
+ BYTE* const oend = ostart + dstSize;
+ BYTE* op = ostart;
+ size_t n;
+ BIT_CStream_t bitC;
+
+ /* init */
+ if (dstSize < 8) return 0; /* not enough space to compress */
+ { size_t const initErr = BIT_initCStream(&bitC, op, (size_t)(oend-op));
+ if (HUF_isError(initErr)) return 0; }
+
+ n = srcSize & ~3; /* join to mod 4 */
+ switch (srcSize & 3)
+ {
+ case 3 : HUF_encodeSymbol(&bitC, ip[n+ 2], CTable);
+ HUF_FLUSHBITS_2(&bitC);
+ /* fall-through */
+ case 2 : HUF_encodeSymbol(&bitC, ip[n+ 1], CTable);
+ HUF_FLUSHBITS_1(&bitC);
+ /* fall-through */
+ case 1 : HUF_encodeSymbol(&bitC, ip[n+ 0], CTable);
+ HUF_FLUSHBITS(&bitC);
+ /* fall-through */
+ case 0 : /* fall-through */
+ default: break;
+ }
+
+ for (; n>0; n-=4) { /* note : n&3==0 at this stage */
+ HUF_encodeSymbol(&bitC, ip[n- 1], CTable);
+ HUF_FLUSHBITS_1(&bitC);
+ HUF_encodeSymbol(&bitC, ip[n- 2], CTable);
+ HUF_FLUSHBITS_2(&bitC);
+ HUF_encodeSymbol(&bitC, ip[n- 3], CTable);
+ HUF_FLUSHBITS_1(&bitC);
+ HUF_encodeSymbol(&bitC, ip[n- 4], CTable);
+ HUF_FLUSHBITS(&bitC);
+ }
+
+ return BIT_closeCStream(&bitC);
+}
+
+#if DYNAMIC_BMI2
+
+static TARGET_ATTRIBUTE("bmi2") 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_internal(dst, dstSize, src, srcSize, CTable, /* bmi2 */ 0);
+}
+
+
+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) return 0;
+ assert(cSize <= 65535);
+ 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) return 0;
+ assert(cSize <= 65535);
+ 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) return 0;
+ assert(cSize <= 65535);
+ 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) 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_internal(dst, dstSize, src, srcSize, CTable, /* bmi2 */ 0);
+}
+
+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_SYMBOLVALUE_MAX + 1];
+ HUF_buildCTable_wksp_tables buildCTable_wksp;
+} HUF_compress_tables_t;
+
+/* HUF_compress_internal() :
+ * `workSpace` must a table of at least HUF_WORKSPACE_SIZE_U32 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)
+{
+ HUF_compress_tables_t* const table = (HUF_compress_tables_t*)workSpace;
+ BYTE* const ostart = (BYTE*)dst;
+ BYTE* const oend = ostart + dstSize;
+ BYTE* op = ostart;
+
+ HUF_STATIC_ASSERT(sizeof(*table) <= HUF_WORKSPACE_SIZE);
+
+ /* checks & inits */
+ if (((size_t)workSpace & 3) != 0) return ERROR(GENERIC); /* must be aligned on 4-bytes boundaries */
+ if (wkspSize < HUF_WORKSPACE_SIZE) 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);
+ }
+
+ /* Scan input and build symbol stats */
+ { CHECK_V_F(largest, HIST_count_wksp (table->count, &maxSymbolValue, (const BYTE*)src, srcSize, workSpace, wkspSize) );
+ 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->buildCTable_wksp, sizeof(table->buildCTable_wksp));
+ CHECK_F(maxBits);
+ huffLog = (U32)maxBits;
+ /* Zero unused symbols in CTable, so we can check it for validity */
+ memset(table->CTable + (maxSymbolValue + 1), 0,
+ sizeof(table->CTable) - ((maxSymbolValue + 1) * sizeof(HUF_CElt)));
+ }
+
+ /* Write table description header */
+ { CHECK_V_F(hSize, HUF_writeCTable (op, dstSize, table->CTable, maxSymbolValue, huffLog) );
+ /* 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)
+ 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*/);
+}
+
+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)
+{
+ return HUF_compress_internal(dst, dstSize, src, srcSize,
+ maxSymbolValue, huffLog, HUF_singleStream,
+ workSpace, wkspSize, hufTable,
+ repeat, preferRepeat, bmi2);
+}
+
+size_t HUF_compress1X (void* dst, size_t dstSize,
+ const void* src, size_t srcSize,
+ unsigned maxSymbolValue, unsigned huffLog)
+{
+ unsigned workSpace[HUF_WORKSPACE_SIZE_U32];
+ return HUF_compress1X_wksp(dst, dstSize, src, srcSize, maxSymbolValue, huffLog, workSpace, sizeof(workSpace));
+}
+
+/* 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*/);
+}
+
+/* HUF_compress4X_repeat():
+ * compress input using 4 streams.
+ * 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)
+{
+ return HUF_compress_internal(dst, dstSize, src, srcSize,
+ maxSymbolValue, huffLog, HUF_fourStreams,
+ workSpace, wkspSize,
+ hufTable, repeat, preferRepeat, bmi2);
+}
+
+size_t HUF_compress2 (void* dst, size_t dstSize,
+ const void* src, size_t srcSize,
+ unsigned maxSymbolValue, unsigned huffLog)
+{
+ unsigned workSpace[HUF_WORKSPACE_SIZE_U32];
+ return HUF_compress4X_wksp(dst, dstSize, src, srcSize, maxSymbolValue, huffLog, workSpace, sizeof(workSpace));
+}
+
+size_t HUF_compress (void* dst, size_t maxDstSize, const void* src, size_t srcSize)
+{
+ return HUF_compress2(dst, maxDstSize, src, srcSize, 255, HUF_TABLELOG_DEFAULT);
+}
diff --git a/src/zstd/lib/compress/zstd_compress.c b/src/zstd/lib/compress/zstd_compress.c
new file mode 100644
index 000000000..3f963b1cf
--- /dev/null
+++ b/src/zstd/lib/compress/zstd_compress.c
@@ -0,0 +1,4278 @@
+/*
+ * Copyright (c) 2016-2020, 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 <limits.h> /* INT_MAX */
+#include <string.h> /* memset */
+#include "../common/cpu.h"
+#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"
+
+
+/*-*************************************
+* 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;
+ 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 */
+}; /* 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);
+ memset(cctx, 0, sizeof(*cctx));
+ cctx->customMem = memManager;
+ cctx->bmi2 = ZSTD_cpuid_bmi2(ZSTD_cpuid());
+ { 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_malloc(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);
+
+ cctx = (ZSTD_CCtx*)ZSTD_cwksp_reserve_object(&ws, sizeof(ZSTD_CCtx));
+ if (cctx == NULL) return NULL;
+
+ 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, HUF_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, HUF_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_free(cctx->localDict.dictBuffer, cctx->customMem);
+ ZSTD_freeCDict(cctx->localDict.cdict);
+ memset(&cctx->localDict, 0, sizeof(cctx->localDict));
+ 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);
+#ifdef ZSTD_MULTITHREAD
+ ZSTDMT_freeCCtx(cctx->mtctx); cctx->mtctx = NULL;
+#endif
+ 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_free(cctx, cctx->customMem);
+ }
+ }
+ return 0;
+}
+
+
+static size_t ZSTD_sizeof_mtctx(const ZSTD_CCtx* cctx)
+{
+#ifdef ZSTD_MULTITHREAD
+ return ZSTDMT_sizeof_CCtx(cctx->mtctx);
+#else
+ (void)cctx;
+ return 0;
+#endif
+}
+
+
+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); }
+
+static ZSTD_CCtx_params ZSTD_makeCCtxParamsFromCParams(
+ ZSTD_compressionParameters cParams)
+{
+ ZSTD_CCtx_params cctxParams;
+ memset(&cctxParams, 0, sizeof(cctxParams));
+ cctxParams.cParams = cParams;
+ cctxParams.compressionLevel = ZSTD_CLEVEL_DEFAULT; /* should not matter, as all cParams are presumed properly defined */
+ assert(!ZSTD_checkCParams(cParams));
+ cctxParams.fParams.contentSizeFlag = 1;
+ 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_calloc(
+ sizeof(ZSTD_CCtx_params), customMem);
+ if (!params) { return NULL; }
+ params->customMem = customMem;
+ params->compressionLevel = ZSTD_CLEVEL_DEFAULT;
+ params->fParams.contentSizeFlag = 1;
+ 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_free(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!");
+ memset(cctxParams, 0, sizeof(*cctxParams));
+ cctxParams->compressionLevel = compressionLevel;
+ cctxParams->fParams.contentSizeFlag = 1;
+ return 0;
+}
+
+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) , "");
+ memset(cctxParams, 0, sizeof(*cctxParams));
+ assert(!ZSTD_checkCParams(params.cParams));
+ cctxParams->cParams = params.cParams;
+ cctxParams->fParams = params.fParams;
+ cctxParams->compressionLevel = ZSTD_CLEVEL_DEFAULT; /* should not matter, as all cParams are presumed properly defined */
+ return 0;
+}
+
+/* ZSTD_assignParamsToCCtxParams() :
+ * params is presumed valid at this stage */
+static ZSTD_CCtx_params ZSTD_assignParamsToCCtxParams(
+ const ZSTD_CCtx_params* cctxParams, const ZSTD_parameters* params)
+{
+ ZSTD_CCtx_params ret = *cctxParams;
+ assert(!ZSTD_checkCParams(params->cParams));
+ ret.cParams = params->cParams;
+ ret.fParams = params->fParams;
+ ret.compressionLevel = ZSTD_CLEVEL_DEFAULT; /* should not matter, as all cParams are presumed properly defined */
+ return ret;
+}
+
+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;
+#ifdef ZSTD_MULTITHREAD
+ bounds.upperBound = ZSTDMT_NBWORKERS_MAX;
+#else
+ bounds.upperBound = 0;
+#endif
+ return bounds;
+
+ case ZSTD_c_jobSize:
+ bounds.lowerBound = 0;
+#ifdef ZSTD_MULTITHREAD
+ bounds.upperBound = ZSTDMT_JOBSIZE_MAX;
+#else
+ bounds.upperBound = 0;
+#endif
+ return bounds;
+
+ case ZSTD_c_overlapLog:
+#ifdef ZSTD_MULTITHREAD
+ bounds.lowerBound = ZSTD_OVERLAPLOG_MIN;
+ bounds.upperBound = ZSTD_OVERLAPLOG_MAX;
+#else
+ bounds.lowerBound = 0;
+ bounds.upperBound = 0;
+#endif
+ 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_dictForceCopy);
+ 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_lcm_auto < ZSTD_lcm_huffman && ZSTD_lcm_huffman < ZSTD_lcm_uncompressed);
+ bounds.lowerBound = ZSTD_lcm_auto;
+ bounds.upperBound = ZSTD_lcm_uncompressed;
+ 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;
+
+ 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_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:
+ 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_enableLongDistanceMatching:
+ case ZSTD_c_ldmHashLog:
+ case ZSTD_c_ldmMinMatch:
+ case ZSTD_c_ldmBucketSizeLog:
+ case ZSTD_c_targetCBlockSize:
+ case ZSTD_c_srcSizeHint:
+ 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 : does not change current level */
+ 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_literalCompressionMode_e lcm = (ZSTD_literalCompressionMode_e)value;
+ BOUNDCHECK(ZSTD_c_literalCompressionMode, lcm);
+ CCtxParams->literalCompressionMode = lcm;
+ return CCtxParams->literalCompressionMode;
+ }
+
+ case ZSTD_c_nbWorkers :
+#ifndef ZSTD_MULTITHREAD
+ RETURN_ERROR_IF(value!=0, parameter_unsupported, "not compiled with multithreading");
+ return 0;
+#else
+ FORWARD_IF_ERROR(ZSTD_cParam_clampBounds(param, &value), "");
+ CCtxParams->nbWorkers = value;
+ return CCtxParams->nbWorkers;
+#endif
+
+ case ZSTD_c_jobSize :
+#ifndef ZSTD_MULTITHREAD
+ RETURN_ERROR_IF(value!=0, parameter_unsupported, "not compiled with multithreading");
+ return 0;
+#else
+ /* Adjust to the minimum non-default value. */
+ if (value != 0 && value < ZSTDMT_JOBSIZE_MIN)
+ value = ZSTDMT_JOBSIZE_MIN;
+ FORWARD_IF_ERROR(ZSTD_cParam_clampBounds(param, &value), "");
+ assert(value >= 0);
+ CCtxParams->jobSize = value;
+ return CCtxParams->jobSize;
+#endif
+
+ case ZSTD_c_overlapLog :
+#ifndef ZSTD_MULTITHREAD
+ RETURN_ERROR_IF(value!=0, parameter_unsupported, "not compiled with multithreading");
+ return 0;
+#else
+ FORWARD_IF_ERROR(ZSTD_cParam_clampBounds(ZSTD_c_overlapLog, &value), "");
+ CCtxParams->overlapLog = value;
+ return CCtxParams->overlapLog;
+#endif
+
+ case ZSTD_c_rsyncable :
+#ifndef ZSTD_MULTITHREAD
+ RETURN_ERROR_IF(value!=0, parameter_unsupported, "not compiled with multithreading");
+ return 0;
+#else
+ FORWARD_IF_ERROR(ZSTD_cParam_clampBounds(ZSTD_c_overlapLog, &value), "");
+ CCtxParams->rsyncable = value;
+ return CCtxParams->rsyncable;
+#endif
+
+ case ZSTD_c_enableLongDistanceMatching :
+ CCtxParams->ldmParams.enableLdm = (value!=0);
+ 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 :
+ RETURN_ERROR_IF(value > ZSTD_WINDOWLOG_MAX - ZSTD_HASHLOG_MIN,
+ parameter_outOfBound, "Param out of bounds!");
+ 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;
+
+ default: RETURN_ERROR(parameter_unsupported, "unknown parameter");
+ }
+}
+
+size_t ZSTD_CCtx_getParameter(ZSTD_CCtx* cctx, ZSTD_cParameter param, int* value)
+{
+ return ZSTD_CCtxParams_getParameter(&cctx->requestedParams, param, value);
+}
+
+size_t ZSTD_CCtxParams_getParameter(
+ ZSTD_CCtx_params* 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 :
+#ifndef ZSTD_MULTITHREAD
+ assert(CCtxParams->nbWorkers == 0);
+#endif
+ *value = CCtxParams->nbWorkers;
+ break;
+ case ZSTD_c_jobSize :
+#ifndef ZSTD_MULTITHREAD
+ RETURN_ERROR(parameter_unsupported, "not compiled with multithreading");
+#else
+ assert(CCtxParams->jobSize <= INT_MAX);
+ *value = (int)CCtxParams->jobSize;
+ break;
+#endif
+ case ZSTD_c_overlapLog :
+#ifndef ZSTD_MULTITHREAD
+ RETURN_ERROR(parameter_unsupported, "not compiled with multithreading");
+#else
+ *value = CCtxParams->overlapLog;
+ break;
+#endif
+ case ZSTD_c_rsyncable :
+#ifndef ZSTD_MULTITHREAD
+ RETURN_ERROR(parameter_unsupported, "not compiled with multithreading");
+#else
+ *value = CCtxParams->rsyncable;
+ break;
+#endif
+ 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;
+ 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;
+}
+
+ZSTDLIB_API 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;
+}
+
+/**
+ * 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;
+ ZSTD_compressionParameters const cParams = ZSTD_getCParamsFromCCtxParams(
+ &cctx->requestedParams, ZSTD_CONTENTSIZE_UNKNOWN, dl->dictSize);
+ 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_advanced(
+ dl->dict,
+ dl->dictSize,
+ ZSTD_dlm_byRef,
+ dl->dictContentType,
+ cParams,
+ 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.");
+ RETURN_ERROR_IF(cctx->staticSize, memory_allocation,
+ "no malloc for static CCtx");
+ 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 = ZSTD_malloc(dictSize, cctx->customMem);
+ RETURN_ERROR_IF(!dictBuffer, memory_allocation, "NULL pointer!");
+ memcpy(dictBuffer, dict, dictSize);
+ cctx->localDict.dictBuffer = dictBuffer;
+ cctx->localDict.dict = dictBuffer;
+ }
+ cctx->localDict.dictSize = dictSize;
+ cctx->localDict.dictContentType = dictContentType;
+ return 0;
+}
+
+ZSTDLIB_API 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);
+}
+
+ZSTDLIB_API 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_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_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.
+ * 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)
+{
+ static const U64 minSrcSize = 513; /* (1<<9) + 1 */
+ static const U64 maxWindowResize = 1ULL << (ZSTD_WINDOWLOG_MAX-1);
+ assert(ZSTD_checkCParams(cPar)==0);
+
+ if (dictSize && srcSize == ZSTD_CONTENTSIZE_UNKNOWN)
+ srcSize = minSrcSize;
+
+ /* 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 (cPar.hashLog > cPar.windowLog+1) cPar.hashLog = cPar.windowLog+1;
+ { U32 const cycleLog = ZSTD_cycleLog(cPar.chainLog, cPar.strategy);
+ if (cycleLog > cPar.windowLog)
+ cPar.chainLog -= (cycleLog - cPar.windowLog);
+ }
+
+ 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);
+}
+
+static ZSTD_compressionParameters ZSTD_getCParams_internal(int compressionLevel, unsigned long long srcSizeHint, size_t dictSize);
+static ZSTD_parameters ZSTD_getParams_internal(int compressionLevel, unsigned long long srcSizeHint, size_t dictSize);
+
+ZSTD_compressionParameters ZSTD_getCParamsFromCCtxParams(
+ const ZSTD_CCtx_params* CCtxParams, U64 srcSizeHint, size_t dictSize)
+{
+ ZSTD_compressionParameters cParams;
+ if (srcSizeHint == ZSTD_CONTENTSIZE_UNKNOWN && CCtxParams->srcSizeHint > 0) {
+ srcSizeHint = CCtxParams->srcSizeHint;
+ }
+ cParams = ZSTD_getCParams_internal(CCtxParams->compressionLevel, srcSizeHint, dictSize);
+ if (CCtxParams->ldmParams.enableLdm) cParams.windowLog = ZSTD_LDM_DEFAULT_WINDOW_LOG;
+ if (CCtxParams->cParams.windowLog) cParams.windowLog = CCtxParams->cParams.windowLog;
+ if (CCtxParams->cParams.hashLog) cParams.hashLog = CCtxParams->cParams.hashLog;
+ if (CCtxParams->cParams.chainLog) cParams.chainLog = CCtxParams->cParams.chainLog;
+ if (CCtxParams->cParams.searchLog) cParams.searchLog = CCtxParams->cParams.searchLog;
+ if (CCtxParams->cParams.minMatch) cParams.minMatch = CCtxParams->cParams.minMatch;
+ if (CCtxParams->cParams.targetLength) cParams.targetLength = CCtxParams->cParams.targetLength;
+ if (CCtxParams->cParams.strategy) cParams.strategy = CCtxParams->cParams.strategy;
+ assert(!ZSTD_checkCParams(cParams));
+ /* srcSizeHint == 0 means 0 */
+ return ZSTD_adjustCParams_internal(cParams, srcSizeHint, dictSize);
+}
+
+static size_t
+ZSTD_sizeof_matchState(const ZSTD_compressionParameters* const cParams,
+ const U32 forCCtx)
+{
+ size_t const chainSize = (cParams->strategy == ZSTD_fast) ? 0 : ((size_t)1 << cParams->chainLog);
+ 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_alloc_size((MaxML+1) * sizeof(U32))
+ + ZSTD_cwksp_alloc_size((MaxLL+1) * sizeof(U32))
+ + ZSTD_cwksp_alloc_size((MaxOff+1) * sizeof(U32))
+ + ZSTD_cwksp_alloc_size((1<<Litbits) * sizeof(U32))
+ + ZSTD_cwksp_alloc_size((ZSTD_OPT_NUM+1) * sizeof(ZSTD_match_t))
+ + ZSTD_cwksp_alloc_size((ZSTD_OPT_NUM+1) * sizeof(ZSTD_optimal_t));
+ size_t const optSpace = (forCCtx && (cParams->strategy >= ZSTD_btopt))
+ ? optPotentialSpace
+ : 0;
+ DEBUGLOG(4, "chainSize: %u - hSize: %u - h3Size: %u",
+ (U32)chainSize, (U32)hSize, (U32)h3Size);
+ return tableSpace + optSpace;
+}
+
+size_t ZSTD_estimateCCtxSize_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);
+ size_t const blockSize = MIN(ZSTD_BLOCKSIZE_MAX, (size_t)1 << cParams.windowLog);
+ 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_alloc_size(maxNbSeq * sizeof(seqDef))
+ + 3 * ZSTD_cwksp_alloc_size(maxNbSeq * sizeof(BYTE));
+ size_t const entropySpace = ZSTD_cwksp_alloc_size(HUF_WORKSPACE_SIZE);
+ size_t const blockStateSpace = 2 * ZSTD_cwksp_alloc_size(sizeof(ZSTD_compressedBlockState_t));
+ size_t const matchStateSize = ZSTD_sizeof_matchState(&cParams, /* forCCtx */ 1);
+
+ size_t const ldmSpace = ZSTD_ldm_getTableSize(params->ldmParams);
+ size_t const ldmSeqSpace = ZSTD_cwksp_alloc_size(ZSTD_ldm_getMaxNbSeq(params->ldmParams, blockSize) * sizeof(rawSeq));
+
+ /* 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. */
+ size_t const bufferSpace = ZSTD_cwksp_alloc_size(0)
+ + ZSTD_cwksp_alloc_size(0);
+
+ size_t const cctxSpace = ZSTD_cwksp_alloc_size(sizeof(ZSTD_CCtx));
+
+ 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_usingCParams(ZSTD_compressionParameters cParams)
+{
+ ZSTD_CCtx_params const params = ZSTD_makeCCtxParamsFromCParams(cParams);
+ return ZSTD_estimateCCtxSize_usingCCtxParams(&params);
+}
+
+static size_t ZSTD_estimateCCtxSize_internal(int compressionLevel)
+{
+ ZSTD_compressionParameters const cParams = ZSTD_getCParams_internal(compressionLevel, ZSTD_CONTENTSIZE_UNKNOWN, 0);
+ return ZSTD_estimateCCtxSize_usingCParams(cParams);
+}
+
+size_t ZSTD_estimateCCtxSize(int compressionLevel)
+{
+ int level;
+ size_t memBudget = 0;
+ for (level=MIN(compressionLevel, 1); level<=compressionLevel; level++) {
+ 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);
+ size_t const CCtxSize = ZSTD_estimateCCtxSize_usingCCtxParams(params);
+ size_t const blockSize = MIN(ZSTD_BLOCKSIZE_MAX, (size_t)1 << cParams.windowLog);
+ size_t const inBuffSize = ((size_t)1 << cParams.windowLog) + blockSize;
+ size_t const outBuffSize = ZSTD_compressBound(blockSize) + 1;
+ size_t const streamingSize = ZSTD_cwksp_alloc_size(inBuffSize)
+ + ZSTD_cwksp_alloc_size(outBuffSize);
+
+ return CCtxSize + streamingSize;
+ }
+}
+
+size_t ZSTD_estimateCStreamSize_usingCParams(ZSTD_compressionParameters cParams)
+{
+ ZSTD_CCtx_params const params = ZSTD_makeCCtxParamsFromCParams(cParams);
+ return ZSTD_estimateCStreamSize_usingCCtxParams(&params);
+}
+
+static size_t ZSTD_estimateCStreamSize_internal(int compressionLevel)
+{
+ ZSTD_compressionParameters const cParams = ZSTD_getCParams_internal(compressionLevel, ZSTD_CONTENTSIZE_UNKNOWN, 0);
+ 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)
+{
+#ifdef ZSTD_MULTITHREAD
+ if (cctx->appliedParams.nbWorkers > 0) {
+ return ZSTDMT_getFrameProgression(cctx->mtctx);
+ }
+#endif
+ { 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)
+{
+#ifdef ZSTD_MULTITHREAD
+ if (cctx->appliedParams.nbWorkers > 0) {
+ return ZSTDMT_toFlushNow(cctx->mtctx);
+ }
+#endif
+ (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;
+}
+
+/**
+ * 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;
+
+/**
+ * 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_compResetPolicy_e crp,
+ const ZSTD_indexResetPolicy_e forceResetIndex,
+ const ZSTD_resetTarget_e forWho)
+{
+ size_t const chainSize = (cParams->strategy == ZSTD_fast) ? 0 : ((size_t)1 << cParams->chainLog);
+ 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);
+ 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));
+ }
+
+ 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_resetCCtx_internal() :
+ note : `params` are assumed fully validated at this stage */
+static size_t ZSTD_resetCCtx_internal(ZSTD_CCtx* zc,
+ ZSTD_CCtx_params params,
+ U64 const pledgedSrcSize,
+ 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",
+ (U32)pledgedSrcSize, params.cParams.windowLog);
+ assert(!ZSTD_isError(ZSTD_checkCParams(params.cParams)));
+
+ zc->isFirstBlock = 1;
+
+ if (params.ldmParams.enableLdm) {
+ /* Adjust long distance matching parameters */
+ ZSTD_ldm_adjustParameters(&params.ldmParams, &params.cParams);
+ assert(params.ldmParams.hashLog >= params.ldmParams.bucketSizeLog);
+ assert(params.ldmParams.hashRateLog < 32);
+ zc->ldmState.hashPower = ZSTD_rollingHash_primePower(params.ldmParams.minMatchLength);
+ }
+
+ { 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 tokenSpace = ZSTD_cwksp_alloc_size(WILDCOPY_OVERLENGTH + blockSize)
+ + ZSTD_cwksp_alloc_size(maxNbSeq * sizeof(seqDef))
+ + 3 * ZSTD_cwksp_alloc_size(maxNbSeq * sizeof(BYTE));
+ size_t const buffOutSize = (zbuff==ZSTDb_buffered) ? ZSTD_compressBound(blockSize)+1 : 0;
+ size_t const buffInSize = (zbuff==ZSTDb_buffered) ? windowSize + blockSize : 0;
+ size_t const matchStateSize = ZSTD_sizeof_matchState(&params.cParams, /* forCCtx */ 1);
+ size_t const maxNbLdmSeq = ZSTD_ldm_getMaxNbSeq(params.ldmParams, blockSize);
+
+ ZSTD_indexResetPolicy_e needsIndexReset = zc->initialized ? ZSTDirp_continue : ZSTDirp_reset;
+
+ if (ZSTD_indexTooCloseToMax(zc->blockState.matchState.window)) {
+ needsIndexReset = ZSTDirp_reset;
+ }
+
+ if (!zc->staticSize) ZSTD_cwksp_bump_oversized_duration(ws, 0);
+
+ /* Check if workspace is large enough, alloc a new one if needed */
+ { size_t const cctxSpace = zc->staticSize ? ZSTD_cwksp_alloc_size(sizeof(ZSTD_CCtx)) : 0;
+ size_t const entropySpace = ZSTD_cwksp_alloc_size(HUF_WORKSPACE_SIZE);
+ size_t const blockStateSpace = 2 * ZSTD_cwksp_alloc_size(sizeof(ZSTD_compressedBlockState_t));
+ size_t const bufferSpace = ZSTD_cwksp_alloc_size(buffInSize) + ZSTD_cwksp_alloc_size(buffOutSize);
+ size_t const ldmSpace = ZSTD_ldm_getTableSize(params.ldmParams);
+ size_t const ldmSeqSpace = ZSTD_cwksp_alloc_size(maxNbLdmSeq * sizeof(rawSeq));
+
+ size_t const neededSpace =
+ cctxSpace +
+ entropySpace +
+ blockStateSpace +
+ ldmSpace +
+ ldmSeqSpace +
+ matchStateSize +
+ tokenSpace +
+ bufferSpace;
+
+ int const workspaceTooSmall = ZSTD_cwksp_sizeof(ws) < neededSpace;
+ int const workspaceWasteful = ZSTD_cwksp_check_wasteful(ws, neededSpace);
+
+ DEBUGLOG(4, "Need %zuKB workspace, including %zuKB for match state, and %zuKB for buffers",
+ neededSpace>>10, matchStateSize>>10, bufferSpace>>10);
+ DEBUGLOG(4, "windowSize: %zu - blockSize: %zu", windowSize, blockSize);
+
+ if (workspaceTooSmall || workspaceWasteful) {
+ 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, HUF_WORKSPACE_SIZE);
+ RETURN_ERROR_IF(zc->blockState.nextCBlock == NULL, memory_allocation, "couldn't allocate entropyWorkspace");
+ } }
+
+ ZSTD_cwksp_clear(ws);
+
+ /* init params */
+ zc->appliedParams = 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;
+
+ 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->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) {
+ /* TODO: avoid memset? */
+ size_t const ldmBucketSize =
+ ((size_t)1) << (params.ldmParams.hashLog -
+ params.ldmParams.bucketSizeLog);
+ zc->ldmState.bucketOffsets = ZSTD_cwksp_reserve_buffer(ws, ldmBucketSize);
+ memset(zc->ldmState.bucketOffsets, 0, ldmBucketSize);
+ }
+
+ /* 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,
+ crp,
+ needsIndexReset,
+ ZSTD_resetTarget_CCtx), "");
+
+ /* ldm hash table */
+ if (params.ldmParams.enableLdm) {
+ /* 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));
+ 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);
+ ZSTD_window_clear(&zc->ldmState.window);
+ zc->ldmState.loadedDictEnd = 0;
+ }
+
+ DEBUGLOG(3, "wksp: finished allocating, %zd bytes remain available", ZSTD_cwksp_available_space(ws));
+ 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];
+ return ( 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)
+{
+ { const ZSTD_compressionParameters* const 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. */
+ /* pledgeSrcSize == 0 means 0! */
+ params.cParams = ZSTD_adjustCParams_internal(*cdict_cParams, pledgedSrcSize, 0);
+ params.cParams.windowLog = windowLog;
+ FORWARD_IF_ERROR(ZSTD_resetCCtx_internal(cctx, params, pledgedSrcSize,
+ ZSTDcrp_makeClean, zbuff), "");
+ assert(cctx->appliedParams.cParams.strategy == 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;
+
+ /* copy block state */
+ 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;
+
+ DEBUGLOG(4, "copying dictionary into context");
+
+ { unsigned const windowLog = params.cParams.windowLog;
+ assert(windowLog != 0);
+ /* Copy only compression parameters related to tables. */
+ params.cParams = *cdict_cParams;
+ params.cParams.windowLog = windowLog;
+ FORWARD_IF_ERROR(ZSTD_resetCCtx_internal(cctx, params, pledgedSrcSize,
+ 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);
+
+ /* copy tables */
+ { size_t const chainSize = (cdict_cParams->strategy == ZSTD_fast) ? 0 : ((size_t)1 << cdict_cParams->chainLog);
+ size_t const hSize = (size_t)1 << cdict_cParams->hashLog;
+
+ memcpy(cctx->blockState.matchState.hashTable,
+ cdict->matchState.hashTable,
+ hSize * sizeof(U32));
+ memcpy(cctx->blockState.matchState.chainTable,
+ cdict->matchState.chainTable,
+ chainSize * sizeof(U32));
+ }
+
+ /* 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);
+ 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;
+
+ /* copy block state */
+ 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)
+{
+ DEBUGLOG(5, "ZSTD_copyCCtx_internal");
+ RETURN_ERROR_IF(srcCCtx->stage!=ZSTDcs_init, stage_wrong,
+ "Can't copy a ctx that's not in init stage.");
+
+ 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;
+ params.fParams = fParams;
+ ZSTD_resetCCtx_internal(dstCCtx, params, pledgedSrcSize,
+ 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 = (srcCCtx->appliedParams.cParams.strategy == ZSTD_fast) ? 0 : ((size_t)1 << srcCCtx->appliedParams.cParams.chainLog);
+ 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;
+
+ memcpy(dstCCtx->blockState.matchState.hashTable,
+ srcCCtx->blockState.matchState.hashTable,
+ hSize * sizeof(U32));
+ memcpy(dstCCtx->blockState.matchState.chainTable,
+ srcCCtx->blockState.matchState.chainTable,
+ chainSize * sizeof(U32));
+ 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;
+
+ /* copy block state */
+ 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 = (ZSTD_buffered_policy_e)(srcCCtx->inBuffSize>0);
+ 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;
+ assert((size & (ZSTD_ROWSIZE-1)) == 0); /* multiple of ZSTD_ROWSIZE */
+ assert(size < (1U<<31)); /* can be casted to int */
+
+#if defined (MEMORY_SANITIZER) && !defined (ZSTD_MSAN_DONT_POISON_WORKSPACE)
+ /* To validate that the table re-use logic is sound, and that we don't
+ * access table space that we haven't cleaned, we re-"poison" the table
+ * space every time we mark it dirty.
+ *
+ * This function however is intended to operate on those dirty tables and
+ * re-clean them. So when this function is used correctly, we can unpoison
+ * the memory it operated on. This introduces a blind spot though, since
+ * if we now try to operate on __actually__ poisoned memory, we will not
+ * detect that. */
+ __msan_unpoison(table, size * sizeof(U32));
+#endif
+
+ for (rowNb=0 ; rowNb < nbRows ; rowNb++) {
+ int column;
+ for (column=0; column<ZSTD_ROWSIZE; column++) {
+ if (preserveMark) {
+ U32 const adder = (table[cellNb] == ZSTD_DUBT_UNSORTED_MARK) ? reducerValue : 0;
+ table[cellNb] += adder;
+ }
+ if (table[cellNb] < reducerValue) table[cellNb] = 0;
+ else table[cellNb] -= reducerValue;
+ 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 (params->cParams.strategy != ZSTD_fast) {
+ 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].matchLength;
+ llCodeTable[u] = (BYTE)ZSTD_LLcode(llv);
+ ofCodeTable[u] = (BYTE)ZSTD_highbit32(sequences[u].offset);
+ mlCodeTable[u] = (BYTE)ZSTD_MLcode(mlv);
+ }
+ if (seqStorePtr->longLengthID==1)
+ llCodeTable[seqStorePtr->longLengthPos] = MaxLL;
+ if (seqStorePtr->longLengthID==2)
+ 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_compressSequences_internal():
+ * actually compresses both literals and sequences */
+MEM_STATIC size_t
+ZSTD_compressSequences_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[MaxSeq+1];
+ FSE_CTable* CTable_LitLength = nextEntropy->fse.litlengthCTable;
+ FSE_CTable* CTable_OffsetBits = nextEntropy->fse.offcodeCTable;
+ FSE_CTable* CTable_MatchLength = nextEntropy->fse.matchlengthCTable;
+ U32 LLtype, Offtype, MLtype; /* compressed, raw or rle */
+ const seqDef* const 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 const nbSeq = (size_t)(seqStorePtr->sequences - seqStorePtr->sequencesStart);
+ BYTE* seqHead;
+ BYTE* lastNCount = NULL;
+
+ DEBUGLOG(5, "ZSTD_compressSequences_internal (nbSeq=%zu)", nbSeq);
+ ZSTD_STATIC_ASSERT(HUF_WORKSPACE_SIZE >= (1<<MAX(MLFSELog,LLFSELog)));
+
+ /* Compress literals */
+ { const BYTE* const literals = seqStorePtr->litStart;
+ size_t const litSize = (size_t)(seqStorePtr->lit - literals);
+ size_t const cSize = ZSTD_compressLiterals(
+ &prevEntropy->huf, &nextEntropy->huf,
+ cctxParams->cParams.strategy,
+ ZSTD_disableLiteralsCompression(cctxParams),
+ op, dstCapacity,
+ literals, litSize,
+ entropyWorkspace, entropyWkspSize,
+ bmi2);
+ 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 */
+ memcpy(&nextEntropy->fse, &prevEntropy->fse, sizeof(prevEntropy->fse));
+ return (size_t)(op - ostart);
+ }
+
+ /* seqHead : flags for FSE encoding type */
+ seqHead = op++;
+ assert(op <= oend);
+
+ /* convert length/distances into codes */
+ ZSTD_seqToCodes(seqStorePtr);
+ /* build CTable for Literal Lengths */
+ { unsigned max = MaxLL;
+ size_t const mostFrequent = HIST_countFast_wksp(count, &max, llCodeTable, nbSeq, entropyWorkspace, entropyWkspSize); /* can't fail */
+ DEBUGLOG(5, "Building LL table");
+ nextEntropy->fse.litlength_repeatMode = prevEntropy->fse.litlength_repeatMode;
+ LLtype = ZSTD_selectEncodingType(&nextEntropy->fse.litlength_repeatMode,
+ count, max, mostFrequent, nbSeq,
+ LLFSELog, prevEntropy->fse.litlengthCTable,
+ LL_defaultNorm, LL_defaultNormLog,
+ ZSTD_defaultAllowed, strategy);
+ assert(set_basic < set_compressed && set_rle < set_compressed);
+ assert(!(LLtype < set_compressed && nextEntropy->fse.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)LLtype,
+ count, max, llCodeTable, nbSeq,
+ LL_defaultNorm, LL_defaultNormLog, MaxLL,
+ prevEntropy->fse.litlengthCTable,
+ sizeof(prevEntropy->fse.litlengthCTable),
+ entropyWorkspace, entropyWkspSize);
+ FORWARD_IF_ERROR(countSize, "ZSTD_buildCTable for LitLens failed");
+ if (LLtype == set_compressed)
+ lastNCount = op;
+ op += countSize;
+ assert(op <= oend);
+ } }
+ /* build CTable for Offsets */
+ { unsigned max = MaxOff;
+ size_t const mostFrequent = HIST_countFast_wksp(
+ count, &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->fse.offcode_repeatMode = prevEntropy->fse.offcode_repeatMode;
+ Offtype = ZSTD_selectEncodingType(&nextEntropy->fse.offcode_repeatMode,
+ count, max, mostFrequent, nbSeq,
+ OffFSELog, prevEntropy->fse.offcodeCTable,
+ OF_defaultNorm, OF_defaultNormLog,
+ defaultPolicy, strategy);
+ assert(!(Offtype < set_compressed && nextEntropy->fse.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)Offtype,
+ count, max, ofCodeTable, nbSeq,
+ OF_defaultNorm, OF_defaultNormLog, DefaultMaxOff,
+ prevEntropy->fse.offcodeCTable,
+ sizeof(prevEntropy->fse.offcodeCTable),
+ entropyWorkspace, entropyWkspSize);
+ FORWARD_IF_ERROR(countSize, "ZSTD_buildCTable for Offsets failed");
+ if (Offtype == set_compressed)
+ lastNCount = op;
+ op += countSize;
+ assert(op <= oend);
+ } }
+ /* build CTable for MatchLengths */
+ { unsigned max = MaxML;
+ size_t const mostFrequent = HIST_countFast_wksp(
+ count, &max, mlCodeTable, nbSeq, entropyWorkspace, entropyWkspSize); /* can't fail */
+ DEBUGLOG(5, "Building ML table (remaining space : %i)", (int)(oend-op));
+ nextEntropy->fse.matchlength_repeatMode = prevEntropy->fse.matchlength_repeatMode;
+ MLtype = ZSTD_selectEncodingType(&nextEntropy->fse.matchlength_repeatMode,
+ count, max, mostFrequent, nbSeq,
+ MLFSELog, prevEntropy->fse.matchlengthCTable,
+ ML_defaultNorm, ML_defaultNormLog,
+ ZSTD_defaultAllowed, strategy);
+ assert(!(MLtype < set_compressed && nextEntropy->fse.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)MLtype,
+ count, max, mlCodeTable, nbSeq,
+ ML_defaultNorm, ML_defaultNormLog, MaxML,
+ prevEntropy->fse.matchlengthCTable,
+ sizeof(prevEntropy->fse.matchlengthCTable),
+ entropyWorkspace, entropyWkspSize);
+ FORWARD_IF_ERROR(countSize, "ZSTD_buildCTable for MatchLengths failed");
+ if (MLtype == set_compressed)
+ lastNCount = op;
+ op += countSize;
+ assert(op <= oend);
+ } }
+
+ *seqHead = (BYTE)((LLtype<<6) + (Offtype<<4) + (MLtype<<2));
+
+ { 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 (lastNCount && (op - lastNCount) < 4) {
+ /* NCountSize >= 2 && bitstreamSize > 0 ==> lastCountSize == 3 */
+ assert(op - lastNCount == 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_compressSequences(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_compressSequences_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_compressSequences_internal failed");
+
+ /* Check compressibility */
+ { size_t const maxCSize = srcSize - ZSTD_minGain(srcSize, cctxParams->cParams.strategy);
+ if (cSize >= maxCSize) return 0; /* block not compressed */
+ }
+
+ 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_dictMode_e dictMode)
+{
+ static const ZSTD_blockCompressor blockCompressor[3][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 }
+ };
+ ZSTD_blockCompressor selectedCompressor;
+ ZSTD_STATIC_ASSERT((unsigned)ZSTD_fast == 1);
+
+ assert(ZSTD_cParam_withinBounds(ZSTD_c_strategy, strat));
+ selectedCompressor = blockCompressor[(int)dictMode][(int)strat];
+ assert(selectedCompressor != NULL);
+ return selectedCompressor;
+}
+
+static void ZSTD_storeLastLiterals(seqStore_t* seqStorePtr,
+ const BYTE* anchor, size_t lastLLSize)
+{
+ memcpy(seqStorePtr->lit, anchor, lastLLSize);
+ seqStorePtr->lit += lastLLSize;
+}
+
+void ZSTD_resetSeqStore(seqStore_t* ssPtr)
+{
+ ssPtr->lit = ssPtr->litStart;
+ ssPtr->sequences = ssPtr->sequencesStart;
+ ssPtr->longLengthID = 0;
+}
+
+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) {
+ 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 current = (U32)(istart-base);
+ if (sizeof(ptrdiff_t)==8) assert(istart - base < (ptrdiff_t)(U32)(-1)); /* ensure no overflow */
+ if (current > ms->nextToUpdate + 384)
+ ms->nextToUpdate = current - MIN(192, (U32)(current - 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);
+ /* Updates ldmSeqStore.pos */
+ lastLLSize =
+ ZSTD_ldm_blockCompress(&zc->externSeqStore,
+ ms, &zc->seqStore,
+ zc->blockState.nextCBlock->rep,
+ src, srcSize);
+ assert(zc->externSeqStore.pos <= zc->externSeqStore.size);
+ } else if (zc->appliedParams.ldmParams.enableLdm) {
+ rawSeqStore_t ldmSeqStore = {NULL, 0, 0, 0};
+
+ 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,
+ src, srcSize);
+ assert(ldmSeqStore.pos == ldmSeqStore.size);
+ } else { /* not long range mode */
+ ZSTD_blockCompressor const blockCompressor = ZSTD_selectBlockCompressor(zc->appliedParams.cParams.strategy, dictMode);
+ 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* seqs = seqStore->sequencesStart;
+ size_t seqsSize = seqStore->sequences - seqs;
+
+ ZSTD_Sequence* outSeqs = &zc->seqCollector.seqStart[zc->seqCollector.seqIndex];
+ size_t i; size_t position; int repIdx;
+
+ assert(zc->seqCollector.seqIndex + 1 < zc->seqCollector.maxSequences);
+ for (i = 0, position = 0; i < seqsSize; ++i) {
+ outSeqs[i].offset = seqs[i].offset;
+ outSeqs[i].litLength = seqs[i].litLength;
+ outSeqs[i].matchLength = seqs[i].matchLength + MINMATCH;
+
+ if (i == seqStore->longLengthPos) {
+ if (seqStore->longLengthID == 1) {
+ outSeqs[i].litLength += 0x10000;
+ } else if (seqStore->longLengthID == 2) {
+ outSeqs[i].matchLength += 0x10000;
+ }
+ }
+
+ if (outSeqs[i].offset <= ZSTD_REP_NUM) {
+ outSeqs[i].rep = outSeqs[i].offset;
+ repIdx = (unsigned int)i - outSeqs[i].offset;
+
+ if (outSeqs[i].litLength == 0) {
+ if (outSeqs[i].offset < 3) {
+ --repIdx;
+ } else {
+ repIdx = (unsigned int)i - 1;
+ }
+ ++outSeqs[i].rep;
+ }
+ assert(repIdx >= -3);
+ outSeqs[i].offset = repIdx >= 0 ? outSeqs[repIdx].offset : repStartValue[-repIdx - 1];
+ if (outSeqs[i].rep == 4) {
+ --outSeqs[i].offset;
+ }
+ } else {
+ outSeqs[i].offset -= ZSTD_REP_NUM;
+ }
+
+ position += outSeqs[i].litLength;
+ outSeqs[i].matchPos = (unsigned int)position;
+ position += outSeqs[i].matchLength;
+ }
+ zc->seqCollector.seqIndex += seqsSize;
+}
+
+size_t ZSTD_getSequences(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_malloc(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_free(dst, ZSTD_defaultCMem);
+ return zc->seqCollector.seqIndex;
+}
+
+/* Returns true if the given block is a RLE block */
+static int ZSTD_isRLE(const BYTE *ip, size_t length) {
+ size_t i;
+ if (length < 2) return 1;
+ for (i = 1; i < length; ++i) {
+ if (ip[0] != ip[i]) 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_confirmRepcodesAndEntropyTables(ZSTD_CCtx* zc)
+{
+ ZSTD_compressedBlockState_t* const tmp = zc->blockState.prevCBlock;
+ zc->blockState.prevCBlock = zc->blockState.nextCBlock;
+ zc->blockState.nextCBlock = tmp;
+}
+
+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);
+ return 0;
+ }
+
+ /* encode sequences and literals */
+ cSize = ZSTD_compressSequences(&zc->seqStore,
+ &zc->blockState.prevCBlock->entropy, &zc->blockState.nextCBlock->entropy,
+ &zc->appliedParams,
+ dst, dstCapacity,
+ srcSize,
+ zc->entropyWorkspace, HUF_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_confirmRepcodesAndEntropyTables(zc);
+ }
+ /* 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_confirmRepcodesAndEntropyTables(zc);
+ 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)
+{
+ if (ZSTD_window_needOverflowCorrection(ms->window, iend)) {
+ U32 const maxDist = (U32)1 << params->cParams.windowLog;
+ U32 const cycleLog = ZSTD_cycleLog(params->cParams.chainLog, params->cParams.strategy);
+ 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(5, "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);
+
+ /* 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 {
+ 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 */
+ 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 */
+ 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_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,
+ parameter_unsupported,
+ "incompatible with ldm");
+ cctx->externSeqStore.seq = seq;
+ cctx->externSeqStore.size = nbSeq;
+ cctx->externSeqStore.capacity = nbSeq;
+ cctx->externSeqStore.pos = 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->nextToUpdate = ms->window.dictLimit;
+ }
+ if (cctx->appliedParams.ldmParams.enableLdm) {
+ ZSTD_window_update(&cctx->ldmState.window, src, srcSize);
+ }
+
+ 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;
+
+ ZSTD_window_update(&ms->window, src, srcSize);
+ ms->loadedDictEnd = params->forceWindow ? 0 : (U32)(iend - ms->window.base);
+
+ if (params->ldmParams.enableLdm && ls != NULL) {
+ ZSTD_window_update(&ls->window, src, srcSize);
+ ls->loadedDictEnd = params->forceWindow ? 0 : (U32)(iend - ls->window.base);
+ }
+
+ /* Assert that we the ms params match the params we're being given */
+ ZSTD_assertEqualCParams(params->cParams, ms->cParams);
+
+ if (srcSize <= HASH_READ_SIZE) return 0;
+
+ while (iend - ip > HASH_READ_SIZE) {
+ size_t const remaining = (size_t)(iend - ip);
+ size_t const chunk = MIN(remaining, ZSTD_CHUNKSIZE_MAX);
+ const BYTE* const ichunk = ip + chunk;
+
+ ZSTD_overflowCorrectIfNeeded(ms, ws, params, ip, ichunk);
+
+ if (params->ldmParams.enableLdm && ls != NULL)
+ ZSTD_ldm_fillHashTable(ls, (const BYTE*)src, (const BYTE*)src + srcSize, &params->ldmParams);
+
+ switch(params->cParams.strategy)
+ {
+ case ZSTD_fast:
+ ZSTD_fillHashTable(ms, ichunk, dtlm);
+ break;
+ case ZSTD_dfast:
+ ZSTD_fillDoubleHashTable(ms, ichunk, dtlm);
+ break;
+
+ case ZSTD_greedy:
+ case ZSTD_lazy:
+ case ZSTD_lazy2:
+ if (chunk >= HASH_READ_SIZE)
+ ZSTD_insertAndFindFirstIndex(ms, ichunk-HASH_READ_SIZE);
+ break;
+
+ case ZSTD_btlazy2: /* we want the dictionary table fully sorted */
+ case ZSTD_btopt:
+ case ZSTD_btultra:
+ case ZSTD_btultra2:
+ if (chunk >= HASH_READ_SIZE)
+ ZSTD_updateTree(ms, ichunk-HASH_READ_SIZE, ichunk);
+ break;
+
+ default:
+ assert(0); /* not possible : not a valid strategy id */
+ }
+
+ ip = ichunk;
+ }
+
+ ms->nextToUpdate = (U32)(iend - ms->window.base);
+ return 0;
+}
+
+
+/* Dictionaries that assign zero probability to symbols that show up causes problems
+ when FSE encoding. Refuse dictionaries that assign zero probability to symbols
+ that we may encounter during compression.
+ NOTE: This behavior is not standard and could be improved in the future. */
+static size_t ZSTD_checkDictNCount(short* normalizedCounter, unsigned dictMaxSymbolValue, unsigned maxSymbolValue) {
+ U32 s;
+ RETURN_ERROR_IF(dictMaxSymbolValue < maxSymbolValue, dictionary_corrupted, "dict fse tables don't have all symbols");
+ for (s = 0; s <= maxSymbolValue; ++s) {
+ RETURN_ERROR_IF(normalizedCounter[s] == 0, dictionary_corrupted, "dict fse tables don't have all symbols");
+ }
+ return 0;
+}
+
+size_t ZSTD_loadCEntropy(ZSTD_compressedBlockState_t* bs, void* workspace,
+ short* offcodeNCount, unsigned* offcodeMaxValue,
+ const void* const dict, size_t dictSize)
+{
+ 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, "");
+ /* Defer checking offcodeMaxValue because we need to know the size of the dictionary content */
+ /* 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, "");
+ 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, "");
+ /* Every match length code must have non-zero probability */
+ FORWARD_IF_ERROR( ZSTD_checkDictNCount(matchlengthNCount, matchlengthMaxValue, MaxML), "");
+ RETURN_ERROR_IF(FSE_isError(FSE_buildCTable_wksp(
+ bs->entropy.fse.matchlengthCTable,
+ matchlengthNCount, matchlengthMaxValue, matchlengthLog,
+ workspace, HUF_WORKSPACE_SIZE)),
+ dictionary_corrupted, "");
+ 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, "");
+ /* Every literal length code must have non-zero probability */
+ FORWARD_IF_ERROR( ZSTD_checkDictNCount(litlengthNCount, litlengthMaxValue, MaxLL), "");
+ RETURN_ERROR_IF(FSE_isError(FSE_buildCTable_wksp(
+ bs->entropy.fse.litlengthCTable,
+ litlengthNCount, litlengthMaxValue, litlengthLog,
+ workspace, HUF_WORKSPACE_SIZE)),
+ dictionary_corrupted, "");
+ 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;
+
+ return dictPtr - (const BYTE*)dict;
+}
+
+/* Dictionary format :
+ * See :
+ * https://github.com/facebook/zstd/blob/master/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;
+ short offcodeNCount[MaxOff+1];
+ unsigned offcodeMaxValue = MaxOff;
+ 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, offcodeNCount, &offcodeMaxValue, dict, dictSize);
+ FORWARD_IF_ERROR(eSize, "ZSTD_loadCEntropy failed");
+ dictPtr += eSize;
+
+ { 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 */
+ FORWARD_IF_ERROR(ZSTD_checkDictNCount(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, "");
+ } }
+
+ bs->entropy.fse.offcode_repeatMode = FSE_repeat_valid;
+ bs->entropy.fse.matchlength_repeatMode = FSE_repeat_valid;
+ bs->entropy.fse.litlength_repeatMode = FSE_repeat_valid;
+ 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 (6)
+
+/*! 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)
+{
+ 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,
+ 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, 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;
+ }
+ 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 const cctxParams =
+ ZSTD_assignParamsToCCtxParams(&cctx->requestedParams, &params);
+ 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_parameters const params = ZSTD_getParams_internal(compressionLevel, ZSTD_CONTENTSIZE_UNKNOWN, dictSize);
+ ZSTD_CCtx_params const cctxParams =
+ ZSTD_assignParamsToCCtxParams(&cctx->requestedParams, &params);
+ 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;
+}
+
+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);
+ }
+ return cSize + endResult;
+}
+
+
+static size_t ZSTD_compress_internal (ZSTD_CCtx* cctx,
+ void* dst, size_t dstCapacity,
+ const void* src, size_t srcSize,
+ const void* dict,size_t dictSize,
+ const ZSTD_parameters* params)
+{
+ ZSTD_CCtx_params const cctxParams =
+ ZSTD_assignParamsToCCtxParams(&cctx->requestedParams, params);
+ DEBUGLOG(4, "ZSTD_compress_internal");
+ return ZSTD_compress_advanced_internal(cctx,
+ dst, dstCapacity,
+ src, srcSize,
+ dict, dictSize,
+ &cctxParams);
+}
+
+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), "");
+ return ZSTD_compress_internal(cctx,
+ dst, dstCapacity,
+ src, srcSize,
+ dict, dictSize,
+ &params);
+}
+
+/* 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_CCtx_params cctxParams = ZSTD_assignParamsToCCtxParams(&cctx->requestedParams, &params);
+ DEBUGLOG(4, "ZSTD_compress_usingDict (srcSize=%u)", (unsigned)srcSize);
+ assert(params.fParams.contentSizeFlag == 1);
+ return ZSTD_compress_advanced_internal(cctx, dst, dstCapacity, src, srcSize, dict, dictSize, &cctxParams);
+}
+
+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 ctxBody;
+ ZSTD_initCCtx(&ctxBody, ZSTD_defaultCMem);
+ result = ZSTD_compressCCtx(&ctxBody, dst, dstCapacity, src, srcSize, compressionLevel);
+ ZSTD_freeCCtxContent(&ctxBody); /* can't free ctxBody itself, as it's on stack; free only heap content */
+ 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)
+ + ZSTD_sizeof_matchState(&cParams, /* 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);
+ 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_compressionParameters cParams)
+{
+ DEBUGLOG(3, "ZSTD_initCDict_internal (dictContentType:%u)", (unsigned)dictContentType);
+ assert(!ZSTD_checkCParams(cParams));
+ cdict->matchState.cParams = cParams;
+ 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;
+ memcpy(internalBuffer, dictBuffer, dictSize);
+ }
+ cdict->dictContentSize = dictSize;
+
+ 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,
+ &cParams,
+ ZSTDcrp_makeClean,
+ ZSTDirp_reset,
+ ZSTD_resetTarget_CDict), "");
+ /* (Maybe) load the dictionary
+ * Skips loading the dictionary if it is < 8 bytes.
+ */
+ { ZSTD_CCtx_params params;
+ memset(&params, 0, sizeof(params));
+ params.compressionLevel = ZSTD_CLEVEL_DEFAULT;
+ params.fParams.contentSizeFlag = 1;
+ params.cParams = cParams;
+ { 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;
+}
+
+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)
+{
+ DEBUGLOG(3, "ZSTD_createCDict_advanced, mode %u", (unsigned)dictContentType);
+ 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, /* forCCtx */ 0) +
+ (dictLoadMethod == ZSTD_dlm_byRef ? 0
+ : ZSTD_cwksp_alloc_size(ZSTD_cwksp_align(dictSize, sizeof(void*))));
+ void* const workspace = ZSTD_malloc(workspaceSize, customMem);
+ ZSTD_cwksp ws;
+ ZSTD_CDict* cdict;
+
+ if (!workspace) {
+ ZSTD_free(workspace, customMem);
+ return NULL;
+ }
+
+ ZSTD_cwksp_init(&ws, workspace, workspaceSize);
+
+ 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 = 0; /* signals advanced API usage */
+
+ if (ZSTD_isError( ZSTD_initCDict_internal(cdict,
+ dictBuffer, dictSize,
+ dictLoadMethod, dictContentType,
+ cParams) )) {
+ 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_CDict* 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);
+ return ZSTD_createCDict_advanced(dict, dictSize,
+ ZSTD_dlm_byRef, ZSTD_dct_auto,
+ cParams, ZSTD_defaultCMem);
+}
+
+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_free(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)
+{
+ size_t const matchStateSize = ZSTD_sizeof_matchState(&cParams, /* 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;
+
+ if ((size_t)workspace & 7) return NULL; /* 8-aligned */
+
+ {
+ ZSTD_cwksp ws;
+ ZSTD_cwksp_init(&ws, workspace, workspaceSize);
+ 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;
+
+ if (ZSTD_isError( ZSTD_initCDict_internal(cdict,
+ dict, dictSize,
+ dictLoadMethod, dictContentType,
+ cParams) ))
+ return NULL;
+
+ return cdict;
+}
+
+ZSTD_compressionParameters ZSTD_getCParamsFromCDict(const ZSTD_CDict* cdict)
+{
+ assert(cdict != NULL);
+ return cdict->matchState.cParams;
+}
+
+/* ZSTD_compressBegin_usingCDict_advanced() :
+ * 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)
+{
+ DEBUGLOG(4, "ZSTD_compressBegin_usingCDict_advanced");
+ RETURN_ERROR_IF(cdict==NULL, dictionary_wrong, "NULL pointer!");
+ { ZSTD_CCtx_params params = cctx->requestedParams;
+ 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 )
+ && (params.attachDictPref != ZSTD_dictForceLoad) ?
+ ZSTD_getCParamsFromCDict(cdict)
+ : ZSTD_getCParams(cdict->compressionLevel,
+ pledgedSrcSize,
+ cdict->dictContentSize);
+ /* 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;
+ params.cParams.windowLog = MAX(params.cParams.windowLog, limitedSrcLog);
+ }
+ params.fParams = fParams;
+ return ZSTD_compressBegin_internal(cctx,
+ NULL, 0, ZSTD_dct_auto, ZSTD_dtlm_fast,
+ cdict,
+ &params, pledgedSrcSize,
+ ZSTDb_not_buffered);
+ }
+}
+
+/* ZSTD_compressBegin_usingCDict() :
+ * pledgedSrcSize=0 means "unknown"
+ * if pledgedSrcSize>0, it will enable contentSizeFlag */
+size_t ZSTD_compressBegin_usingCDict(ZSTD_CCtx* cctx, const ZSTD_CDict* cdict)
+{
+ ZSTD_frameParameters const fParams = { 0 /*content*/, 0 /*checksum*/, 0 /*noDictID*/ };
+ DEBUGLOG(4, "ZSTD_compressBegin_usingCDict : dictIDFlag == %u", !fParams.noDictIDFlag);
+ return ZSTD_compressBegin_usingCDict_advanced(cctx, cdict, fParams, ZSTD_CONTENTSIZE_UNKNOWN);
+}
+
+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)
+{
+ FORWARD_IF_ERROR(ZSTD_compressBegin_usingCDict_advanced(cctx, cdict, fParams, srcSize), ""); /* will check if cdict != NULL */
+ return ZSTD_compressEnd(cctx, dst, dstCapacity, src, srcSize);
+}
+
+/*! 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_advanced(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 size_t ZSTD_resetCStream_internal(ZSTD_CStream* cctx,
+ const void* const dict, size_t const dictSize, ZSTD_dictContentType_e const dictContentType,
+ const ZSTD_CDict* const cdict,
+ ZSTD_CCtx_params params, unsigned long long const pledgedSrcSize)
+{
+ DEBUGLOG(4, "ZSTD_resetCStream_internal");
+ /* Finalize the compression parameters */
+ params.cParams = ZSTD_getCParamsFromCCtxParams(&params, pledgedSrcSize, dictSize);
+ /* 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 */
+
+ FORWARD_IF_ERROR( ZSTD_compressBegin_internal(cctx,
+ dict, dictSize, dictContentType, ZSTD_dtlm_fast,
+ cdict,
+ &params, pledgedSrcSize,
+ ZSTDb_buffered) , "");
+
+ cctx->inToCompress = 0;
+ cctx->inBuffPos = 0;
+ cctx->inBuffTarget = cctx->blockSize
+ + (cctx->blockSize == pledgedSrcSize); /* 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->outBuffContentSize = cctx->outBuffFlushedSize = 0;
+ cctx->streamStage = zcss_load;
+ cctx->frameEnded = 0;
+ return 0; /* ready to go */
+}
+
+/* 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) , "");
+ zcs->requestedParams = ZSTD_assignParamsToCCtxParams(&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);
+ assert(zcs->inBuff != NULL);
+ assert(zcs->inBuffSize > 0);
+ assert(zcs->outBuff != NULL);
+ assert(zcs->outBuffSize > 0);
+ assert(output->pos <= output->size);
+ assert(input->pos <= input->size);
+
+ 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 dstCapacity */
+ && (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 */
+ { 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);
+ { void* cDst;
+ size_t cSize;
+ size_t const iSize = zcs->inBuffPos - zcs->inToCompress;
+ size_t oSize = oend-op;
+ unsigned const lastBlock = (flushMode == ZSTD_e_end) && (ip==iend);
+ if (oSize >= ZSTD_compressBound(iSize))
+ cDst = op; /* compress into output buffer, to skip flush stage */
+ else
+ cDst = zcs->outBuff, oSize = zcs->outBuffSize;
+ 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;
+ 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 */
+ }
+ /* fall-through */
+ case zcss_flush:
+ DEBUGLOG(5, "flush stage");
+ { 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)
+{
+#ifdef ZSTD_MULTITHREAD
+ if (cctx->appliedParams.nbWorkers >= 1) {
+ assert(cctx->mtctx != NULL);
+ return ZSTDMT_nextInputSizeHint(cctx->mtctx);
+ }
+#endif
+ 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);
+}
+
+
+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, GENERIC, "invalid buffer");
+ RETURN_ERROR_IF(input->pos > input->size, GENERIC, "invalid buffer");
+ assert(cctx!=NULL);
+
+ /* transparent initialization stage */
+ if (cctx->streamStage == zcss_init) {
+ ZSTD_CCtx_params params = cctx->requestedParams;
+ ZSTD_prefixDict const prefixDict = cctx->prefixDict;
+ FORWARD_IF_ERROR( ZSTD_initLocalDict(cctx) , ""); /* Init the local dict if present. */
+ memset(&cctx->prefixDict, 0, sizeof(cctx->prefixDict)); /* single usage */
+ assert(prefixDict.dict==NULL || cctx->cdict==NULL); /* only one can be set */
+ DEBUGLOG(4, "ZSTD_compressStream2 : transparent init stage");
+ if (endOp == ZSTD_e_end) cctx->pledgedSrcSizePlusOne = input->size + 1; /* auto-fix pledgedSrcSize */
+ params.cParams = ZSTD_getCParamsFromCCtxParams(
+ &cctx->requestedParams, cctx->pledgedSrcSizePlusOne-1, 0 /*dictSize*/);
+
+
+#ifdef ZSTD_MULTITHREAD
+ if ((cctx->pledgedSrcSizePlusOne-1) <= ZSTDMT_JOBSIZE_MIN) {
+ params.nbWorkers = 0; /* do not invoke multi-threading when src size is too small */
+ }
+ if (params.nbWorkers > 0) {
+ /* mt context creation */
+ if (cctx->mtctx == NULL) {
+ DEBUGLOG(4, "ZSTD_compressStream2: creating new mtctx for nbWorkers=%u",
+ params.nbWorkers);
+ cctx->mtctx = ZSTDMT_createCCtx_advanced((U32)params.nbWorkers, cctx->customMem);
+ RETURN_ERROR_IF(cctx->mtctx == NULL, memory_allocation, "NULL pointer!");
+ }
+ /* mt compression */
+ DEBUGLOG(4, "call ZSTDMT_initCStream_internal as nbWorkers=%u", params.nbWorkers);
+ FORWARD_IF_ERROR( ZSTDMT_initCStream_internal(
+ cctx->mtctx,
+ prefixDict.dict, prefixDict.dictSize, prefixDict.dictContentType,
+ cctx->cdict, params, cctx->pledgedSrcSizePlusOne-1) , "");
+ cctx->streamStage = zcss_load;
+ cctx->appliedParams.nbWorkers = params.nbWorkers;
+ } else
+#endif
+ { FORWARD_IF_ERROR( ZSTD_resetCStream_internal(cctx,
+ prefixDict.dict, prefixDict.dictSize, prefixDict.dictContentType,
+ cctx->cdict,
+ params, cctx->pledgedSrcSizePlusOne-1) , "");
+ assert(cctx->streamStage == zcss_load);
+ assert(cctx->appliedParams.nbWorkers == 0);
+ } }
+ /* end of transparent initialization stage */
+
+ /* compression stage */
+#ifdef ZSTD_MULTITHREAD
+ if (cctx->appliedParams.nbWorkers > 0) {
+ int const forceMaxProgress = (endOp == ZSTD_e_flush || endOp == ZSTD_e_end);
+ size_t flushMin;
+ assert(forceMaxProgress || endOp == ZSTD_e_continue /* Protection for a new flush type */);
+ if (cctx->cParamsChanged) {
+ ZSTDMT_updateCParams_whileCompressing(cctx->mtctx, &cctx->requestedParams);
+ cctx->cParamsChanged = 0;
+ }
+ do {
+ flushMin = ZSTDMT_compressStream_generic(cctx->mtctx, output, input, endOp);
+ if ( ZSTD_isError(flushMin)
+ || (endOp == ZSTD_e_end && flushMin == 0) ) { /* compression completed */
+ ZSTD_CCtx_reset(cctx, ZSTD_reset_session_only);
+ }
+ FORWARD_IF_ERROR(flushMin, "ZSTDMT_compressStream_generic failed");
+ } while (forceMaxProgress && flushMin != 0 && output->pos < output->size);
+ DEBUGLOG(5, "completed ZSTD_compressStream2 delegating to ZSTDMT_compressStream_generic");
+ /* Either we don't require maximum forward progress, we've finished the
+ * flush, or we are out of output space.
+ */
+ assert(!forceMaxProgress || flushMin == 0 || output->pos == output->size);
+ return flushMin;
+ }
+#endif
+ FORWARD_IF_ERROR( ZSTD_compressStream_generic(cctx, output, input, endOp) , "");
+ DEBUGLOG(5, "completed ZSTD_compressStream2");
+ 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)
+{
+ DEBUGLOG(4, "ZSTD_compress2 (srcSize=%u)", (unsigned)srcSize);
+ ZSTD_CCtx_reset(cctx, ZSTD_reset_session_only);
+ { 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);
+ 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;
+ }
+}
+
+/*====== 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 =====-*/
+
+#define ZSTD_MAX_CLEVEL 22
+int ZSTD_maxCLevel(void) { return ZSTD_MAX_CLEVEL; }
+int ZSTD_minCLevel(void) { return (int)-ZSTD_TARGETLENGTH_MAX; }
+
+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, 2, 5, 2, ZSTD_greedy }, /* level 5 */
+ { 21, 19, 19, 3, 5, 4, ZSTD_greedy }, /* level 6 */
+ { 21, 19, 19, 3, 5, 8, ZSTD_lazy }, /* level 7 */
+ { 21, 19, 19, 3, 5, 16, ZSTD_lazy2 }, /* level 8 */
+ { 21, 19, 20, 4, 5, 16, ZSTD_lazy2 }, /* level 9 */
+ { 22, 20, 21, 4, 5, 16, ZSTD_lazy2 }, /* level 10 */
+ { 22, 21, 22, 4, 5, 16, ZSTD_lazy2 }, /* level 11 */
+ { 22, 21, 22, 5, 5, 16, ZSTD_lazy2 }, /* level 12 */
+ { 22, 21, 22, 5, 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, 2, 5, 2, ZSTD_greedy }, /* level 4.*/
+ { 18, 18, 18, 3, 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, 17, 17, 3, 4, 4, ZSTD_lazy }, /* level 6 */
+ { 17, 17, 17, 3, 4, 8, ZSTD_lazy2 }, /* level 7 */
+ { 17, 17, 17, 4, 4, 8, ZSTD_lazy2 }, /* level 8 */
+ { 17, 17, 17, 5, 4, 8, ZSTD_lazy2 }, /* level 9 */
+ { 17, 17, 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.*/
+},
+};
+
+/*! 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. */
+static ZSTD_compressionParameters ZSTD_getCParams_internal(int compressionLevel, unsigned long long srcSizeHint, size_t dictSize)
+{
+ int const unknown = srcSizeHint == ZSTD_CONTENTSIZE_UNKNOWN;
+ size_t const addedSize = unknown && dictSize > 0 ? 500 : 0;
+ U64 const rSize = unknown && dictSize == 0 ? ZSTD_CONTENTSIZE_UNKNOWN : srcSizeHint+dictSize+addedSize;
+ U32 const tableID = (rSize <= 256 KB) + (rSize <= 128 KB) + (rSize <= 16 KB);
+ int row = compressionLevel;
+ DEBUGLOG(5, "ZSTD_getCParams_internal (cLevel=%i)", compressionLevel);
+ if (compressionLevel == 0) row = ZSTD_CLEVEL_DEFAULT; /* 0 == default */
+ if (compressionLevel < 0) row = 0; /* entry 0 is baseline for fast mode */
+ if (compressionLevel > ZSTD_MAX_CLEVEL) row = ZSTD_MAX_CLEVEL;
+ { ZSTD_compressionParameters cp = ZSTD_defaultCParameters[tableID][row];
+ if (compressionLevel < 0) cp.targetLength = (unsigned)(-compressionLevel); /* acceleration factor */
+ /* refine parameters based on srcSize & dictSize */
+ return ZSTD_adjustCParams_internal(cp, srcSizeHint, dictSize);
+ }
+}
+
+/*! 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_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_parameters params;
+ ZSTD_compressionParameters const cParams = ZSTD_getCParams_internal(compressionLevel, srcSizeHint, dictSize);
+ DEBUGLOG(5, "ZSTD_getParams (cLevel=%i)", compressionLevel);
+ 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);
+}
diff --git a/src/zstd/lib/compress/zstd_compress_internal.h b/src/zstd/lib/compress/zstd_compress_internal.h
new file mode 100644
index 000000000..db73f6ce2
--- /dev/null
+++ b/src/zstd/lib/compress/zstd_compress_internal.h
@@ -0,0 +1,1125 @@
+/*
+ * Copyright (c) 2016-2020, 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"
+#ifdef ZSTD_MULTITHREAD
+# include "zstdmt_compress.h"
+#endif
+
+#if defined (__cplusplus)
+extern "C" {
+#endif
+
+
+/*-*************************************
+* 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 {
+ U32 CTable[HUF_CTABLE_SIZE_U32(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;
+
+typedef struct {
+ U32 off;
+ U32 len;
+} ZSTD_match_t;
+
+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_literalCompressionMode_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 */
+} ZSTD_window_t;
+
+typedef struct ZSTD_matchState_t ZSTD_matchState_t;
+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* hashTable;
+ U32* hashTable3;
+ U32* chainTable;
+ optState_t opt; /* optimal parser state */
+ const ZSTD_matchState_t* dictMatchState;
+ ZSTD_compressionParameters cParams;
+};
+
+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 {
+ 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 */
+ U64 hashPower; /* Used to compute the rolling hash.
+ * Depends on ldmParams.minMatchLength */
+} ldmState_t;
+
+typedef struct {
+ U32 enableLdm; /* 1 if enable long distance matching */
+ 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 {
+ U32 offset;
+ U32 litLength;
+ U32 matchLength;
+} rawSeq;
+
+typedef struct {
+ rawSeq* seq; /* The start of the sequences */
+ size_t pos; /* The position where reading stopped. <= size. */
+ size_t size; /* The number of sequences. <= capacity. */
+ size_t capacity; /* The capacity starting from `seq` pointer */
+} rawSeqStore_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_literalCompressionMode_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;
+
+ /* Internal use, for createCCtxParams() and freeCCtxParams() only */
+ ZSTD_customMem customMem;
+}; /* typedef'd to ZSTD_CCtx_params within "zstd.h" */
+
+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;
+ U32 dictID;
+
+ 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;
+ XXH64_state_t xxhState;
+ ZSTD_customMem customMem;
+ 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 HUF_WORKSPACE_SIZE bytes */
+
+ /* 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;
+
+ /* Dictionary */
+ ZSTD_localDict localDict;
+ const ZSTD_CDict* cdict;
+ ZSTD_prefixDict prefixDict; /* single-usage dictionary */
+
+ /* Multi-threading */
+#ifdef ZSTD_MULTITHREAD
+ ZSTDMT_CCtx* mtctx;
+#endif
+};
+
+typedef enum { ZSTD_dtlm_fast, ZSTD_dtlm_full } ZSTD_dictTableLoadMethod_e;
+
+typedef enum { ZSTD_noDict = 0, ZSTD_extDict = 1, ZSTD_dictMatchState = 2 } ZSTD_dictMode_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_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];
+}
+
+typedef struct repcodes_s {
+ U32 rep[3];
+} repcodes_t;
+
+MEM_STATIC repcodes_t ZSTD_updateRep(U32 const rep[3], U32 const offset, U32 const ll0)
+{
+ repcodes_t newReps;
+ if (offset >= ZSTD_REP_NUM) { /* full offset */
+ newReps.rep[2] = rep[1];
+ newReps.rep[1] = rep[0];
+ newReps.rep[0] = offset - ZSTD_REP_MOVE;
+ } else { /* repcode */
+ U32 const repCode = offset + ll0;
+ if (repCode > 0) { /* note : if repCode==0, no change */
+ U32 const currentOffset = (repCode==ZSTD_REP_NUM) ? (rep[0] - 1) : rep[repCode];
+ newReps.rep[2] = (repCode >= 2) ? rep[1] : rep[2];
+ newReps.rep[1] = rep[0];
+ newReps.rep[0] = currentOffset;
+ } else { /* repCode == 0 */
+ memcpy(&newReps, rep, sizeof(newReps));
+ }
+ }
+ return newReps;
+}
+
+/* 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);
+ 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_disableLiteralsCompression(const ZSTD_CCtx_params* cctxParams)
+{
+ switch (cctxParams->literalCompressionMode) {
+ case ZSTD_lcm_huffman:
+ return 0;
+ case ZSTD_lcm_uncompressed:
+ return 1;
+ default:
+ assert(0 /* impossible: pre-validated */);
+ /* fall-through */
+ case ZSTD_lcm_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++;
+}
+
+/*! ZSTD_storeSeq() :
+ * Store a sequence (litlen, litPtr, offCode and mlBase) into seqStore_t.
+ * `offCode` : distance to match + ZSTD_REP_MOVE (values <= ZSTD_REP_MOVE are repCodes).
+ * `mlBase` : matchLength - 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 offCode, size_t mlBase)
+{
+ 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)mlBase+MINMATCH, (U32)offCode);
+ }
+#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->longLengthID == 0); /* there can only be a single long length */
+ seqStorePtr->longLengthID = 1;
+ seqStorePtr->longLengthPos = (U32)(seqStorePtr->sequences - seqStorePtr->sequencesStart);
+ }
+ seqStorePtr->sequences[0].litLength = (U16)litLength;
+
+ /* match offset */
+ seqStorePtr->sequences[0].offset = offCode + 1;
+
+ /* match Length */
+ if (mlBase>0xFFFF) {
+ assert(seqStorePtr->longLengthID == 0); /* there can only be a single long length */
+ seqStorePtr->longLengthID = 2;
+ seqStorePtr->longLengthPos = (U32)(seqStorePtr->sequences - seqStorePtr->sequencesStart);
+ }
+ seqStorePtr->sequences[0].matchLength = (U16)mlBase;
+
+ seqStorePtr->sequences++;
+}
+
+
+/*-*************************************
+* Match length counter
+***************************************/
+static unsigned ZSTD_NbCommonBytes (size_t val)
+{
+ if (MEM_isLittleEndian()) {
+ if (MEM_64bits()) {
+# if defined(_MSC_VER) && defined(_WIN64)
+ unsigned long r = 0;
+ return _BitScanForward64( &r, (U64)val ) ? (unsigned)(r >> 3) : 0;
+# elif defined(__GNUC__) && (__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 defined(_MSC_VER)
+ unsigned long r=0;
+ return _BitScanForward( &r, (U32)val ) ? (unsigned)(r >> 3) : 0;
+# elif defined(__GNUC__) && (__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 defined(_MSC_VER) && defined(_WIN64)
+ unsigned long r = 0;
+ return _BitScanReverse64( &r, val ) ? (unsigned)(r >> 3) : 0;
+# elif defined(__GNUC__) && (__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 defined(_MSC_VER)
+ unsigned long r = 0;
+ return _BitScanReverse( &r, (unsigned long)val ) ? (unsigned)(r >> 3) : 0;
+# elif defined(__GNUC__) && (__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 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;
+}
+
+/**
+ * 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 ?
+ ZSTD_dictMatchState :
+ ZSTD_noDict;
+}
+
+/**
+ * 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,
+ void const* srcEnd)
+{
+ U32 const current = (U32)((BYTE const*)srcEnd - window.base);
+ return current > 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.
+ * NOTE: (maxDist & cycleMask) must be zero.
+ */
+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 cycleMask = (1U << cycleLog) - 1;
+ U32 const current = (U32)((BYTE const*)src - window->base);
+ U32 const currentCycle0 = current & cycleMask;
+ /* Exclude zero so that newCurrent - maxDist >= 1. */
+ U32 const currentCycle1 = currentCycle0 == 0 ? (1U << cycleLog) : currentCycle0;
+ U32 const newCurrent = currentCycle1 + maxDist;
+ U32 const correction = current - newCurrent;
+ assert((maxDist & cycleMask) == 0);
+ assert(current > newCurrent);
+ /* Loose bound, should be around 1<<29 (see above) */
+ assert(correction > 1<<28);
+
+ window->base += correction;
+ window->dictBase += correction;
+ if (window->lowLimit <= correction) window->lowLimit = 1;
+ else window->lowLimit -= correction;
+ if (window->dictLimit <= correction) window->dictLimit = 1;
+ else window->dictLimit -= correction;
+
+ /* Ensure we can still reference the full window. */
+ assert(newCurrent >= maxDist);
+ assert(newCurrent - maxDist >= 1);
+ /* Ensure that lowLimit and dictLimit didn't underflow. */
+ assert(window->lowLimit <= newCurrent);
+ assert(window->dictLimit <= newCurrent);
+
+ 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) {
+ memset(window, 0, sizeof(*window));
+ window->base = (BYTE const*)"";
+ window->dictBase = (BYTE const*)"";
+ window->dictLimit = 1; /* start from 1, so that 1st position is valid */
+ window->lowLimit = 1; /* it ensures first and later CCtx usages compress the same */
+ window->nextSrc = window->base + 1; /* see issue #1241 */
+}
+
+/**
+ * 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)
+{
+ 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) {
+ /* 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 current, unsigned windowLog)
+{
+ U32 const maxDistance = 1U << windowLog;
+ U32 const lowestValid = ms->window.lowLimit;
+ U32 const withinWindow = (current - lowestValid > maxDistance) ? current - maxDistance : lowestValid;
+ U32 const isDictionary = (ms->loadedDictEnd != 0);
+ 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 current, unsigned windowLog)
+{
+ U32 const maxDistance = 1U << windowLog;
+ U32 const lowestValid = ms->window.dictLimit;
+ U32 const withinWindow = (current - lowestValid > maxDistance) ? current - maxDistance : lowestValid;
+ U32 const isDictionary = (ms->loadedDictEnd != 0);
+ 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
+
+
+#if defined (__cplusplus)
+}
+#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,
+ short* offcodeNCount, unsigned* offcodeMaxValue,
+ 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_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);
+
+#endif /* ZSTD_COMPRESS_H */
diff --git a/src/zstd/lib/compress/zstd_compress_literals.c b/src/zstd/lib/compress/zstd_compress_literals.c
new file mode 100644
index 000000000..17e7168d8
--- /dev/null
+++ b/src/zstd/lib/compress/zstd_compress_literals.c
@@ -0,0 +1,158 @@
+/*
+ * Copyright (c) 2016-2020, 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* const)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);
+ }
+
+ 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* const)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)
+{
+ 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 */
+ 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) :
+ HUF_compress4X_repeat(
+ ostart+lhSize, dstCapacity-lhSize, src, srcSize,
+ HUF_SYMBOLVALUE_MAX, HUF_TABLELOG_DEFAULT, entropyWorkspace, entropyWorkspaceSize,
+ (HUF_CElt*)nextHuf->CTable, &repeat, preferRepeat, bmi2);
+ 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)) {
+ memcpy(nextHuf, prevHuf, sizeof(*prevHuf));
+ return ZSTD_noCompressLiterals(dst, dstCapacity, src, srcSize);
+ }
+ if (cLitSize==1) {
+ 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/src/zstd/lib/compress/zstd_compress_literals.h b/src/zstd/lib/compress/zstd_compress_literals.h
new file mode 100644
index 000000000..8b0870574
--- /dev/null
+++ b/src/zstd/lib/compress/zstd_compress_literals.h
@@ -0,0 +1,29 @@
+/*
+ * Copyright (c) 2016-2020, 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);
+
+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);
+
+#endif /* ZSTD_COMPRESS_LITERALS_H */
diff --git a/src/zstd/lib/compress/zstd_compress_sequences.c b/src/zstd/lib/compress/zstd_compress_sequences.c
new file mode 100644
index 000000000..f9f8097c8
--- /dev/null
+++ b/src/zstd/lib/compress/zstd_compress_sequences.c
@@ -0,0 +1,419 @@
+/*
+ * Copyright (c) 2016-2020, 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 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), "");
+ 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;
+ 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;
+}
+
+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:
+ 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: {
+ S16 norm[MaxSeq + 1];
+ 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);
+ FORWARD_IF_ERROR(FSE_normalizeCount(norm, tableLog, count, nbSeq_1, max), "");
+ { size_t const NCountSize = FSE_writeNCount(op, oend - op, norm, max, tableLog); /* overflow protected */
+ FORWARD_IF_ERROR(NCountSize, "FSE_writeNCount failed");
+ FORWARD_IF_ERROR(FSE_buildCTable_wksp(nextCTable, norm, max, tableLog, entropyWorkspace, entropyWorkspaceSize), "");
+ 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].matchLength, 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].offset, extraBits);
+ BIT_flushBits(&blockStream);
+ }
+ BIT_addBits(&blockStream, sequences[nbSeq-1].offset >> extraBits,
+ ofBits - extraBits);
+ } else {
+ BIT_addBits(&blockStream, sequences[nbSeq-1].offset, 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].matchLength + MINMATCH,
+ (unsigned)sequences[n].offset);
+ /* 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].matchLength, 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].offset, extraBits);
+ BIT_flushBits(&blockStream); /* (7)*/
+ }
+ BIT_addBits(&blockStream, sequences[n].offset >> extraBits,
+ ofBits - extraBits); /* 31 */
+ } else {
+ BIT_addBits(&blockStream, sequences[n].offset, 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 TARGET_ATTRIBUTE("bmi2") 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/src/zstd/lib/compress/zstd_compress_sequences.h b/src/zstd/lib/compress/zstd_compress_sequences.h
new file mode 100644
index 000000000..68c6f9a5a
--- /dev/null
+++ b/src/zstd/lib/compress/zstd_compress_sequences.h
@@ -0,0 +1,54 @@
+/*
+ * Copyright (c) 2016-2020, 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/src/zstd/lib/compress/zstd_compress_superblock.c b/src/zstd/lib/compress/zstd_compress_superblock.c
new file mode 100644
index 000000000..b693866c0
--- /dev/null
+++ b/src/zstd/lib/compress/zstd_compress_superblock.c
@@ -0,0 +1,845 @@
+/*
+ * Copyright (c) 2016-2020, 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"
+#include "zstd_compress_sequences.h"
+#include "zstd_compress_literals.h"
+
+/*-*************************************
+* Superblock entropy buffer structs
+***************************************/
+/** 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_buildSuperBlockEntropy_literal() */
+typedef struct {
+ symbolEncodingType_e hType;
+ BYTE hufDesBuffer[500]; /* TODO give name to this value */
+ 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_buildSuperBlockEntropy_sequences() */
+typedef struct {
+ symbolEncodingType_e llType;
+ symbolEncodingType_e ofType;
+ symbolEncodingType_e mlType;
+ BYTE fseTablesBuffer[500]; /* TODO give name to this value */
+ size_t fseTablesSize;
+ size_t lastCountSize; /* This is to account for bug in 1.3.4. More detail in ZSTD_compressSubBlock_sequences() */
+} ZSTD_fseCTablesMetadata_t;
+
+typedef struct {
+ ZSTD_hufCTablesMetadata_t hufMetadata;
+ ZSTD_fseCTablesMetadata_t fseMetadata;
+} ZSTD_entropyCTablesMetadata_t;
+
+
+/** ZSTD_buildSuperBlockEntropy_literal() :
+ * Builds entropy for the super-block literals.
+ * Stores literals block type (raw, rle, compressed, repeat) and
+ * huffman description table to hufMetadata.
+ * @return : size of huffman description table or error code */
+static size_t ZSTD_buildSuperBlockEntropy_literal(void* const src, size_t srcSize,
+ const ZSTD_hufCTables_t* prevHuf,
+ ZSTD_hufCTables_t* nextHuf,
+ ZSTD_hufCTablesMetadata_t* hufMetadata,
+ const int disableLiteralsCompression,
+ 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 = 255;
+ unsigned huffLog = HUF_TABLELOG_DEFAULT;
+ HUF_repeat repeat = prevHuf->repeatMode;
+
+ DEBUGLOG(5, "ZSTD_buildSuperBlockEntropy_literal (srcSize=%zu)", srcSize);
+
+ /* Prepare nextEntropy assuming reusing the existing table */
+ memcpy(nextHuf, prevHuf, sizeof(*prevHuf));
+
+ if (disableLiteralsCompression) {
+ DEBUGLOG(5, "set_basic - disabled");
+ hufMetadata->hType = set_basic;
+ return 0;
+ }
+
+ /* 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) {
+ 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 */
+ 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(
+ hufMetadata->hufDesBuffer, sizeof(hufMetadata->hufDesBuffer),
+ (HUF_CElt*)nextHuf->CTable, maxSymbolValue, huffLog);
+ /* 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");
+ memcpy(nextHuf, prevHuf, sizeof(*prevHuf));
+ hufMetadata->hType = set_repeat;
+ return 0;
+ }
+ }
+ if (newCSize + hSize >= srcSize) {
+ DEBUGLOG(5, "set_basic - no gains");
+ 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_buildSuperBlockEntropy_sequences() :
+ * Builds entropy for the super-block sequences.
+ * Stores symbol compression modes and fse table to fseMetadata.
+ * @return : size of fse tables or error code */
+static size_t ZSTD_buildSuperBlockEntropy_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)
+{
+ BYTE* const wkspStart = (BYTE*)workspace;
+ BYTE* const wkspEnd = wkspStart + wkspSize;
+ BYTE* const countWkspStart = wkspStart;
+ unsigned* const countWksp = (unsigned*)workspace;
+ const size_t countWkspSize = (MaxSeq + 1) * sizeof(unsigned);
+ BYTE* const cTableWksp = countWkspStart + countWkspSize;
+ const size_t cTableWkspSize = wkspEnd-cTableWksp;
+ ZSTD_strategy const strategy = cctxParams->cParams.strategy;
+ 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;
+ size_t const nbSeq = seqStorePtr->sequences - seqStorePtr->sequencesStart;
+ BYTE* const ostart = fseMetadata->fseTablesBuffer;
+ BYTE* const oend = ostart + sizeof(fseMetadata->fseTablesBuffer);
+ BYTE* op = ostart;
+
+ assert(cTableWkspSize >= (1 << MaxFSELog) * sizeof(FSE_FUNCTION_TYPE));
+ DEBUGLOG(5, "ZSTD_buildSuperBlockEntropy_sequences (nbSeq=%zu)", nbSeq);
+ memset(workspace, 0, wkspSize);
+
+ fseMetadata->lastCountSize = 0;
+ /* convert length/distances into codes */
+ ZSTD_seqToCodes(seqStorePtr);
+ /* build CTable for Literal Lengths */
+ { U32 LLtype;
+ unsigned max = MaxLL;
+ size_t const mostFrequent = HIST_countFast_wksp(countWksp, &max, llCodeTable, nbSeq, workspace, wkspSize); /* can't fail */
+ DEBUGLOG(5, "Building LL table");
+ nextEntropy->litlength_repeatMode = prevEntropy->litlength_repeatMode;
+ LLtype = ZSTD_selectEncodingType(&nextEntropy->litlength_repeatMode,
+ countWksp, max, mostFrequent, nbSeq,
+ LLFSELog, prevEntropy->litlengthCTable,
+ LL_defaultNorm, LL_defaultNormLog,
+ ZSTD_defaultAllowed, strategy);
+ assert(set_basic < set_compressed && set_rle < set_compressed);
+ assert(!(LLtype < set_compressed && nextEntropy->litlength_repeatMode != FSE_repeat_none)); /* We don't copy tables */
+ { size_t const countSize = ZSTD_buildCTable(op, oend - op, CTable_LitLength, LLFSELog, (symbolEncodingType_e)LLtype,
+ countWksp, max, llCodeTable, nbSeq, LL_defaultNorm, LL_defaultNormLog, MaxLL,
+ prevEntropy->litlengthCTable, sizeof(prevEntropy->litlengthCTable),
+ cTableWksp, cTableWkspSize);
+ FORWARD_IF_ERROR(countSize, "ZSTD_buildCTable for LitLens failed");
+ if (LLtype == set_compressed)
+ fseMetadata->lastCountSize = countSize;
+ op += countSize;
+ fseMetadata->llType = (symbolEncodingType_e) LLtype;
+ } }
+ /* build CTable for Offsets */
+ { U32 Offtype;
+ unsigned max = MaxOff;
+ size_t const mostFrequent = HIST_countFast_wksp(countWksp, &max, ofCodeTable, nbSeq, workspace, wkspSize); /* 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;
+ Offtype = ZSTD_selectEncodingType(&nextEntropy->offcode_repeatMode,
+ countWksp, max, mostFrequent, nbSeq,
+ OffFSELog, prevEntropy->offcodeCTable,
+ OF_defaultNorm, OF_defaultNormLog,
+ defaultPolicy, strategy);
+ assert(!(Offtype < set_compressed && nextEntropy->offcode_repeatMode != FSE_repeat_none)); /* We don't copy tables */
+ { size_t const countSize = ZSTD_buildCTable(op, oend - op, CTable_OffsetBits, OffFSELog, (symbolEncodingType_e)Offtype,
+ countWksp, max, ofCodeTable, nbSeq, OF_defaultNorm, OF_defaultNormLog, DefaultMaxOff,
+ prevEntropy->offcodeCTable, sizeof(prevEntropy->offcodeCTable),
+ cTableWksp, cTableWkspSize);
+ FORWARD_IF_ERROR(countSize, "ZSTD_buildCTable for Offsets failed");
+ if (Offtype == set_compressed)
+ fseMetadata->lastCountSize = countSize;
+ op += countSize;
+ fseMetadata->ofType = (symbolEncodingType_e) Offtype;
+ } }
+ /* build CTable for MatchLengths */
+ { U32 MLtype;
+ unsigned max = MaxML;
+ size_t const mostFrequent = HIST_countFast_wksp(countWksp, &max, mlCodeTable, nbSeq, workspace, wkspSize); /* can't fail */
+ DEBUGLOG(5, "Building ML table (remaining space : %i)", (int)(oend-op));
+ nextEntropy->matchlength_repeatMode = prevEntropy->matchlength_repeatMode;
+ MLtype = ZSTD_selectEncodingType(&nextEntropy->matchlength_repeatMode,
+ countWksp, max, mostFrequent, nbSeq,
+ MLFSELog, prevEntropy->matchlengthCTable,
+ ML_defaultNorm, ML_defaultNormLog,
+ ZSTD_defaultAllowed, strategy);
+ assert(!(MLtype < set_compressed && nextEntropy->matchlength_repeatMode != FSE_repeat_none)); /* We don't copy tables */
+ { size_t const countSize = ZSTD_buildCTable(op, oend - op, CTable_MatchLength, MLFSELog, (symbolEncodingType_e)MLtype,
+ countWksp, max, mlCodeTable, nbSeq, ML_defaultNorm, ML_defaultNormLog, MaxML,
+ prevEntropy->matchlengthCTable, sizeof(prevEntropy->matchlengthCTable),
+ cTableWksp, cTableWkspSize);
+ FORWARD_IF_ERROR(countSize, "ZSTD_buildCTable for MatchLengths failed");
+ if (MLtype == set_compressed)
+ fseMetadata->lastCountSize = countSize;
+ op += countSize;
+ fseMetadata->mlType = (symbolEncodingType_e) MLtype;
+ } }
+ assert((size_t) (op-ostart) <= sizeof(fseMetadata->fseTablesBuffer));
+ return op-ostart;
+}
+
+
+/** ZSTD_buildSuperBlockEntropy() :
+ * Builds entropy for the super-block.
+ * @return : 0 on success or error code */
+static size_t
+ZSTD_buildSuperBlockEntropy(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;
+ DEBUGLOG(5, "ZSTD_buildSuperBlockEntropy");
+ entropyMetadata->hufMetadata.hufDesSize =
+ ZSTD_buildSuperBlockEntropy_literal(seqStorePtr->litStart, litSize,
+ &prevEntropy->huf, &nextEntropy->huf,
+ &entropyMetadata->hufMetadata,
+ ZSTD_disableLiteralsCompression(cctxParams),
+ workspace, wkspSize);
+ FORWARD_IF_ERROR(entropyMetadata->hufMetadata.hufDesSize, "ZSTD_buildSuperBlockEntropy_literal failed");
+ entropyMetadata->fseMetadata.fseTablesSize =
+ ZSTD_buildSuperBlockEntropy_sequences(seqStorePtr,
+ &prevEntropy->fse, &nextEntropy->fse,
+ cctxParams,
+ &entropyMetadata->fseMetadata,
+ workspace, wkspSize);
+ FORWARD_IF_ERROR(entropyMetadata->fseMetadata.fseTablesSize, "ZSTD_buildSuperBlockEntropy_sequences failed");
+ return 0;
+}
+
+/** 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 entropyWrriten=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) {
+ 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;
+ 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));
+ 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 U32* additionalBits,
+ short const* defaultNorm, U32 defaultNormLog,
+ 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) {
+ 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 / 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 sequencesSectionHeaderSize = 3; /* Use hard coded size of 3 bytes */
+ size_t cSeqSizeEstimate = 0;
+ cSeqSizeEstimate += ZSTD_estimateSubBlockSize_symbolType(fseMetadata->ofType, ofCodeTable, MaxOff,
+ nbSeq, fseTables->offcodeCTable, NULL,
+ OF_defaultNorm, OF_defaultNormLog,
+ workspace, wkspSize);
+ cSeqSizeEstimate += ZSTD_estimateSubBlockSize_symbolType(fseMetadata->llType, llCodeTable, MaxLL,
+ nbSeq, fseTables->litlengthCTable, LL_bits,
+ LL_defaultNorm, LL_defaultNormLog,
+ workspace, wkspSize);
+ cSeqSizeEstimate += ZSTD_estimateSubBlockSize_symbolType(fseMetadata->mlType, mlCodeTable, MaxML,
+ nbSeq, fseTables->matchlengthCTable, ML_bits,
+ ML_defaultNorm, ML_defaultNormLog,
+ 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 everytime.
+ */
+ 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");
+ 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;
+ memcpy(&rep, prevCBlock->rep, sizeof(rep));
+ for (seq = sstart; seq < sp; ++seq) {
+ rep = ZSTD_updateRep(rep.rep, seq->offset - 1, ZSTD_getSequenceLength(seqStorePtr, seq).litLength == 0);
+ }
+ 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_buildSuperBlockEntropy(&zc->seqStore,
+ &zc->blockState.prevCBlock->entropy,
+ &zc->blockState.nextCBlock->entropy,
+ &zc->appliedParams,
+ &entropyMetadata,
+ zc->entropyWorkspace, HUF_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, HUF_WORKSPACE_SIZE /* statically allocated in resetCCtx */);
+}
diff --git a/src/zstd/lib/compress/zstd_compress_superblock.h b/src/zstd/lib/compress/zstd_compress_superblock.h
new file mode 100644
index 000000000..07f4cb1dc
--- /dev/null
+++ b/src/zstd/lib/compress/zstd_compress_superblock.h
@@ -0,0 +1,32 @@
+/*
+ * Copyright (c) 2016-2020, 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 "../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/src/zstd/lib/compress/zstd_cwksp.h b/src/zstd/lib/compress/zstd_cwksp.h
new file mode 100644
index 000000000..a25c9263b
--- /dev/null
+++ b/src/zstd/lib/compress/zstd_cwksp.h
@@ -0,0 +1,525 @@
+/*
+ * Copyright (c) 2016-2020, 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"
+
+#if defined (__cplusplus)
+extern "C" {
+#endif
+
+/*-*************************************
+* 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
+
+/*-*************************************
+* Structures
+***************************************/
+typedef enum {
+ ZSTD_cwksp_alloc_objects,
+ ZSTD_cwksp_alloc_buffers,
+ ZSTD_cwksp_alloc_aligned
+} ZSTD_cwksp_alloc_phase_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_free{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.
+ *
+ * - Aligned: these buffers are used for various purposes that require 4 byte
+ * alignment, but don't require any initialization before they're used.
+ *
+ * - 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
+ * 4. 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;
+
+ int allocFailed;
+ int workspaceOversizedDuration;
+ ZSTD_cwksp_alloc_phase_e phase;
+} 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.
+ */
+MEM_STATIC size_t ZSTD_cwksp_alloc_size(size_t size) {
+#if defined (ADDRESS_SANITIZER) && !defined (ZSTD_ASAN_DONT_POISON_WORKSPACE)
+ return size + 2 * ZSTD_CWKSP_ASAN_REDZONE_SIZE;
+#else
+ return size;
+#endif
+}
+
+MEM_STATIC void ZSTD_cwksp_internal_advance_phase(
+ ZSTD_cwksp* ws, ZSTD_cwksp_alloc_phase_e phase) {
+ assert(phase >= ws->phase);
+ if (phase > ws->phase) {
+ if (ws->phase < ZSTD_cwksp_alloc_buffers &&
+ phase >= ZSTD_cwksp_alloc_buffers) {
+ ws->tableValidEnd = ws->objectEnd;
+ }
+ if (ws->phase < ZSTD_cwksp_alloc_aligned &&
+ phase >= ZSTD_cwksp_alloc_aligned) {
+ /* If unaligned allocations down from a too-large top have left us
+ * unaligned, we need to realign our alloc ptr. Technically, this
+ * can consume space that is unaccounted for in the neededSpace
+ * calculation. However, I believe this can only happen when the
+ * workspace is too large, and specifically when it is too large
+ * by a larger margin than the space that will be consumed. */
+ /* TODO: cleaner, compiler warning friendly way to do this??? */
+ ws->allocStart = (BYTE*)ws->allocStart - ((size_t)ws->allocStart & (sizeof(U32)-1));
+ if (ws->allocStart < ws->tableValidEnd) {
+ ws->tableValidEnd = ws->allocStart;
+ }
+ }
+ ws->phase = phase;
+ }
+}
+
+/**
+ * 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;
+ void* bottom = ws->tableEnd;
+ ZSTD_cwksp_internal_advance_phase(ws, phase);
+ alloc = (BYTE *)ws->allocStart - bytes;
+
+#if defined (ADDRESS_SANITIZER) && !defined (ZSTD_ASAN_DONT_POISON_WORKSPACE)
+ /* over-reserve space */
+ alloc = (BYTE *)alloc - 2 * ZSTD_CWKSP_ASAN_REDZONE_SIZE;
+#endif
+
+ 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;
+ }
+ if (alloc < ws->tableValidEnd) {
+ ws->tableValidEnd = alloc;
+ }
+ ws->allocStart = alloc;
+
+#if defined (ADDRESS_SANITIZER) && !defined (ZSTD_ASAN_DONT_POISON_WORKSPACE)
+ /* Move alloc so there's ZSTD_CWKSP_ASAN_REDZONE_SIZE unused space on
+ * either size. */
+ alloc = (BYTE *)alloc + ZSTD_CWKSP_ASAN_REDZONE_SIZE;
+ __asan_unpoison_memory_region(alloc, bytes);
+#endif
+
+ 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 sizeof(unsigned).
+ */
+MEM_STATIC void* ZSTD_cwksp_reserve_aligned(ZSTD_cwksp* ws, size_t bytes) {
+ assert((bytes & (sizeof(U32)-1)) == 0);
+ return ZSTD_cwksp_reserve_internal(ws, ZSTD_cwksp_align(bytes, sizeof(U32)), ZSTD_cwksp_alloc_aligned);
+}
+
+/**
+ * Aligned on sizeof(unsigned). 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 = ws->tableEnd;
+ void* end = (BYTE *)alloc + bytes;
+ void* 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_internal_advance_phase(ws, phase);
+ 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;
+
+#if defined (ADDRESS_SANITIZER) && !defined (ZSTD_ASAN_DONT_POISON_WORKSPACE)
+ __asan_unpoison_memory_region(alloc, bytes);
+#endif
+
+ return alloc;
+}
+
+/**
+ * Aligned on sizeof(void*).
+ */
+MEM_STATIC void* ZSTD_cwksp_reserve_object(ZSTD_cwksp* ws, size_t bytes) {
+ size_t roundedBytes = ZSTD_cwksp_align(bytes, sizeof(void*));
+ void* alloc = ws->objectEnd;
+ void* end = (BYTE*)alloc + roundedBytes;
+
+#if defined (ADDRESS_SANITIZER) && !defined (ZSTD_ASAN_DONT_POISON_WORKSPACE)
+ /* over-reserve space */
+ end = (BYTE *)end + 2 * ZSTD_CWKSP_ASAN_REDZONE_SIZE;
+#endif
+
+ DEBUGLOG(5,
+ "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 & (sizeof(void*)-1)) == 0);
+ assert((bytes & (sizeof(void*)-1)) == 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(4, "cwksp: object alloc failed!");
+ ws->allocFailed = 1;
+ return NULL;
+ }
+ ws->objectEnd = end;
+ ws->tableEnd = end;
+ ws->tableValidEnd = end;
+
+#if defined (ADDRESS_SANITIZER) && !defined (ZSTD_ASAN_DONT_POISON_WORKSPACE)
+ /* Move alloc so there's ZSTD_CWKSP_ASAN_REDZONE_SIZE unused space on
+ * either size. */
+ alloc = (BYTE *)alloc + ZSTD_CWKSP_ASAN_REDZONE_SIZE;
+ __asan_unpoison_memory_region(alloc, bytes);
+#endif
+
+ return alloc;
+}
+
+MEM_STATIC void ZSTD_cwksp_mark_tables_dirty(ZSTD_cwksp* ws) {
+ DEBUGLOG(4, "cwksp: ZSTD_cwksp_mark_tables_dirty");
+
+#if defined (MEMORY_SANITIZER) && !defined (ZSTD_MSAN_DONT_POISON_WORKSPACE)
+ /* To validate that the table re-use logic is sound, and that we don't
+ * access table space that we haven't cleaned, we re-"poison" the table
+ * space every time we mark it dirty. */
+ {
+ size_t size = (BYTE*)ws->tableValidEnd - (BYTE*)ws->objectEnd;
+ assert(__msan_test_shadow(ws->objectEnd, size) == -1);
+ __msan_poison(ws->objectEnd, size);
+ }
+#endif
+
+ 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) {
+ 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!");
+
+#if defined (ADDRESS_SANITIZER) && !defined (ZSTD_ASAN_DONT_POISON_WORKSPACE)
+ {
+ size_t size = (BYTE*)ws->tableValidEnd - (BYTE*)ws->objectEnd;
+ __asan_poison_memory_region(ws->objectEnd, size);
+ }
+#endif
+
+ 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!");
+
+#if defined (MEMORY_SANITIZER) && !defined (ZSTD_MSAN_DONT_POISON_WORKSPACE)
+ /* To validate that the context re-use logic is sound, and that we don't
+ * access stuff that this compression hasn't initialized, we re-"poison"
+ * the workspace (or at least the non-static, non-table parts of it)
+ * every time we start a new compression. */
+ {
+ size_t size = (BYTE*)ws->workspaceEnd - (BYTE*)ws->tableValidEnd;
+ __msan_poison(ws->tableValidEnd, size);
+ }
+#endif
+
+#if defined (ADDRESS_SANITIZER) && !defined (ZSTD_ASAN_DONT_POISON_WORKSPACE)
+ {
+ size_t size = (BYTE*)ws->workspaceEnd - (BYTE*)ws->objectEnd;
+ __asan_poison_memory_region(ws->objectEnd, size);
+ }
+#endif
+
+ 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) {
+ 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;
+ 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_malloc(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);
+ return 0;
+}
+
+MEM_STATIC void ZSTD_cwksp_free(ZSTD_cwksp* ws, ZSTD_customMem customMem) {
+ void *ptr = ws->workspace;
+ DEBUGLOG(4, "cwksp: freeing workspace");
+ memset(ws, 0, sizeof(ZSTD_cwksp));
+ ZSTD_free(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 its used again).
+ */
+MEM_STATIC void ZSTD_cwksp_move(ZSTD_cwksp* dst, ZSTD_cwksp* src) {
+ *dst = *src;
+ 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 int ZSTD_cwksp_reserve_failed(const ZSTD_cwksp* ws) {
+ return ws->allocFailed;
+}
+
+/*-*************************************
+* Functions Checking Free Space
+***************************************/
+
+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;
+ }
+}
+
+#if defined (__cplusplus)
+}
+#endif
+
+#endif /* ZSTD_CWKSP_H */
diff --git a/src/zstd/lib/compress/zstd_double_fast.c b/src/zstd/lib/compress/zstd_double_fast.c
new file mode 100644
index 000000000..27eed66cf
--- /dev/null
+++ b/src/zstd/lib/compress/zstd_double_fast.c
@@ -0,0 +1,521 @@
+/*
+ * Copyright (c) 2016-2020, 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 current = (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] = current + i;
+ if (i == 0 || hashLarge[lgHash] == 0)
+ hashLarge[lgHash] = current + i;
+ /* Only load extra positions for ZSTD_dtlm_full */
+ if (dtlm == ZSTD_dtlm_fast)
+ break;
+ } }
+}
+
+
+FORCE_INLINE_TEMPLATE
+size_t ZSTD_compressBlock_doubleFast_generic(
+ ZSTD_matchState_t* ms, seqStore_t* seqStore, U32 rep[ZSTD_REP_NUM],
+ void const* src, size_t srcSize,
+ U32 const mls /* template */, ZSTD_dictMode_e const dictMode)
+{
+ 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 =
+ dictMode == ZSTD_dictMatchState ?
+ &dms->cParams : NULL;
+ const U32* const dictHashLong = dictMode == ZSTD_dictMatchState ?
+ dms->hashTable : NULL;
+ const U32* const dictHashSmall = dictMode == ZSTD_dictMatchState ?
+ dms->chainTable : NULL;
+ const U32 dictStartIndex = dictMode == ZSTD_dictMatchState ?
+ dms->window.dictLimit : 0;
+ const BYTE* const dictBase = dictMode == ZSTD_dictMatchState ?
+ dms->window.base : NULL;
+ const BYTE* const dictStart = dictMode == ZSTD_dictMatchState ?
+ dictBase + dictStartIndex : NULL;
+ const BYTE* const dictEnd = dictMode == ZSTD_dictMatchState ?
+ dms->window.nextSrc : NULL;
+ const U32 dictIndexDelta = dictMode == ZSTD_dictMatchState ?
+ prefixLowestIndex - (U32)(dictEnd - dictBase) :
+ 0;
+ const U32 dictHBitsL = dictMode == ZSTD_dictMatchState ?
+ dictCParams->hashLog : hBitsL;
+ const U32 dictHBitsS = dictMode == ZSTD_dictMatchState ?
+ dictCParams->chainLog : hBitsS;
+ const U32 dictAndPrefixLength = (U32)((ip - prefixLowest) + (dictEnd - dictStart));
+
+ DEBUGLOG(5, "ZSTD_compressBlock_doubleFast_generic");
+
+ assert(dictMode == ZSTD_noDict || dictMode == ZSTD_dictMatchState);
+
+ /* if a dictionary is attached, it must be within window range */
+ if (dictMode == ZSTD_dictMatchState) {
+ assert(ms->window.dictLimit + (1U << cParams->windowLog) >= endIndex);
+ }
+
+ /* init */
+ ip += (dictAndPrefixLength == 0);
+ if (dictMode == ZSTD_noDict) {
+ 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;
+ }
+ if (dictMode == ZSTD_dictMatchState) {
+ /* 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 current = (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 = current + 1 - offset_1;
+ const BYTE* repMatch = (dictMode == ZSTD_dictMatchState
+ && repIndex < prefixLowestIndex) ?
+ dictBase + (repIndex - dictIndexDelta) :
+ base + repIndex;
+ hashLong[h2] = hashSmall[h] = current; /* update hash tables */
+
+ /* check dictMatchState repcode */
+ if (dictMode == ZSTD_dictMatchState
+ && ((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, 0, mLength-MINMATCH);
+ goto _match_stored;
+ }
+
+ /* check noDict repcode */
+ if ( dictMode == ZSTD_noDict
+ && ((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, 0, mLength-MINMATCH);
+ 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 if (dictMode == ZSTD_dictMatchState) {
+ /* 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)(current - 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 if (dictMode == ZSTD_dictMatchState) {
+ /* 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] = current + 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 if (dictMode == ZSTD_dictMatchState) {
+ /* 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)(current + 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 (dictMode == ZSTD_dictMatchState && matchIndexS < prefixLowestIndex) {
+ mLength = ZSTD_count_2segments(ip+4, match+4, iend, dictEnd, prefixLowest) + 4;
+ offset = (U32)(current - 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 */
+ }
+
+ /* fall-through */
+
+_match_found:
+ offset_2 = offset_1;
+ offset_1 = offset;
+
+ ZSTD_storeSeq(seqStore, (size_t)(ip-anchor), anchor, iend, offset + ZSTD_REP_MOVE, mLength-MINMATCH);
+
+_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 = current+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 */
+ if (dictMode == ZSTD_dictMatchState) {
+ while (ip <= ilimit) {
+ U32 const current2 = (U32)(ip-base);
+ U32 const repIndex2 = current2 - offset_2;
+ const BYTE* repMatch2 = dictMode == ZSTD_dictMatchState
+ && 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, 0, repLength2-MINMATCH);
+ hashSmall[ZSTD_hashPtr(ip, hBitsS, mls)] = current2;
+ hashLong[ZSTD_hashPtr(ip, hBitsL, 8)] = current2;
+ ip += repLength2;
+ anchor = ip;
+ continue;
+ }
+ break;
+ } }
+
+ if (dictMode == ZSTD_noDict) {
+ 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, 0, rLength-MINMATCH);
+ ip += rLength;
+ anchor = ip;
+ continue; /* faster when present ... (?) */
+ } } }
+ } /* 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);
+}
+
+
+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_generic(ms, seqStore, rep, src, srcSize, 4, ZSTD_noDict);
+ case 5 :
+ return ZSTD_compressBlock_doubleFast_generic(ms, seqStore, rep, src, srcSize, 5, ZSTD_noDict);
+ case 6 :
+ return ZSTD_compressBlock_doubleFast_generic(ms, seqStore, rep, src, srcSize, 6, ZSTD_noDict);
+ case 7 :
+ return ZSTD_compressBlock_doubleFast_generic(ms, seqStore, rep, src, srcSize, 7, ZSTD_noDict);
+ }
+}
+
+
+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_generic(ms, seqStore, rep, src, srcSize, 4, ZSTD_dictMatchState);
+ case 5 :
+ return ZSTD_compressBlock_doubleFast_generic(ms, seqStore, rep, src, srcSize, 5, ZSTD_dictMatchState);
+ case 6 :
+ return ZSTD_compressBlock_doubleFast_generic(ms, seqStore, rep, src, srcSize, 6, ZSTD_dictMatchState);
+ case 7 :
+ return ZSTD_compressBlock_doubleFast_generic(ms, seqStore, rep, src, srcSize, 7, ZSTD_dictMatchState);
+ }
+}
+
+
+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_generic(ms, seqStore, rep, src, srcSize, mls, ZSTD_noDict);
+
+ /* 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 current = (U32)(ip-base);
+ const U32 repIndex = current + 1 - offset_1; /* offset_1 expected <= current +1 */
+ const BYTE* const repBase = repIndex < prefixStartIndex ? dictBase : base;
+ const BYTE* const repMatch = repBase + repIndex;
+ size_t mLength;
+ hashSmall[hSmall] = hashLong[hLong] = current; /* update hash table */
+
+ if ((((U32)((prefixStartIndex-1) - repIndex) >= 3) /* intentional underflow : ensure repIndex doesn't overlap dict + prefix */
+ & (repIndex > dictStartIndex))
+ && (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, 0, mLength-MINMATCH);
+ } 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 = current - 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, offset + ZSTD_REP_MOVE, mLength-MINMATCH);
+
+ } 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] = current + 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 = current+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 = current - 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, offset + ZSTD_REP_MOVE, mLength-MINMATCH);
+
+ } 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 = current+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 */
+ & (repIndex2 > 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, 0, repLength2-MINMATCH);
+ 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);
+}
+
+
+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_generic(ms, seqStore, rep, src, srcSize, 4);
+ case 5 :
+ return ZSTD_compressBlock_doubleFast_extDict_generic(ms, seqStore, rep, src, srcSize, 5);
+ case 6 :
+ return ZSTD_compressBlock_doubleFast_extDict_generic(ms, seqStore, rep, src, srcSize, 6);
+ case 7 :
+ return ZSTD_compressBlock_doubleFast_extDict_generic(ms, seqStore, rep, src, srcSize, 7);
+ }
+}
diff --git a/src/zstd/lib/compress/zstd_double_fast.h b/src/zstd/lib/compress/zstd_double_fast.h
new file mode 100644
index 000000000..14d944d69
--- /dev/null
+++ b/src/zstd/lib/compress/zstd_double_fast.h
@@ -0,0 +1,38 @@
+/*
+ * Copyright (c) 2016-2020, 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
+
+#if defined (__cplusplus)
+extern "C" {
+#endif
+
+#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);
+
+
+#if defined (__cplusplus)
+}
+#endif
+
+#endif /* ZSTD_DOUBLE_FAST_H */
diff --git a/src/zstd/lib/compress/zstd_fast.c b/src/zstd/lib/compress/zstd_fast.c
new file mode 100644
index 000000000..85a3a7a91
--- /dev/null
+++ b/src/zstd/lib/compress/zstd_fast.c
@@ -0,0 +1,496 @@
+/*
+ * Copyright (c) 2016-2020, 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 current = (U32)(ip - base);
+ size_t const hash0 = ZSTD_hashPtr(ip, hBits, mls);
+ hashTable[hash0] = current;
+ 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] = current + p;
+ } } } }
+}
+
+
+FORCE_INLINE_TEMPLATE size_t
+ZSTD_compressBlock_fast_generic(
+ ZSTD_matchState_t* ms, seqStore_t* seqStore, U32 rep[ZSTD_REP_NUM],
+ void const* src, size_t srcSize,
+ U32 const mls)
+{
+ 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 = cParams->targetLength + !(cParams->targetLength) + 1;
+ const BYTE* const base = ms->window.base;
+ const BYTE* const istart = (const BYTE*)src;
+ /* We check ip0 (ip + 0) and ip1 (ip + 1) each loop */
+ const BYTE* ip0 = istart;
+ const BYTE* ip1;
+ const BYTE* anchor = istart;
+ 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;
+ U32 offset_1=rep[0], offset_2=rep[1];
+ U32 offsetSaved = 0;
+
+ /* init */
+ DEBUGLOG(5, "ZSTD_compressBlock_fast_generic");
+ ip0 += (ip0 == prefixStart);
+ ip1 = ip0 + 1;
+ { U32 const current = (U32)(ip0 - 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;
+ }
+
+ /* Main Search Loop */
+#ifdef __INTEL_COMPILER
+ /* From intel 'The vector pragma indicates that the loop should be
+ * vectorized if it is legal to do so'. Can be used together with
+ * #pragma ivdep (but have opted to exclude that because intel
+ * warns against using it).*/
+ #pragma vector always
+#endif
+ while (ip1 < ilimit) { /* < instead of <=, because check at ip0+2 */
+ size_t mLength;
+ BYTE const* ip2 = ip0 + 2;
+ size_t const h0 = ZSTD_hashPtr(ip0, hlog, mls);
+ U32 const val0 = MEM_read32(ip0);
+ size_t const h1 = ZSTD_hashPtr(ip1, hlog, mls);
+ U32 const val1 = MEM_read32(ip1);
+ U32 const current0 = (U32)(ip0-base);
+ U32 const current1 = (U32)(ip1-base);
+ U32 const matchIndex0 = hashTable[h0];
+ U32 const matchIndex1 = hashTable[h1];
+ BYTE const* repMatch = ip2 - offset_1;
+ const BYTE* match0 = base + matchIndex0;
+ const BYTE* match1 = base + matchIndex1;
+ U32 offcode;
+
+#if defined(__aarch64__)
+ PREFETCH_L1(ip0+256);
+#endif
+
+ hashTable[h0] = current0; /* update hash table */
+ hashTable[h1] = current1; /* update hash table */
+
+ assert(ip0 + 1 == ip1);
+
+ if ((offset_1 > 0) & (MEM_read32(repMatch) == MEM_read32(ip2))) {
+ mLength = (ip2[-1] == repMatch[-1]) ? 1 : 0;
+ ip0 = ip2 - mLength;
+ match0 = repMatch - mLength;
+ mLength += 4;
+ offcode = 0;
+ goto _match;
+ }
+ if ((matchIndex0 > prefixStartIndex) && MEM_read32(match0) == val0) {
+ /* found a regular match */
+ goto _offset;
+ }
+ if ((matchIndex1 > prefixStartIndex) && MEM_read32(match1) == val1) {
+ /* found a regular match after one literal */
+ ip0 = ip1;
+ match0 = match1;
+ goto _offset;
+ }
+ { size_t const step = ((size_t)(ip0-anchor) >> (kSearchStrength - 1)) + stepSize;
+ assert(step >= 2);
+ ip0 += step;
+ ip1 += step;
+ continue;
+ }
+_offset: /* Requires: ip0, match0 */
+ /* Compute the offset code */
+ offset_2 = offset_1;
+ offset_1 = (U32)(ip0-match0);
+ offcode = offset_1 + ZSTD_REP_MOVE;
+ mLength = 4;
+ /* Count the backwards match length */
+ while (((ip0>anchor) & (match0>prefixStart))
+ && (ip0[-1] == match0[-1])) { ip0--; match0--; mLength++; } /* catch up */
+
+_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-MINMATCH);
+ /* match found */
+ ip0 += mLength;
+ anchor = ip0;
+
+ 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 (offset_2 > 0) { /* offset_2==0 means offset_2 is invalidated */
+ while ( (ip0 <= ilimit) && (MEM_read32(ip0) == MEM_read32(ip0 - offset_2)) ) {
+ /* store sequence */
+ size_t const rLength = ZSTD_count(ip0+4, ip0+4-offset_2, iend) + 4;
+ { U32 const tmpOff = offset_2; offset_2 = offset_1; offset_1 = tmpOff; } /* swap offset_2 <=> offset_1 */
+ hashTable[ZSTD_hashPtr(ip0, hlog, mls)] = (U32)(ip0-base);
+ ip0 += rLength;
+ ZSTD_storeSeq(seqStore, 0 /*litLen*/, anchor, iend, 0 /*offCode*/, rLength-MINMATCH);
+ anchor = ip0;
+ continue; /* faster when present (confirmed on gcc-8) ... (?) */
+ } } }
+ ip1 = ip0 + 1;
+ }
+
+ /* 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);
+}
+
+
+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);
+ switch(mls)
+ {
+ default: /* includes case 3 */
+ case 4 :
+ return ZSTD_compressBlock_fast_generic(ms, seqStore, rep, src, srcSize, 4);
+ case 5 :
+ return ZSTD_compressBlock_fast_generic(ms, seqStore, rep, src, srcSize, 5);
+ case 6 :
+ return ZSTD_compressBlock_fast_generic(ms, seqStore, rep, src, srcSize, 6);
+ case 7 :
+ return ZSTD_compressBlock_fast_generic(ms, seqStore, rep, src, srcSize, 7);
+ }
+}
+
+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)
+{
+ 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 */
+
+ /* ensure there will be no 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 current = (U32)(ip-base);
+ U32 const matchIndex = hashTable[h];
+ const BYTE* match = base + matchIndex;
+ const U32 repIndex = current + 1 - offset_1;
+ const BYTE* repMatch = (repIndex < prefixStartIndex) ?
+ dictBase + (repIndex - dictIndexDelta) :
+ base + repIndex;
+ hashTable[h] = current; /* 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, 0, mLength-MINMATCH);
+ } 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)(current-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, offset + ZSTD_REP_MOVE, mLength-MINMATCH);
+ }
+ } 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, offset + ZSTD_REP_MOVE, mLength-MINMATCH);
+ }
+
+ /* match found */
+ ip += mLength;
+ anchor = ip;
+
+ if (ip <= ilimit) {
+ /* Fill Table */
+ assert(base+current+2 > istart); /* check base overflow */
+ hashTable[ZSTD_hashPtr(base+current+2, hlog, mls)] = current+2; /* here because current+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, 0, repLength2-MINMATCH);
+ 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);
+}
+
+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_generic(ms, seqStore, rep, src, srcSize, 4);
+ case 5 :
+ return ZSTD_compressBlock_fast_dictMatchState_generic(ms, seqStore, rep, src, srcSize, 5);
+ case 6 :
+ return ZSTD_compressBlock_fast_dictMatchState_generic(ms, seqStore, rep, src, srcSize, 6);
+ case 7 :
+ return ZSTD_compressBlock_fast_dictMatchState_generic(ms, seqStore, rep, src, srcSize, 7);
+ }
+}
+
+
+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)
+{
+ 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];
+
+ 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_generic(ms, seqStore, rep, src, srcSize, mls);
+
+ /* 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 current = (U32)(ip-base);
+ const U32 repIndex = current + 1 - offset_1;
+ const BYTE* const repBase = repIndex < prefixStartIndex ? dictBase : base;
+ const BYTE* const repMatch = repBase + repIndex;
+ hashTable[h] = current; /* update hash table */
+ DEBUGLOG(7, "offset_1 = %u , current = %u", offset_1, current);
+ assert(offset_1 <= current +1); /* check repIndex */
+
+ if ( (((U32)((prefixStartIndex-1) - repIndex) >= 3) /* intentional underflow */ & (repIndex > dictStartIndex))
+ && (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, 0, rLength-MINMATCH);
+ 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 = current - 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, offset + ZSTD_REP_MOVE, mLength-MINMATCH);
+ ip += mLength;
+ anchor = ip;
+ } }
+
+ if (ip <= ilimit) {
+ /* Fill Table */
+ hashTable[ZSTD_hashPtr(base+current+2, hlog, mls)] = current+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) & (repIndex2 > 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, 0 /*offcode*/, repLength2-MINMATCH);
+ 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);
+}
+
+
+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_generic(ms, seqStore, rep, src, srcSize, 4);
+ case 5 :
+ return ZSTD_compressBlock_fast_extDict_generic(ms, seqStore, rep, src, srcSize, 5);
+ case 6 :
+ return ZSTD_compressBlock_fast_extDict_generic(ms, seqStore, rep, src, srcSize, 6);
+ case 7 :
+ return ZSTD_compressBlock_fast_extDict_generic(ms, seqStore, rep, src, srcSize, 7);
+ }
+}
diff --git a/src/zstd/lib/compress/zstd_fast.h b/src/zstd/lib/compress/zstd_fast.h
new file mode 100644
index 000000000..cf6aaa8e6
--- /dev/null
+++ b/src/zstd/lib/compress/zstd_fast.h
@@ -0,0 +1,37 @@
+/*
+ * Copyright (c) 2016-2020, 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
+
+#if defined (__cplusplus)
+extern "C" {
+#endif
+
+#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);
+
+#if defined (__cplusplus)
+}
+#endif
+
+#endif /* ZSTD_FAST_H */
diff --git a/src/zstd/lib/compress/zstd_lazy.c b/src/zstd/lib/compress/zstd_lazy.c
new file mode 100644
index 000000000..4cf5c88b5
--- /dev/null
+++ b/src/zstd/lib/compress/zstd_lazy.c
@@ -0,0 +1,1138 @@
+/*
+ * Copyright (c) 2016-2020, 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 : current >= btlow == (current - btmask)
+ * doesn't fail */
+static void
+ZSTD_insertDUBT1(ZSTD_matchState_t* ms,
+ U32 current, 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 = (current>=dictLimit) ? base + current : dictBase + current;
+ const BYTE* const iend = (current>=dictLimit) ? inputEnd : dictBase + dictLimit;
+ const BYTE* const dictEnd = dictBase + dictLimit;
+ const BYTE* const prefixStart = base + dictLimit;
+ const BYTE* match;
+ U32* smallerPtr = bt + 2*(current&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 = (current - windowValid > maxDistance) ? current - maxDistance : windowValid;
+
+
+ DEBUGLOG(8, "ZSTD_insertDUBT1(%u) (dictLimit=%u, lowLimit=%u)",
+ current, dictLimit, windowLow);
+ assert(current >= btLow);
+ assert(ip < iend); /* condition for ZSTD_count */
+
+ while (nbCompares-- && (matchIndex > windowLow)) {
+ U32* const nextPtr = bt + 2*(matchIndex & btMask);
+ size_t matchLength = MIN(commonLengthSmaller, commonLengthLarger); /* guaranteed minimum nb of common bytes */
+ assert(matchIndex < current);
+ /* 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*/
+ || (current < 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 */
+ || (current < 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 ",
+ current, 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 (
+ 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 current = (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);
+
+ while (nbCompares-- && (dictMatchIndex > dictLowLimit)) {
+ 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(current-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)",
+ current, (U32)bestLength, (U32)matchLength, (U32)*offsetPtr, ZSTD_REP_MOVE + current - matchIndex, dictMatchIndex, matchIndex);
+ bestLength = matchLength, *offsetPtr = ZSTD_REP_MOVE + current - 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 = current - ((U32)*offsetPtr - ZSTD_REP_MOVE); (void)mIndex;
+ DEBUGLOG(8, "ZSTD_DUBT_findBetterDictMatch(%u) : found match of length %u and offsetCode %u (pos %u)",
+ current, (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 current = (U32)(ip-base);
+ U32 const windowLow = ZSTD_getLowestMatchIndex(ms, current, cParams->windowLog);
+
+ U32* const bt = ms->chainTable;
+ U32 const btLog = cParams->chainLog - 1;
+ U32 const btMask = (1 << btLog) - 1;
+ U32 const btLow = (btMask >= current) ? 0 : current - 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) ", current);
+ assert(ip <= iend-8); /* required for h calculation */
+
+ /* 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*(current&btMask);
+ U32* largerPtr = bt + 2*(current&btMask) + 1;
+ U32 matchEndIdx = current + 8 + 1;
+ U32 dummy32; /* to be nullified at the end */
+ size_t bestLength = 0;
+
+ matchIndex = hashTable[h];
+ hashTable[h] = current; /* Update Hash Table */
+
+ while (nbCompares-- && (matchIndex > windowLow)) {
+ 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(current-matchIndex+1) - ZSTD_highbit32((U32)offsetPtr[0]+1)) )
+ bestLength = matchLength, *offsetPtr = ZSTD_REP_MOVE + current - 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;
+
+ if (dictMode == ZSTD_dictMatchState && nbCompares) {
+ bestLength = ZSTD_DUBT_findBetterDictMatch(
+ ms, ip, iend,
+ offsetPtr, bestLength, nbCompares,
+ mls, dictMode);
+ }
+
+ assert(matchEndIdx > current+8); /* ensure nextToUpdate is increased */
+ ms->nextToUpdate = matchEndIdx - 8; /* skip repetitive patterns */
+ if (bestLength >= MINMATCH) {
+ U32 const mIndex = current - ((U32)*offsetPtr - ZSTD_REP_MOVE); (void)mIndex;
+ DEBUGLOG(8, "ZSTD_DUBT_findBestMatch(%u) : found match of length %u and offsetCode %u (pos %u)",
+ current, (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);
+}
+
+
+static size_t
+ZSTD_BtFindBestMatch_selectMLS ( ZSTD_matchState_t* ms,
+ const BYTE* ip, const BYTE* const iLimit,
+ size_t* offsetPtr)
+{
+ switch(ms->cParams.minMatch)
+ {
+ default : /* includes case 3 */
+ case 4 : return ZSTD_BtFindBestMatch(ms, ip, iLimit, offsetPtr, 4, ZSTD_noDict);
+ case 5 : return ZSTD_BtFindBestMatch(ms, ip, iLimit, offsetPtr, 5, ZSTD_noDict);
+ case 7 :
+ case 6 : return ZSTD_BtFindBestMatch(ms, ip, iLimit, offsetPtr, 6, ZSTD_noDict);
+ }
+}
+
+
+static size_t ZSTD_BtFindBestMatch_dictMatchState_selectMLS (
+ ZSTD_matchState_t* ms,
+ const BYTE* ip, const BYTE* const iLimit,
+ size_t* offsetPtr)
+{
+ switch(ms->cParams.minMatch)
+ {
+ default : /* includes case 3 */
+ case 4 : return ZSTD_BtFindBestMatch(ms, ip, iLimit, offsetPtr, 4, ZSTD_dictMatchState);
+ case 5 : return ZSTD_BtFindBestMatch(ms, ip, iLimit, offsetPtr, 5, ZSTD_dictMatchState);
+ case 7 :
+ case 6 : return ZSTD_BtFindBestMatch(ms, ip, iLimit, offsetPtr, 6, ZSTD_dictMatchState);
+ }
+}
+
+
+static size_t ZSTD_BtFindBestMatch_extDict_selectMLS (
+ ZSTD_matchState_t* ms,
+ const BYTE* ip, const BYTE* const iLimit,
+ size_t* offsetPtr)
+{
+ switch(ms->cParams.minMatch)
+ {
+ default : /* includes case 3 */
+ case 4 : return ZSTD_BtFindBestMatch(ms, ip, iLimit, offsetPtr, 4, ZSTD_extDict);
+ case 5 : return ZSTD_BtFindBestMatch(ms, ip, iLimit, offsetPtr, 5, ZSTD_extDict);
+ case 7 :
+ case 6 : return ZSTD_BtFindBestMatch(ms, ip, iLimit, offsetPtr, 6, ZSTD_extDict);
+ }
+}
+
+
+
+/* *********************************
+* 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) */
+static 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_generic (
+ 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 current = (U32)(ip-base);
+ const U32 maxDistance = 1U << cParams->windowLog;
+ const U32 lowestValid = ms->window.lowLimit;
+ const U32 withinMaxDistance = (current - lowestValid > maxDistance) ? current - maxDistance : lowestValid;
+ const U32 isDictionary = (ms->loadedDictEnd != 0);
+ const U32 lowLimit = isDictionary ? lowestValid : withinMaxDistance;
+ const U32 minChain = current > chainSize ? current - chainSize : 0;
+ U32 nbAttempts = 1U << cParams->searchLog;
+ size_t ml=4-1;
+
+ /* HC4 match finder */
+ U32 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 = current - matchIndex + ZSTD_REP_MOVE;
+ if (ip+currentMl == iLimit) break; /* best possible, avoids read overflow on next attempt */
+ }
+
+ if (matchIndex <= minChain) break;
+ matchIndex = NEXT_IN_CHAIN(matchIndex, chainMask);
+ }
+
+ if (dictMode == ZSTD_dictMatchState) {
+ const ZSTD_matchState_t* const dms = ms->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;
+ *offsetPtr = current - (matchIndex + dmsIndexDelta) + ZSTD_REP_MOVE;
+ if (ip+currentMl == iLimit) break; /* best possible, avoids read overflow on next attempt */
+ }
+
+ if (matchIndex <= dmsMinChain) break;
+ matchIndex = dmsChainTable[matchIndex & dmsChainMask];
+ }
+ }
+
+ return ml;
+}
+
+
+FORCE_INLINE_TEMPLATE size_t ZSTD_HcFindBestMatch_selectMLS (
+ ZSTD_matchState_t* ms,
+ const BYTE* ip, const BYTE* const iLimit,
+ size_t* offsetPtr)
+{
+ switch(ms->cParams.minMatch)
+ {
+ default : /* includes case 3 */
+ case 4 : return ZSTD_HcFindBestMatch_generic(ms, ip, iLimit, offsetPtr, 4, ZSTD_noDict);
+ case 5 : return ZSTD_HcFindBestMatch_generic(ms, ip, iLimit, offsetPtr, 5, ZSTD_noDict);
+ case 7 :
+ case 6 : return ZSTD_HcFindBestMatch_generic(ms, ip, iLimit, offsetPtr, 6, ZSTD_noDict);
+ }
+}
+
+
+static size_t ZSTD_HcFindBestMatch_dictMatchState_selectMLS (
+ ZSTD_matchState_t* ms,
+ const BYTE* ip, const BYTE* const iLimit,
+ size_t* offsetPtr)
+{
+ switch(ms->cParams.minMatch)
+ {
+ default : /* includes case 3 */
+ case 4 : return ZSTD_HcFindBestMatch_generic(ms, ip, iLimit, offsetPtr, 4, ZSTD_dictMatchState);
+ case 5 : return ZSTD_HcFindBestMatch_generic(ms, ip, iLimit, offsetPtr, 5, ZSTD_dictMatchState);
+ case 7 :
+ case 6 : return ZSTD_HcFindBestMatch_generic(ms, ip, iLimit, offsetPtr, 6, ZSTD_dictMatchState);
+ }
+}
+
+
+FORCE_INLINE_TEMPLATE size_t ZSTD_HcFindBestMatch_extDict_selectMLS (
+ ZSTD_matchState_t* ms,
+ const BYTE* ip, const BYTE* const iLimit,
+ size_t* offsetPtr)
+{
+ switch(ms->cParams.minMatch)
+ {
+ default : /* includes case 3 */
+ case 4 : return ZSTD_HcFindBestMatch_generic(ms, ip, iLimit, offsetPtr, 4, ZSTD_extDict);
+ case 5 : return ZSTD_HcFindBestMatch_generic(ms, ip, iLimit, offsetPtr, 5, ZSTD_extDict);
+ case 7 :
+ case 6 : return ZSTD_HcFindBestMatch_generic(ms, ip, iLimit, offsetPtr, 6, ZSTD_extDict);
+ }
+}
+
+
+/* *******************************
+* Common parser - lazy strategy
+*********************************/
+typedef enum { search_hashChain, search_binaryTree } searchMethod_e;
+
+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 = iend - 8;
+ const BYTE* const base = ms->window.base;
+ const U32 prefixLowestIndex = ms->window.dictLimit;
+ const BYTE* const prefixLowest = base + prefixLowestIndex;
+
+ typedef size_t (*searchMax_f)(
+ ZSTD_matchState_t* ms,
+ const BYTE* ip, const BYTE* iLimit, size_t* offsetPtr);
+ searchMax_f const searchMax = dictMode == ZSTD_dictMatchState ?
+ (searchMethod==search_binaryTree ? ZSTD_BtFindBestMatch_dictMatchState_selectMLS
+ : ZSTD_HcFindBestMatch_dictMatchState_selectMLS) :
+ (searchMethod==search_binaryTree ? ZSTD_BtFindBestMatch_selectMLS
+ : ZSTD_HcFindBestMatch_selectMLS);
+ U32 offset_1 = rep[0], offset_2 = rep[1], savedOffset=0;
+
+ const ZSTD_matchState_t* const dms = ms->dictMatchState;
+ const U32 dictLowestIndex = dictMode == ZSTD_dictMatchState ?
+ dms->window.dictLimit : 0;
+ const BYTE* const dictBase = dictMode == ZSTD_dictMatchState ?
+ dms->window.base : NULL;
+ const BYTE* const dictLowest = dictMode == ZSTD_dictMatchState ?
+ dictBase + dictLowestIndex : NULL;
+ const BYTE* const dictEnd = dictMode == ZSTD_dictMatchState ?
+ dms->window.nextSrc : NULL;
+ const U32 dictIndexDelta = dictMode == ZSTD_dictMatchState ?
+ prefixLowestIndex - (U32)(dictEnd - dictBase) :
+ 0;
+ const U32 dictAndPrefixLength = (U32)((ip - prefixLowest) + (dictEnd - dictLowest));
+
+ DEBUGLOG(5, "ZSTD_compressBlock_lazy_generic (dictMode=%u)", (U32)dictMode);
+
+ /* init */
+ ip += (dictAndPrefixLength == 0);
+ if (dictMode == ZSTD_noDict) {
+ U32 const current = (U32)(ip - base);
+ U32 const windowLow = ZSTD_getLowestPrefixIndex(ms, current, ms->cParams.windowLog);
+ U32 const maxRep = current - windowLow;
+ if (offset_2 > maxRep) savedOffset = offset_2, offset_2 = 0;
+ if (offset_1 > maxRep) savedOffset = offset_1, offset_1 = 0;
+ }
+ if (dictMode == ZSTD_dictMatchState) {
+ /* dictMatchState repCode checks don't currently handle repCode == 0
+ * disabling. */
+ assert(offset_1 <= dictAndPrefixLength);
+ assert(offset_2 <= dictAndPrefixLength);
+ }
+
+ /* Match Loop */
+#if defined(__GNUC__) && 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 offset=0;
+ const BYTE* start=ip+1;
+
+ /* check repCode */
+ if (dictMode == ZSTD_dictMatchState) {
+ const U32 repIndex = (U32)(ip - base) + 1 - offset_1;
+ const BYTE* repMatch = (dictMode == ZSTD_dictMatchState
+ && 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 = searchMax(ms, ip, iend, &offsetFound);
+ if (ml2 > matchLength)
+ matchLength = ml2, start = ip, offset=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 ++;
+ if ( (dictMode == ZSTD_noDict)
+ && (offset) && ((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)offset+1) + 1);
+ if ((mlRep >= 4) && (gain2 > gain1))
+ matchLength = mlRep, offset = 0, start = ip;
+ }
+ if (dictMode == ZSTD_dictMatchState) {
+ 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)offset+1) + 1);
+ if ((mlRep >= 4) && (gain2 > gain1))
+ matchLength = mlRep, offset = 0, start = ip;
+ }
+ }
+ { size_t offset2=999999999;
+ size_t const ml2 = searchMax(ms, ip, iend, &offset2);
+ int const gain2 = (int)(ml2*4 - ZSTD_highbit32((U32)offset2+1)); /* raw approx */
+ int const gain1 = (int)(matchLength*4 - ZSTD_highbit32((U32)offset+1) + 4);
+ if ((ml2 >= 4) && (gain2 > gain1)) {
+ matchLength = ml2, offset = offset2, start = ip;
+ continue; /* search a better one */
+ } }
+
+ /* let's find an even better one */
+ if ((depth==2) && (ip<ilimit)) {
+ ip ++;
+ if ( (dictMode == ZSTD_noDict)
+ && (offset) && ((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)offset+1) + 1);
+ if ((mlRep >= 4) && (gain2 > gain1))
+ matchLength = mlRep, offset = 0, start = ip;
+ }
+ if (dictMode == ZSTD_dictMatchState) {
+ 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)offset+1) + 1);
+ if ((mlRep >= 4) && (gain2 > gain1))
+ matchLength = mlRep, offset = 0, start = ip;
+ }
+ }
+ { size_t offset2=999999999;
+ size_t const ml2 = searchMax(ms, ip, iend, &offset2);
+ int const gain2 = (int)(ml2*4 - ZSTD_highbit32((U32)offset2+1)); /* raw approx */
+ int const gain1 = (int)(matchLength*4 - ZSTD_highbit32((U32)offset+1) + 7);
+ if ((ml2 >= 4) && (gain2 > gain1)) {
+ matchLength = ml2, offset = offset2, start = ip;
+ continue;
+ } } }
+ break; /* nothing found : store previous solution */
+ }
+
+ /* NOTE:
+ * start[-offset+ZSTD_REP_MOVE-1] is undefined behavior.
+ * (-offset+ZSTD_REP_MOVE-1) is unsigned, and is added to start, which
+ * overflows the pointer, which is undefined behavior.
+ */
+ /* catch up */
+ if (offset) {
+ if (dictMode == ZSTD_noDict) {
+ while ( ((start > anchor) & (start - (offset-ZSTD_REP_MOVE) > prefixLowest))
+ && (start[-1] == (start-(offset-ZSTD_REP_MOVE))[-1]) ) /* only search for offset within prefix */
+ { start--; matchLength++; }
+ }
+ if (dictMode == ZSTD_dictMatchState) {
+ U32 const matchIndex = (U32)((start-base) - (offset - ZSTD_REP_MOVE));
+ 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)(offset - ZSTD_REP_MOVE);
+ }
+ /* store sequence */
+_storeSequence:
+ { size_t const litLength = start - anchor;
+ ZSTD_storeSeq(seqStore, litLength, anchor, iend, (U32)offset, matchLength-MINMATCH);
+ anchor = ip = start + matchLength;
+ }
+
+ /* check immediate repcode */
+ if (dictMode == ZSTD_dictMatchState) {
+ while (ip <= ilimit) {
+ U32 const current2 = (U32)(ip-base);
+ U32 const repIndex = current2 - offset_2;
+ const BYTE* repMatch = dictMode == ZSTD_dictMatchState
+ && 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;
+ offset = offset_2; offset_2 = offset_1; offset_1 = (U32)offset; /* swap offset_2 <=> offset_1 */
+ ZSTD_storeSeq(seqStore, 0, anchor, iend, 0, matchLength-MINMATCH);
+ 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;
+ offset = offset_2; offset_2 = offset_1; offset_1 = (U32)offset; /* swap repcodes */
+ ZSTD_storeSeq(seqStore, 0, anchor, iend, 0, matchLength-MINMATCH);
+ 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);
+}
+
+
+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 = 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;
+
+ typedef size_t (*searchMax_f)(
+ ZSTD_matchState_t* ms,
+ const BYTE* ip, const BYTE* iLimit, size_t* offsetPtr);
+ searchMax_f searchMax = searchMethod==search_binaryTree ? ZSTD_BtFindBestMatch_extDict_selectMLS : ZSTD_HcFindBestMatch_extDict_selectMLS;
+
+ U32 offset_1 = rep[0], offset_2 = rep[1];
+
+ DEBUGLOG(5, "ZSTD_compressBlock_lazy_extDict_generic");
+
+ /* init */
+ ip += (ip == prefixStart);
+
+ /* Match Loop */
+#if defined(__GNUC__) && 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 offset=0;
+ const BYTE* start=ip+1;
+ U32 current = (U32)(ip-base);
+
+ /* check repCode */
+ { const U32 windowLow = ZSTD_getLowestMatchIndex(ms, current+1, windowLog);
+ const U32 repIndex = (U32)(current+1 - offset_1);
+ const BYTE* const repBase = repIndex < dictLimit ? dictBase : base;
+ const BYTE* const repMatch = repBase + repIndex;
+ if (((U32)((dictLimit-1) - repIndex) >= 3) & (repIndex > windowLow)) /* intentional overflow */
+ 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 = searchMax(ms, ip, iend, &offsetFound);
+ if (ml2 > matchLength)
+ matchLength = ml2, start = ip, offset=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 ++;
+ current++;
+ /* check repCode */
+ if (offset) {
+ const U32 windowLow = ZSTD_getLowestMatchIndex(ms, current, windowLog);
+ const U32 repIndex = (U32)(current - offset_1);
+ const BYTE* const repBase = repIndex < dictLimit ? dictBase : base;
+ const BYTE* const repMatch = repBase + repIndex;
+ if (((U32)((dictLimit-1) - repIndex) >= 3) & (repIndex > windowLow)) /* intentional overflow */
+ 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)offset+1) + 1);
+ if ((repLength >= 4) && (gain2 > gain1))
+ matchLength = repLength, offset = 0, start = ip;
+ } }
+
+ /* search match, depth 1 */
+ { size_t offset2=999999999;
+ size_t const ml2 = searchMax(ms, ip, iend, &offset2);
+ int const gain2 = (int)(ml2*4 - ZSTD_highbit32((U32)offset2+1)); /* raw approx */
+ int const gain1 = (int)(matchLength*4 - ZSTD_highbit32((U32)offset+1) + 4);
+ if ((ml2 >= 4) && (gain2 > gain1)) {
+ matchLength = ml2, offset = offset2, start = ip;
+ continue; /* search a better one */
+ } }
+
+ /* let's find an even better one */
+ if ((depth==2) && (ip<ilimit)) {
+ ip ++;
+ current++;
+ /* check repCode */
+ if (offset) {
+ const U32 windowLow = ZSTD_getLowestMatchIndex(ms, current, windowLog);
+ const U32 repIndex = (U32)(current - offset_1);
+ const BYTE* const repBase = repIndex < dictLimit ? dictBase : base;
+ const BYTE* const repMatch = repBase + repIndex;
+ if (((U32)((dictLimit-1) - repIndex) >= 3) & (repIndex > windowLow)) /* intentional overflow */
+ 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)offset+1) + 1);
+ if ((repLength >= 4) && (gain2 > gain1))
+ matchLength = repLength, offset = 0, start = ip;
+ } }
+
+ /* search match, depth 2 */
+ { size_t offset2=999999999;
+ size_t const ml2 = searchMax(ms, ip, iend, &offset2);
+ int const gain2 = (int)(ml2*4 - ZSTD_highbit32((U32)offset2+1)); /* raw approx */
+ int const gain1 = (int)(matchLength*4 - ZSTD_highbit32((U32)offset+1) + 7);
+ if ((ml2 >= 4) && (gain2 > gain1)) {
+ matchLength = ml2, offset = offset2, start = ip;
+ continue;
+ } } }
+ break; /* nothing found : store previous solution */
+ }
+
+ /* catch up */
+ if (offset) {
+ U32 const matchIndex = (U32)((start-base) - (offset - ZSTD_REP_MOVE));
+ 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)(offset - ZSTD_REP_MOVE);
+ }
+
+ /* store sequence */
+_storeSequence:
+ { size_t const litLength = start - anchor;
+ ZSTD_storeSeq(seqStore, litLength, anchor, iend, (U32)offset, matchLength-MINMATCH);
+ 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) & (repIndex > windowLow)) /* intentional overflow */
+ 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;
+ offset = offset_2; offset_2 = offset_1; offset_1 = (U32)offset; /* swap offset history */
+ ZSTD_storeSeq(seqStore, 0, anchor, iend, 0, matchLength-MINMATCH);
+ 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);
+}
diff --git a/src/zstd/lib/compress/zstd_lazy.h b/src/zstd/lib/compress/zstd_lazy.h
new file mode 100644
index 000000000..581936f03
--- /dev/null
+++ b/src/zstd/lib/compress/zstd_lazy.h
@@ -0,0 +1,67 @@
+/*
+ * Copyright (c) 2016-2020, 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
+
+#if defined (__cplusplus)
+extern "C" {
+#endif
+
+#include "zstd_compress_internal.h"
+
+U32 ZSTD_insertAndFindFirstIndex(ZSTD_matchState_t* ms, const BYTE* 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_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_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_btlazy2_extDict(
+ ZSTD_matchState_t* ms, seqStore_t* seqStore, U32 rep[ZSTD_REP_NUM],
+ void const* src, size_t srcSize);
+
+#if defined (__cplusplus)
+}
+#endif
+
+#endif /* ZSTD_LAZY_H */
diff --git a/src/zstd/lib/compress/zstd_ldm.c b/src/zstd/lib/compress/zstd_ldm.c
new file mode 100644
index 000000000..8c4794835
--- /dev/null
+++ b/src/zstd/lib/compress/zstd_ldm.c
@@ -0,0 +1,619 @@
+/*
+ * Copyright (c) 2016-2020, 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 "zstd_fast.h" /* ZSTD_fillHashTable() */
+#include "zstd_double_fast.h" /* ZSTD_fillDoubleHashTable() */
+
+#define LDM_BUCKET_SIZE_LOG 3
+#define LDM_MIN_MATCH_LENGTH 64
+#define LDM_HASH_RLOG 7
+#define LDM_HASH_CHAR_OFFSET 10
+
+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 (cParams->strategy >= ZSTD_btopt) {
+ /* Get out of the way of the optimal parser */
+ U32 const minMatch = MAX(cParams->targetLength, params->minMatchLength);
+ assert(minMatch >= ZSTD_LDM_MINMATCH_MIN);
+ assert(minMatch <= ZSTD_LDM_MINMATCH_MAX);
+ params->minMatchLength = minMatch;
+ }
+ 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 ? totalSize : 0;
+}
+
+size_t ZSTD_ldm_getMaxNbSeq(ldmParams_t params, size_t maxChunkSize)
+{
+ return params.enableLdm ? (maxChunkSize / params.minMatchLength) : 0;
+}
+
+/** ZSTD_ldm_getSmallHash() :
+ * numBits should be <= 32
+ * If numBits==0, returns 0.
+ * @return : the most significant numBits of value. */
+static U32 ZSTD_ldm_getSmallHash(U64 value, U32 numBits)
+{
+ assert(numBits <= 32);
+ return numBits == 0 ? 0 : (U32)(value >> (64 - numBits));
+}
+
+/** ZSTD_ldm_getChecksum() :
+ * numBitsToDiscard should be <= 32
+ * @return : the next most significant 32 bits after numBitsToDiscard */
+static U32 ZSTD_ldm_getChecksum(U64 hash, U32 numBitsToDiscard)
+{
+ assert(numBitsToDiscard <= 32);
+ return (hash >> (64 - 32 - numBitsToDiscard)) & 0xFFFFFFFF;
+}
+
+/** ZSTD_ldm_getTag() ;
+ * Given the hash, returns the most significant numTagBits bits
+ * after (32 + hbits) bits.
+ *
+ * If there are not enough bits remaining, return the last
+ * numTagBits bits. */
+static U32 ZSTD_ldm_getTag(U64 hash, U32 hbits, U32 numTagBits)
+{
+ assert(numTagBits < 32 && hbits <= 32);
+ if (32 - hbits < numTagBits) {
+ return hash & (((U32)1 << numTagBits) - 1);
+ } else {
+ return (hash >> (32 - hbits - numTagBits)) & (((U32)1 << numTagBits) - 1);
+ }
+}
+
+/** 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 bucketOffsets = ldmState->bucketOffsets;
+ *(ZSTD_ldm_getBucket(ldmState, hash, ldmParams) + bucketOffsets[hash]) = entry;
+ bucketOffsets[hash]++;
+ bucketOffsets[hash] &= ((U32)1 << ldmParams.bucketSizeLog) - 1;
+}
+
+/** ZSTD_ldm_makeEntryAndInsertByTag() :
+ *
+ * Gets the small hash, checksum, and tag from the rollingHash.
+ *
+ * If the tag matches (1 << ldmParams.hashRateLog)-1, then
+ * creates an ldmEntry from the offset, and inserts it into the hash table.
+ *
+ * hBits is the length of the small hash, which is the most significant hBits
+ * of rollingHash. The checksum is the next 32 most significant bits, followed
+ * by ldmParams.hashRateLog bits that make up the tag. */
+static void ZSTD_ldm_makeEntryAndInsertByTag(ldmState_t* ldmState,
+ U64 const rollingHash,
+ U32 const hBits,
+ U32 const offset,
+ ldmParams_t const ldmParams)
+{
+ U32 const tag = ZSTD_ldm_getTag(rollingHash, hBits, ldmParams.hashRateLog);
+ U32 const tagMask = ((U32)1 << ldmParams.hashRateLog) - 1;
+ if (tag == tagMask) {
+ U32 const hash = ZSTD_ldm_getSmallHash(rollingHash, hBits);
+ U32 const checksum = ZSTD_ldm_getChecksum(rollingHash, hBits);
+ ldmEntry_t entry;
+ entry.offset = offset;
+ entry.checksum = checksum;
+ ZSTD_ldm_insertEntry(ldmState, hash, entry, ldmParams);
+ }
+}
+
+/** 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* pBase)
+{
+ size_t matchLength = 0;
+ while (pIn > pAnchor && pMatch > pBase && pIn[-1] == pMatch[-1]) {
+ pIn--;
+ pMatch--;
+ 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;
+}
+
+/** ZSTD_ldm_fillLdmHashTable() :
+ *
+ * Fills hashTable from (lastHashed + 1) to iend (non-inclusive).
+ * lastHash is the rolling hash that corresponds to lastHashed.
+ *
+ * Returns the rolling hash corresponding to position iend-1. */
+static U64 ZSTD_ldm_fillLdmHashTable(ldmState_t* state,
+ U64 lastHash, const BYTE* lastHashed,
+ const BYTE* iend, const BYTE* base,
+ U32 hBits, ldmParams_t const ldmParams)
+{
+ U64 rollingHash = lastHash;
+ const BYTE* cur = lastHashed + 1;
+
+ while (cur < iend) {
+ rollingHash = ZSTD_rollingHash_rotate(rollingHash, cur[-1],
+ cur[ldmParams.minMatchLength-1],
+ state->hashPower);
+ ZSTD_ldm_makeEntryAndInsertByTag(state,
+ rollingHash, hBits,
+ (U32)(cur - base), ldmParams);
+ ++cur;
+ }
+ return rollingHash;
+}
+
+void ZSTD_ldm_fillHashTable(
+ ldmState_t* state, const BYTE* ip,
+ const BYTE* iend, ldmParams_t const* params)
+{
+ DEBUGLOG(5, "ZSTD_ldm_fillHashTable");
+ if ((size_t)(iend - ip) >= params->minMatchLength) {
+ U64 startingHash = ZSTD_rollingHash_compute(ip, params->minMatchLength);
+ ZSTD_ldm_fillLdmHashTable(
+ state, startingHash, ip, iend - params->minMatchLength, state->window.base,
+ params->hashLog - params->bucketSizeLog,
+ *params);
+ }
+}
+
+
+/** 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 current = (U32)(anchor - ms->window.base);
+ if (current > ms->nextToUpdate + 1024) {
+ ms->nextToUpdate =
+ current - MIN(512, current - 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;
+ U64 const hashPower = ldmState->hashPower;
+ U32 const hBits = params->hashLog - params->bucketSizeLog;
+ U32 const ldmBucketSize = 1U << params->bucketSizeLog;
+ U32 const hashRateLog = params->hashRateLog;
+ U32 const ldmTagMask = (1U << params->hashRateLog) - 1;
+ /* 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 - MAX(minMatchLength, HASH_READ_SIZE);
+ /* Input positions */
+ BYTE const* anchor = istart;
+ BYTE const* ip = istart;
+ /* Rolling hash */
+ BYTE const* lastHashed = NULL;
+ U64 rollingHash = 0;
+
+ while (ip <= ilimit) {
+ size_t mLength;
+ U32 const current = (U32)(ip - base);
+ size_t forwardMatchLength = 0, backwardMatchLength = 0;
+ ldmEntry_t* bestEntry = NULL;
+ if (ip != istart) {
+ rollingHash = ZSTD_rollingHash_rotate(rollingHash, lastHashed[0],
+ lastHashed[minMatchLength],
+ hashPower);
+ } else {
+ rollingHash = ZSTD_rollingHash_compute(ip, minMatchLength);
+ }
+ lastHashed = ip;
+
+ /* Do not insert and do not look for a match */
+ if (ZSTD_ldm_getTag(rollingHash, hBits, hashRateLog) != ldmTagMask) {
+ ip++;
+ continue;
+ }
+
+ /* Get the best entry and compute the match lengths */
+ {
+ ldmEntry_t* const bucket =
+ ZSTD_ldm_getBucket(ldmState,
+ ZSTD_ldm_getSmallHash(rollingHash, hBits),
+ *params);
+ ldmEntry_t* cur;
+ size_t bestMatchLength = 0;
+ U32 const checksum = ZSTD_ldm_getChecksum(rollingHash, hBits);
+
+ for (cur = bucket; cur < bucket + ldmBucketSize; ++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(
+ ip, pMatch, iend,
+ matchEnd, lowPrefixPtr);
+ if (curForwardMatchLength < minMatchLength) {
+ continue;
+ }
+ curBackwardMatchLength =
+ ZSTD_ldm_countBackwardsMatch(ip, anchor, pMatch,
+ lowMatchPtr);
+ curTotalMatchLength = curForwardMatchLength +
+ curBackwardMatchLength;
+ } else { /* !extDict */
+ BYTE const* const pMatch = base + cur->offset;
+ curForwardMatchLength = ZSTD_count(ip, pMatch, iend);
+ if (curForwardMatchLength < minMatchLength) {
+ continue;
+ }
+ curBackwardMatchLength =
+ ZSTD_ldm_countBackwardsMatch(ip, anchor, pMatch,
+ lowPrefixPtr);
+ curTotalMatchLength = curForwardMatchLength +
+ curBackwardMatchLength;
+ }
+
+ if (curTotalMatchLength > bestMatchLength) {
+ bestMatchLength = curTotalMatchLength;
+ forwardMatchLength = curForwardMatchLength;
+ backwardMatchLength = curBackwardMatchLength;
+ bestEntry = cur;
+ }
+ }
+ }
+
+ /* No match found -- continue searching */
+ if (bestEntry == NULL) {
+ ZSTD_ldm_makeEntryAndInsertByTag(ldmState, rollingHash,
+ hBits, current,
+ *params);
+ ip++;
+ continue;
+ }
+
+ /* Match found */
+ mLength = forwardMatchLength + backwardMatchLength;
+ ip -= backwardMatchLength;
+
+ {
+ /* Store the sequence:
+ * ip = current - backwardMatchLength
+ * The match is at (bestEntry->offset - backwardMatchLength)
+ */
+ U32 const matchIndex = bestEntry->offset;
+ U32 const offset = current - matchIndex;
+ rawSeq* const seq = rawSeqStore->seq + rawSeqStore->size;
+
+ /* Out of sequence storage */
+ if (rawSeqStore->size == rawSeqStore->capacity)
+ return ERROR(dstSize_tooSmall);
+ seq->litLength = (U32)(ip - anchor);
+ seq->matchLength = (U32)mLength;
+ seq->offset = offset;
+ rawSeqStore->size++;
+ }
+
+ /* Insert the current entry into the hash table */
+ ZSTD_ldm_makeEntryAndInsertByTag(ldmState, rollingHash, hBits,
+ (U32)(lastHashed - base),
+ *params);
+
+ assert(ip + backwardMatchLength == lastHashed);
+
+ /* Fill the hash table from lastHashed+1 to ip+mLength*/
+ /* Heuristic: don't need to fill the entire table at end of block */
+ if (ip + mLength <= ilimit) {
+ rollingHash = ZSTD_ldm_fillLdmHashTable(
+ ldmState, rollingHash, lastHashed,
+ ip + mLength, base, hBits, *params);
+ lastHashed = ip + mLength - 1;
+ }
+ ip += mLength;
+ anchor = ip;
+ }
+ 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, 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;
+}
+
+size_t ZSTD_ldm_blockCompress(rawSeqStore_t* rawSeqStore,
+ ZSTD_matchState_t* ms, seqStore_t* seqStore, U32 rep[ZSTD_REP_NUM],
+ 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, 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);
+ assert(rawSeqStore->pos <= rawSeqStore->size);
+ assert(rawSeqStore->size <= rawSeqStore->capacity);
+ /* Loop through each sequence and apply the block compressor to the lits */
+ 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,
+ sequence.offset + ZSTD_REP_MOVE,
+ sequence.matchLength - MINMATCH);
+ 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/src/zstd/lib/compress/zstd_ldm.h b/src/zstd/lib/compress/zstd_ldm.h
new file mode 100644
index 000000000..229ea05a9
--- /dev/null
+++ b/src/zstd/lib/compress/zstd_ldm.h
@@ -0,0 +1,110 @@
+/*
+ * Copyright (c) 2016-2020, 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
+
+#if defined (__cplusplus)
+extern "C" {
+#endif
+
+#include "zstd_compress_internal.h" /* ldmParams_t, U32 */
+#include "../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],
+ 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 with is not passed to ZSTD_ldm_blockCompress().
+ */
+void ZSTD_ldm_skipSequences(rawSeqStore_t* rawSeqStore, size_t srcSize,
+ U32 const minMatch);
+
+
+/** 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);
+
+#if defined (__cplusplus)
+}
+#endif
+
+#endif /* ZSTD_FAST_H */
diff --git a/src/zstd/lib/compress/zstd_opt.c b/src/zstd/lib/compress/zstd_opt.c
new file mode 100644
index 000000000..36fff050c
--- /dev/null
+++ b/src/zstd/lib/compress/zstd_opt.c
@@ -0,0 +1,1200 @@
+/*
+ * Copyright (c) 2016-2020, 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_FREQ_DIV 4 /* log factor when using previous stats to init next 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 */
+# define BITCOST_ACCURACY 0
+# define BITCOST_MULTIPLIER (1 << BITCOST_ACCURACY)
+# define WEIGHT(stat) ((void)opt, ZSTD_bitWeight(stat))
+#elif 0 /* fractional bit accuracy */
+# 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_lcm_uncompressed;
+}
+
+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);
+}
+
+
+/* ZSTD_downscaleStat() :
+ * reduce all elements in table by a factor 2^(ZSTD_FREQ_DIV+malus)
+ * return the resulting sum of elements */
+static U32 ZSTD_downscaleStat(unsigned* table, U32 lastEltIndex, int malus)
+{
+ U32 s, sum=0;
+ DEBUGLOG(5, "ZSTD_downscaleStat (nbElts=%u)", (unsigned)lastEltIndex+1);
+ assert(ZSTD_FREQ_DIV+malus > 0 && ZSTD_FREQ_DIV+malus < 31);
+ for (s=0; s<lastEltIndex+1; s++) {
+ table[s] = 1 + (table[s] >> (ZSTD_FREQ_DIV+malus));
+ sum += table[s];
+ }
+ return sum;
+}
+
+/* ZSTD_rescaleFreqs() :
+ * if first block (detected by optPtr->litLengthSum == 0) : init statistics
+ * take hints from dictionary if there is one
+ * or init from zero, using src for literals stats, or flat 1 for match 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_getNbBits(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_downscaleStat(optPtr->litFreq, MaxLit, 1);
+ }
+
+ { unsigned ll;
+ for (ll=0; ll<=MaxLL; ll++)
+ optPtr->litLengthFreq[ll] = 1;
+ }
+ optPtr->litLengthSum = MaxLL+1;
+
+ { unsigned ml;
+ for (ml=0; ml<=MaxML; ml++)
+ optPtr->matchLengthFreq[ml] = 1;
+ }
+ optPtr->matchLengthSum = MaxML+1;
+
+ { unsigned of;
+ for (of=0; of<=MaxOff; of++)
+ optPtr->offCodeFreq[of] = 1;
+ }
+ optPtr->offCodeSum = MaxOff+1;
+
+ }
+
+ } else { /* new block : re-use previous statistics, scaled down */
+
+ if (compressedLiterals)
+ optPtr->litSum = ZSTD_downscaleStat(optPtr->litFreq, MaxLit, 1);
+ optPtr->litLengthSum = ZSTD_downscaleStat(optPtr->litLengthFreq, MaxLL, 0);
+ optPtr->matchLengthSum = ZSTD_downscaleStat(optPtr->matchLengthFreq, MaxML, 0);
+ optPtr->offCodeSum = ZSTD_downscaleStat(optPtr->offCodeFreq, MaxOff, 0);
+ }
+
+ 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)
+{
+ if (optPtr->priceType == zop_predef) return WEIGHT(litLength, 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.
+ * optLevel: when <2, favors small offset for decompression speed (improved cache efficiency) */
+FORCE_INLINE_TEMPLATE U32
+ZSTD_getMatchPrice(U32 const offset,
+ U32 const matchLength,
+ const optState_t* const optPtr,
+ int const optLevel)
+{
+ U32 price;
+ U32 const offCode = ZSTD_highbit32(offset+1);
+ 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++;
+ }
+
+ /* match offset code (0-2=>repCode; 3+=>offset+2) */
+ { U32 const offCode = ZSTD_highbit32(offsetCode+1);
+ 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 (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.
+ * ip : assumed <= iend-8 .
+ * @return : nb of positions added */
+static U32 ZSTD_insertBt1(
+ ZSTD_matchState_t* ms,
+ const BYTE* const ip, const BYTE* const iend,
+ 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 current = (U32)(ip-base);
+ const U32 btLow = btMask >= current ? 0 : current - btMask;
+ U32* smallerPtr = bt + 2*(current&btMask);
+ U32* largerPtr = smallerPtr + 1;
+ U32 dummy32; /* to be nullified at the end */
+ U32 const windowLow = ms->window.lowLimit;
+ U32 matchEndIdx = current+8+1;
+ size_t bestLength = 8;
+ U32 nbCompares = 1U << cParams->searchLog;
+#ifdef ZSTD_C_PREDICT
+ U32 predictedSmall = *(bt + 2*((current-1)&btMask) + 0);
+ U32 predictedLarge = *(bt + 2*((current-1)&btMask) + 1);
+ predictedSmall += (predictedSmall>0);
+ predictedLarge += (predictedLarge>0);
+#endif /* ZSTD_C_PREDICT */
+
+ DEBUGLOG(8, "ZSTD_insertBt1 (%u)", current);
+
+ assert(ip <= iend-8); /* required for h calculation */
+ hashTable[h] = current; /* Update Hash Table */
+
+ assert(windowLow > 0);
+ while (nbCompares-- && (matchIndex >= windowLow)) {
+ U32* const nextPtr = bt + 2*(matchIndex & btMask);
+ size_t matchLength = MIN(commonLengthSmaller, commonLengthLarger); /* guaranteed minimum nb of common bytes */
+ assert(matchIndex < current);
+
+#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 > current + 8);
+ return MAX(positions, matchEndIdx - (current + 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, 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 current = (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 >= current) ? 0 : current - btMask;
+ U32 const windowLow = ZSTD_getLowestMatchIndex(ms, current, cParams->windowLog);
+ U32 const matchLow = windowLow ? windowLow : 1;
+ U32* smallerPtr = bt + 2*(current&btMask);
+ U32* largerPtr = bt + 2*(current&btMask) + 1;
+ U32 matchEndIdx = current+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", current);
+
+ /* 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 = current - repOffset;
+ U32 repLen = 0;
+ assert(current >= dictLimit);
+ if (repOffset-1 /* intentional overflow, discards 0 and -1 */ < current-dictLimit) { /* equivalent to `current > 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 >= current */
+ const BYTE* const repMatch = dictMode == ZSTD_dictMatchState ?
+ dmsBase + repIndex - dmsIndexDelta :
+ dictBase + repIndex;
+ assert(current >= windowLow);
+ if ( dictMode == ZSTD_extDict
+ && ( ((repOffset-1) /*intentional overflow*/ < current - windowLow) /* equivalent to `current > 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*/ < current - (dmsLowLimit + dmsIndexDelta)) /* equivalent to `current > 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 = repCode - ll0;
+ 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)
+ & (current - 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(current > matchIndex3);
+ assert(mnum==0); /* no prior solution */
+ matches[0].off = (current - matchIndex3) + ZSTD_REP_MOVE;
+ matches[0].len = (U32)mlen;
+ mnum = 1;
+ if ( (mlen > sufficient_len) |
+ (ip+mlen == iLimit) ) { /* best possible length */
+ ms->nextToUpdate = current+1; /* skip insertion */
+ return 1;
+ } } }
+ /* no dictMatchState lookup: dicts don't have a populated HC3 table */
+ }
+
+ hashTable[h] = current; /* Update Hash Table */
+
+ while (nbCompares-- && (matchIndex >= matchLow)) {
+ U32* const nextPtr = bt + 2*(matchIndex & btMask);
+ const BYTE* match;
+ size_t matchLength = MIN(commonLengthSmaller, commonLengthLarger); /* guaranteed minimum nb of common bytes */
+ assert(current > 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, current - matchIndex, current - matchIndex + ZSTD_REP_MOVE);
+ assert(matchEndIdx > matchIndex);
+ if (matchLength > matchEndIdx - matchIndex)
+ matchEndIdx = matchIndex + (U32)matchLength;
+ bestLength = matchLength;
+ matches[mnum].off = (current - matchIndex) + ZSTD_REP_MOVE;
+ 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;
+
+ 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;
+ while (nbCompares-- && (dictMatchIndex > dmsLowLimit)) {
+ 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, current - matchIndex, current - matchIndex + ZSTD_REP_MOVE);
+ if (matchLength > matchEndIdx - matchIndex)
+ matchEndIdx = matchIndex + (U32)matchLength;
+ bestLength = matchLength;
+ matches[mnum].off = (current - matchIndex) + ZSTD_REP_MOVE;
+ 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];
+ }
+ }
+ }
+
+ assert(matchEndIdx > current+8);
+ ms->nextToUpdate = matchEndIdx - 8; /* skip repetitive patterns */
+ return mnum;
+}
+
+
+FORCE_INLINE_TEMPLATE U32 ZSTD_BtGetAllMatches (
+ ZSTD_match_t* matches, /* store result (match found, increasing size) in this table */
+ ZSTD_matchState_t* ms,
+ U32* nextToUpdate3,
+ const BYTE* ip, const BYTE* const iHighLimit, const ZSTD_dictMode_e dictMode,
+ const U32 rep[ZSTD_REP_NUM],
+ U32 const ll0,
+ U32 const lengthToBeat)
+{
+ const ZSTD_compressionParameters* const cParams = &ms->cParams;
+ U32 const matchLengthSearch = cParams->minMatch;
+ DEBUGLOG(8, "ZSTD_BtGetAllMatches");
+ if (ip < ms->window.base + ms->nextToUpdate) return 0; /* skipped area */
+ ZSTD_updateTree_internal(ms, ip, iHighLimit, matchLengthSearch, dictMode);
+ switch(matchLengthSearch)
+ {
+ case 3 : return ZSTD_insertBtAndGetAllMatches(matches, ms, nextToUpdate3, ip, iHighLimit, dictMode, rep, ll0, lengthToBeat, 3);
+ default :
+ case 4 : return ZSTD_insertBtAndGetAllMatches(matches, ms, nextToUpdate3, ip, iHighLimit, dictMode, rep, ll0, lengthToBeat, 4);
+ case 5 : return ZSTD_insertBtAndGetAllMatches(matches, ms, nextToUpdate3, ip, iHighLimit, dictMode, rep, ll0, lengthToBeat, 5);
+ case 7 :
+ case 6 : return ZSTD_insertBtAndGetAllMatches(matches, ms, nextToUpdate3, ip, iHighLimit, dictMode, rep, ll0, lengthToBeat, 6);
+ }
+}
+
+
+/*-*******************************
+* 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;
+
+ 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;
+
+ /* 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 const nbMatches = ZSTD_BtGetAllMatches(matches, ms, &nextToUpdate3, ip, iend, dictMode, rep, ll0, minMatch);
+ 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 = ZSTD_litLengthPrice(litlen, optStatePtr, optLevel);
+
+ /* large match -> immediate encoding */
+ { U32 const maxML = matches[nbMatches-1].len;
+ U32 const maxOffset = matches[nbMatches-1].off;
+ DEBUGLOG(6, "found %u matches of maxLength=%u and maxOffCode=%u at cPos=%u => start new series",
+ nbMatches, maxML, maxOffset, (U32)(ip-prefixStart));
+
+ if (maxML > sufficient_len) {
+ lastSequence.litlen = litlen;
+ lastSequence.mlen = maxML;
+ lastSequence.off = maxOffset;
+ 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 */
+ { U32 const literalsPrice = 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 offset = matches[matchNb].off;
+ U32 const end = matches[matchNb].len;
+ for ( ; pos <= end ; pos++ ) {
+ U32 const matchPrice = ZSTD_getMatchPrice(offset, 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 = offset;
+ opt[pos].litlen = litlen;
+ opt[pos].price = 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
+ + ZSTD_rawLiteralsCost(ip+cur-1, 1, optStatePtr, optLevel)
+ + ZSTD_litLengthPrice(litlen, optStatePtr, optLevel)
+ - 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 newReps = ZSTD_updateRep(opt[prev].rep, opt[cur].off, opt[cur].litlen==0);
+ memcpy(opt[cur].rep, &newReps, sizeof(repcodes_t));
+ } else {
+ 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 */
+ }
+
+ { U32 const ll0 = (opt[cur].mlen != 0);
+ U32 const litlen = (opt[cur].mlen == 0) ? opt[cur].litlen : 0;
+ U32 const previousPrice = opt[cur].price;
+ U32 const basePrice = previousPrice + ZSTD_litLengthPrice(0, optStatePtr, optLevel);
+ U32 const nbMatches = ZSTD_BtGetAllMatches(matches, ms, &nextToUpdate3, inr, iend, dictMode, opt[cur].rep, ll0, minMatch);
+ U32 matchNb;
+ 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 = basePrice + 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 reps = ZSTD_updateRep(opt[cur].rep, lastSequence.off, lastSequence.litlen==0);
+ memcpy(rep, &reps, sizeof(reps));
+ } else {
+ 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-MINMATCH);
+ anchor += advance;
+ ip = anchor;
+ } }
+ ZSTD_setBasePrices(optStatePtr, optLevel);
+ }
+ } /* while (ip < ilimit) */
+
+ /* Return the last literals size */
+ return (size_t)(iend - anchor);
+}
+
+
+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_opt_generic(ms, seqStore, rep, src, srcSize, 0 /*optLevel*/, ZSTD_noDict);
+}
+
+
+/* used in 2-pass strategy */
+static U32 ZSTD_upscaleStat(unsigned* table, U32 lastEltIndex, int bonus)
+{
+ U32 s, sum=0;
+ assert(ZSTD_FREQ_DIV+bonus >= 0);
+ for (s=0; s<lastEltIndex+1; s++) {
+ table[s] <<= ZSTD_FREQ_DIV+bonus;
+ table[s]--;
+ sum += table[s];
+ }
+ return sum;
+}
+
+/* used in 2-pass strategy */
+MEM_STATIC void ZSTD_upscaleStats(optState_t* optPtr)
+{
+ if (ZSTD_compressedLiterals(optPtr))
+ optPtr->litSum = ZSTD_upscaleStat(optPtr->litFreq, MaxLit, 0);
+ optPtr->litLengthSum = ZSTD_upscaleStat(optPtr->litLengthFreq, MaxLL, 0);
+ optPtr->matchLengthSum = ZSTD_upscaleStat(optPtr->matchLengthFreq, MaxML, 0);
+ optPtr->offCodeSum = ZSTD_upscaleStat(optPtr->offCodeFreq, MaxOff, 0);
+}
+
+/* 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 */
+ 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_opt_generic(ms, seqStore, tmpRep, src, srcSize, 2 /*optLevel*/, 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;
+
+ /* re-inforce weight of collected statistics */
+ ZSTD_upscaleStats(&ms->opt);
+}
+
+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_opt_generic(ms, seqStore, rep, src, srcSize, 2 /*optLevel*/, 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 current = (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 */
+ && (current == 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_opt_generic(ms, seqStore, rep, src, srcSize, 2 /*optLevel*/, 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_opt_generic(ms, seqStore, rep, src, srcSize, 0 /*optLevel*/, 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_opt_generic(ms, seqStore, rep, src, srcSize, 2 /*optLevel*/, 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_opt_generic(ms, seqStore, rep, src, srcSize, 0 /*optLevel*/, 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_opt_generic(ms, seqStore, rep, src, srcSize, 2 /*optLevel*/, 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/src/zstd/lib/compress/zstd_opt.h b/src/zstd/lib/compress/zstd_opt.h
new file mode 100644
index 000000000..9aba8a901
--- /dev/null
+++ b/src/zstd/lib/compress/zstd_opt.h
@@ -0,0 +1,56 @@
+/*
+ * Copyright (c) 2016-2020, 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
+
+#if defined (__cplusplus)
+extern "C" {
+#endif
+
+#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) */
+
+#if defined (__cplusplus)
+}
+#endif
+
+#endif /* ZSTD_OPT_H */
diff --git a/src/zstd/lib/compress/zstdmt_compress.c b/src/zstd/lib/compress/zstdmt_compress.c
new file mode 100644
index 000000000..1e3c8fdbe
--- /dev/null
+++ b/src/zstd/lib/compress/zstdmt_compress.c
@@ -0,0 +1,2143 @@
+/*
+ * Copyright (c) 2016-2020, 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.
+ */
+
+
+/* ====== Compiler specifics ====== */
+#if defined(_MSC_VER)
+# pragma warning(disable : 4204) /* disable: C4204: non-constant aggregate initializer */
+#endif
+
+
+/* ====== Constants ====== */
+#define ZSTDMT_OVERLAPLOG_DEFAULT 0
+
+
+/* ====== Dependencies ====== */
+#include <string.h> /* memcpy, memset */
+#include <limits.h> /* INT_MAX, UINT_MAX */
+#include "../common/mem.h" /* MEM_STATIC */
+#include "../common/pool.h" /* threadpool */
+#include "../common/threading.h" /* mutex */
+#include "zstd_compress_internal.h" /* MIN, ERROR, ZSTD_*, ZSTD_highbit32 */
+#include "zstd_ldm.h"
+#include "zstdmt_compress.h"
+
+/* Guards code to support resizing the SeqPool.
+ * We will want to resize the SeqPool to save memory in the future.
+ * Until then, comment the code out since it is unused.
+ */
+#define ZSTD_RESIZE_SEQPOOL 0
+
+/* ====== Debug ====== */
+#if defined(DEBUGLEVEL) && (DEBUGLEVEL>=2) \
+ && !defined(_MSC_VER) \
+ && !defined(__MINGW32__)
+
+# include <stdio.h>
+# include <unistd.h>
+# include <sys/times.h>
+
+# define DEBUG_PRINTHEX(l,p,n) { \
+ unsigned debug_u; \
+ for (debug_u=0; debug_u<(n); debug_u++) \
+ RAWLOG(l, "%02X ", ((const unsigned char*)(p))[debug_u]); \
+ RAWLOG(l, " \n"); \
+}
+
+static unsigned long long GetCurrentClockTimeMicroseconds(void)
+{
+ static clock_t _ticksPerSecond = 0;
+ if (_ticksPerSecond <= 0) _ticksPerSecond = sysconf(_SC_CLK_TCK);
+
+ { struct tms junk; clock_t newTicks = (clock_t) times(&junk);
+ return ((((unsigned long long)newTicks)*(1000000))/_ticksPerSecond);
+} }
+
+#define MUTEX_WAIT_TIME_DLEVEL 6
+#define ZSTD_PTHREAD_MUTEX_LOCK(mutex) { \
+ if (DEBUGLEVEL >= MUTEX_WAIT_TIME_DLEVEL) { \
+ unsigned long long const beforeTime = GetCurrentClockTimeMicroseconds(); \
+ ZSTD_pthread_mutex_lock(mutex); \
+ { unsigned long long const afterTime = GetCurrentClockTimeMicroseconds(); \
+ unsigned long long const elapsedTime = (afterTime-beforeTime); \
+ if (elapsedTime > 1000) { /* or whatever threshold you like; I'm using 1 millisecond here */ \
+ DEBUGLOG(MUTEX_WAIT_TIME_DLEVEL, "Thread took %llu microseconds to acquire mutex %s \n", \
+ elapsedTime, #mutex); \
+ } } \
+ } else { \
+ ZSTD_pthread_mutex_lock(mutex); \
+ } \
+}
+
+#else
+
+# define ZSTD_PTHREAD_MUTEX_LOCK(m) ZSTD_pthread_mutex_lock(m)
+# define DEBUG_PRINTHEX(l,p,n) {}
+
+#endif
+
+
+/* ===== Buffer Pool ===== */
+/* a single Buffer Pool can be invoked from multiple threads in parallel */
+
+typedef struct buffer_s {
+ void* start;
+ size_t capacity;
+} buffer_t;
+
+static const buffer_t g_nullBuffer = { NULL, 0 };
+
+typedef struct ZSTDMT_bufferPool_s {
+ ZSTD_pthread_mutex_t poolMutex;
+ size_t bufferSize;
+ unsigned totalBuffers;
+ unsigned nbBuffers;
+ ZSTD_customMem cMem;
+ buffer_t bTable[1]; /* variable size */
+} ZSTDMT_bufferPool;
+
+static ZSTDMT_bufferPool* ZSTDMT_createBufferPool(unsigned nbWorkers, ZSTD_customMem cMem)
+{
+ unsigned const maxNbBuffers = 2*nbWorkers + 3;
+ ZSTDMT_bufferPool* const bufPool = (ZSTDMT_bufferPool*)ZSTD_calloc(
+ sizeof(ZSTDMT_bufferPool) + (maxNbBuffers-1) * sizeof(buffer_t), cMem);
+ if (bufPool==NULL) return NULL;
+ if (ZSTD_pthread_mutex_init(&bufPool->poolMutex, NULL)) {
+ ZSTD_free(bufPool, cMem);
+ return NULL;
+ }
+ bufPool->bufferSize = 64 KB;
+ bufPool->totalBuffers = maxNbBuffers;
+ bufPool->nbBuffers = 0;
+ bufPool->cMem = cMem;
+ return bufPool;
+}
+
+static void ZSTDMT_freeBufferPool(ZSTDMT_bufferPool* bufPool)
+{
+ unsigned u;
+ DEBUGLOG(3, "ZSTDMT_freeBufferPool (address:%08X)", (U32)(size_t)bufPool);
+ if (!bufPool) return; /* compatibility with free on NULL */
+ for (u=0; u<bufPool->totalBuffers; u++) {
+ DEBUGLOG(4, "free buffer %2u (address:%08X)", u, (U32)(size_t)bufPool->bTable[u].start);
+ ZSTD_free(bufPool->bTable[u].start, bufPool->cMem);
+ }
+ ZSTD_pthread_mutex_destroy(&bufPool->poolMutex);
+ ZSTD_free(bufPool, bufPool->cMem);
+}
+
+/* only works at initialization, not during compression */
+static size_t ZSTDMT_sizeof_bufferPool(ZSTDMT_bufferPool* bufPool)
+{
+ size_t const poolSize = sizeof(*bufPool)
+ + (bufPool->totalBuffers - 1) * sizeof(buffer_t);
+ unsigned u;
+ size_t totalBufferSize = 0;
+ ZSTD_pthread_mutex_lock(&bufPool->poolMutex);
+ for (u=0; u<bufPool->totalBuffers; u++)
+ totalBufferSize += bufPool->bTable[u].capacity;
+ ZSTD_pthread_mutex_unlock(&bufPool->poolMutex);
+
+ return poolSize + totalBufferSize;
+}
+
+/* ZSTDMT_setBufferSize() :
+ * all future buffers provided by this buffer pool will have _at least_ this size
+ * note : it's better for all buffers to have same size,
+ * as they become freely interchangeable, reducing malloc/free usages and memory fragmentation */
+static void ZSTDMT_setBufferSize(ZSTDMT_bufferPool* const bufPool, size_t const bSize)
+{
+ ZSTD_pthread_mutex_lock(&bufPool->poolMutex);
+ DEBUGLOG(4, "ZSTDMT_setBufferSize: bSize = %u", (U32)bSize);
+ bufPool->bufferSize = bSize;
+ ZSTD_pthread_mutex_unlock(&bufPool->poolMutex);
+}
+
+
+static ZSTDMT_bufferPool* ZSTDMT_expandBufferPool(ZSTDMT_bufferPool* srcBufPool, U32 nbWorkers)
+{
+ unsigned const maxNbBuffers = 2*nbWorkers + 3;
+ if (srcBufPool==NULL) return NULL;
+ if (srcBufPool->totalBuffers >= maxNbBuffers) /* good enough */
+ return srcBufPool;
+ /* need a larger buffer pool */
+ { ZSTD_customMem const cMem = srcBufPool->cMem;
+ size_t const bSize = srcBufPool->bufferSize; /* forward parameters */
+ ZSTDMT_bufferPool* newBufPool;
+ ZSTDMT_freeBufferPool(srcBufPool);
+ newBufPool = ZSTDMT_createBufferPool(nbWorkers, cMem);
+ if (newBufPool==NULL) return newBufPool;
+ ZSTDMT_setBufferSize(newBufPool, bSize);
+ return newBufPool;
+ }
+}
+
+/** ZSTDMT_getBuffer() :
+ * assumption : bufPool must be valid
+ * @return : a buffer, with start pointer and size
+ * note: allocation may fail, in this case, start==NULL and size==0 */
+static buffer_t ZSTDMT_getBuffer(ZSTDMT_bufferPool* bufPool)
+{
+ size_t const bSize = bufPool->bufferSize;
+ DEBUGLOG(5, "ZSTDMT_getBuffer: bSize = %u", (U32)bufPool->bufferSize);
+ ZSTD_pthread_mutex_lock(&bufPool->poolMutex);
+ if (bufPool->nbBuffers) { /* try to use an existing buffer */
+ buffer_t const buf = bufPool->bTable[--(bufPool->nbBuffers)];
+ size_t const availBufferSize = buf.capacity;
+ bufPool->bTable[bufPool->nbBuffers] = g_nullBuffer;
+ if ((availBufferSize >= bSize) & ((availBufferSize>>3) <= bSize)) {
+ /* large enough, but not too much */
+ DEBUGLOG(5, "ZSTDMT_getBuffer: provide buffer %u of size %u",
+ bufPool->nbBuffers, (U32)buf.capacity);
+ ZSTD_pthread_mutex_unlock(&bufPool->poolMutex);
+ return buf;
+ }
+ /* size conditions not respected : scratch this buffer, create new one */
+ DEBUGLOG(5, "ZSTDMT_getBuffer: existing buffer does not meet size conditions => freeing");
+ ZSTD_free(buf.start, bufPool->cMem);
+ }
+ ZSTD_pthread_mutex_unlock(&bufPool->poolMutex);
+ /* create new buffer */
+ DEBUGLOG(5, "ZSTDMT_getBuffer: create a new buffer");
+ { buffer_t buffer;
+ void* const start = ZSTD_malloc(bSize, bufPool->cMem);
+ buffer.start = start; /* note : start can be NULL if malloc fails ! */
+ buffer.capacity = (start==NULL) ? 0 : bSize;
+ if (start==NULL) {
+ DEBUGLOG(5, "ZSTDMT_getBuffer: buffer allocation failure !!");
+ } else {
+ DEBUGLOG(5, "ZSTDMT_getBuffer: created buffer of size %u", (U32)bSize);
+ }
+ return buffer;
+ }
+}
+
+#if ZSTD_RESIZE_SEQPOOL
+/** ZSTDMT_resizeBuffer() :
+ * assumption : bufPool must be valid
+ * @return : a buffer that is at least the buffer pool buffer size.
+ * If a reallocation happens, the data in the input buffer is copied.
+ */
+static buffer_t ZSTDMT_resizeBuffer(ZSTDMT_bufferPool* bufPool, buffer_t buffer)
+{
+ size_t const bSize = bufPool->bufferSize;
+ if (buffer.capacity < bSize) {
+ void* const start = ZSTD_malloc(bSize, bufPool->cMem);
+ buffer_t newBuffer;
+ newBuffer.start = start;
+ newBuffer.capacity = start == NULL ? 0 : bSize;
+ if (start != NULL) {
+ assert(newBuffer.capacity >= buffer.capacity);
+ memcpy(newBuffer.start, buffer.start, buffer.capacity);
+ DEBUGLOG(5, "ZSTDMT_resizeBuffer: created buffer of size %u", (U32)bSize);
+ return newBuffer;
+ }
+ DEBUGLOG(5, "ZSTDMT_resizeBuffer: buffer allocation failure !!");
+ }
+ return buffer;
+}
+#endif
+
+/* store buffer for later re-use, up to pool capacity */
+static void ZSTDMT_releaseBuffer(ZSTDMT_bufferPool* bufPool, buffer_t buf)
+{
+ DEBUGLOG(5, "ZSTDMT_releaseBuffer");
+ if (buf.start == NULL) return; /* compatible with release on NULL */
+ ZSTD_pthread_mutex_lock(&bufPool->poolMutex);
+ if (bufPool->nbBuffers < bufPool->totalBuffers) {
+ bufPool->bTable[bufPool->nbBuffers++] = buf; /* stored for later use */
+ DEBUGLOG(5, "ZSTDMT_releaseBuffer: stored buffer of size %u in slot %u",
+ (U32)buf.capacity, (U32)(bufPool->nbBuffers-1));
+ ZSTD_pthread_mutex_unlock(&bufPool->poolMutex);
+ return;
+ }
+ ZSTD_pthread_mutex_unlock(&bufPool->poolMutex);
+ /* Reached bufferPool capacity (should not happen) */
+ DEBUGLOG(5, "ZSTDMT_releaseBuffer: pool capacity reached => freeing ");
+ ZSTD_free(buf.start, bufPool->cMem);
+}
+
+
+/* ===== Seq Pool Wrapper ====== */
+
+static rawSeqStore_t kNullRawSeqStore = {NULL, 0, 0, 0};
+
+typedef ZSTDMT_bufferPool ZSTDMT_seqPool;
+
+static size_t ZSTDMT_sizeof_seqPool(ZSTDMT_seqPool* seqPool)
+{
+ return ZSTDMT_sizeof_bufferPool(seqPool);
+}
+
+static rawSeqStore_t bufferToSeq(buffer_t buffer)
+{
+ rawSeqStore_t seq = {NULL, 0, 0, 0};
+ seq.seq = (rawSeq*)buffer.start;
+ seq.capacity = buffer.capacity / sizeof(rawSeq);
+ return seq;
+}
+
+static buffer_t seqToBuffer(rawSeqStore_t seq)
+{
+ buffer_t buffer;
+ buffer.start = seq.seq;
+ buffer.capacity = seq.capacity * sizeof(rawSeq);
+ return buffer;
+}
+
+static rawSeqStore_t ZSTDMT_getSeq(ZSTDMT_seqPool* seqPool)
+{
+ if (seqPool->bufferSize == 0) {
+ return kNullRawSeqStore;
+ }
+ return bufferToSeq(ZSTDMT_getBuffer(seqPool));
+}
+
+#if ZSTD_RESIZE_SEQPOOL
+static rawSeqStore_t ZSTDMT_resizeSeq(ZSTDMT_seqPool* seqPool, rawSeqStore_t seq)
+{
+ return bufferToSeq(ZSTDMT_resizeBuffer(seqPool, seqToBuffer(seq)));
+}
+#endif
+
+static void ZSTDMT_releaseSeq(ZSTDMT_seqPool* seqPool, rawSeqStore_t seq)
+{
+ ZSTDMT_releaseBuffer(seqPool, seqToBuffer(seq));
+}
+
+static void ZSTDMT_setNbSeq(ZSTDMT_seqPool* const seqPool, size_t const nbSeq)
+{
+ ZSTDMT_setBufferSize(seqPool, nbSeq * sizeof(rawSeq));
+}
+
+static ZSTDMT_seqPool* ZSTDMT_createSeqPool(unsigned nbWorkers, ZSTD_customMem cMem)
+{
+ ZSTDMT_seqPool* const seqPool = ZSTDMT_createBufferPool(nbWorkers, cMem);
+ if (seqPool == NULL) return NULL;
+ ZSTDMT_setNbSeq(seqPool, 0);
+ return seqPool;
+}
+
+static void ZSTDMT_freeSeqPool(ZSTDMT_seqPool* seqPool)
+{
+ ZSTDMT_freeBufferPool(seqPool);
+}
+
+static ZSTDMT_seqPool* ZSTDMT_expandSeqPool(ZSTDMT_seqPool* pool, U32 nbWorkers)
+{
+ return ZSTDMT_expandBufferPool(pool, nbWorkers);
+}
+
+
+/* ===== CCtx Pool ===== */
+/* a single CCtx Pool can be invoked from multiple threads in parallel */
+
+typedef struct {
+ ZSTD_pthread_mutex_t poolMutex;
+ int totalCCtx;
+ int availCCtx;
+ ZSTD_customMem cMem;
+ ZSTD_CCtx* cctx[1]; /* variable size */
+} ZSTDMT_CCtxPool;
+
+/* note : all CCtx borrowed from the pool should be released back to the pool _before_ freeing the pool */
+static void ZSTDMT_freeCCtxPool(ZSTDMT_CCtxPool* pool)
+{
+ int cid;
+ for (cid=0; cid<pool->totalCCtx; cid++)
+ ZSTD_freeCCtx(pool->cctx[cid]); /* note : compatible with free on NULL */
+ ZSTD_pthread_mutex_destroy(&pool->poolMutex);
+ ZSTD_free(pool, pool->cMem);
+}
+
+/* ZSTDMT_createCCtxPool() :
+ * implies nbWorkers >= 1 , checked by caller ZSTDMT_createCCtx() */
+static ZSTDMT_CCtxPool* ZSTDMT_createCCtxPool(int nbWorkers,
+ ZSTD_customMem cMem)
+{
+ ZSTDMT_CCtxPool* const cctxPool = (ZSTDMT_CCtxPool*) ZSTD_calloc(
+ sizeof(ZSTDMT_CCtxPool) + (nbWorkers-1)*sizeof(ZSTD_CCtx*), cMem);
+ assert(nbWorkers > 0);
+ if (!cctxPool) return NULL;
+ if (ZSTD_pthread_mutex_init(&cctxPool->poolMutex, NULL)) {
+ ZSTD_free(cctxPool, cMem);
+ return NULL;
+ }
+ cctxPool->cMem = cMem;
+ cctxPool->totalCCtx = nbWorkers;
+ cctxPool->availCCtx = 1; /* at least one cctx for single-thread mode */
+ cctxPool->cctx[0] = ZSTD_createCCtx_advanced(cMem);
+ if (!cctxPool->cctx[0]) { ZSTDMT_freeCCtxPool(cctxPool); return NULL; }
+ DEBUGLOG(3, "cctxPool created, with %u workers", nbWorkers);
+ return cctxPool;
+}
+
+static ZSTDMT_CCtxPool* ZSTDMT_expandCCtxPool(ZSTDMT_CCtxPool* srcPool,
+ int nbWorkers)
+{
+ if (srcPool==NULL) return NULL;
+ if (nbWorkers <= srcPool->totalCCtx) return srcPool; /* good enough */
+ /* need a larger cctx pool */
+ { ZSTD_customMem const cMem = srcPool->cMem;
+ ZSTDMT_freeCCtxPool(srcPool);
+ return ZSTDMT_createCCtxPool(nbWorkers, cMem);
+ }
+}
+
+/* only works during initialization phase, not during compression */
+static size_t ZSTDMT_sizeof_CCtxPool(ZSTDMT_CCtxPool* cctxPool)
+{
+ ZSTD_pthread_mutex_lock(&cctxPool->poolMutex);
+ { unsigned const nbWorkers = cctxPool->totalCCtx;
+ size_t const poolSize = sizeof(*cctxPool)
+ + (nbWorkers-1) * sizeof(ZSTD_CCtx*);
+ unsigned u;
+ size_t totalCCtxSize = 0;
+ for (u=0; u<nbWorkers; u++) {
+ totalCCtxSize += ZSTD_sizeof_CCtx(cctxPool->cctx[u]);
+ }
+ ZSTD_pthread_mutex_unlock(&cctxPool->poolMutex);
+ assert(nbWorkers > 0);
+ return poolSize + totalCCtxSize;
+ }
+}
+
+static ZSTD_CCtx* ZSTDMT_getCCtx(ZSTDMT_CCtxPool* cctxPool)
+{
+ DEBUGLOG(5, "ZSTDMT_getCCtx");
+ ZSTD_pthread_mutex_lock(&cctxPool->poolMutex);
+ if (cctxPool->availCCtx) {
+ cctxPool->availCCtx--;
+ { ZSTD_CCtx* const cctx = cctxPool->cctx[cctxPool->availCCtx];
+ ZSTD_pthread_mutex_unlock(&cctxPool->poolMutex);
+ return cctx;
+ } }
+ ZSTD_pthread_mutex_unlock(&cctxPool->poolMutex);
+ DEBUGLOG(5, "create one more CCtx");
+ return ZSTD_createCCtx_advanced(cctxPool->cMem); /* note : can be NULL, when creation fails ! */
+}
+
+static void ZSTDMT_releaseCCtx(ZSTDMT_CCtxPool* pool, ZSTD_CCtx* cctx)
+{
+ if (cctx==NULL) return; /* compatibility with release on NULL */
+ ZSTD_pthread_mutex_lock(&pool->poolMutex);
+ if (pool->availCCtx < pool->totalCCtx)
+ pool->cctx[pool->availCCtx++] = cctx;
+ else {
+ /* pool overflow : should not happen, since totalCCtx==nbWorkers */
+ DEBUGLOG(4, "CCtx pool overflow : free cctx");
+ ZSTD_freeCCtx(cctx);
+ }
+ ZSTD_pthread_mutex_unlock(&pool->poolMutex);
+}
+
+/* ==== Serial State ==== */
+
+typedef struct {
+ void const* start;
+ size_t size;
+} range_t;
+
+typedef struct {
+ /* All variables in the struct are protected by mutex. */
+ ZSTD_pthread_mutex_t mutex;
+ ZSTD_pthread_cond_t cond;
+ ZSTD_CCtx_params params;
+ ldmState_t ldmState;
+ XXH64_state_t xxhState;
+ unsigned nextJobID;
+ /* Protects ldmWindow.
+ * Must be acquired after the main mutex when acquiring both.
+ */
+ ZSTD_pthread_mutex_t ldmWindowMutex;
+ ZSTD_pthread_cond_t ldmWindowCond; /* Signaled when ldmWindow is updated */
+ ZSTD_window_t ldmWindow; /* A thread-safe copy of ldmState.window */
+} serialState_t;
+
+static int
+ZSTDMT_serialState_reset(serialState_t* serialState,
+ ZSTDMT_seqPool* seqPool,
+ ZSTD_CCtx_params params,
+ size_t jobSize,
+ const void* dict, size_t const dictSize,
+ ZSTD_dictContentType_e dictContentType)
+{
+ /* Adjust parameters */
+ if (params.ldmParams.enableLdm) {
+ DEBUGLOG(4, "LDM window size = %u KB", (1U << params.cParams.windowLog) >> 10);
+ ZSTD_ldm_adjustParameters(&params.ldmParams, &params.cParams);
+ assert(params.ldmParams.hashLog >= params.ldmParams.bucketSizeLog);
+ assert(params.ldmParams.hashRateLog < 32);
+ serialState->ldmState.hashPower =
+ ZSTD_rollingHash_primePower(params.ldmParams.minMatchLength);
+ } else {
+ memset(&params.ldmParams, 0, sizeof(params.ldmParams));
+ }
+ serialState->nextJobID = 0;
+ if (params.fParams.checksumFlag)
+ XXH64_reset(&serialState->xxhState, 0);
+ if (params.ldmParams.enableLdm) {
+ ZSTD_customMem cMem = params.customMem;
+ unsigned const hashLog = params.ldmParams.hashLog;
+ size_t const hashSize = ((size_t)1 << hashLog) * sizeof(ldmEntry_t);
+ unsigned const bucketLog =
+ params.ldmParams.hashLog - params.ldmParams.bucketSizeLog;
+ size_t const bucketSize = (size_t)1 << bucketLog;
+ unsigned const prevBucketLog =
+ serialState->params.ldmParams.hashLog -
+ serialState->params.ldmParams.bucketSizeLog;
+ /* Size the seq pool tables */
+ ZSTDMT_setNbSeq(seqPool, ZSTD_ldm_getMaxNbSeq(params.ldmParams, jobSize));
+ /* Reset the window */
+ ZSTD_window_init(&serialState->ldmState.window);
+ /* Resize tables and output space if necessary. */
+ if (serialState->ldmState.hashTable == NULL || serialState->params.ldmParams.hashLog < hashLog) {
+ ZSTD_free(serialState->ldmState.hashTable, cMem);
+ serialState->ldmState.hashTable = (ldmEntry_t*)ZSTD_malloc(hashSize, cMem);
+ }
+ if (serialState->ldmState.bucketOffsets == NULL || prevBucketLog < bucketLog) {
+ ZSTD_free(serialState->ldmState.bucketOffsets, cMem);
+ serialState->ldmState.bucketOffsets = (BYTE*)ZSTD_malloc(bucketSize, cMem);
+ }
+ if (!serialState->ldmState.hashTable || !serialState->ldmState.bucketOffsets)
+ return 1;
+ /* Zero the tables */
+ memset(serialState->ldmState.hashTable, 0, hashSize);
+ memset(serialState->ldmState.bucketOffsets, 0, bucketSize);
+
+ /* Update window state and fill hash table with dict */
+ serialState->ldmState.loadedDictEnd = 0;
+ if (dictSize > 0) {
+ if (dictContentType == ZSTD_dct_rawContent) {
+ BYTE const* const dictEnd = (const BYTE*)dict + dictSize;
+ ZSTD_window_update(&serialState->ldmState.window, dict, dictSize);
+ ZSTD_ldm_fillHashTable(&serialState->ldmState, (const BYTE*)dict, dictEnd, &params.ldmParams);
+ serialState->ldmState.loadedDictEnd = params.forceWindow ? 0 : (U32)(dictEnd - serialState->ldmState.window.base);
+ } else {
+ /* don't even load anything */
+ }
+ }
+
+ /* Initialize serialState's copy of ldmWindow. */
+ serialState->ldmWindow = serialState->ldmState.window;
+ }
+
+ serialState->params = params;
+ serialState->params.jobSize = (U32)jobSize;
+ return 0;
+}
+
+static int ZSTDMT_serialState_init(serialState_t* serialState)
+{
+ int initError = 0;
+ memset(serialState, 0, sizeof(*serialState));
+ initError |= ZSTD_pthread_mutex_init(&serialState->mutex, NULL);
+ initError |= ZSTD_pthread_cond_init(&serialState->cond, NULL);
+ initError |= ZSTD_pthread_mutex_init(&serialState->ldmWindowMutex, NULL);
+ initError |= ZSTD_pthread_cond_init(&serialState->ldmWindowCond, NULL);
+ return initError;
+}
+
+static void ZSTDMT_serialState_free(serialState_t* serialState)
+{
+ ZSTD_customMem cMem = serialState->params.customMem;
+ ZSTD_pthread_mutex_destroy(&serialState->mutex);
+ ZSTD_pthread_cond_destroy(&serialState->cond);
+ ZSTD_pthread_mutex_destroy(&serialState->ldmWindowMutex);
+ ZSTD_pthread_cond_destroy(&serialState->ldmWindowCond);
+ ZSTD_free(serialState->ldmState.hashTable, cMem);
+ ZSTD_free(serialState->ldmState.bucketOffsets, cMem);
+}
+
+static void ZSTDMT_serialState_update(serialState_t* serialState,
+ ZSTD_CCtx* jobCCtx, rawSeqStore_t seqStore,
+ range_t src, unsigned jobID)
+{
+ /* Wait for our turn */
+ ZSTD_PTHREAD_MUTEX_LOCK(&serialState->mutex);
+ while (serialState->nextJobID < jobID) {
+ DEBUGLOG(5, "wait for serialState->cond");
+ ZSTD_pthread_cond_wait(&serialState->cond, &serialState->mutex);
+ }
+ /* A future job may error and skip our job */
+ if (serialState->nextJobID == jobID) {
+ /* It is now our turn, do any processing necessary */
+ if (serialState->params.ldmParams.enableLdm) {
+ size_t error;
+ assert(seqStore.seq != NULL && seqStore.pos == 0 &&
+ seqStore.size == 0 && seqStore.capacity > 0);
+ assert(src.size <= serialState->params.jobSize);
+ ZSTD_window_update(&serialState->ldmState.window, src.start, src.size);
+ error = ZSTD_ldm_generateSequences(
+ &serialState->ldmState, &seqStore,
+ &serialState->params.ldmParams, src.start, src.size);
+ /* We provide a large enough buffer to never fail. */
+ assert(!ZSTD_isError(error)); (void)error;
+ /* Update ldmWindow to match the ldmState.window and signal the main
+ * thread if it is waiting for a buffer.
+ */
+ ZSTD_PTHREAD_MUTEX_LOCK(&serialState->ldmWindowMutex);
+ serialState->ldmWindow = serialState->ldmState.window;
+ ZSTD_pthread_cond_signal(&serialState->ldmWindowCond);
+ ZSTD_pthread_mutex_unlock(&serialState->ldmWindowMutex);
+ }
+ if (serialState->params.fParams.checksumFlag && src.size > 0)
+ XXH64_update(&serialState->xxhState, src.start, src.size);
+ }
+ /* Now it is the next jobs turn */
+ serialState->nextJobID++;
+ ZSTD_pthread_cond_broadcast(&serialState->cond);
+ ZSTD_pthread_mutex_unlock(&serialState->mutex);
+
+ if (seqStore.size > 0) {
+ size_t const err = ZSTD_referenceExternalSequences(
+ jobCCtx, seqStore.seq, seqStore.size);
+ assert(serialState->params.ldmParams.enableLdm);
+ assert(!ZSTD_isError(err));
+ (void)err;
+ }
+}
+
+static void ZSTDMT_serialState_ensureFinished(serialState_t* serialState,
+ unsigned jobID, size_t cSize)
+{
+ ZSTD_PTHREAD_MUTEX_LOCK(&serialState->mutex);
+ if (serialState->nextJobID <= jobID) {
+ assert(ZSTD_isError(cSize)); (void)cSize;
+ DEBUGLOG(5, "Skipping past job %u because of error", jobID);
+ serialState->nextJobID = jobID + 1;
+ ZSTD_pthread_cond_broadcast(&serialState->cond);
+
+ ZSTD_PTHREAD_MUTEX_LOCK(&serialState->ldmWindowMutex);
+ ZSTD_window_clear(&serialState->ldmWindow);
+ ZSTD_pthread_cond_signal(&serialState->ldmWindowCond);
+ ZSTD_pthread_mutex_unlock(&serialState->ldmWindowMutex);
+ }
+ ZSTD_pthread_mutex_unlock(&serialState->mutex);
+
+}
+
+
+/* ------------------------------------------ */
+/* ===== Worker thread ===== */
+/* ------------------------------------------ */
+
+static const range_t kNullRange = { NULL, 0 };
+
+typedef struct {
+ size_t consumed; /* SHARED - set0 by mtctx, then modified by worker AND read by mtctx */
+ size_t cSize; /* SHARED - set0 by mtctx, then modified by worker AND read by mtctx, then set0 by mtctx */
+ ZSTD_pthread_mutex_t job_mutex; /* Thread-safe - used by mtctx and worker */
+ ZSTD_pthread_cond_t job_cond; /* Thread-safe - used by mtctx and worker */
+ ZSTDMT_CCtxPool* cctxPool; /* Thread-safe - used by mtctx and (all) workers */
+ ZSTDMT_bufferPool* bufPool; /* Thread-safe - used by mtctx and (all) workers */
+ ZSTDMT_seqPool* seqPool; /* Thread-safe - used by mtctx and (all) workers */
+ serialState_t* serial; /* Thread-safe - used by mtctx and (all) workers */
+ buffer_t dstBuff; /* set by worker (or mtctx), then read by worker & mtctx, then modified by mtctx => no barrier */
+ range_t prefix; /* set by mtctx, then read by worker & mtctx => no barrier */
+ range_t src; /* set by mtctx, then read by worker & mtctx => no barrier */
+ unsigned jobID; /* set by mtctx, then read by worker => no barrier */
+ unsigned firstJob; /* set by mtctx, then read by worker => no barrier */
+ unsigned lastJob; /* set by mtctx, then read by worker => no barrier */
+ ZSTD_CCtx_params params; /* set by mtctx, then read by worker => no barrier */
+ const ZSTD_CDict* cdict; /* set by mtctx, then read by worker => no barrier */
+ unsigned long long fullFrameSize; /* set by mtctx, then read by worker => no barrier */
+ size_t dstFlushed; /* used only by mtctx */
+ unsigned frameChecksumNeeded; /* used only by mtctx */
+} ZSTDMT_jobDescription;
+
+#define JOB_ERROR(e) { \
+ ZSTD_PTHREAD_MUTEX_LOCK(&job->job_mutex); \
+ job->cSize = e; \
+ ZSTD_pthread_mutex_unlock(&job->job_mutex); \
+ goto _endJob; \
+}
+
+/* ZSTDMT_compressionJob() is a POOL_function type */
+static void ZSTDMT_compressionJob(void* jobDescription)
+{
+ ZSTDMT_jobDescription* const job = (ZSTDMT_jobDescription*)jobDescription;
+ ZSTD_CCtx_params jobParams = job->params; /* do not modify job->params ! copy it, modify the copy */
+ ZSTD_CCtx* const cctx = ZSTDMT_getCCtx(job->cctxPool);
+ rawSeqStore_t rawSeqStore = ZSTDMT_getSeq(job->seqPool);
+ buffer_t dstBuff = job->dstBuff;
+ size_t lastCBlockSize = 0;
+
+ /* resources */
+ if (cctx==NULL) JOB_ERROR(ERROR(memory_allocation));
+ if (dstBuff.start == NULL) { /* streaming job : doesn't provide a dstBuffer */
+ dstBuff = ZSTDMT_getBuffer(job->bufPool);
+ if (dstBuff.start==NULL) JOB_ERROR(ERROR(memory_allocation));
+ job->dstBuff = dstBuff; /* this value can be read in ZSTDMT_flush, when it copies the whole job */
+ }
+ if (jobParams.ldmParams.enableLdm && rawSeqStore.seq == NULL)
+ JOB_ERROR(ERROR(memory_allocation));
+
+ /* Don't compute the checksum for chunks, since we compute it externally,
+ * but write it in the header.
+ */
+ if (job->jobID != 0) jobParams.fParams.checksumFlag = 0;
+ /* Don't run LDM for the chunks, since we handle it externally */
+ jobParams.ldmParams.enableLdm = 0;
+
+
+ /* init */
+ if (job->cdict) {
+ size_t const initError = ZSTD_compressBegin_advanced_internal(cctx, NULL, 0, ZSTD_dct_auto, ZSTD_dtlm_fast, job->cdict, &jobParams, job->fullFrameSize);
+ assert(job->firstJob); /* only allowed for first job */
+ if (ZSTD_isError(initError)) JOB_ERROR(initError);
+ } else { /* srcStart points at reloaded section */
+ U64 const pledgedSrcSize = job->firstJob ? job->fullFrameSize : job->src.size;
+ { size_t const forceWindowError = ZSTD_CCtxParams_setParameter(&jobParams, ZSTD_c_forceMaxWindow, !job->firstJob);
+ if (ZSTD_isError(forceWindowError)) JOB_ERROR(forceWindowError);
+ }
+ { size_t const initError = ZSTD_compressBegin_advanced_internal(cctx,
+ job->prefix.start, job->prefix.size, ZSTD_dct_rawContent, /* load dictionary in "content-only" mode (no header analysis) */
+ ZSTD_dtlm_fast,
+ NULL, /*cdict*/
+ &jobParams, pledgedSrcSize);
+ if (ZSTD_isError(initError)) JOB_ERROR(initError);
+ } }
+
+ /* Perform serial step as early as possible, but after CCtx initialization */
+ ZSTDMT_serialState_update(job->serial, cctx, rawSeqStore, job->src, job->jobID);
+
+ if (!job->firstJob) { /* flush and overwrite frame header when it's not first job */
+ size_t const hSize = ZSTD_compressContinue(cctx, dstBuff.start, dstBuff.capacity, job->src.start, 0);
+ if (ZSTD_isError(hSize)) JOB_ERROR(hSize);
+ DEBUGLOG(5, "ZSTDMT_compressionJob: flush and overwrite %u bytes of frame header (not first job)", (U32)hSize);
+ ZSTD_invalidateRepCodes(cctx);
+ }
+
+ /* compress */
+ { size_t const chunkSize = 4*ZSTD_BLOCKSIZE_MAX;
+ int const nbChunks = (int)((job->src.size + (chunkSize-1)) / chunkSize);
+ const BYTE* ip = (const BYTE*) job->src.start;
+ BYTE* const ostart = (BYTE*)dstBuff.start;
+ BYTE* op = ostart;
+ BYTE* oend = op + dstBuff.capacity;
+ int chunkNb;
+ if (sizeof(size_t) > sizeof(int)) assert(job->src.size < ((size_t)INT_MAX) * chunkSize); /* check overflow */
+ DEBUGLOG(5, "ZSTDMT_compressionJob: compress %u bytes in %i blocks", (U32)job->src.size, nbChunks);
+ assert(job->cSize == 0);
+ for (chunkNb = 1; chunkNb < nbChunks; chunkNb++) {
+ size_t const cSize = ZSTD_compressContinue(cctx, op, oend-op, ip, chunkSize);
+ if (ZSTD_isError(cSize)) JOB_ERROR(cSize);
+ ip += chunkSize;
+ op += cSize; assert(op < oend);
+ /* stats */
+ ZSTD_PTHREAD_MUTEX_LOCK(&job->job_mutex);
+ job->cSize += cSize;
+ job->consumed = chunkSize * chunkNb;
+ DEBUGLOG(5, "ZSTDMT_compressionJob: compress new block : cSize==%u bytes (total: %u)",
+ (U32)cSize, (U32)job->cSize);
+ ZSTD_pthread_cond_signal(&job->job_cond); /* warns some more data is ready to be flushed */
+ ZSTD_pthread_mutex_unlock(&job->job_mutex);
+ }
+ /* last block */
+ assert(chunkSize > 0);
+ assert((chunkSize & (chunkSize - 1)) == 0); /* chunkSize must be power of 2 for mask==(chunkSize-1) to work */
+ if ((nbChunks > 0) | job->lastJob /*must output a "last block" flag*/ ) {
+ size_t const lastBlockSize1 = job->src.size & (chunkSize-1);
+ size_t const lastBlockSize = ((lastBlockSize1==0) & (job->src.size>=chunkSize)) ? chunkSize : lastBlockSize1;
+ size_t const cSize = (job->lastJob) ?
+ ZSTD_compressEnd (cctx, op, oend-op, ip, lastBlockSize) :
+ ZSTD_compressContinue(cctx, op, oend-op, ip, lastBlockSize);
+ if (ZSTD_isError(cSize)) JOB_ERROR(cSize);
+ lastCBlockSize = cSize;
+ } }
+
+_endJob:
+ ZSTDMT_serialState_ensureFinished(job->serial, job->jobID, job->cSize);
+ if (job->prefix.size > 0)
+ DEBUGLOG(5, "Finished with prefix: %zx", (size_t)job->prefix.start);
+ DEBUGLOG(5, "Finished with source: %zx", (size_t)job->src.start);
+ /* release resources */
+ ZSTDMT_releaseSeq(job->seqPool, rawSeqStore);
+ ZSTDMT_releaseCCtx(job->cctxPool, cctx);
+ /* report */
+ ZSTD_PTHREAD_MUTEX_LOCK(&job->job_mutex);
+ if (ZSTD_isError(job->cSize)) assert(lastCBlockSize == 0);
+ job->cSize += lastCBlockSize;
+ job->consumed = job->src.size; /* when job->consumed == job->src.size , compression job is presumed completed */
+ ZSTD_pthread_cond_signal(&job->job_cond);
+ ZSTD_pthread_mutex_unlock(&job->job_mutex);
+}
+
+
+/* ------------------------------------------ */
+/* ===== Multi-threaded compression ===== */
+/* ------------------------------------------ */
+
+typedef struct {
+ range_t prefix; /* read-only non-owned prefix buffer */
+ buffer_t buffer;
+ size_t filled;
+} inBuff_t;
+
+typedef struct {
+ BYTE* buffer; /* The round input buffer. All jobs get references
+ * to pieces of the buffer. ZSTDMT_tryGetInputRange()
+ * handles handing out job input buffers, and makes
+ * sure it doesn't overlap with any pieces still in use.
+ */
+ size_t capacity; /* The capacity of buffer. */
+ size_t pos; /* The position of the current inBuff in the round
+ * buffer. Updated past the end if the inBuff once
+ * the inBuff is sent to the worker thread.
+ * pos <= capacity.
+ */
+} roundBuff_t;
+
+static const roundBuff_t kNullRoundBuff = {NULL, 0, 0};
+
+#define RSYNC_LENGTH 32
+
+typedef struct {
+ U64 hash;
+ U64 hitMask;
+ U64 primePower;
+} rsyncState_t;
+
+struct ZSTDMT_CCtx_s {
+ POOL_ctx* factory;
+ ZSTDMT_jobDescription* jobs;
+ ZSTDMT_bufferPool* bufPool;
+ ZSTDMT_CCtxPool* cctxPool;
+ ZSTDMT_seqPool* seqPool;
+ ZSTD_CCtx_params params;
+ size_t targetSectionSize;
+ size_t targetPrefixSize;
+ int jobReady; /* 1 => one job is already prepared, but pool has shortage of workers. Don't create a new job. */
+ inBuff_t inBuff;
+ roundBuff_t roundBuff;
+ serialState_t serial;
+ rsyncState_t rsync;
+ unsigned singleBlockingThread;
+ unsigned jobIDMask;
+ unsigned doneJobID;
+ unsigned nextJobID;
+ unsigned frameEnded;
+ unsigned allJobsCompleted;
+ unsigned long long frameContentSize;
+ unsigned long long consumed;
+ unsigned long long produced;
+ ZSTD_customMem cMem;
+ ZSTD_CDict* cdictLocal;
+ const ZSTD_CDict* cdict;
+};
+
+static void ZSTDMT_freeJobsTable(ZSTDMT_jobDescription* jobTable, U32 nbJobs, ZSTD_customMem cMem)
+{
+ U32 jobNb;
+ if (jobTable == NULL) return;
+ for (jobNb=0; jobNb<nbJobs; jobNb++) {
+ ZSTD_pthread_mutex_destroy(&jobTable[jobNb].job_mutex);
+ ZSTD_pthread_cond_destroy(&jobTable[jobNb].job_cond);
+ }
+ ZSTD_free(jobTable, cMem);
+}
+
+/* ZSTDMT_allocJobsTable()
+ * allocate and init a job table.
+ * update *nbJobsPtr to next power of 2 value, as size of table */
+static ZSTDMT_jobDescription* ZSTDMT_createJobsTable(U32* nbJobsPtr, ZSTD_customMem cMem)
+{
+ U32 const nbJobsLog2 = ZSTD_highbit32(*nbJobsPtr) + 1;
+ U32 const nbJobs = 1 << nbJobsLog2;
+ U32 jobNb;
+ ZSTDMT_jobDescription* const jobTable = (ZSTDMT_jobDescription*)
+ ZSTD_calloc(nbJobs * sizeof(ZSTDMT_jobDescription), cMem);
+ int initError = 0;
+ if (jobTable==NULL) return NULL;
+ *nbJobsPtr = nbJobs;
+ for (jobNb=0; jobNb<nbJobs; jobNb++) {
+ initError |= ZSTD_pthread_mutex_init(&jobTable[jobNb].job_mutex, NULL);
+ initError |= ZSTD_pthread_cond_init(&jobTable[jobNb].job_cond, NULL);
+ }
+ if (initError != 0) {
+ ZSTDMT_freeJobsTable(jobTable, nbJobs, cMem);
+ return NULL;
+ }
+ return jobTable;
+}
+
+static size_t ZSTDMT_expandJobsTable (ZSTDMT_CCtx* mtctx, U32 nbWorkers) {
+ U32 nbJobs = nbWorkers + 2;
+ if (nbJobs > mtctx->jobIDMask+1) { /* need more job capacity */
+ ZSTDMT_freeJobsTable(mtctx->jobs, mtctx->jobIDMask+1, mtctx->cMem);
+ mtctx->jobIDMask = 0;
+ mtctx->jobs = ZSTDMT_createJobsTable(&nbJobs, mtctx->cMem);
+ if (mtctx->jobs==NULL) return ERROR(memory_allocation);
+ assert((nbJobs != 0) && ((nbJobs & (nbJobs - 1)) == 0)); /* ensure nbJobs is a power of 2 */
+ mtctx->jobIDMask = nbJobs - 1;
+ }
+ return 0;
+}
+
+
+/* ZSTDMT_CCtxParam_setNbWorkers():
+ * Internal use only */
+size_t ZSTDMT_CCtxParam_setNbWorkers(ZSTD_CCtx_params* params, unsigned nbWorkers)
+{
+ return ZSTD_CCtxParams_setParameter(params, ZSTD_c_nbWorkers, (int)nbWorkers);
+}
+
+MEM_STATIC ZSTDMT_CCtx* ZSTDMT_createCCtx_advanced_internal(unsigned nbWorkers, ZSTD_customMem cMem)
+{
+ ZSTDMT_CCtx* mtctx;
+ U32 nbJobs = nbWorkers + 2;
+ int initError;
+ DEBUGLOG(3, "ZSTDMT_createCCtx_advanced (nbWorkers = %u)", nbWorkers);
+
+ if (nbWorkers < 1) return NULL;
+ nbWorkers = MIN(nbWorkers , ZSTDMT_NBWORKERS_MAX);
+ if ((cMem.customAlloc!=NULL) ^ (cMem.customFree!=NULL))
+ /* invalid custom allocator */
+ return NULL;
+
+ mtctx = (ZSTDMT_CCtx*) ZSTD_calloc(sizeof(ZSTDMT_CCtx), cMem);
+ if (!mtctx) return NULL;
+ ZSTDMT_CCtxParam_setNbWorkers(&mtctx->params, nbWorkers);
+ mtctx->cMem = cMem;
+ mtctx->allJobsCompleted = 1;
+ mtctx->factory = POOL_create_advanced(nbWorkers, 0, cMem);
+ mtctx->jobs = ZSTDMT_createJobsTable(&nbJobs, cMem);
+ assert(nbJobs > 0); assert((nbJobs & (nbJobs - 1)) == 0); /* ensure nbJobs is a power of 2 */
+ mtctx->jobIDMask = nbJobs - 1;
+ mtctx->bufPool = ZSTDMT_createBufferPool(nbWorkers, cMem);
+ mtctx->cctxPool = ZSTDMT_createCCtxPool(nbWorkers, cMem);
+ mtctx->seqPool = ZSTDMT_createSeqPool(nbWorkers, cMem);
+ initError = ZSTDMT_serialState_init(&mtctx->serial);
+ mtctx->roundBuff = kNullRoundBuff;
+ if (!mtctx->factory | !mtctx->jobs | !mtctx->bufPool | !mtctx->cctxPool | !mtctx->seqPool | initError) {
+ ZSTDMT_freeCCtx(mtctx);
+ return NULL;
+ }
+ DEBUGLOG(3, "mt_cctx created, for %u threads", nbWorkers);
+ return mtctx;
+}
+
+ZSTDMT_CCtx* ZSTDMT_createCCtx_advanced(unsigned nbWorkers, ZSTD_customMem cMem)
+{
+#ifdef ZSTD_MULTITHREAD
+ return ZSTDMT_createCCtx_advanced_internal(nbWorkers, cMem);
+#else
+ (void)nbWorkers;
+ (void)cMem;
+ return NULL;
+#endif
+}
+
+ZSTDMT_CCtx* ZSTDMT_createCCtx(unsigned nbWorkers)
+{
+ return ZSTDMT_createCCtx_advanced(nbWorkers, ZSTD_defaultCMem);
+}
+
+
+/* ZSTDMT_releaseAllJobResources() :
+ * note : ensure all workers are killed first ! */
+static void ZSTDMT_releaseAllJobResources(ZSTDMT_CCtx* mtctx)
+{
+ unsigned jobID;
+ DEBUGLOG(3, "ZSTDMT_releaseAllJobResources");
+ for (jobID=0; jobID <= mtctx->jobIDMask; jobID++) {
+ /* Copy the mutex/cond out */
+ ZSTD_pthread_mutex_t const mutex = mtctx->jobs[jobID].job_mutex;
+ ZSTD_pthread_cond_t const cond = mtctx->jobs[jobID].job_cond;
+
+ DEBUGLOG(4, "job%02u: release dst address %08X", jobID, (U32)(size_t)mtctx->jobs[jobID].dstBuff.start);
+ ZSTDMT_releaseBuffer(mtctx->bufPool, mtctx->jobs[jobID].dstBuff);
+
+ /* Clear the job description, but keep the mutex/cond */
+ memset(&mtctx->jobs[jobID], 0, sizeof(mtctx->jobs[jobID]));
+ mtctx->jobs[jobID].job_mutex = mutex;
+ mtctx->jobs[jobID].job_cond = cond;
+ }
+ mtctx->inBuff.buffer = g_nullBuffer;
+ mtctx->inBuff.filled = 0;
+ mtctx->allJobsCompleted = 1;
+}
+
+static void ZSTDMT_waitForAllJobsCompleted(ZSTDMT_CCtx* mtctx)
+{
+ DEBUGLOG(4, "ZSTDMT_waitForAllJobsCompleted");
+ while (mtctx->doneJobID < mtctx->nextJobID) {
+ unsigned const jobID = mtctx->doneJobID & mtctx->jobIDMask;
+ ZSTD_PTHREAD_MUTEX_LOCK(&mtctx->jobs[jobID].job_mutex);
+ while (mtctx->jobs[jobID].consumed < mtctx->jobs[jobID].src.size) {
+ DEBUGLOG(4, "waiting for jobCompleted signal from job %u", mtctx->doneJobID); /* we want to block when waiting for data to flush */
+ ZSTD_pthread_cond_wait(&mtctx->jobs[jobID].job_cond, &mtctx->jobs[jobID].job_mutex);
+ }
+ ZSTD_pthread_mutex_unlock(&mtctx->jobs[jobID].job_mutex);
+ mtctx->doneJobID++;
+ }
+}
+
+size_t ZSTDMT_freeCCtx(ZSTDMT_CCtx* mtctx)
+{
+ if (mtctx==NULL) return 0; /* compatible with free on NULL */
+ POOL_free(mtctx->factory); /* stop and free worker threads */
+ ZSTDMT_releaseAllJobResources(mtctx); /* release job resources into pools first */
+ ZSTDMT_freeJobsTable(mtctx->jobs, mtctx->jobIDMask+1, mtctx->cMem);
+ ZSTDMT_freeBufferPool(mtctx->bufPool);
+ ZSTDMT_freeCCtxPool(mtctx->cctxPool);
+ ZSTDMT_freeSeqPool(mtctx->seqPool);
+ ZSTDMT_serialState_free(&mtctx->serial);
+ ZSTD_freeCDict(mtctx->cdictLocal);
+ if (mtctx->roundBuff.buffer)
+ ZSTD_free(mtctx->roundBuff.buffer, mtctx->cMem);
+ ZSTD_free(mtctx, mtctx->cMem);
+ return 0;
+}
+
+size_t ZSTDMT_sizeof_CCtx(ZSTDMT_CCtx* mtctx)
+{
+ if (mtctx == NULL) return 0; /* supports sizeof NULL */
+ return sizeof(*mtctx)
+ + POOL_sizeof(mtctx->factory)
+ + ZSTDMT_sizeof_bufferPool(mtctx->bufPool)
+ + (mtctx->jobIDMask+1) * sizeof(ZSTDMT_jobDescription)
+ + ZSTDMT_sizeof_CCtxPool(mtctx->cctxPool)
+ + ZSTDMT_sizeof_seqPool(mtctx->seqPool)
+ + ZSTD_sizeof_CDict(mtctx->cdictLocal)
+ + mtctx->roundBuff.capacity;
+}
+
+/* Internal only */
+size_t
+ZSTDMT_CCtxParam_setMTCtxParameter(ZSTD_CCtx_params* params,
+ ZSTDMT_parameter parameter,
+ int value)
+{
+ DEBUGLOG(4, "ZSTDMT_CCtxParam_setMTCtxParameter");
+ switch(parameter)
+ {
+ case ZSTDMT_p_jobSize :
+ DEBUGLOG(4, "ZSTDMT_CCtxParam_setMTCtxParameter : set jobSize to %i", value);
+ return ZSTD_CCtxParams_setParameter(params, ZSTD_c_jobSize, value);
+ case ZSTDMT_p_overlapLog :
+ DEBUGLOG(4, "ZSTDMT_p_overlapLog : %i", value);
+ return ZSTD_CCtxParams_setParameter(params, ZSTD_c_overlapLog, value);
+ case ZSTDMT_p_rsyncable :
+ DEBUGLOG(4, "ZSTD_p_rsyncable : %i", value);
+ return ZSTD_CCtxParams_setParameter(params, ZSTD_c_rsyncable, value);
+ default :
+ return ERROR(parameter_unsupported);
+ }
+}
+
+size_t ZSTDMT_setMTCtxParameter(ZSTDMT_CCtx* mtctx, ZSTDMT_parameter parameter, int value)
+{
+ DEBUGLOG(4, "ZSTDMT_setMTCtxParameter");
+ return ZSTDMT_CCtxParam_setMTCtxParameter(&mtctx->params, parameter, value);
+}
+
+size_t ZSTDMT_getMTCtxParameter(ZSTDMT_CCtx* mtctx, ZSTDMT_parameter parameter, int* value)
+{
+ switch (parameter) {
+ case ZSTDMT_p_jobSize:
+ return ZSTD_CCtxParams_getParameter(&mtctx->params, ZSTD_c_jobSize, value);
+ case ZSTDMT_p_overlapLog:
+ return ZSTD_CCtxParams_getParameter(&mtctx->params, ZSTD_c_overlapLog, value);
+ case ZSTDMT_p_rsyncable:
+ return ZSTD_CCtxParams_getParameter(&mtctx->params, ZSTD_c_rsyncable, value);
+ default:
+ return ERROR(parameter_unsupported);
+ }
+}
+
+/* Sets parameters relevant to the compression job,
+ * initializing others to default values. */
+static ZSTD_CCtx_params ZSTDMT_initJobCCtxParams(const ZSTD_CCtx_params* params)
+{
+ ZSTD_CCtx_params jobParams = *params;
+ /* Clear parameters related to multithreading */
+ jobParams.forceWindow = 0;
+ jobParams.nbWorkers = 0;
+ jobParams.jobSize = 0;
+ jobParams.overlapLog = 0;
+ jobParams.rsyncable = 0;
+ memset(&jobParams.ldmParams, 0, sizeof(ldmParams_t));
+ memset(&jobParams.customMem, 0, sizeof(ZSTD_customMem));
+ return jobParams;
+}
+
+
+/* ZSTDMT_resize() :
+ * @return : error code if fails, 0 on success */
+static size_t ZSTDMT_resize(ZSTDMT_CCtx* mtctx, unsigned nbWorkers)
+{
+ if (POOL_resize(mtctx->factory, nbWorkers)) return ERROR(memory_allocation);
+ FORWARD_IF_ERROR( ZSTDMT_expandJobsTable(mtctx, nbWorkers) , "");
+ mtctx->bufPool = ZSTDMT_expandBufferPool(mtctx->bufPool, nbWorkers);
+ if (mtctx->bufPool == NULL) return ERROR(memory_allocation);
+ mtctx->cctxPool = ZSTDMT_expandCCtxPool(mtctx->cctxPool, nbWorkers);
+ if (mtctx->cctxPool == NULL) return ERROR(memory_allocation);
+ mtctx->seqPool = ZSTDMT_expandSeqPool(mtctx->seqPool, nbWorkers);
+ if (mtctx->seqPool == NULL) return ERROR(memory_allocation);
+ ZSTDMT_CCtxParam_setNbWorkers(&mtctx->params, nbWorkers);
+ return 0;
+}
+
+
+/*! ZSTDMT_updateCParams_whileCompressing() :
+ * Updates a selected set of compression parameters, remaining compatible with currently active frame.
+ * New parameters will be applied to next compression job. */
+void ZSTDMT_updateCParams_whileCompressing(ZSTDMT_CCtx* mtctx, const ZSTD_CCtx_params* cctxParams)
+{
+ U32 const saved_wlog = mtctx->params.cParams.windowLog; /* Do not modify windowLog while compressing */
+ int const compressionLevel = cctxParams->compressionLevel;
+ DEBUGLOG(5, "ZSTDMT_updateCParams_whileCompressing (level:%i)",
+ compressionLevel);
+ mtctx->params.compressionLevel = compressionLevel;
+ { ZSTD_compressionParameters cParams = ZSTD_getCParamsFromCCtxParams(cctxParams, ZSTD_CONTENTSIZE_UNKNOWN, 0);
+ cParams.windowLog = saved_wlog;
+ mtctx->params.cParams = cParams;
+ }
+}
+
+/* ZSTDMT_getFrameProgression():
+ * tells how much data has been consumed (input) and produced (output) for current frame.
+ * able to count progression inside worker threads.
+ * Note : mutex will be acquired during statistics collection inside workers. */
+ZSTD_frameProgression ZSTDMT_getFrameProgression(ZSTDMT_CCtx* mtctx)
+{
+ ZSTD_frameProgression fps;
+ DEBUGLOG(5, "ZSTDMT_getFrameProgression");
+ fps.ingested = mtctx->consumed + mtctx->inBuff.filled;
+ fps.consumed = mtctx->consumed;
+ fps.produced = fps.flushed = mtctx->produced;
+ fps.currentJobID = mtctx->nextJobID;
+ fps.nbActiveWorkers = 0;
+ { unsigned jobNb;
+ unsigned lastJobNb = mtctx->nextJobID + mtctx->jobReady; assert(mtctx->jobReady <= 1);
+ DEBUGLOG(6, "ZSTDMT_getFrameProgression: jobs: from %u to <%u (jobReady:%u)",
+ mtctx->doneJobID, lastJobNb, mtctx->jobReady)
+ for (jobNb = mtctx->doneJobID ; jobNb < lastJobNb ; jobNb++) {
+ unsigned const wJobID = jobNb & mtctx->jobIDMask;
+ ZSTDMT_jobDescription* jobPtr = &mtctx->jobs[wJobID];
+ ZSTD_pthread_mutex_lock(&jobPtr->job_mutex);
+ { size_t const cResult = jobPtr->cSize;
+ size_t const produced = ZSTD_isError(cResult) ? 0 : cResult;
+ size_t const flushed = ZSTD_isError(cResult) ? 0 : jobPtr->dstFlushed;
+ assert(flushed <= produced);
+ fps.ingested += jobPtr->src.size;
+ fps.consumed += jobPtr->consumed;
+ fps.produced += produced;
+ fps.flushed += flushed;
+ fps.nbActiveWorkers += (jobPtr->consumed < jobPtr->src.size);
+ }
+ ZSTD_pthread_mutex_unlock(&mtctx->jobs[wJobID].job_mutex);
+ }
+ }
+ return fps;
+}
+
+
+size_t ZSTDMT_toFlushNow(ZSTDMT_CCtx* mtctx)
+{
+ size_t toFlush;
+ unsigned const jobID = mtctx->doneJobID;
+ assert(jobID <= mtctx->nextJobID);
+ if (jobID == mtctx->nextJobID) return 0; /* no active job => nothing to flush */
+
+ /* look into oldest non-fully-flushed job */
+ { unsigned const wJobID = jobID & mtctx->jobIDMask;
+ ZSTDMT_jobDescription* const jobPtr = &mtctx->jobs[wJobID];
+ ZSTD_pthread_mutex_lock(&jobPtr->job_mutex);
+ { size_t const cResult = jobPtr->cSize;
+ size_t const produced = ZSTD_isError(cResult) ? 0 : cResult;
+ size_t const flushed = ZSTD_isError(cResult) ? 0 : jobPtr->dstFlushed;
+ assert(flushed <= produced);
+ assert(jobPtr->consumed <= jobPtr->src.size);
+ toFlush = produced - flushed;
+ /* if toFlush==0, nothing is available to flush.
+ * However, jobID is expected to still be active:
+ * if jobID was already completed and fully flushed,
+ * ZSTDMT_flushProduced() should have already moved onto next job.
+ * Therefore, some input has not yet been consumed. */
+ if (toFlush==0) {
+ assert(jobPtr->consumed < jobPtr->src.size);
+ }
+ }
+ ZSTD_pthread_mutex_unlock(&mtctx->jobs[wJobID].job_mutex);
+ }
+
+ return toFlush;
+}
+
+
+/* ------------------------------------------ */
+/* ===== Multi-threaded compression ===== */
+/* ------------------------------------------ */
+
+static unsigned ZSTDMT_computeTargetJobLog(const ZSTD_CCtx_params* params)
+{
+ unsigned jobLog;
+ if (params->ldmParams.enableLdm) {
+ /* In Long Range Mode, the windowLog is typically oversized.
+ * In which case, it's preferable to determine the jobSize
+ * based on chainLog instead. */
+ jobLog = MAX(21, params->cParams.chainLog + 4);
+ } else {
+ jobLog = MAX(20, params->cParams.windowLog + 2);
+ }
+ return MIN(jobLog, (unsigned)ZSTDMT_JOBLOG_MAX);
+}
+
+static int ZSTDMT_overlapLog_default(ZSTD_strategy strat)
+{
+ switch(strat)
+ {
+ case ZSTD_btultra2:
+ return 9;
+ case ZSTD_btultra:
+ case ZSTD_btopt:
+ return 8;
+ case ZSTD_btlazy2:
+ case ZSTD_lazy2:
+ return 7;
+ case ZSTD_lazy:
+ case ZSTD_greedy:
+ case ZSTD_dfast:
+ case ZSTD_fast:
+ default:;
+ }
+ return 6;
+}
+
+static int ZSTDMT_overlapLog(int ovlog, ZSTD_strategy strat)
+{
+ assert(0 <= ovlog && ovlog <= 9);
+ if (ovlog == 0) return ZSTDMT_overlapLog_default(strat);
+ return ovlog;
+}
+
+static size_t ZSTDMT_computeOverlapSize(const ZSTD_CCtx_params* params)
+{
+ int const overlapRLog = 9 - ZSTDMT_overlapLog(params->overlapLog, params->cParams.strategy);
+ int ovLog = (overlapRLog >= 8) ? 0 : (params->cParams.windowLog - overlapRLog);
+ assert(0 <= overlapRLog && overlapRLog <= 8);
+ if (params->ldmParams.enableLdm) {
+ /* In Long Range Mode, the windowLog is typically oversized.
+ * In which case, it's preferable to determine the jobSize
+ * based on chainLog instead.
+ * Then, ovLog becomes a fraction of the jobSize, rather than windowSize */
+ ovLog = MIN(params->cParams.windowLog, ZSTDMT_computeTargetJobLog(params) - 2)
+ - overlapRLog;
+ }
+ assert(0 <= ovLog && ovLog <= ZSTD_WINDOWLOG_MAX);
+ DEBUGLOG(4, "overlapLog : %i", params->overlapLog);
+ DEBUGLOG(4, "overlap size : %i", 1 << ovLog);
+ return (ovLog==0) ? 0 : (size_t)1 << ovLog;
+}
+
+static unsigned
+ZSTDMT_computeNbJobs(const ZSTD_CCtx_params* params, size_t srcSize, unsigned nbWorkers)
+{
+ assert(nbWorkers>0);
+ { size_t const jobSizeTarget = (size_t)1 << ZSTDMT_computeTargetJobLog(params);
+ size_t const jobMaxSize = jobSizeTarget << 2;
+ size_t const passSizeMax = jobMaxSize * nbWorkers;
+ unsigned const multiplier = (unsigned)(srcSize / passSizeMax) + 1;
+ unsigned const nbJobsLarge = multiplier * nbWorkers;
+ unsigned const nbJobsMax = (unsigned)(srcSize / jobSizeTarget) + 1;
+ unsigned const nbJobsSmall = MIN(nbJobsMax, nbWorkers);
+ return (multiplier>1) ? nbJobsLarge : nbJobsSmall;
+} }
+
+/* ZSTDMT_compress_advanced_internal() :
+ * This is a blocking function : it will only give back control to caller after finishing its compression job.
+ */
+static size_t
+ZSTDMT_compress_advanced_internal(
+ ZSTDMT_CCtx* mtctx,
+ void* dst, size_t dstCapacity,
+ const void* src, size_t srcSize,
+ const ZSTD_CDict* cdict,
+ ZSTD_CCtx_params params)
+{
+ ZSTD_CCtx_params const jobParams = ZSTDMT_initJobCCtxParams(&params);
+ size_t const overlapSize = ZSTDMT_computeOverlapSize(&params);
+ unsigned const nbJobs = ZSTDMT_computeNbJobs(&params, srcSize, params.nbWorkers);
+ size_t const proposedJobSize = (srcSize + (nbJobs-1)) / nbJobs;
+ size_t const avgJobSize = (((proposedJobSize-1) & 0x1FFFF) < 0x7FFF) ? proposedJobSize + 0xFFFF : proposedJobSize; /* avoid too small last block */
+ const char* const srcStart = (const char*)src;
+ size_t remainingSrcSize = srcSize;
+ unsigned const compressWithinDst = (dstCapacity >= ZSTD_compressBound(srcSize)) ? nbJobs : (unsigned)(dstCapacity / ZSTD_compressBound(avgJobSize)); /* presumes avgJobSize >= 256 KB, which should be the case */
+ size_t frameStartPos = 0, dstBufferPos = 0;
+ assert(jobParams.nbWorkers == 0);
+ assert(mtctx->cctxPool->totalCCtx == params.nbWorkers);
+
+ params.jobSize = (U32)avgJobSize;
+ DEBUGLOG(4, "ZSTDMT_compress_advanced_internal: nbJobs=%2u (rawSize=%u bytes; fixedSize=%u) ",
+ nbJobs, (U32)proposedJobSize, (U32)avgJobSize);
+
+ if ((nbJobs==1) | (params.nbWorkers<=1)) { /* fallback to single-thread mode : this is a blocking invocation anyway */
+ ZSTD_CCtx* const cctx = mtctx->cctxPool->cctx[0];
+ DEBUGLOG(4, "ZSTDMT_compress_advanced_internal: fallback to single-thread mode");
+ if (cdict) return ZSTD_compress_usingCDict_advanced(cctx, dst, dstCapacity, src, srcSize, cdict, jobParams.fParams);
+ return ZSTD_compress_advanced_internal(cctx, dst, dstCapacity, src, srcSize, NULL, 0, &jobParams);
+ }
+
+ assert(avgJobSize >= 256 KB); /* condition for ZSTD_compressBound(A) + ZSTD_compressBound(B) <= ZSTD_compressBound(A+B), required to compress directly into Dst (no additional buffer) */
+ ZSTDMT_setBufferSize(mtctx->bufPool, ZSTD_compressBound(avgJobSize) );
+ /* LDM doesn't even try to load the dictionary in single-ingestion mode */
+ if (ZSTDMT_serialState_reset(&mtctx->serial, mtctx->seqPool, params, avgJobSize, NULL, 0, ZSTD_dct_auto))
+ return ERROR(memory_allocation);
+
+ FORWARD_IF_ERROR( ZSTDMT_expandJobsTable(mtctx, nbJobs) , ""); /* only expands if necessary */
+
+ { unsigned u;
+ for (u=0; u<nbJobs; u++) {
+ size_t const jobSize = MIN(remainingSrcSize, avgJobSize);
+ size_t const dstBufferCapacity = ZSTD_compressBound(jobSize);
+ buffer_t const dstAsBuffer = { (char*)dst + dstBufferPos, dstBufferCapacity };
+ buffer_t const dstBuffer = u < compressWithinDst ? dstAsBuffer : g_nullBuffer;
+ size_t dictSize = u ? overlapSize : 0;
+
+ mtctx->jobs[u].prefix.start = srcStart + frameStartPos - dictSize;
+ mtctx->jobs[u].prefix.size = dictSize;
+ mtctx->jobs[u].src.start = srcStart + frameStartPos;
+ mtctx->jobs[u].src.size = jobSize; assert(jobSize > 0); /* avoid job.src.size == 0 */
+ mtctx->jobs[u].consumed = 0;
+ mtctx->jobs[u].cSize = 0;
+ mtctx->jobs[u].cdict = (u==0) ? cdict : NULL;
+ mtctx->jobs[u].fullFrameSize = srcSize;
+ mtctx->jobs[u].params = jobParams;
+ /* do not calculate checksum within sections, but write it in header for first section */
+ mtctx->jobs[u].dstBuff = dstBuffer;
+ mtctx->jobs[u].cctxPool = mtctx->cctxPool;
+ mtctx->jobs[u].bufPool = mtctx->bufPool;
+ mtctx->jobs[u].seqPool = mtctx->seqPool;
+ mtctx->jobs[u].serial = &mtctx->serial;
+ mtctx->jobs[u].jobID = u;
+ mtctx->jobs[u].firstJob = (u==0);
+ mtctx->jobs[u].lastJob = (u==nbJobs-1);
+
+ DEBUGLOG(5, "ZSTDMT_compress_advanced_internal: posting job %u (%u bytes)", u, (U32)jobSize);
+ DEBUG_PRINTHEX(6, mtctx->jobs[u].prefix.start, 12);
+ POOL_add(mtctx->factory, ZSTDMT_compressionJob, &mtctx->jobs[u]);
+
+ frameStartPos += jobSize;
+ dstBufferPos += dstBufferCapacity;
+ remainingSrcSize -= jobSize;
+ } }
+
+ /* collect result */
+ { size_t error = 0, dstPos = 0;
+ unsigned jobID;
+ for (jobID=0; jobID<nbJobs; jobID++) {
+ DEBUGLOG(5, "waiting for job %u ", jobID);
+ ZSTD_PTHREAD_MUTEX_LOCK(&mtctx->jobs[jobID].job_mutex);
+ while (mtctx->jobs[jobID].consumed < mtctx->jobs[jobID].src.size) {
+ DEBUGLOG(5, "waiting for jobCompleted signal from job %u", jobID);
+ ZSTD_pthread_cond_wait(&mtctx->jobs[jobID].job_cond, &mtctx->jobs[jobID].job_mutex);
+ }
+ ZSTD_pthread_mutex_unlock(&mtctx->jobs[jobID].job_mutex);
+ DEBUGLOG(5, "ready to write job %u ", jobID);
+
+ { size_t const cSize = mtctx->jobs[jobID].cSize;
+ if (ZSTD_isError(cSize)) error = cSize;
+ if ((!error) && (dstPos + cSize > dstCapacity)) error = ERROR(dstSize_tooSmall);
+ if (jobID) { /* note : job 0 is written directly at dst, which is correct position */
+ if (!error)
+ memmove((char*)dst + dstPos, mtctx->jobs[jobID].dstBuff.start, cSize); /* may overlap when job compressed within dst */
+ if (jobID >= compressWithinDst) { /* job compressed into its own buffer, which must be released */
+ DEBUGLOG(5, "releasing buffer %u>=%u", jobID, compressWithinDst);
+ ZSTDMT_releaseBuffer(mtctx->bufPool, mtctx->jobs[jobID].dstBuff);
+ } }
+ mtctx->jobs[jobID].dstBuff = g_nullBuffer;
+ mtctx->jobs[jobID].cSize = 0;
+ dstPos += cSize ;
+ }
+ } /* for (jobID=0; jobID<nbJobs; jobID++) */
+
+ DEBUGLOG(4, "checksumFlag : %u ", params.fParams.checksumFlag);
+ if (params.fParams.checksumFlag) {
+ U32 const checksum = (U32)XXH64_digest(&mtctx->serial.xxhState);
+ if (dstPos + 4 > dstCapacity) {
+ error = ERROR(dstSize_tooSmall);
+ } else {
+ DEBUGLOG(4, "writing checksum : %08X \n", checksum);
+ MEM_writeLE32((char*)dst + dstPos, checksum);
+ dstPos += 4;
+ } }
+
+ if (!error) DEBUGLOG(4, "compressed size : %u ", (U32)dstPos);
+ return error ? error : dstPos;
+ }
+}
+
+size_t ZSTDMT_compress_advanced(ZSTDMT_CCtx* mtctx,
+ void* dst, size_t dstCapacity,
+ const void* src, size_t srcSize,
+ const ZSTD_CDict* cdict,
+ ZSTD_parameters params,
+ int overlapLog)
+{
+ ZSTD_CCtx_params cctxParams = mtctx->params;
+ cctxParams.cParams = params.cParams;
+ cctxParams.fParams = params.fParams;
+ assert(ZSTD_OVERLAPLOG_MIN <= overlapLog && overlapLog <= ZSTD_OVERLAPLOG_MAX);
+ cctxParams.overlapLog = overlapLog;
+ return ZSTDMT_compress_advanced_internal(mtctx,
+ dst, dstCapacity,
+ src, srcSize,
+ cdict, cctxParams);
+}
+
+
+size_t ZSTDMT_compressCCtx(ZSTDMT_CCtx* mtctx,
+ void* dst, size_t dstCapacity,
+ const void* src, size_t srcSize,
+ int compressionLevel)
+{
+ ZSTD_parameters params = ZSTD_getParams(compressionLevel, srcSize, 0);
+ int const overlapLog = ZSTDMT_overlapLog_default(params.cParams.strategy);
+ params.fParams.contentSizeFlag = 1;
+ return ZSTDMT_compress_advanced(mtctx, dst, dstCapacity, src, srcSize, NULL, params, overlapLog);
+}
+
+
+/* ====================================== */
+/* ======= Streaming API ======= */
+/* ====================================== */
+
+size_t ZSTDMT_initCStream_internal(
+ ZSTDMT_CCtx* mtctx,
+ const void* dict, size_t dictSize, ZSTD_dictContentType_e dictContentType,
+ const ZSTD_CDict* cdict, ZSTD_CCtx_params params,
+ unsigned long long pledgedSrcSize)
+{
+ DEBUGLOG(4, "ZSTDMT_initCStream_internal (pledgedSrcSize=%u, nbWorkers=%u, cctxPool=%u)",
+ (U32)pledgedSrcSize, params.nbWorkers, mtctx->cctxPool->totalCCtx);
+
+ /* params supposed partially fully validated at this point */
+ assert(!ZSTD_isError(ZSTD_checkCParams(params.cParams)));
+ assert(!((dict) && (cdict))); /* either dict or cdict, not both */
+
+ /* init */
+ if (params.nbWorkers != mtctx->params.nbWorkers)
+ FORWARD_IF_ERROR( ZSTDMT_resize(mtctx, params.nbWorkers) , "");
+
+ if (params.jobSize != 0 && params.jobSize < ZSTDMT_JOBSIZE_MIN) params.jobSize = ZSTDMT_JOBSIZE_MIN;
+ if (params.jobSize > (size_t)ZSTDMT_JOBSIZE_MAX) params.jobSize = (size_t)ZSTDMT_JOBSIZE_MAX;
+
+ mtctx->singleBlockingThread = (pledgedSrcSize <= ZSTDMT_JOBSIZE_MIN); /* do not trigger multi-threading when srcSize is too small */
+ if (mtctx->singleBlockingThread) {
+ ZSTD_CCtx_params const singleThreadParams = ZSTDMT_initJobCCtxParams(&params);
+ DEBUGLOG(5, "ZSTDMT_initCStream_internal: switch to single blocking thread mode");
+ assert(singleThreadParams.nbWorkers == 0);
+ return ZSTD_initCStream_internal(mtctx->cctxPool->cctx[0],
+ dict, dictSize, cdict,
+ &singleThreadParams, pledgedSrcSize);
+ }
+
+ DEBUGLOG(4, "ZSTDMT_initCStream_internal: %u workers", params.nbWorkers);
+
+ if (mtctx->allJobsCompleted == 0) { /* previous compression not correctly finished */
+ ZSTDMT_waitForAllJobsCompleted(mtctx);
+ ZSTDMT_releaseAllJobResources(mtctx);
+ mtctx->allJobsCompleted = 1;
+ }
+
+ mtctx->params = params;
+ mtctx->frameContentSize = pledgedSrcSize;
+ if (dict) {
+ ZSTD_freeCDict(mtctx->cdictLocal);
+ mtctx->cdictLocal = ZSTD_createCDict_advanced(dict, dictSize,
+ ZSTD_dlm_byCopy, dictContentType, /* note : a loadPrefix becomes an internal CDict */
+ params.cParams, mtctx->cMem);
+ mtctx->cdict = mtctx->cdictLocal;
+ if (mtctx->cdictLocal == NULL) return ERROR(memory_allocation);
+ } else {
+ ZSTD_freeCDict(mtctx->cdictLocal);
+ mtctx->cdictLocal = NULL;
+ mtctx->cdict = cdict;
+ }
+
+ mtctx->targetPrefixSize = ZSTDMT_computeOverlapSize(&params);
+ DEBUGLOG(4, "overlapLog=%i => %u KB", params.overlapLog, (U32)(mtctx->targetPrefixSize>>10));
+ mtctx->targetSectionSize = params.jobSize;
+ if (mtctx->targetSectionSize == 0) {
+ mtctx->targetSectionSize = 1ULL << ZSTDMT_computeTargetJobLog(&params);
+ }
+ assert(mtctx->targetSectionSize <= (size_t)ZSTDMT_JOBSIZE_MAX);
+
+ if (params.rsyncable) {
+ /* Aim for the targetsectionSize as the average job size. */
+ U32 const jobSizeMB = (U32)(mtctx->targetSectionSize >> 20);
+ U32 const rsyncBits = ZSTD_highbit32(jobSizeMB) + 20;
+ assert(jobSizeMB >= 1);
+ DEBUGLOG(4, "rsyncLog = %u", rsyncBits);
+ mtctx->rsync.hash = 0;
+ mtctx->rsync.hitMask = (1ULL << rsyncBits) - 1;
+ mtctx->rsync.primePower = ZSTD_rollingHash_primePower(RSYNC_LENGTH);
+ }
+ if (mtctx->targetSectionSize < mtctx->targetPrefixSize) mtctx->targetSectionSize = mtctx->targetPrefixSize; /* job size must be >= overlap size */
+ DEBUGLOG(4, "Job Size : %u KB (note : set to %u)", (U32)(mtctx->targetSectionSize>>10), (U32)params.jobSize);
+ DEBUGLOG(4, "inBuff Size : %u KB", (U32)(mtctx->targetSectionSize>>10));
+ ZSTDMT_setBufferSize(mtctx->bufPool, ZSTD_compressBound(mtctx->targetSectionSize));
+ {
+ /* If ldm is enabled we need windowSize space. */
+ size_t const windowSize = mtctx->params.ldmParams.enableLdm ? (1U << mtctx->params.cParams.windowLog) : 0;
+ /* Two buffers of slack, plus extra space for the overlap
+ * This is the minimum slack that LDM works with. One extra because
+ * flush might waste up to targetSectionSize-1 bytes. Another extra
+ * for the overlap (if > 0), then one to fill which doesn't overlap
+ * with the LDM window.
+ */
+ size_t const nbSlackBuffers = 2 + (mtctx->targetPrefixSize > 0);
+ size_t const slackSize = mtctx->targetSectionSize * nbSlackBuffers;
+ /* Compute the total size, and always have enough slack */
+ size_t const nbWorkers = MAX(mtctx->params.nbWorkers, 1);
+ size_t const sectionsSize = mtctx->targetSectionSize * nbWorkers;
+ size_t const capacity = MAX(windowSize, sectionsSize) + slackSize;
+ if (mtctx->roundBuff.capacity < capacity) {
+ if (mtctx->roundBuff.buffer)
+ ZSTD_free(mtctx->roundBuff.buffer, mtctx->cMem);
+ mtctx->roundBuff.buffer = (BYTE*)ZSTD_malloc(capacity, mtctx->cMem);
+ if (mtctx->roundBuff.buffer == NULL) {
+ mtctx->roundBuff.capacity = 0;
+ return ERROR(memory_allocation);
+ }
+ mtctx->roundBuff.capacity = capacity;
+ }
+ }
+ DEBUGLOG(4, "roundBuff capacity : %u KB", (U32)(mtctx->roundBuff.capacity>>10));
+ mtctx->roundBuff.pos = 0;
+ mtctx->inBuff.buffer = g_nullBuffer;
+ mtctx->inBuff.filled = 0;
+ mtctx->inBuff.prefix = kNullRange;
+ mtctx->doneJobID = 0;
+ mtctx->nextJobID = 0;
+ mtctx->frameEnded = 0;
+ mtctx->allJobsCompleted = 0;
+ mtctx->consumed = 0;
+ mtctx->produced = 0;
+ if (ZSTDMT_serialState_reset(&mtctx->serial, mtctx->seqPool, params, mtctx->targetSectionSize,
+ dict, dictSize, dictContentType))
+ return ERROR(memory_allocation);
+ return 0;
+}
+
+size_t ZSTDMT_initCStream_advanced(ZSTDMT_CCtx* mtctx,
+ const void* dict, size_t dictSize,
+ ZSTD_parameters params,
+ unsigned long long pledgedSrcSize)
+{
+ ZSTD_CCtx_params cctxParams = mtctx->params; /* retrieve sticky params */
+ DEBUGLOG(4, "ZSTDMT_initCStream_advanced (pledgedSrcSize=%u)", (U32)pledgedSrcSize);
+ cctxParams.cParams = params.cParams;
+ cctxParams.fParams = params.fParams;
+ return ZSTDMT_initCStream_internal(mtctx, dict, dictSize, ZSTD_dct_auto, NULL,
+ cctxParams, pledgedSrcSize);
+}
+
+size_t ZSTDMT_initCStream_usingCDict(ZSTDMT_CCtx* mtctx,
+ const ZSTD_CDict* cdict,
+ ZSTD_frameParameters fParams,
+ unsigned long long pledgedSrcSize)
+{
+ ZSTD_CCtx_params cctxParams = mtctx->params;
+ if (cdict==NULL) return ERROR(dictionary_wrong); /* method incompatible with NULL cdict */
+ cctxParams.cParams = ZSTD_getCParamsFromCDict(cdict);
+ cctxParams.fParams = fParams;
+ return ZSTDMT_initCStream_internal(mtctx, NULL, 0 /*dictSize*/, ZSTD_dct_auto, cdict,
+ cctxParams, pledgedSrcSize);
+}
+
+
+/* ZSTDMT_resetCStream() :
+ * pledgedSrcSize can be zero == unknown (for the time being)
+ * prefer using ZSTD_CONTENTSIZE_UNKNOWN,
+ * as `0` might mean "empty" in the future */
+size_t ZSTDMT_resetCStream(ZSTDMT_CCtx* mtctx, unsigned long long pledgedSrcSize)
+{
+ if (!pledgedSrcSize) pledgedSrcSize = ZSTD_CONTENTSIZE_UNKNOWN;
+ return ZSTDMT_initCStream_internal(mtctx, NULL, 0, ZSTD_dct_auto, 0, mtctx->params,
+ pledgedSrcSize);
+}
+
+size_t ZSTDMT_initCStream(ZSTDMT_CCtx* mtctx, int compressionLevel) {
+ ZSTD_parameters const params = ZSTD_getParams(compressionLevel, ZSTD_CONTENTSIZE_UNKNOWN, 0);
+ ZSTD_CCtx_params cctxParams = mtctx->params; /* retrieve sticky params */
+ DEBUGLOG(4, "ZSTDMT_initCStream (cLevel=%i)", compressionLevel);
+ cctxParams.cParams = params.cParams;
+ cctxParams.fParams = params.fParams;
+ return ZSTDMT_initCStream_internal(mtctx, NULL, 0, ZSTD_dct_auto, NULL, cctxParams, ZSTD_CONTENTSIZE_UNKNOWN);
+}
+
+
+/* ZSTDMT_writeLastEmptyBlock()
+ * Write a single empty block with an end-of-frame to finish a frame.
+ * Job must be created from streaming variant.
+ * This function is always successful if expected conditions are fulfilled.
+ */
+static void ZSTDMT_writeLastEmptyBlock(ZSTDMT_jobDescription* job)
+{
+ assert(job->lastJob == 1);
+ assert(job->src.size == 0); /* last job is empty -> will be simplified into a last empty block */
+ assert(job->firstJob == 0); /* cannot be first job, as it also needs to create frame header */
+ assert(job->dstBuff.start == NULL); /* invoked from streaming variant only (otherwise, dstBuff might be user's output) */
+ job->dstBuff = ZSTDMT_getBuffer(job->bufPool);
+ if (job->dstBuff.start == NULL) {
+ job->cSize = ERROR(memory_allocation);
+ return;
+ }
+ assert(job->dstBuff.capacity >= ZSTD_blockHeaderSize); /* no buffer should ever be that small */
+ job->src = kNullRange;
+ job->cSize = ZSTD_writeLastEmptyBlock(job->dstBuff.start, job->dstBuff.capacity);
+ assert(!ZSTD_isError(job->cSize));
+ assert(job->consumed == 0);
+}
+
+static size_t ZSTDMT_createCompressionJob(ZSTDMT_CCtx* mtctx, size_t srcSize, ZSTD_EndDirective endOp)
+{
+ unsigned const jobID = mtctx->nextJobID & mtctx->jobIDMask;
+ int const endFrame = (endOp == ZSTD_e_end);
+
+ if (mtctx->nextJobID > mtctx->doneJobID + mtctx->jobIDMask) {
+ DEBUGLOG(5, "ZSTDMT_createCompressionJob: will not create new job : table is full");
+ assert((mtctx->nextJobID & mtctx->jobIDMask) == (mtctx->doneJobID & mtctx->jobIDMask));
+ return 0;
+ }
+
+ if (!mtctx->jobReady) {
+ BYTE const* src = (BYTE const*)mtctx->inBuff.buffer.start;
+ DEBUGLOG(5, "ZSTDMT_createCompressionJob: preparing job %u to compress %u bytes with %u preload ",
+ mtctx->nextJobID, (U32)srcSize, (U32)mtctx->inBuff.prefix.size);
+ mtctx->jobs[jobID].src.start = src;
+ mtctx->jobs[jobID].src.size = srcSize;
+ assert(mtctx->inBuff.filled >= srcSize);
+ mtctx->jobs[jobID].prefix = mtctx->inBuff.prefix;
+ mtctx->jobs[jobID].consumed = 0;
+ mtctx->jobs[jobID].cSize = 0;
+ mtctx->jobs[jobID].params = mtctx->params;
+ mtctx->jobs[jobID].cdict = mtctx->nextJobID==0 ? mtctx->cdict : NULL;
+ mtctx->jobs[jobID].fullFrameSize = mtctx->frameContentSize;
+ mtctx->jobs[jobID].dstBuff = g_nullBuffer;
+ mtctx->jobs[jobID].cctxPool = mtctx->cctxPool;
+ mtctx->jobs[jobID].bufPool = mtctx->bufPool;
+ mtctx->jobs[jobID].seqPool = mtctx->seqPool;
+ mtctx->jobs[jobID].serial = &mtctx->serial;
+ mtctx->jobs[jobID].jobID = mtctx->nextJobID;
+ mtctx->jobs[jobID].firstJob = (mtctx->nextJobID==0);
+ mtctx->jobs[jobID].lastJob = endFrame;
+ mtctx->jobs[jobID].frameChecksumNeeded = mtctx->params.fParams.checksumFlag && endFrame && (mtctx->nextJobID>0);
+ mtctx->jobs[jobID].dstFlushed = 0;
+
+ /* Update the round buffer pos and clear the input buffer to be reset */
+ mtctx->roundBuff.pos += srcSize;
+ mtctx->inBuff.buffer = g_nullBuffer;
+ mtctx->inBuff.filled = 0;
+ /* Set the prefix */
+ if (!endFrame) {
+ size_t const newPrefixSize = MIN(srcSize, mtctx->targetPrefixSize);
+ mtctx->inBuff.prefix.start = src + srcSize - newPrefixSize;
+ mtctx->inBuff.prefix.size = newPrefixSize;
+ } else { /* endFrame==1 => no need for another input buffer */
+ mtctx->inBuff.prefix = kNullRange;
+ mtctx->frameEnded = endFrame;
+ if (mtctx->nextJobID == 0) {
+ /* single job exception : checksum is already calculated directly within worker thread */
+ mtctx->params.fParams.checksumFlag = 0;
+ } }
+
+ if ( (srcSize == 0)
+ && (mtctx->nextJobID>0)/*single job must also write frame header*/ ) {
+ DEBUGLOG(5, "ZSTDMT_createCompressionJob: creating a last empty block to end frame");
+ assert(endOp == ZSTD_e_end); /* only possible case : need to end the frame with an empty last block */
+ ZSTDMT_writeLastEmptyBlock(mtctx->jobs + jobID);
+ mtctx->nextJobID++;
+ return 0;
+ }
+ }
+
+ DEBUGLOG(5, "ZSTDMT_createCompressionJob: posting job %u : %u bytes (end:%u, jobNb == %u (mod:%u))",
+ mtctx->nextJobID,
+ (U32)mtctx->jobs[jobID].src.size,
+ mtctx->jobs[jobID].lastJob,
+ mtctx->nextJobID,
+ jobID);
+ if (POOL_tryAdd(mtctx->factory, ZSTDMT_compressionJob, &mtctx->jobs[jobID])) {
+ mtctx->nextJobID++;
+ mtctx->jobReady = 0;
+ } else {
+ DEBUGLOG(5, "ZSTDMT_createCompressionJob: no worker available for job %u", mtctx->nextJobID);
+ mtctx->jobReady = 1;
+ }
+ return 0;
+}
+
+
+/*! ZSTDMT_flushProduced() :
+ * flush whatever data has been produced but not yet flushed in current job.
+ * move to next job if current one is fully flushed.
+ * `output` : `pos` will be updated with amount of data flushed .
+ * `blockToFlush` : if >0, the function will block and wait if there is no data available to flush .
+ * @return : amount of data remaining within internal buffer, 0 if no more, 1 if unknown but > 0, or an error code */
+static size_t ZSTDMT_flushProduced(ZSTDMT_CCtx* mtctx, ZSTD_outBuffer* output, unsigned blockToFlush, ZSTD_EndDirective end)
+{
+ unsigned const wJobID = mtctx->doneJobID & mtctx->jobIDMask;
+ DEBUGLOG(5, "ZSTDMT_flushProduced (blocking:%u , job %u <= %u)",
+ blockToFlush, mtctx->doneJobID, mtctx->nextJobID);
+ assert(output->size >= output->pos);
+
+ ZSTD_PTHREAD_MUTEX_LOCK(&mtctx->jobs[wJobID].job_mutex);
+ if ( blockToFlush
+ && (mtctx->doneJobID < mtctx->nextJobID) ) {
+ assert(mtctx->jobs[wJobID].dstFlushed <= mtctx->jobs[wJobID].cSize);
+ while (mtctx->jobs[wJobID].dstFlushed == mtctx->jobs[wJobID].cSize) { /* nothing to flush */
+ if (mtctx->jobs[wJobID].consumed == mtctx->jobs[wJobID].src.size) {
+ DEBUGLOG(5, "job %u is completely consumed (%u == %u) => don't wait for cond, there will be none",
+ mtctx->doneJobID, (U32)mtctx->jobs[wJobID].consumed, (U32)mtctx->jobs[wJobID].src.size);
+ break;
+ }
+ DEBUGLOG(5, "waiting for something to flush from job %u (currently flushed: %u bytes)",
+ mtctx->doneJobID, (U32)mtctx->jobs[wJobID].dstFlushed);
+ ZSTD_pthread_cond_wait(&mtctx->jobs[wJobID].job_cond, &mtctx->jobs[wJobID].job_mutex); /* block when nothing to flush but some to come */
+ } }
+
+ /* try to flush something */
+ { size_t cSize = mtctx->jobs[wJobID].cSize; /* shared */
+ size_t const srcConsumed = mtctx->jobs[wJobID].consumed; /* shared */
+ size_t const srcSize = mtctx->jobs[wJobID].src.size; /* read-only, could be done after mutex lock, but no-declaration-after-statement */
+ ZSTD_pthread_mutex_unlock(&mtctx->jobs[wJobID].job_mutex);
+ if (ZSTD_isError(cSize)) {
+ DEBUGLOG(5, "ZSTDMT_flushProduced: job %u : compression error detected : %s",
+ mtctx->doneJobID, ZSTD_getErrorName(cSize));
+ ZSTDMT_waitForAllJobsCompleted(mtctx);
+ ZSTDMT_releaseAllJobResources(mtctx);
+ return cSize;
+ }
+ /* add frame checksum if necessary (can only happen once) */
+ assert(srcConsumed <= srcSize);
+ if ( (srcConsumed == srcSize) /* job completed -> worker no longer active */
+ && mtctx->jobs[wJobID].frameChecksumNeeded ) {
+ U32 const checksum = (U32)XXH64_digest(&mtctx->serial.xxhState);
+ DEBUGLOG(4, "ZSTDMT_flushProduced: writing checksum : %08X \n", checksum);
+ MEM_writeLE32((char*)mtctx->jobs[wJobID].dstBuff.start + mtctx->jobs[wJobID].cSize, checksum);
+ cSize += 4;
+ mtctx->jobs[wJobID].cSize += 4; /* can write this shared value, as worker is no longer active */
+ mtctx->jobs[wJobID].frameChecksumNeeded = 0;
+ }
+
+ if (cSize > 0) { /* compression is ongoing or completed */
+ size_t const toFlush = MIN(cSize - mtctx->jobs[wJobID].dstFlushed, output->size - output->pos);
+ DEBUGLOG(5, "ZSTDMT_flushProduced: Flushing %u bytes from job %u (completion:%u/%u, generated:%u)",
+ (U32)toFlush, mtctx->doneJobID, (U32)srcConsumed, (U32)srcSize, (U32)cSize);
+ assert(mtctx->doneJobID < mtctx->nextJobID);
+ assert(cSize >= mtctx->jobs[wJobID].dstFlushed);
+ assert(mtctx->jobs[wJobID].dstBuff.start != NULL);
+ if (toFlush > 0) {
+ memcpy((char*)output->dst + output->pos,
+ (const char*)mtctx->jobs[wJobID].dstBuff.start + mtctx->jobs[wJobID].dstFlushed,
+ toFlush);
+ }
+ output->pos += toFlush;
+ mtctx->jobs[wJobID].dstFlushed += toFlush; /* can write : this value is only used by mtctx */
+
+ if ( (srcConsumed == srcSize) /* job is completed */
+ && (mtctx->jobs[wJobID].dstFlushed == cSize) ) { /* output buffer fully flushed => free this job position */
+ DEBUGLOG(5, "Job %u completed (%u bytes), moving to next one",
+ mtctx->doneJobID, (U32)mtctx->jobs[wJobID].dstFlushed);
+ ZSTDMT_releaseBuffer(mtctx->bufPool, mtctx->jobs[wJobID].dstBuff);
+ DEBUGLOG(5, "dstBuffer released");
+ mtctx->jobs[wJobID].dstBuff = g_nullBuffer;
+ mtctx->jobs[wJobID].cSize = 0; /* ensure this job slot is considered "not started" in future check */
+ mtctx->consumed += srcSize;
+ mtctx->produced += cSize;
+ mtctx->doneJobID++;
+ } }
+
+ /* return value : how many bytes left in buffer ; fake it to 1 when unknown but >0 */
+ if (cSize > mtctx->jobs[wJobID].dstFlushed) return (cSize - mtctx->jobs[wJobID].dstFlushed);
+ if (srcSize > srcConsumed) return 1; /* current job not completely compressed */
+ }
+ if (mtctx->doneJobID < mtctx->nextJobID) return 1; /* some more jobs ongoing */
+ if (mtctx->jobReady) return 1; /* one job is ready to push, just not yet in the list */
+ if (mtctx->inBuff.filled > 0) return 1; /* input is not empty, and still needs to be converted into a job */
+ mtctx->allJobsCompleted = mtctx->frameEnded; /* all jobs are entirely flushed => if this one is last one, frame is completed */
+ if (end == ZSTD_e_end) return !mtctx->frameEnded; /* for ZSTD_e_end, question becomes : is frame completed ? instead of : are internal buffers fully flushed ? */
+ return 0; /* internal buffers fully flushed */
+}
+
+/**
+ * Returns the range of data used by the earliest job that is not yet complete.
+ * If the data of the first job is broken up into two segments, we cover both
+ * sections.
+ */
+static range_t ZSTDMT_getInputDataInUse(ZSTDMT_CCtx* mtctx)
+{
+ unsigned const firstJobID = mtctx->doneJobID;
+ unsigned const lastJobID = mtctx->nextJobID;
+ unsigned jobID;
+
+ for (jobID = firstJobID; jobID < lastJobID; ++jobID) {
+ unsigned const wJobID = jobID & mtctx->jobIDMask;
+ size_t consumed;
+
+ ZSTD_PTHREAD_MUTEX_LOCK(&mtctx->jobs[wJobID].job_mutex);
+ consumed = mtctx->jobs[wJobID].consumed;
+ ZSTD_pthread_mutex_unlock(&mtctx->jobs[wJobID].job_mutex);
+
+ if (consumed < mtctx->jobs[wJobID].src.size) {
+ range_t range = mtctx->jobs[wJobID].prefix;
+ if (range.size == 0) {
+ /* Empty prefix */
+ range = mtctx->jobs[wJobID].src;
+ }
+ /* Job source in multiple segments not supported yet */
+ assert(range.start <= mtctx->jobs[wJobID].src.start);
+ return range;
+ }
+ }
+ return kNullRange;
+}
+
+/**
+ * Returns non-zero iff buffer and range overlap.
+ */
+static int ZSTDMT_isOverlapped(buffer_t buffer, range_t range)
+{
+ BYTE const* const bufferStart = (BYTE const*)buffer.start;
+ BYTE const* const bufferEnd = bufferStart + buffer.capacity;
+ BYTE const* const rangeStart = (BYTE const*)range.start;
+ BYTE const* const rangeEnd = range.size != 0 ? rangeStart + range.size : rangeStart;
+
+ if (rangeStart == NULL || bufferStart == NULL)
+ return 0;
+ /* Empty ranges cannot overlap */
+ if (bufferStart == bufferEnd || rangeStart == rangeEnd)
+ return 0;
+
+ return bufferStart < rangeEnd && rangeStart < bufferEnd;
+}
+
+static int ZSTDMT_doesOverlapWindow(buffer_t buffer, ZSTD_window_t window)
+{
+ range_t extDict;
+ range_t prefix;
+
+ DEBUGLOG(5, "ZSTDMT_doesOverlapWindow");
+ extDict.start = window.dictBase + window.lowLimit;
+ extDict.size = window.dictLimit - window.lowLimit;
+
+ prefix.start = window.base + window.dictLimit;
+ prefix.size = window.nextSrc - (window.base + window.dictLimit);
+ DEBUGLOG(5, "extDict [0x%zx, 0x%zx)",
+ (size_t)extDict.start,
+ (size_t)extDict.start + extDict.size);
+ DEBUGLOG(5, "prefix [0x%zx, 0x%zx)",
+ (size_t)prefix.start,
+ (size_t)prefix.start + prefix.size);
+
+ return ZSTDMT_isOverlapped(buffer, extDict)
+ || ZSTDMT_isOverlapped(buffer, prefix);
+}
+
+static void ZSTDMT_waitForLdmComplete(ZSTDMT_CCtx* mtctx, buffer_t buffer)
+{
+ if (mtctx->params.ldmParams.enableLdm) {
+ ZSTD_pthread_mutex_t* mutex = &mtctx->serial.ldmWindowMutex;
+ DEBUGLOG(5, "ZSTDMT_waitForLdmComplete");
+ DEBUGLOG(5, "source [0x%zx, 0x%zx)",
+ (size_t)buffer.start,
+ (size_t)buffer.start + buffer.capacity);
+ ZSTD_PTHREAD_MUTEX_LOCK(mutex);
+ while (ZSTDMT_doesOverlapWindow(buffer, mtctx->serial.ldmWindow)) {
+ DEBUGLOG(5, "Waiting for LDM to finish...");
+ ZSTD_pthread_cond_wait(&mtctx->serial.ldmWindowCond, mutex);
+ }
+ DEBUGLOG(6, "Done waiting for LDM to finish");
+ ZSTD_pthread_mutex_unlock(mutex);
+ }
+}
+
+/**
+ * Attempts to set the inBuff to the next section to fill.
+ * If any part of the new section is still in use we give up.
+ * Returns non-zero if the buffer is filled.
+ */
+static int ZSTDMT_tryGetInputRange(ZSTDMT_CCtx* mtctx)
+{
+ range_t const inUse = ZSTDMT_getInputDataInUse(mtctx);
+ size_t const spaceLeft = mtctx->roundBuff.capacity - mtctx->roundBuff.pos;
+ size_t const target = mtctx->targetSectionSize;
+ buffer_t buffer;
+
+ DEBUGLOG(5, "ZSTDMT_tryGetInputRange");
+ assert(mtctx->inBuff.buffer.start == NULL);
+ assert(mtctx->roundBuff.capacity >= target);
+
+ if (spaceLeft < target) {
+ /* ZSTD_invalidateRepCodes() doesn't work for extDict variants.
+ * Simply copy the prefix to the beginning in that case.
+ */
+ BYTE* const start = (BYTE*)mtctx->roundBuff.buffer;
+ size_t const prefixSize = mtctx->inBuff.prefix.size;
+
+ buffer.start = start;
+ buffer.capacity = prefixSize;
+ if (ZSTDMT_isOverlapped(buffer, inUse)) {
+ DEBUGLOG(5, "Waiting for buffer...");
+ return 0;
+ }
+ ZSTDMT_waitForLdmComplete(mtctx, buffer);
+ memmove(start, mtctx->inBuff.prefix.start, prefixSize);
+ mtctx->inBuff.prefix.start = start;
+ mtctx->roundBuff.pos = prefixSize;
+ }
+ buffer.start = mtctx->roundBuff.buffer + mtctx->roundBuff.pos;
+ buffer.capacity = target;
+
+ if (ZSTDMT_isOverlapped(buffer, inUse)) {
+ DEBUGLOG(5, "Waiting for buffer...");
+ return 0;
+ }
+ assert(!ZSTDMT_isOverlapped(buffer, mtctx->inBuff.prefix));
+
+ ZSTDMT_waitForLdmComplete(mtctx, buffer);
+
+ DEBUGLOG(5, "Using prefix range [%zx, %zx)",
+ (size_t)mtctx->inBuff.prefix.start,
+ (size_t)mtctx->inBuff.prefix.start + mtctx->inBuff.prefix.size);
+ DEBUGLOG(5, "Using source range [%zx, %zx)",
+ (size_t)buffer.start,
+ (size_t)buffer.start + buffer.capacity);
+
+
+ mtctx->inBuff.buffer = buffer;
+ mtctx->inBuff.filled = 0;
+ assert(mtctx->roundBuff.pos + buffer.capacity <= mtctx->roundBuff.capacity);
+ return 1;
+}
+
+typedef struct {
+ size_t toLoad; /* The number of bytes to load from the input. */
+ int flush; /* Boolean declaring if we must flush because we found a synchronization point. */
+} syncPoint_t;
+
+/**
+ * Searches through the input for a synchronization point. If one is found, we
+ * will instruct the caller to flush, and return the number of bytes to load.
+ * Otherwise, we will load as many bytes as possible and instruct the caller
+ * to continue as normal.
+ */
+static syncPoint_t
+findSynchronizationPoint(ZSTDMT_CCtx const* mtctx, ZSTD_inBuffer const input)
+{
+ BYTE const* const istart = (BYTE const*)input.src + input.pos;
+ U64 const primePower = mtctx->rsync.primePower;
+ U64 const hitMask = mtctx->rsync.hitMask;
+
+ syncPoint_t syncPoint;
+ U64 hash;
+ BYTE const* prev;
+ size_t pos;
+
+ syncPoint.toLoad = MIN(input.size - input.pos, mtctx->targetSectionSize - mtctx->inBuff.filled);
+ syncPoint.flush = 0;
+ if (!mtctx->params.rsyncable)
+ /* Rsync is disabled. */
+ return syncPoint;
+ if (mtctx->inBuff.filled + syncPoint.toLoad < RSYNC_LENGTH)
+ /* Not enough to compute the hash.
+ * We will miss any synchronization points in this RSYNC_LENGTH byte
+ * window. However, since it depends only in the internal buffers, if the
+ * state is already synchronized, we will remain synchronized.
+ * Additionally, the probability that we miss a synchronization point is
+ * low: RSYNC_LENGTH / targetSectionSize.
+ */
+ return syncPoint;
+ /* Initialize the loop variables. */
+ if (mtctx->inBuff.filled >= RSYNC_LENGTH) {
+ /* We have enough bytes buffered to initialize the hash.
+ * Start scanning at the beginning of the input.
+ */
+ pos = 0;
+ prev = (BYTE const*)mtctx->inBuff.buffer.start + mtctx->inBuff.filled - RSYNC_LENGTH;
+ hash = ZSTD_rollingHash_compute(prev, RSYNC_LENGTH);
+ } else {
+ /* We don't have enough bytes buffered to initialize the hash, but
+ * we know we have at least RSYNC_LENGTH bytes total.
+ * Start scanning after the first RSYNC_LENGTH bytes less the bytes
+ * already buffered.
+ */
+ pos = RSYNC_LENGTH - mtctx->inBuff.filled;
+ prev = (BYTE const*)mtctx->inBuff.buffer.start - pos;
+ hash = ZSTD_rollingHash_compute(mtctx->inBuff.buffer.start, mtctx->inBuff.filled);
+ hash = ZSTD_rollingHash_append(hash, istart, pos);
+ }
+ /* Starting with the hash of the previous RSYNC_LENGTH bytes, roll
+ * through the input. If we hit a synchronization point, then cut the
+ * job off, and tell the compressor to flush the job. Otherwise, load
+ * all the bytes and continue as normal.
+ * If we go too long without a synchronization point (targetSectionSize)
+ * then a block will be emitted anyways, but this is okay, since if we
+ * are already synchronized we will remain synchronized.
+ */
+ for (; pos < syncPoint.toLoad; ++pos) {
+ BYTE const toRemove = pos < RSYNC_LENGTH ? prev[pos] : istart[pos - RSYNC_LENGTH];
+ /* if (pos >= RSYNC_LENGTH) assert(ZSTD_rollingHash_compute(istart + pos - RSYNC_LENGTH, RSYNC_LENGTH) == hash); */
+ hash = ZSTD_rollingHash_rotate(hash, toRemove, istart[pos], primePower);
+ if ((hash & hitMask) == hitMask) {
+ syncPoint.toLoad = pos + 1;
+ syncPoint.flush = 1;
+ break;
+ }
+ }
+ return syncPoint;
+}
+
+size_t ZSTDMT_nextInputSizeHint(const ZSTDMT_CCtx* mtctx)
+{
+ size_t hintInSize = mtctx->targetSectionSize - mtctx->inBuff.filled;
+ if (hintInSize==0) hintInSize = mtctx->targetSectionSize;
+ return hintInSize;
+}
+
+/** ZSTDMT_compressStream_generic() :
+ * internal use only - exposed to be invoked from zstd_compress.c
+ * assumption : output and input are valid (pos <= size)
+ * @return : minimum amount of data remaining to flush, 0 if none */
+size_t ZSTDMT_compressStream_generic(ZSTDMT_CCtx* mtctx,
+ ZSTD_outBuffer* output,
+ ZSTD_inBuffer* input,
+ ZSTD_EndDirective endOp)
+{
+ unsigned forwardInputProgress = 0;
+ DEBUGLOG(5, "ZSTDMT_compressStream_generic (endOp=%u, srcSize=%u)",
+ (U32)endOp, (U32)(input->size - input->pos));
+ assert(output->pos <= output->size);
+ assert(input->pos <= input->size);
+
+ if (mtctx->singleBlockingThread) { /* delegate to single-thread (synchronous) */
+ return ZSTD_compressStream2(mtctx->cctxPool->cctx[0], output, input, endOp);
+ }
+
+ if ((mtctx->frameEnded) && (endOp==ZSTD_e_continue)) {
+ /* current frame being ended. Only flush/end are allowed */
+ return ERROR(stage_wrong);
+ }
+
+ /* single-pass shortcut (note : synchronous-mode) */
+ if ( (!mtctx->params.rsyncable) /* rsyncable mode is disabled */
+ && (mtctx->nextJobID == 0) /* just started */
+ && (mtctx->inBuff.filled == 0) /* nothing buffered */
+ && (!mtctx->jobReady) /* no job already created */
+ && (endOp == ZSTD_e_end) /* end order */
+ && (output->size - output->pos >= ZSTD_compressBound(input->size - input->pos)) ) { /* enough space in dst */
+ size_t const cSize = ZSTDMT_compress_advanced_internal(mtctx,
+ (char*)output->dst + output->pos, output->size - output->pos,
+ (const char*)input->src + input->pos, input->size - input->pos,
+ mtctx->cdict, mtctx->params);
+ if (ZSTD_isError(cSize)) return cSize;
+ input->pos = input->size;
+ output->pos += cSize;
+ mtctx->allJobsCompleted = 1;
+ mtctx->frameEnded = 1;
+ return 0;
+ }
+
+ /* fill input buffer */
+ if ( (!mtctx->jobReady)
+ && (input->size > input->pos) ) { /* support NULL input */
+ if (mtctx->inBuff.buffer.start == NULL) {
+ assert(mtctx->inBuff.filled == 0); /* Can't fill an empty buffer */
+ if (!ZSTDMT_tryGetInputRange(mtctx)) {
+ /* It is only possible for this operation to fail if there are
+ * still compression jobs ongoing.
+ */
+ DEBUGLOG(5, "ZSTDMT_tryGetInputRange failed");
+ assert(mtctx->doneJobID != mtctx->nextJobID);
+ } else
+ DEBUGLOG(5, "ZSTDMT_tryGetInputRange completed successfully : mtctx->inBuff.buffer.start = %p", mtctx->inBuff.buffer.start);
+ }
+ if (mtctx->inBuff.buffer.start != NULL) {
+ syncPoint_t const syncPoint = findSynchronizationPoint(mtctx, *input);
+ if (syncPoint.flush && endOp == ZSTD_e_continue) {
+ endOp = ZSTD_e_flush;
+ }
+ assert(mtctx->inBuff.buffer.capacity >= mtctx->targetSectionSize);
+ DEBUGLOG(5, "ZSTDMT_compressStream_generic: adding %u bytes on top of %u to buffer of size %u",
+ (U32)syncPoint.toLoad, (U32)mtctx->inBuff.filled, (U32)mtctx->targetSectionSize);
+ memcpy((char*)mtctx->inBuff.buffer.start + mtctx->inBuff.filled, (const char*)input->src + input->pos, syncPoint.toLoad);
+ input->pos += syncPoint.toLoad;
+ mtctx->inBuff.filled += syncPoint.toLoad;
+ forwardInputProgress = syncPoint.toLoad>0;
+ }
+ if ((input->pos < input->size) && (endOp == ZSTD_e_end))
+ endOp = ZSTD_e_flush; /* can't end now : not all input consumed */
+ }
+
+ if ( (mtctx->jobReady)
+ || (mtctx->inBuff.filled >= mtctx->targetSectionSize) /* filled enough : let's compress */
+ || ((endOp != ZSTD_e_continue) && (mtctx->inBuff.filled > 0)) /* something to flush : let's go */
+ || ((endOp == ZSTD_e_end) && (!mtctx->frameEnded)) ) { /* must finish the frame with a zero-size block */
+ size_t const jobSize = mtctx->inBuff.filled;
+ assert(mtctx->inBuff.filled <= mtctx->targetSectionSize);
+ FORWARD_IF_ERROR( ZSTDMT_createCompressionJob(mtctx, jobSize, endOp) , "");
+ }
+
+ /* check for potential compressed data ready to be flushed */
+ { size_t const remainingToFlush = ZSTDMT_flushProduced(mtctx, output, !forwardInputProgress, endOp); /* block if there was no forward input progress */
+ if (input->pos < input->size) return MAX(remainingToFlush, 1); /* input not consumed : do not end flush yet */
+ DEBUGLOG(5, "end of ZSTDMT_compressStream_generic: remainingToFlush = %u", (U32)remainingToFlush);
+ return remainingToFlush;
+ }
+}
+
+
+size_t ZSTDMT_compressStream(ZSTDMT_CCtx* mtctx, ZSTD_outBuffer* output, ZSTD_inBuffer* input)
+{
+ FORWARD_IF_ERROR( ZSTDMT_compressStream_generic(mtctx, output, input, ZSTD_e_continue) , "");
+
+ /* recommended next input size : fill current input buffer */
+ return mtctx->targetSectionSize - mtctx->inBuff.filled; /* note : could be zero when input buffer is fully filled and no more availability to create new job */
+}
+
+
+static size_t ZSTDMT_flushStream_internal(ZSTDMT_CCtx* mtctx, ZSTD_outBuffer* output, ZSTD_EndDirective endFrame)
+{
+ size_t const srcSize = mtctx->inBuff.filled;
+ DEBUGLOG(5, "ZSTDMT_flushStream_internal");
+
+ if ( mtctx->jobReady /* one job ready for a worker to pick up */
+ || (srcSize > 0) /* still some data within input buffer */
+ || ((endFrame==ZSTD_e_end) && !mtctx->frameEnded)) { /* need a last 0-size block to end frame */
+ DEBUGLOG(5, "ZSTDMT_flushStream_internal : create a new job (%u bytes, end:%u)",
+ (U32)srcSize, (U32)endFrame);
+ FORWARD_IF_ERROR( ZSTDMT_createCompressionJob(mtctx, srcSize, endFrame) , "");
+ }
+
+ /* check if there is any data available to flush */
+ return ZSTDMT_flushProduced(mtctx, output, 1 /* blockToFlush */, endFrame);
+}
+
+
+size_t ZSTDMT_flushStream(ZSTDMT_CCtx* mtctx, ZSTD_outBuffer* output)
+{
+ DEBUGLOG(5, "ZSTDMT_flushStream");
+ if (mtctx->singleBlockingThread)
+ return ZSTD_flushStream(mtctx->cctxPool->cctx[0], output);
+ return ZSTDMT_flushStream_internal(mtctx, output, ZSTD_e_flush);
+}
+
+size_t ZSTDMT_endStream(ZSTDMT_CCtx* mtctx, ZSTD_outBuffer* output)
+{
+ DEBUGLOG(4, "ZSTDMT_endStream");
+ if (mtctx->singleBlockingThread)
+ return ZSTD_endStream(mtctx->cctxPool->cctx[0], output);
+ return ZSTDMT_flushStream_internal(mtctx, output, ZSTD_e_end);
+}
diff --git a/src/zstd/lib/compress/zstdmt_compress.h b/src/zstd/lib/compress/zstdmt_compress.h
new file mode 100644
index 000000000..89914eb7f
--- /dev/null
+++ b/src/zstd/lib/compress/zstdmt_compress.h
@@ -0,0 +1,192 @@
+/*
+ * Copyright (c) 2016-2020, 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 ZSTDMT_COMPRESS_H
+ #define ZSTDMT_COMPRESS_H
+
+ #if defined (__cplusplus)
+ extern "C" {
+ #endif
+
+
+/* Note : This is an internal API.
+ * These APIs used to be exposed with ZSTDLIB_API,
+ * because it used to be the only way to invoke MT compression.
+ * Now, it's recommended to use ZSTD_compress2 and ZSTD_compressStream2()
+ * instead.
+ *
+ * If you depend on these APIs and can't switch, then define
+ * ZSTD_LEGACY_MULTITHREADED_API when making the dynamic library.
+ * However, we may completely remove these functions in a future
+ * release, so please switch soon.
+ *
+ * This API requires ZSTD_MULTITHREAD to be defined during compilation,
+ * otherwise ZSTDMT_createCCtx*() will fail.
+ */
+
+#ifdef ZSTD_LEGACY_MULTITHREADED_API
+# define ZSTDMT_API ZSTDLIB_API
+#else
+# define ZSTDMT_API
+#endif
+
+/* === Dependencies === */
+#include <stddef.h> /* size_t */
+#define ZSTD_STATIC_LINKING_ONLY /* ZSTD_parameters */
+#include "../zstd.h" /* ZSTD_inBuffer, ZSTD_outBuffer, ZSTDLIB_API */
+
+
+/* === Constants === */
+#ifndef ZSTDMT_NBWORKERS_MAX
+# define ZSTDMT_NBWORKERS_MAX 200
+#endif
+#ifndef ZSTDMT_JOBSIZE_MIN
+# define ZSTDMT_JOBSIZE_MIN (1 MB)
+#endif
+#define ZSTDMT_JOBLOG_MAX (MEM_32bits() ? 29 : 30)
+#define ZSTDMT_JOBSIZE_MAX (MEM_32bits() ? (512 MB) : (1024 MB))
+
+
+/* === Memory management === */
+typedef struct ZSTDMT_CCtx_s ZSTDMT_CCtx;
+/* Requires ZSTD_MULTITHREAD to be defined during compilation, otherwise it will return NULL. */
+ZSTDMT_API ZSTDMT_CCtx* ZSTDMT_createCCtx(unsigned nbWorkers);
+/* Requires ZSTD_MULTITHREAD to be defined during compilation, otherwise it will return NULL. */
+ZSTDMT_API ZSTDMT_CCtx* ZSTDMT_createCCtx_advanced(unsigned nbWorkers,
+ ZSTD_customMem cMem);
+ZSTDMT_API size_t ZSTDMT_freeCCtx(ZSTDMT_CCtx* mtctx);
+
+ZSTDMT_API size_t ZSTDMT_sizeof_CCtx(ZSTDMT_CCtx* mtctx);
+
+
+/* === Simple one-pass compression function === */
+
+ZSTDMT_API size_t ZSTDMT_compressCCtx(ZSTDMT_CCtx* mtctx,
+ void* dst, size_t dstCapacity,
+ const void* src, size_t srcSize,
+ int compressionLevel);
+
+
+
+/* === Streaming functions === */
+
+ZSTDMT_API size_t ZSTDMT_initCStream(ZSTDMT_CCtx* mtctx, int compressionLevel);
+ZSTDMT_API size_t ZSTDMT_resetCStream(ZSTDMT_CCtx* mtctx, unsigned long long pledgedSrcSize); /**< if srcSize is not known at reset time, use ZSTD_CONTENTSIZE_UNKNOWN. Note: for compatibility with older programs, 0 means the same as ZSTD_CONTENTSIZE_UNKNOWN, but it will change in the future to mean "empty" */
+
+ZSTDMT_API size_t ZSTDMT_nextInputSizeHint(const ZSTDMT_CCtx* mtctx);
+ZSTDMT_API size_t ZSTDMT_compressStream(ZSTDMT_CCtx* mtctx, ZSTD_outBuffer* output, ZSTD_inBuffer* input);
+
+ZSTDMT_API size_t ZSTDMT_flushStream(ZSTDMT_CCtx* mtctx, ZSTD_outBuffer* output); /**< @return : 0 == all flushed; >0 : still some data to be flushed; or an error code (ZSTD_isError()) */
+ZSTDMT_API size_t ZSTDMT_endStream(ZSTDMT_CCtx* mtctx, ZSTD_outBuffer* output); /**< @return : 0 == all flushed; >0 : still some data to be flushed; or an error code (ZSTD_isError()) */
+
+
+/* === Advanced functions and parameters === */
+
+ZSTDMT_API size_t ZSTDMT_compress_advanced(ZSTDMT_CCtx* mtctx,
+ void* dst, size_t dstCapacity,
+ const void* src, size_t srcSize,
+ const ZSTD_CDict* cdict,
+ ZSTD_parameters params,
+ int overlapLog);
+
+ZSTDMT_API size_t ZSTDMT_initCStream_advanced(ZSTDMT_CCtx* mtctx,
+ const void* dict, size_t dictSize, /* dict can be released after init, a local copy is preserved within zcs */
+ ZSTD_parameters params,
+ unsigned long long pledgedSrcSize); /* pledgedSrcSize is optional and can be zero == unknown */
+
+ZSTDMT_API size_t ZSTDMT_initCStream_usingCDict(ZSTDMT_CCtx* mtctx,
+ const ZSTD_CDict* cdict,
+ ZSTD_frameParameters fparams,
+ unsigned long long pledgedSrcSize); /* note : zero means empty */
+
+/* ZSTDMT_parameter :
+ * List of parameters that can be set using ZSTDMT_setMTCtxParameter() */
+typedef enum {
+ ZSTDMT_p_jobSize, /* Each job is compressed in parallel. By default, this value is dynamically determined depending on compression parameters. Can be set explicitly here. */
+ ZSTDMT_p_overlapLog, /* Each job may reload a part of previous job to enhance compression ratio; 0 == no overlap, 6(default) == use 1/8th of window, >=9 == use full window. This is a "sticky" parameter : its value will be re-used on next compression job */
+ ZSTDMT_p_rsyncable /* Enables rsyncable mode. */
+} ZSTDMT_parameter;
+
+/* ZSTDMT_setMTCtxParameter() :
+ * allow setting individual parameters, one at a time, among a list of enums defined in ZSTDMT_parameter.
+ * The function must be called typically after ZSTD_createCCtx() but __before ZSTDMT_init*() !__
+ * Parameters not explicitly reset by ZSTDMT_init*() remain the same in consecutive compression sessions.
+ * @return : 0, or an error code (which can be tested using ZSTD_isError()) */
+ZSTDMT_API size_t ZSTDMT_setMTCtxParameter(ZSTDMT_CCtx* mtctx, ZSTDMT_parameter parameter, int value);
+
+/* ZSTDMT_getMTCtxParameter() :
+ * Query the ZSTDMT_CCtx for a parameter value.
+ * @return : 0, or an error code (which can be tested using ZSTD_isError()) */
+ZSTDMT_API size_t ZSTDMT_getMTCtxParameter(ZSTDMT_CCtx* mtctx, ZSTDMT_parameter parameter, int* value);
+
+
+/*! ZSTDMT_compressStream_generic() :
+ * Combines ZSTDMT_compressStream() with optional ZSTDMT_flushStream() or ZSTDMT_endStream()
+ * depending on flush directive.
+ * @return : minimum amount of data still to be flushed
+ * 0 if fully flushed
+ * or an error code
+ * note : needs to be init using any ZSTD_initCStream*() variant */
+ZSTDMT_API size_t ZSTDMT_compressStream_generic(ZSTDMT_CCtx* mtctx,
+ ZSTD_outBuffer* output,
+ ZSTD_inBuffer* input,
+ ZSTD_EndDirective endOp);
+
+
+/* ========================================================
+ * === Private interface, for use by ZSTD_compress.c ===
+ * === Not exposed in libzstd. Never invoke directly ===
+ * ======================================================== */
+
+ /*! ZSTDMT_toFlushNow()
+ * Tell how many bytes are ready to be flushed immediately.
+ * Probe the oldest active job (not yet entirely flushed) and check its output buffer.
+ * If return 0, it means there is no active job,
+ * or, it means oldest job is still active, but everything produced has been flushed so far,
+ * therefore flushing is limited by speed of oldest job. */
+size_t ZSTDMT_toFlushNow(ZSTDMT_CCtx* mtctx);
+
+/*! ZSTDMT_CCtxParam_setMTCtxParameter()
+ * like ZSTDMT_setMTCtxParameter(), but into a ZSTD_CCtx_Params */
+size_t ZSTDMT_CCtxParam_setMTCtxParameter(ZSTD_CCtx_params* params, ZSTDMT_parameter parameter, int value);
+
+/*! ZSTDMT_CCtxParam_setNbWorkers()
+ * Set nbWorkers, and clamp it.
+ * Also reset jobSize and overlapLog */
+size_t ZSTDMT_CCtxParam_setNbWorkers(ZSTD_CCtx_params* params, unsigned nbWorkers);
+
+/*! ZSTDMT_updateCParams_whileCompressing() :
+ * Updates only a selected set of compression parameters, to remain compatible with current frame.
+ * New parameters will be applied to next compression job. */
+void ZSTDMT_updateCParams_whileCompressing(ZSTDMT_CCtx* mtctx, const ZSTD_CCtx_params* cctxParams);
+
+/*! ZSTDMT_getFrameProgression():
+ * tells how much data has been consumed (input) and produced (output) for current frame.
+ * able to count progression inside worker threads.
+ */
+ZSTD_frameProgression ZSTDMT_getFrameProgression(ZSTDMT_CCtx* mtctx);
+
+
+/*! ZSTDMT_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 ZSTDMT_initCStream_internal(ZSTDMT_CCtx* zcs,
+ const void* dict, size_t dictSize, ZSTD_dictContentType_e dictContentType,
+ const ZSTD_CDict* cdict,
+ ZSTD_CCtx_params params, unsigned long long pledgedSrcSize);
+
+
+#if defined (__cplusplus)
+}
+#endif
+
+#endif /* ZSTDMT_COMPRESS_H */