summaryrefslogtreecommitdiffstats
path: root/src/liblzma/rangecoder/range_decoder.h
diff options
context:
space:
mode:
Diffstat (limited to 'src/liblzma/rangecoder/range_decoder.h')
-rw-r--r--src/liblzma/rangecoder/range_decoder.h839
1 files changed, 810 insertions, 29 deletions
diff --git a/src/liblzma/rangecoder/range_decoder.h b/src/liblzma/rangecoder/range_decoder.h
index e0b051f..31a58d1 100644
--- a/src/liblzma/rangecoder/range_decoder.h
+++ b/src/liblzma/rangecoder/range_decoder.h
@@ -1,3 +1,5 @@
+// SPDX-License-Identifier: 0BSD
+
///////////////////////////////////////////////////////////////////////////////
//
/// \file range_decoder.h
@@ -6,9 +8,6 @@
// Authors: Igor Pavlov
// Lasse Collin
//
-// This file has been put into the public domain.
-// You can do whatever you want with this file.
-//
///////////////////////////////////////////////////////////////////////////////
#ifndef LZMA_RANGE_DECODER_H
@@ -17,6 +16,55 @@
#include "range_common.h"
+// Choose the range decoder variants to use using a bitmask.
+// If no bits are set, only the basic version is used.
+// If more than one version is selected for the same feature,
+// the last one on the list below is used.
+//
+// Bitwise-or of the following enable branchless C versions:
+// 0x01 normal bittrees
+// 0x02 fixed-sized reverse bittrees
+// 0x04 variable-sized reverse bittrees (not faster)
+// 0x08 matched literal (not faster)
+//
+// GCC & Clang compatible x86-64 inline assembly:
+// 0x010 normal bittrees
+// 0x020 fixed-sized reverse bittrees
+// 0x040 variable-sized reverse bittrees
+// 0x080 matched literal
+// 0x100 direct bits
+//
+// The default can be overridden at build time by defining
+// LZMA_RANGE_DECODER_CONFIG to the desired mask.
+//
+// 2024-02-22: Feedback from benchmarks:
+// - Brancless C (0x003) can be better than basic on x86-64 but often it's
+// slightly worse on other archs. Since asm is much better on x86-64,
+// branchless C is not used at all.
+// - With x86-64 asm, there are slight differences between GCC and Clang
+// and different processors. Overall 0x1F0 seems to be the best choice.
+#ifndef LZMA_RANGE_DECODER_CONFIG
+# if defined(__x86_64__) && !defined(__ILP32__) \
+ && !defined(__NVCOMPILER) \
+ && (defined(__GNUC__) || defined(__clang__))
+# define LZMA_RANGE_DECODER_CONFIG 0x1F0
+# else
+# define LZMA_RANGE_DECODER_CONFIG 0
+# endif
+#endif
+
+
+// Negative RC_BIT_MODEL_TOTAL but the lowest RC_MOVE_BITS are flipped.
+// This is useful for updating probability variables in branchless decoding:
+//
+// uint32_t decoded_bit = ...;
+// probability tmp = RC_BIT_MODEL_OFFSET;
+// tmp &= decoded_bit - 1;
+// prob -= (prob + tmp) >> RC_MOVE_BITS;
+#define RC_BIT_MODEL_OFFSET \
+ ((UINT32_C(1) << RC_MOVE_BITS) - 1 - RC_BIT_MODEL_TOTAL)
+
+
typedef struct {
uint32_t range;
uint32_t code;
@@ -50,18 +98,28 @@ rc_read_init(lzma_range_decoder *rc, const uint8_t *restrict in,
/// Makes local copies of range decoder and *in_pos variables. Doing this
/// improves speed significantly. The range decoder macros expect also
-/// variables `in' and `in_size' to be defined.
-#define rc_to_local(range_decoder, in_pos) \
+/// variables 'in' and 'in_size' to be defined.
+#define rc_to_local(range_decoder, in_pos, fast_mode_in_required) \
lzma_range_decoder rc = range_decoder; \
- size_t rc_in_pos = (in_pos); \
+ const uint8_t *rc_in_ptr = in + (in_pos); \
+ const uint8_t *rc_in_end = in + in_size; \
+ const uint8_t *rc_in_fast_end \
+ = (rc_in_end - rc_in_ptr) <= (fast_mode_in_required) \
+ ? rc_in_ptr \
+ : rc_in_end - (fast_mode_in_required); \
+ (void)rc_in_fast_end; /* Silence a warning with HAVE_SMALL. */ \
uint32_t rc_bound
+/// Evaluates to true if there is enough input remaining to use fast mode.
+#define rc_is_fast_allowed() (rc_in_ptr < rc_in_fast_end)
+
+
/// Stores the local copes back to the range decoder structure.
#define rc_from_local(range_decoder, in_pos) \
do { \
range_decoder = rc; \
- in_pos = rc_in_pos; \
+ in_pos = (size_t)(rc_in_ptr - in); \
} while (0)
@@ -81,18 +139,30 @@ do { \
((range_decoder).code == 0)
-/// Read the next input byte if needed. If more input is needed but there is
+// Read the next input byte if needed.
+#define rc_normalize() \
+do { \
+ if (rc.range < RC_TOP_VALUE) { \
+ rc.range <<= RC_SHIFT_BITS; \
+ rc.code = (rc.code << RC_SHIFT_BITS) | *rc_in_ptr++; \
+ } \
+} while (0)
+
+
+/// If more input is needed but there is
/// no more input available, "goto out" is used to jump out of the main
-/// decoder loop.
-#define rc_normalize(seq) \
+/// decoder loop. The "_safe" macros are used in the Resumable decoder
+/// mode in order to save the sequence to continue decoding from that
+/// point later.
+#define rc_normalize_safe(seq) \
do { \
if (rc.range < RC_TOP_VALUE) { \
- if (unlikely(rc_in_pos == in_size)) { \
+ if (rc_in_ptr == rc_in_end) { \
coder->sequence = seq; \
goto out; \
} \
rc.range <<= RC_SHIFT_BITS; \
- rc.code = (rc.code << RC_SHIFT_BITS) | in[rc_in_pos++]; \
+ rc.code = (rc.code << RC_SHIFT_BITS) | *rc_in_ptr++; \
} \
} while (0)
@@ -100,7 +170,7 @@ do { \
/// Start decoding a bit. This must be used together with rc_update_0()
/// and rc_update_1():
///
-/// rc_if_0(prob, seq) {
+/// rc_if_0(prob) {
/// rc_update_0(prob);
/// // Do something
/// } else {
@@ -108,18 +178,28 @@ do { \
/// // Do something else
/// }
///
-#define rc_if_0(prob, seq) \
- rc_normalize(seq); \
+#define rc_if_0(prob) \
+ rc_normalize(); \
+ rc_bound = (rc.range >> RC_BIT_MODEL_TOTAL_BITS) * (prob); \
+ if (rc.code < rc_bound)
+
+
+#define rc_if_0_safe(prob, seq) \
+ rc_normalize_safe(seq); \
rc_bound = (rc.range >> RC_BIT_MODEL_TOTAL_BITS) * (prob); \
if (rc.code < rc_bound)
/// Update the range decoder state and the used probability variable to
/// match a decoded bit of 0.
+///
+/// The x86-64 assembly uses the commented method but it seems that,
+/// at least on x86-64, the first version is slightly faster as C code.
#define rc_update_0(prob) \
do { \
rc.range = rc_bound; \
prob += (RC_BIT_MODEL_TOTAL - (prob)) >> RC_MOVE_BITS; \
+ /* prob -= ((prob) + RC_BIT_MODEL_OFFSET) >> RC_MOVE_BITS; */ \
} while (0)
@@ -137,9 +217,21 @@ do { \
/// This macro is used as the last step in bittree reverse decoders since
/// those don't use "symbol" for anything else than indexing the probability
/// arrays.
-#define rc_bit_last(prob, action0, action1, seq) \
+#define rc_bit_last(prob, action0, action1) \
+do { \
+ rc_if_0(prob) { \
+ rc_update_0(prob); \
+ action0; \
+ } else { \
+ rc_update_1(prob); \
+ action1; \
+ } \
+} while (0)
+
+
+#define rc_bit_last_safe(prob, action0, action1, seq) \
do { \
- rc_if_0(prob, seq) { \
+ rc_if_0_safe(prob, seq) { \
rc_update_0(prob); \
action0; \
} else { \
@@ -151,35 +243,724 @@ do { \
/// Decodes one bit, updates "symbol", and runs action0 or action1 depending
/// on the decoded bit.
-#define rc_bit(prob, action0, action1, seq) \
+#define rc_bit(prob, action0, action1) \
rc_bit_last(prob, \
symbol <<= 1; action0, \
+ symbol = (symbol << 1) + 1; action1);
+
+
+#define rc_bit_safe(prob, action0, action1, seq) \
+ rc_bit_last_safe(prob, \
+ symbol <<= 1; action0, \
symbol = (symbol << 1) + 1; action1, \
seq);
+// Unroll fixed-sized bittree decoding.
+//
+// A compile-time constant in final_add can be used to get rid of the high bit
+// from symbol that is used for the array indexing (1U << bittree_bits).
+// final_add may also be used to add offset to the result (LZMA length
+// decoder does that).
+//
+// The reason to have final_add here is that in the asm code the addition
+// can be done for free: in x86-64 there is SBB instruction with -1 as
+// the immediate value, and final_add is combined with that value.
+#define rc_bittree_bit(prob) \
+ rc_bit(prob, , )
+
+#define rc_bittree3(probs, final_add) \
+do { \
+ symbol = 1; \
+ rc_bittree_bit(probs[symbol]); \
+ rc_bittree_bit(probs[symbol]); \
+ rc_bittree_bit(probs[symbol]); \
+ symbol += (uint32_t)(final_add); \
+} while (0)
-/// Like rc_bit() but add "case seq:" as a prefix. This makes the unrolled
-/// loops more readable because the code isn't littered with "case"
-/// statements. On the other hand this also makes it less readable, since
-/// spotting the places where the decoder loop may be restarted is less
-/// obvious.
-#define rc_bit_case(prob, action0, action1, seq) \
- case seq: rc_bit(prob, action0, action1, seq)
+#define rc_bittree6(probs, final_add) \
+do { \
+ symbol = 1; \
+ rc_bittree_bit(probs[symbol]); \
+ rc_bittree_bit(probs[symbol]); \
+ rc_bittree_bit(probs[symbol]); \
+ rc_bittree_bit(probs[symbol]); \
+ rc_bittree_bit(probs[symbol]); \
+ rc_bittree_bit(probs[symbol]); \
+ symbol += (uint32_t)(final_add); \
+} while (0)
+
+#define rc_bittree8(probs, final_add) \
+do { \
+ symbol = 1; \
+ rc_bittree_bit(probs[symbol]); \
+ rc_bittree_bit(probs[symbol]); \
+ rc_bittree_bit(probs[symbol]); \
+ rc_bittree_bit(probs[symbol]); \
+ rc_bittree_bit(probs[symbol]); \
+ rc_bittree_bit(probs[symbol]); \
+ rc_bittree_bit(probs[symbol]); \
+ rc_bittree_bit(probs[symbol]); \
+ symbol += (uint32_t)(final_add); \
+} while (0)
+
+
+// Fixed-sized reverse bittree
+#define rc_bittree_rev4(probs) \
+do { \
+ symbol = 0; \
+ rc_bit_last(probs[symbol + 1], , symbol += 1); \
+ rc_bit_last(probs[symbol + 2], , symbol += 2); \
+ rc_bit_last(probs[symbol + 4], , symbol += 4); \
+ rc_bit_last(probs[symbol + 8], , symbol += 8); \
+} while (0)
+
+
+// Decode one bit from variable-sized reverse bittree. The loop is done
+// in the code that uses this macro. This could be changed if the assembly
+// version benefited from having the loop done in assembly but it didn't
+// seem so in early 2024.
+//
+// Also, if the loop was done here, the loop counter would likely be local
+// to the macro so that it wouldn't modify yet another input variable.
+// If a _safe version of a macro with a loop was done then a modifiable
+// input variable couldn't be avoided though.
+#define rc_bit_add_if_1(probs, dest, value_to_add_if_1) \
+ rc_bit(probs[symbol], \
+ , \
+ dest += value_to_add_if_1);
+
+
+// Matched literal
+#define decode_with_match_bit \
+ t_match_byte <<= 1; \
+ t_match_bit = t_match_byte & t_offset; \
+ t_subcoder_index = t_offset + t_match_bit + symbol; \
+ rc_bit(probs[t_subcoder_index], \
+ t_offset &= ~t_match_bit, \
+ t_offset &= t_match_bit)
+
+#define rc_matched_literal(probs_base_var, match_byte) \
+do { \
+ uint32_t t_match_byte = (match_byte); \
+ uint32_t t_match_bit; \
+ uint32_t t_subcoder_index; \
+ uint32_t t_offset = 0x100; \
+ symbol = 1; \
+ decode_with_match_bit; \
+ decode_with_match_bit; \
+ decode_with_match_bit; \
+ decode_with_match_bit; \
+ decode_with_match_bit; \
+ decode_with_match_bit; \
+ decode_with_match_bit; \
+ decode_with_match_bit; \
+} while (0)
/// Decode a bit without using a probability.
-#define rc_direct(dest, seq) \
+//
+// NOTE: GCC 13 and Clang/LLVM 16 can, at least on x86-64, optimize the bound
+// calculation to use an arithmetic right shift so there's no need to provide
+// the alternative code which, according to C99/C11/C23 6.3.1.3-p3 isn't
+// perfectly portable: rc_bound = (uint32_t)((int32_t)rc.code >> 31);
+#define rc_direct(dest, count_var) \
do { \
- rc_normalize(seq); \
+ dest = (dest << 1) + 1; \
+ rc_normalize(); \
+ rc.range >>= 1; \
+ rc.code -= rc.range; \
+ rc_bound = UINT32_C(0) - (rc.code >> 31); \
+ dest += rc_bound; \
+ rc.code += rc.range & rc_bound; \
+} while (--count_var > 0)
+
+
+
+#define rc_direct_safe(dest, count_var, seq) \
+do { \
+ rc_normalize_safe(seq); \
rc.range >>= 1; \
rc.code -= rc.range; \
rc_bound = UINT32_C(0) - (rc.code >> 31); \
rc.code += rc.range & rc_bound; \
dest = (dest << 1) + (rc_bound + 1); \
+} while (--count_var > 0)
+
+
+//////////////////
+// Branchless C //
+//////////////////
+
+/// Decode a bit using a branchless method. This reduces the number of
+/// mispredicted branches and thus can improve speed.
+#define rc_c_bit(prob, action_bit, action_neg) \
+do { \
+ probability *p = &(prob); \
+ rc_normalize(); \
+ rc_bound = (rc.range >> RC_BIT_MODEL_TOTAL_BITS) * *p; \
+ uint32_t rc_mask = rc.code >= rc_bound; /* rc_mask = decoded bit */ \
+ action_bit; /* action when rc_mask is 0 or 1 */ \
+ /* rc_mask becomes 0 if bit is 0 and 0xFFFFFFFF if bit is 1: */ \
+ rc_mask = 0U - rc_mask; \
+ rc.range &= rc_mask; /* If bit 0: set rc.range = 0 */ \
+ rc_bound ^= rc_mask; \
+ rc_bound -= rc_mask; /* If bit 1: rc_bound = 0U - rc_bound */ \
+ rc.range += rc_bound; \
+ rc_bound &= rc_mask; \
+ rc.code += rc_bound; \
+ action_neg; /* action when rc_mask is 0 or 0xFFFFFFFF */ \
+ rc_mask = ~rc_mask; /* If bit 0: all bits are set in rc_mask */ \
+ rc_mask &= RC_BIT_MODEL_OFFSET; \
+ *p -= (*p + rc_mask) >> RC_MOVE_BITS; \
} while (0)
-// NOTE: No macros are provided for bittree decoding. It seems to be simpler
-// to just write them open in the code.
+// Testing on x86-64 give an impression that only the normal bittrees and
+// the fixed-sized reverse bittrees are worth the branchless C code.
+// It should be tested on other archs for which there isn't assembly code
+// in this file.
+
+// Using addition in "(symbol << 1) + rc_mask" allows use of x86 LEA
+// or RISC-V SH1ADD instructions. Compilers might infer it from
+// "(symbol << 1) | rc_mask" too if they see that mask is 0 or 1 but
+// the use of addition doesn't require such analysis from compilers.
+#if LZMA_RANGE_DECODER_CONFIG & 0x01
+#undef rc_bittree_bit
+#define rc_bittree_bit(prob) \
+ rc_c_bit(prob, \
+ symbol = (symbol << 1) + rc_mask, \
+ )
+#endif // LZMA_RANGE_DECODER_CONFIG & 0x01
+
+#if LZMA_RANGE_DECODER_CONFIG & 0x02
+#undef rc_bittree_rev4
+#define rc_bittree_rev4(probs) \
+do { \
+ symbol = 0; \
+ rc_c_bit(probs[symbol + 1], symbol += rc_mask, ); \
+ rc_c_bit(probs[symbol + 2], symbol += rc_mask << 1, ); \
+ rc_c_bit(probs[symbol + 4], symbol += rc_mask << 2, ); \
+ rc_c_bit(probs[symbol + 8], symbol += rc_mask << 3, ); \
+} while (0)
+#endif // LZMA_RANGE_DECODER_CONFIG & 0x02
+
+#if LZMA_RANGE_DECODER_CONFIG & 0x04
+#undef rc_bit_add_if_1
+#define rc_bit_add_if_1(probs, dest, value_to_add_if_1) \
+ rc_c_bit(probs[symbol], \
+ symbol = (symbol << 1) + rc_mask, \
+ dest += (value_to_add_if_1) & rc_mask)
+#endif // LZMA_RANGE_DECODER_CONFIG & 0x04
+
+
+#if LZMA_RANGE_DECODER_CONFIG & 0x08
+#undef decode_with_match_bit
+#define decode_with_match_bit \
+ t_match_byte <<= 1; \
+ t_match_bit = t_match_byte & t_offset; \
+ t_subcoder_index = t_offset + t_match_bit + symbol; \
+ rc_c_bit(probs[t_subcoder_index], \
+ symbol = (symbol << 1) + rc_mask, \
+ t_offset &= ~t_match_bit ^ rc_mask)
+#endif // LZMA_RANGE_DECODER_CONFIG & 0x08
+
+
+////////////
+// x86-64 //
+////////////
+
+#if LZMA_RANGE_DECODER_CONFIG & 0x1F0
+
+// rc_asm_y and rc_asm_n are used as arguments to macros to control which
+// strings to include or omit.
+#define rc_asm_y(str) str
+#define rc_asm_n(str)
+
+// There are a few possible variations for normalization.
+// This is the smallest variant which is also used by LZMA SDK.
+//
+// - This has partial register write (the MOV from (%[in_ptr])).
+//
+// - INC saves one byte in code size over ADD. False dependency on
+// partial flags from INC shouldn't become a problem on any processor
+// because the instructions after normalization don't read the flags
+// until SUB which sets all flags.
+//
+#define rc_asm_normalize \
+ "cmp %[top_value], %[range]\n\t" \
+ "jae 1f\n\t" \
+ "shl %[shift_bits], %[code]\n\t" \
+ "mov (%[in_ptr]), %b[code]\n\t" \
+ "shl %[shift_bits], %[range]\n\t" \
+ "inc %[in_ptr]\n" \
+ "1:\n"
+
+// rc_asm_calc(prob) is roughly equivalent to the C version of rc_if_0(prob)...
+//
+// rc_bound = (rc.range >> RC_BIT_MODEL_TOTAL_BITS) * (prob);
+// if (rc.code < rc_bound)
+//
+// ...but the bound is stored in "range":
+//
+// t0 = range;
+// range = (range >> RC_BIT_MODEL_TOTAL_BITS) * (prob);
+// t0 -= range;
+// t1 = code;
+// code -= range;
+//
+// The carry flag (CF) from the last subtraction holds the negation of
+// the decoded bit (if CF==0 then the decoded bit is 1).
+// The values in t0 and t1 are needed for rc_update_0(prob) and
+// rc_update_1(prob). If the bit is 0, rc_update_0(prob)...
+//
+// rc.range = rc_bound;
+//
+// ...has already been done but the "code -= range" has to be reverted using
+// the old value stored in t1. (Also, prob needs to be updated.)
+//
+// If the bit is 1, rc_update_1(prob)...
+//
+// rc.range -= rc_bound;
+// rc.code -= rc_bound;
+//
+// ...is already done for "code" but the value for "range" needs to be taken
+// from t0. (Also, prob needs to be updated here as well.)
+//
+// The assignments from t0 and t1 can be done in a branchless manner with CMOV
+// after the instructions from this macro. The CF from SUB tells which moves
+// are needed.
+#define rc_asm_calc(prob) \
+ "mov %[range], %[t0]\n\t" \
+ "shr %[bit_model_total_bits], %[range]\n\t" \
+ "imul %[" prob "], %[range]\n\t" \
+ "sub %[range], %[t0]\n\t" \
+ "mov %[code], %[t1]\n\t" \
+ "sub %[range], %[code]\n\t"
+
+// Also, prob needs to be updated: The update math depends on the decoded bit.
+// It can be expressed in a few slightly different ways but this is fairly
+// convenient here:
+//
+// prob -= (prob + (bit ? 0 : RC_BIT_MODEL_OFFSET)) >> RC_MOVE_BITS;
+//
+// To do it in branchless way when the negation of the decoded bit is in CF,
+// both "prob" and "prob + RC_BIT_MODEL_OFFSET" are needed. Then the desired
+// value can be picked with CMOV. The addition can be done using LEA without
+// affecting CF.
+//
+// (This prob update method is a tiny bit different from LZMA SDK 23.01.
+// In the LZMA SDK a single register is reserved solely for a constant to
+// be used with CMOV when updating prob. That is fine since there are enough
+// free registers to do so. The method used here uses one fewer register,
+// which is valuable with inline assembly.)
+//
+// * * *
+//
+// In bittree decoding, each (unrolled) loop iteration decodes one bit
+// and needs one prob variable. To make it faster, the prob variable of
+// the iteration N+1 is loaded during iteration N. There are two possible
+// prob variables to choose from for N+1. Both are loaded from memory and
+// the correct one is chosen with CMOV using the same CF as is used for
+// other things described above.
+//
+// This preloading/prefetching requires an extra register. To avoid
+// useless moves from "preloaded prob register" to "current prob register",
+// the macros swap between the two registers for odd and even iterations.
+//
+// * * *
+//
+// Finally, the decoded bit has to be stored in "symbol". Since the negation
+// of the bit is in CF, this can be done with SBB: symbol -= CF - 1. That is,
+// if the decoded bit is 0 (CF==1) the operation is a no-op "symbol -= 0"
+// and when bit is 1 (CF==0) the operation is "symbol -= 0 - 1" which is
+// the same as "symbol += 1".
+//
+// The instructions for all things are intertwined for a few reasons:
+// - freeing temporary registers for new use
+// - not modifying CF too early
+// - instruction scheduling
+//
+// The first and last iterations can cheat a little. For example,
+// on the first iteration "symbol" is known to start from 1 so it
+// doesn't need to be read; it can even be immediately initialized
+// to 2 to prepare for the second iteration of the loop.
+//
+// * * *
+//
+// a = number of the current prob variable (0 or 1)
+// b = number of the next prob variable (1 or 0)
+// *_only = rc_asm_y or _n to include or exclude code marked with them
+#define rc_asm_bittree(a, b, first_only, middle_only, last_only) \
+ first_only( \
+ "movzw 2(%[probs_base]), %[prob" #a "]\n\t" \
+ "mov $2, %[symbol]\n\t" \
+ "movzw 4(%[probs_base]), %[prob" #b "]\n\t" \
+ ) \
+ middle_only( \
+ /* Note the scaling of 4 instead of 2: */ \
+ "movzw (%[probs_base], %q[symbol], 4), %[prob" #b "]\n\t" \
+ ) \
+ last_only( \
+ "add %[symbol], %[symbol]\n\t" \
+ ) \
+ \
+ rc_asm_normalize \
+ rc_asm_calc("prob" #a) \
+ \
+ "cmovae %[t0], %[range]\n\t" \
+ \
+ first_only( \
+ "movzw 6(%[probs_base]), %[t0]\n\t" \
+ "cmovae %[t0], %[prob" #b "]\n\t" \
+ ) \
+ middle_only( \
+ "movzw 2(%[probs_base], %q[symbol], 4), %[t0]\n\t" \
+ "lea (%q[symbol], %q[symbol]), %[symbol]\n\t" \
+ "cmovae %[t0], %[prob" #b "]\n\t" \
+ ) \
+ \
+ "lea %c[bit_model_offset](%q[prob" #a "]), %[t0]\n\t" \
+ "cmovb %[t1], %[code]\n\t" \
+ "mov %[symbol], %[t1]\n\t" \
+ "cmovae %[prob" #a "], %[t0]\n\t" \
+ \
+ first_only( \
+ "sbb $-1, %[symbol]\n\t" \
+ ) \
+ middle_only( \
+ "sbb $-1, %[symbol]\n\t" \
+ ) \
+ last_only( \
+ "sbb %[last_sbb], %[symbol]\n\t" \
+ ) \
+ \
+ "shr %[move_bits], %[t0]\n\t" \
+ "sub %[t0], %[prob" #a "]\n\t" \
+ /* Scaling of 1 instead of 2 because symbol <<= 1. */ \
+ "mov %w[prob" #a "], (%[probs_base], %q[t1], 1)\n\t"
+
+// NOTE: The order of variables in __asm__ can affect speed and code size.
+#define rc_asm_bittree_n(probs_base_var, final_add, asm_str) \
+do { \
+ uint32_t t0; \
+ uint32_t t1; \
+ uint32_t t_prob0; \
+ uint32_t t_prob1; \
+ \
+ __asm__( \
+ asm_str \
+ : \
+ [range] "+&r"(rc.range), \
+ [code] "+&r"(rc.code), \
+ [t0] "=&r"(t0), \
+ [t1] "=&r"(t1), \
+ [prob0] "=&r"(t_prob0), \
+ [prob1] "=&r"(t_prob1), \
+ [symbol] "=&r"(symbol), \
+ [in_ptr] "+&r"(rc_in_ptr) \
+ : \
+ [probs_base] "r"(probs_base_var), \
+ [last_sbb] "n"(-1 - (final_add)), \
+ [top_value] "n"(RC_TOP_VALUE), \
+ [shift_bits] "n"(RC_SHIFT_BITS), \
+ [bit_model_total_bits] "n"(RC_BIT_MODEL_TOTAL_BITS), \
+ [bit_model_offset] "n"(RC_BIT_MODEL_OFFSET), \
+ [move_bits] "n"(RC_MOVE_BITS) \
+ : \
+ "cc", "memory"); \
+} while (0)
+
+
+#if LZMA_RANGE_DECODER_CONFIG & 0x010
+#undef rc_bittree3
+#define rc_bittree3(probs_base_var, final_add) \
+ rc_asm_bittree_n(probs_base_var, final_add, \
+ rc_asm_bittree(0, 1, rc_asm_y, rc_asm_n, rc_asm_n) \
+ rc_asm_bittree(1, 0, rc_asm_n, rc_asm_y, rc_asm_n) \
+ rc_asm_bittree(0, 1, rc_asm_n, rc_asm_n, rc_asm_y) \
+ )
+
+#undef rc_bittree6
+#define rc_bittree6(probs_base_var, final_add) \
+ rc_asm_bittree_n(probs_base_var, final_add, \
+ rc_asm_bittree(0, 1, rc_asm_y, rc_asm_n, rc_asm_n) \
+ rc_asm_bittree(1, 0, rc_asm_n, rc_asm_y, rc_asm_n) \
+ rc_asm_bittree(0, 1, rc_asm_n, rc_asm_y, rc_asm_n) \
+ rc_asm_bittree(1, 0, rc_asm_n, rc_asm_y, rc_asm_n) \
+ rc_asm_bittree(0, 1, rc_asm_n, rc_asm_y, rc_asm_n) \
+ rc_asm_bittree(1, 0, rc_asm_n, rc_asm_n, rc_asm_y) \
+ )
+
+#undef rc_bittree8
+#define rc_bittree8(probs_base_var, final_add) \
+ rc_asm_bittree_n(probs_base_var, final_add, \
+ rc_asm_bittree(0, 1, rc_asm_y, rc_asm_n, rc_asm_n) \
+ rc_asm_bittree(1, 0, rc_asm_n, rc_asm_y, rc_asm_n) \
+ rc_asm_bittree(0, 1, rc_asm_n, rc_asm_y, rc_asm_n) \
+ rc_asm_bittree(1, 0, rc_asm_n, rc_asm_y, rc_asm_n) \
+ rc_asm_bittree(0, 1, rc_asm_n, rc_asm_y, rc_asm_n) \
+ rc_asm_bittree(1, 0, rc_asm_n, rc_asm_y, rc_asm_n) \
+ rc_asm_bittree(0, 1, rc_asm_n, rc_asm_y, rc_asm_n) \
+ rc_asm_bittree(1, 0, rc_asm_n, rc_asm_n, rc_asm_y) \
+ )
+#endif // LZMA_RANGE_DECODER_CONFIG & 0x010
+
+
+// Fixed-sized reverse bittree
+//
+// This uses the indexing that constructs the final value in symbol directly.
+// add = 1, 2, 4, 8
+// dcur = -, 4, 8, 16
+// dnext0 = 4, 8, 16, -
+// dnext0 = 6, 12, 24, -
+#define rc_asm_bittree_rev(a, b, add, dcur, dnext0, dnext1, \
+ first_only, middle_only, last_only) \
+ first_only( \
+ "movzw 2(%[probs_base]), %[prob" #a "]\n\t" \
+ "xor %[symbol], %[symbol]\n\t" \
+ "movzw 4(%[probs_base]), %[prob" #b "]\n\t" \
+ ) \
+ middle_only( \
+ "movzw " #dnext0 "(%[probs_base], %q[symbol], 2), " \
+ "%[prob" #b "]\n\t" \
+ ) \
+ \
+ rc_asm_normalize \
+ rc_asm_calc("prob" #a) \
+ \
+ "cmovae %[t0], %[range]\n\t" \
+ \
+ first_only( \
+ "movzw 6(%[probs_base]), %[t0]\n\t" \
+ "cmovae %[t0], %[prob" #b "]\n\t" \
+ ) \
+ middle_only( \
+ "movzw " #dnext1 "(%[probs_base], %q[symbol], 2), %[t0]\n\t" \
+ "cmovae %[t0], %[prob" #b "]\n\t" \
+ ) \
+ \
+ "lea " #add "(%q[symbol]), %[t0]\n\t" \
+ "cmovb %[t1], %[code]\n\t" \
+ middle_only( \
+ "mov %[symbol], %[t1]\n\t" \
+ ) \
+ last_only( \
+ "mov %[symbol], %[t1]\n\t" \
+ ) \
+ "cmovae %[t0], %[symbol]\n\t" \
+ "lea %c[bit_model_offset](%q[prob" #a "]), %[t0]\n\t" \
+ "cmovae %[prob" #a "], %[t0]\n\t" \
+ \
+ "shr %[move_bits], %[t0]\n\t" \
+ "sub %[t0], %[prob" #a "]\n\t" \
+ first_only( \
+ "mov %w[prob" #a "], 2(%[probs_base])\n\t" \
+ ) \
+ middle_only( \
+ "mov %w[prob" #a "], " \
+ #dcur "(%[probs_base], %q[t1], 2)\n\t" \
+ ) \
+ last_only( \
+ "mov %w[prob" #a "], " \
+ #dcur "(%[probs_base], %q[t1], 2)\n\t" \
+ )
+
+#if LZMA_RANGE_DECODER_CONFIG & 0x020
+#undef rc_bittree_rev4
+#define rc_bittree_rev4(probs_base_var) \
+rc_asm_bittree_n(probs_base_var, 4, \
+ rc_asm_bittree_rev(0, 1, 1, -, 4, 6, rc_asm_y, rc_asm_n, rc_asm_n) \
+ rc_asm_bittree_rev(1, 0, 2, 4, 8, 12, rc_asm_n, rc_asm_y, rc_asm_n) \
+ rc_asm_bittree_rev(0, 1, 4, 8, 16, 24, rc_asm_n, rc_asm_y, rc_asm_n) \
+ rc_asm_bittree_rev(1, 0, 8, 16, -, -, rc_asm_n, rc_asm_n, rc_asm_y) \
+)
+#endif // LZMA_RANGE_DECODER_CONFIG & 0x020
+
+
+#if LZMA_RANGE_DECODER_CONFIG & 0x040
+#undef rc_bit_add_if_1
+#define rc_bit_add_if_1(probs_base_var, dest_var, value_to_add_if_1) \
+do { \
+ uint32_t t0; \
+ uint32_t t1; \
+ uint32_t t2 = (value_to_add_if_1); \
+ uint32_t t_prob; \
+ uint32_t t_index; \
+ \
+ __asm__( \
+ "movzw (%[probs_base], %q[symbol], 2), %[prob]\n\t" \
+ "mov %[symbol], %[index]\n\t" \
+ \
+ "add %[dest], %[t2]\n\t" \
+ "add %[symbol], %[symbol]\n\t" \
+ \
+ rc_asm_normalize \
+ rc_asm_calc("prob") \
+ \
+ "cmovae %[t0], %[range]\n\t" \
+ "lea %c[bit_model_offset](%q[prob]), %[t0]\n\t" \
+ "cmovb %[t1], %[code]\n\t" \
+ "cmovae %[prob], %[t0]\n\t" \
+ \
+ "cmovae %[t2], %[dest]\n\t" \
+ "sbb $-1, %[symbol]\n\t" \
+ \
+ "sar %[move_bits], %[t0]\n\t" \
+ "sub %[t0], %[prob]\n\t" \
+ "mov %w[prob], (%[probs_base], %q[index], 2)" \
+ : \
+ [range] "+&r"(rc.range), \
+ [code] "+&r"(rc.code), \
+ [t0] "=&r"(t0), \
+ [t1] "=&r"(t1), \
+ [prob] "=&r"(t_prob), \
+ [index] "=&r"(t_index), \
+ [symbol] "+&r"(symbol), \
+ [t2] "+&r"(t2), \
+ [dest] "+&r"(dest_var), \
+ [in_ptr] "+&r"(rc_in_ptr) \
+ : \
+ [probs_base] "r"(probs_base_var), \
+ [top_value] "n"(RC_TOP_VALUE), \
+ [shift_bits] "n"(RC_SHIFT_BITS), \
+ [bit_model_total_bits] "n"(RC_BIT_MODEL_TOTAL_BITS), \
+ [bit_model_offset] "n"(RC_BIT_MODEL_OFFSET), \
+ [move_bits] "n"(RC_MOVE_BITS) \
+ : \
+ "cc", "memory"); \
+} while (0)
+#endif // LZMA_RANGE_DECODER_CONFIG & 0x040
+
+
+// Literal decoding uses a normal 8-bit bittree but literal with match byte
+// is more complex in picking the probability variable from the correct
+// subtree. This doesn't use preloading/prefetching of the next prob because
+// there are four choices instead of two.
+//
+// FIXME? The first iteration starts with symbol = 1 so it could be optimized
+// by a tiny amount.
+#define rc_asm_matched_literal(nonlast_only) \
+ "add %[offset], %[symbol]\n\t" \
+ "and %[offset], %[match_bit]\n\t" \
+ "add %[match_bit], %[symbol]\n\t" \
+ \
+ "movzw (%[probs_base], %q[symbol], 2), %[prob]\n\t" \
+ \
+ "add %[symbol], %[symbol]\n\t" \
+ \
+ nonlast_only( \
+ "xor %[match_bit], %[offset]\n\t" \
+ "add %[match_byte], %[match_byte]\n\t" \
+ ) \
+ \
+ rc_asm_normalize \
+ rc_asm_calc("prob") \
+ \
+ "cmovae %[t0], %[range]\n\t" \
+ "lea %c[bit_model_offset](%q[prob]), %[t0]\n\t" \
+ "cmovb %[t1], %[code]\n\t" \
+ "mov %[symbol], %[t1]\n\t" \
+ "cmovae %[prob], %[t0]\n\t" \
+ \
+ nonlast_only( \
+ "cmovae %[match_bit], %[offset]\n\t" \
+ "mov %[match_byte], %[match_bit]\n\t" \
+ ) \
+ \
+ "sbb $-1, %[symbol]\n\t" \
+ \
+ "shr %[move_bits], %[t0]\n\t" \
+ /* Undo symbol += match_bit + offset: */ \
+ "and $0x1FF, %[symbol]\n\t" \
+ "sub %[t0], %[prob]\n\t" \
+ \
+ /* Scaling of 1 instead of 2 because symbol <<= 1. */ \
+ "mov %w[prob], (%[probs_base], %q[t1], 1)\n\t"
+
+
+#if LZMA_RANGE_DECODER_CONFIG & 0x080
+#undef rc_matched_literal
+#define rc_matched_literal(probs_base_var, match_byte_value) \
+do { \
+ uint32_t t0; \
+ uint32_t t1; \
+ uint32_t t_prob; \
+ uint32_t t_match_byte = (uint32_t)(match_byte_value) << 1; \
+ uint32_t t_match_bit = t_match_byte; \
+ uint32_t t_offset = 0x100; \
+ symbol = 1; \
+ \
+ __asm__( \
+ rc_asm_matched_literal(rc_asm_y) \
+ rc_asm_matched_literal(rc_asm_y) \
+ rc_asm_matched_literal(rc_asm_y) \
+ rc_asm_matched_literal(rc_asm_y) \
+ rc_asm_matched_literal(rc_asm_y) \
+ rc_asm_matched_literal(rc_asm_y) \
+ rc_asm_matched_literal(rc_asm_y) \
+ rc_asm_matched_literal(rc_asm_n) \
+ : \
+ [range] "+&r"(rc.range), \
+ [code] "+&r"(rc.code), \
+ [t0] "=&r"(t0), \
+ [t1] "=&r"(t1), \
+ [prob] "=&r"(t_prob), \
+ [match_bit] "+&r"(t_match_bit), \
+ [symbol] "+&r"(symbol), \
+ [match_byte] "+&r"(t_match_byte), \
+ [offset] "+&r"(t_offset), \
+ [in_ptr] "+&r"(rc_in_ptr) \
+ : \
+ [probs_base] "r"(probs_base_var), \
+ [top_value] "n"(RC_TOP_VALUE), \
+ [shift_bits] "n"(RC_SHIFT_BITS), \
+ [bit_model_total_bits] "n"(RC_BIT_MODEL_TOTAL_BITS), \
+ [bit_model_offset] "n"(RC_BIT_MODEL_OFFSET), \
+ [move_bits] "n"(RC_MOVE_BITS) \
+ : \
+ "cc", "memory"); \
+} while (0)
+#endif // LZMA_RANGE_DECODER_CONFIG & 0x080
+
+
+// Doing the loop in asm instead of C seems to help a little.
+#if LZMA_RANGE_DECODER_CONFIG & 0x100
+#undef rc_direct
+#define rc_direct(dest_var, count_var) \
+do { \
+ uint32_t t0; \
+ uint32_t t1; \
+ \
+ __asm__( \
+ "2:\n\t" \
+ "add %[dest], %[dest]\n\t" \
+ "lea 1(%q[dest]), %[t1]\n\t" \
+ \
+ rc_asm_normalize \
+ \
+ "shr $1, %[range]\n\t" \
+ "mov %[code], %[t0]\n\t" \
+ "sub %[range], %[code]\n\t" \
+ "cmovns %[t1], %[dest]\n\t" \
+ "cmovs %[t0], %[code]\n\t" \
+ "dec %[count]\n\t" \
+ "jnz 2b\n\t" \
+ : \
+ [range] "+&r"(rc.range), \
+ [code] "+&r"(rc.code), \
+ [t0] "=&r"(t0), \
+ [t1] "=&r"(t1), \
+ [dest] "+&r"(dest_var), \
+ [count] "+&r"(count_var), \
+ [in_ptr] "+&r"(rc_in_ptr) \
+ : \
+ [top_value] "n"(RC_TOP_VALUE), \
+ [shift_bits] "n"(RC_SHIFT_BITS) \
+ : \
+ "cc", "memory"); \
+} while (0)
+#endif // LZMA_RANGE_DECODER_CONFIG & 0x100
+
+#endif // x86_64
#endif