diff options
Diffstat (limited to 'thirdparty/zstd/compress/zstd_lazy.c')
| -rw-r--r-- | thirdparty/zstd/compress/zstd_lazy.c | 45 |
1 files changed, 34 insertions, 11 deletions
diff --git a/thirdparty/zstd/compress/zstd_lazy.c b/thirdparty/zstd/compress/zstd_lazy.c index 9ad7e03b54..4cf5c88b53 100644 --- a/thirdparty/zstd/compress/zstd_lazy.c +++ b/thirdparty/zstd/compress/zstd_lazy.c @@ -1,5 +1,5 @@ /* - * Copyright (c) 2016-present, Yann Collet, Facebook, Inc. + * 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 @@ -660,12 +660,16 @@ ZSTD_compressBlock_lazy_generic( const U32 dictIndexDelta = dictMode == ZSTD_dictMatchState ? prefixLowestIndex - (U32)(dictEnd - dictBase) : 0; - const U32 dictAndPrefixLength = (U32)(ip - prefixLowest + dictEnd - dictLowest); + 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 maxRep = (U32)(ip - prefixLowest); + 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; } @@ -677,6 +681,12 @@ ZSTD_compressBlock_lazy_generic( } /* 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; @@ -929,11 +939,11 @@ size_t ZSTD_compressBlock_lazy_extDict_generic( const BYTE* const ilimit = iend - 8; const BYTE* const base = ms->window.base; const U32 dictLimit = ms->window.dictLimit; - const U32 lowestIndex = ms->window.lowLimit; const BYTE* const prefixStart = base + dictLimit; const BYTE* const dictBase = ms->window.dictBase; const BYTE* const dictEnd = dictBase + dictLimit; - const BYTE* const dictStart = dictBase + lowestIndex; + const BYTE* const dictStart = dictBase + ms->window.lowLimit; + const U32 windowLog = ms->cParams.windowLog; typedef size_t (*searchMax_f)( ZSTD_matchState_t* ms, @@ -942,10 +952,18 @@ size_t ZSTD_compressBlock_lazy_extDict_generic( 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; @@ -953,10 +971,11 @@ size_t ZSTD_compressBlock_lazy_extDict_generic( U32 current = (U32)(ip-base); /* check repCode */ - { const U32 repIndex = (U32)(current+1 - offset_1); + { 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 > lowestIndex)) /* intentional overflow */ + 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; @@ -983,10 +1002,11 @@ size_t ZSTD_compressBlock_lazy_extDict_generic( 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 > lowestIndex)) /* intentional overflow */ + 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; @@ -1013,10 +1033,11 @@ size_t ZSTD_compressBlock_lazy_extDict_generic( 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 > lowestIndex)) /* intentional overflow */ + 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; @@ -1057,10 +1078,12 @@ _storeSequence: /* check immediate repcode */ while (ip <= ilimit) { - const U32 repIndex = (U32)((ip-base) - offset_2); + 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 > lowestIndex)) /* intentional overflow */ + 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; |