diff options
Diffstat (limited to 'thirdparty/zstd/common/zstd_internal.h')
| -rw-r--r-- | thirdparty/zstd/common/zstd_internal.h | 127 |
1 files changed, 112 insertions, 15 deletions
diff --git a/thirdparty/zstd/common/zstd_internal.h b/thirdparty/zstd/common/zstd_internal.h index dcdcbdb81c..3bc7e55a0a 100644 --- a/thirdparty/zstd/common/zstd_internal.h +++ b/thirdparty/zstd/common/zstd_internal.h @@ -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 @@ -19,12 +19,15 @@ /*-************************************* * Dependencies ***************************************/ +#ifdef __aarch64__ +#include <arm_neon.h> +#endif #include "compiler.h" #include "mem.h" #include "debug.h" /* assert, DEBUGLOG, RAWLOG, g_debuglevel */ #include "error_private.h" #define ZSTD_STATIC_LINKING_ONLY -#include "zstd.h" +#include "../zstd.h" #define FSE_STATIC_LINKING_ONLY #include "fse.h" #define HUF_STATIC_LINKING_ONLY @@ -54,6 +57,31 @@ extern "C" { #define MAX(a,b) ((a)>(b) ? (a) : (b)) /** + * Ignore: this is an internal helper. + * + * This is a helper function to help force C99-correctness during compilation. + * Under strict compilation modes, variadic macro arguments can't be empty. + * However, variadic function arguments can be. Using a function therefore lets + * us statically check that at least one (string) argument was passed, + * independent of the compilation flags. + */ +static INLINE_KEYWORD UNUSED_ATTR +void _force_has_format_string(const char *format, ...) { + (void)format; +} + +/** + * Ignore: this is an internal helper. + * + * We want to force this function invocation to be syntactically correct, but + * we don't want to force runtime evaluation of its arguments. + */ +#define _FORCE_HAS_FORMAT_STRING(...) \ + if (0) { \ + _force_has_format_string(__VA_ARGS__); \ + } + +/** * Return the specified error if the condition evaluates to true. * * In debug modes, prints additional information. @@ -62,7 +90,9 @@ extern "C" { */ #define RETURN_ERROR_IF(cond, err, ...) \ if (cond) { \ - RAWLOG(3, "%s:%d: ERROR!: check %s failed, returning %s", __FILE__, __LINE__, ZSTD_QUOTE(cond), ZSTD_QUOTE(ERROR(err))); \ + RAWLOG(3, "%s:%d: ERROR!: check %s failed, returning %s", \ + __FILE__, __LINE__, ZSTD_QUOTE(cond), ZSTD_QUOTE(ERROR(err))); \ + _FORCE_HAS_FORMAT_STRING(__VA_ARGS__); \ RAWLOG(3, ": " __VA_ARGS__); \ RAWLOG(3, "\n"); \ return ERROR(err); \ @@ -75,7 +105,9 @@ extern "C" { */ #define RETURN_ERROR(err, ...) \ do { \ - RAWLOG(3, "%s:%d: ERROR!: unconditional check failed, returning %s", __FILE__, __LINE__, ZSTD_QUOTE(ERROR(err))); \ + RAWLOG(3, "%s:%d: ERROR!: unconditional check failed, returning %s", \ + __FILE__, __LINE__, ZSTD_QUOTE(ERROR(err))); \ + _FORCE_HAS_FORMAT_STRING(__VA_ARGS__); \ RAWLOG(3, ": " __VA_ARGS__); \ RAWLOG(3, "\n"); \ return ERROR(err); \ @@ -90,7 +122,9 @@ extern "C" { do { \ size_t const err_code = (err); \ if (ERR_isError(err_code)) { \ - RAWLOG(3, "%s:%d: ERROR!: forwarding error in %s: %s", __FILE__, __LINE__, ZSTD_QUOTE(err), ERR_getErrorName(err_code)); \ + RAWLOG(3, "%s:%d: ERROR!: forwarding error in %s: %s", \ + __FILE__, __LINE__, ZSTD_QUOTE(err), ERR_getErrorName(err_code)); \ + _FORCE_HAS_FORMAT_STRING(__VA_ARGS__); \ RAWLOG(3, ": " __VA_ARGS__); \ RAWLOG(3, "\n"); \ return err_code; \ @@ -128,6 +162,8 @@ static const size_t ZSTD_did_fieldSize[4] = { 0, 1, 2, 4 }; static const size_t ZSTD_blockHeaderSize = ZSTD_BLOCKHEADERSIZE; typedef enum { bt_raw, bt_rle, bt_compressed, bt_reserved } blockType_e; +#define ZSTD_FRAMECHECKSUMSIZE 4 + #define MIN_SEQUENCES_SIZE 1 /* nbSeq==0 */ #define MIN_CBLOCK_SIZE (1 /*litCSize*/ + 1 /* RLE or RAW */ + MIN_SEQUENCES_SIZE /* nbSeq==0 */) /* for a non-null block */ @@ -191,10 +227,22 @@ static const U32 OF_defaultNormLog = OF_DEFAULTNORMLOG; /*-******************************************* * Shared functions to include for inlining *********************************************/ -static void ZSTD_copy8(void* dst, const void* src) { memcpy(dst, src, 8); } +static void ZSTD_copy8(void* dst, const void* src) { +#ifdef __aarch64__ + vst1_u8((uint8_t*)dst, vld1_u8((const uint8_t*)src)); +#else + memcpy(dst, src, 8); +#endif +} #define COPY8(d,s) { ZSTD_copy8(d,s); d+=8; s+=8; } -static void ZSTD_copy16(void* dst, const void* src) { memcpy(dst, src, 16); } +static void ZSTD_copy16(void* dst, const void* src) { +#ifdef __aarch64__ + vst1q_u8((uint8_t*)dst, vld1q_u8((const uint8_t*)src)); +#else + memcpy(dst, src, 16); +#endif +} #define COPY16(d,s) { ZSTD_copy16(d,s); d+=16; s+=16; } #define WILDCOPY_OVERLENGTH 32 @@ -213,7 +261,7 @@ typedef enum { * - ZSTD_overlap_src_before_dst: The src and dst may overlap, but they MUST be at least 8 bytes apart. * The src buffer must be before the dst buffer. */ -MEM_STATIC FORCE_INLINE_ATTR DONT_VECTORIZE +MEM_STATIC FORCE_INLINE_ATTR void ZSTD_wildcopy(void* dst, const void* src, ptrdiff_t length, ZSTD_overlap_e const ovtype) { ptrdiff_t diff = (BYTE*)dst - (const BYTE*)src; @@ -230,13 +278,18 @@ void ZSTD_wildcopy(void* dst, const void* src, ptrdiff_t length, ZSTD_overlap_e } while (op < oend); } else { assert(diff >= WILDCOPY_VECLEN || diff <= -WILDCOPY_VECLEN); - /* Separate out the first two COPY16() calls because the copy length is + /* Separate out the first COPY16() call because the copy length is * almost certain to be short, so the branches have different - * probabilities. - * On gcc-9 unrolling once is +1.6%, twice is +2%, thrice is +1.8%. - * On clang-8 unrolling once is +1.4%, twice is +3.3%, thrice is +3%. + * probabilities. Since it is almost certain to be short, only do + * one COPY16() in the first call. Then, do two calls per loop since + * at that point it is more likely to have a high trip count. */ - COPY16(op, ip); +#ifndef __aarch64__ + do { + COPY16(op, ip); + } + while (op < oend); +#else COPY16(op, ip); if (op >= oend) return; do { @@ -244,9 +297,29 @@ void ZSTD_wildcopy(void* dst, const void* src, ptrdiff_t length, ZSTD_overlap_e COPY16(op, ip); } while (op < oend); +#endif + } +} + +MEM_STATIC size_t ZSTD_limitCopy(void* dst, size_t dstCapacity, const void* src, size_t srcSize) +{ + size_t const length = MIN(dstCapacity, srcSize); + if (length > 0) { + memcpy(dst, src, length); } + return length; } +/* define "workspace is too large" as this number of times larger than needed */ +#define ZSTD_WORKSPACETOOLARGE_FACTOR 3 + +/* when workspace is continuously too large + * during at least this number of times, + * context's memory usage is considered wasteful, + * because it's sized to handle a worst case scenario which rarely happens. + * In which case, resize it down to free some memory */ +#define ZSTD_WORKSPACETOOLARGE_MAXDURATION 128 + /*-******************************************* * Private declarations @@ -271,6 +344,31 @@ typedef struct { U32 longLengthPos; } seqStore_t; +typedef struct { + U32 litLength; + U32 matchLength; +} ZSTD_sequenceLength; + +/** + * Returns the ZSTD_sequenceLength for the given sequences. It handles the decoding of long sequences + * indicated by longLengthPos and longLengthID, and adds MINMATCH back to matchLength. + */ +MEM_STATIC ZSTD_sequenceLength ZSTD_getSequenceLength(seqStore_t const* seqStore, seqDef const* seq) +{ + ZSTD_sequenceLength seqLen; + seqLen.litLength = seq->litLength; + seqLen.matchLength = seq->matchLength + MINMATCH; + if (seqStore->longLengthPos == (U32)(seq - seqStore->sequencesStart)) { + if (seqStore->longLengthID == 1) { + seqLen.litLength += 0xFFFF; + } + if (seqStore->longLengthID == 2) { + seqLen.matchLength += 0xFFFF; + } + } + return seqLen; +} + /** * Contains the compressed frame size and an upper-bound for the decompressed frame size. * Note: before using `compressedSize`, check for errors using ZSTD_isError(). @@ -297,8 +395,7 @@ MEM_STATIC U32 ZSTD_highbit32(U32 val) /* compress, dictBuilder, decodeCorpus { # if defined(_MSC_VER) /* Visual */ unsigned long r=0; - _BitScanReverse(&r, val); - return (unsigned)r; + return _BitScanReverse(&r, val) ? (unsigned)r : 0; # elif defined(__GNUC__) && (__GNUC__ >= 3) /* GCC Intrinsic */ return __builtin_clz (val) ^ 31; # elif defined(__ICCARM__) /* IAR Intrinsic */ |