ddidderr/zstd-rs
1
0
Fork 0
Code Issues Pull Requests Packages Projects Releases Wiki Activity

Merge branch 'dev' into http-to-https

Browse Source
This commit is contained in:
daniellerozenblit
2022-12-15 10:46:13 -05:00
committed by GitHub
parent 4dffc35f2e 728e73ebb4
commit e2fc93340f
34 changed files with 574 additions and 534 deletions
Download Patch File Download Diff File Expand all files Collapse all files
lib/common/mem.h
+16 -33
View File
@@ -133,21 +133,15 @@ MEM_STATIC size_t MEM_swapST(size_t in);
/*-**************************************************************
* Memory I/O Implementation
*****************************************************************/
/* MEM_FORCE_MEMORY_ACCESS :
* By default, access to unaligned memory is controlled by `memcpy()`, which is safe and portable.
* Unfortunately, on some target/compiler combinations, the generated assembly is sub-optimal.
* The below switch allow to select different access method for improved performance.
* Method 0 (default) : use `memcpy()`. Safe and portable.
* Method 1 : `__packed` statement. It depends on compiler extension (i.e., not portable).
* This method is safe if your compiler supports it, and *generally* as fast or faster than `memcpy`.
/* MEM_FORCE_MEMORY_ACCESS : For accessing unaligned memory:
* Method 0 : always use `memcpy()`. Safe and portable.
* Method 1 : Use compiler extension to set unaligned access.
* Method 2 : direct access. This method is portable but violate C standard.
* It can generate buggy code on targets depending on alignment.
* In some circumstances, it's the only known way to get the most performance (i.e. GCC + ARMv6)
* See https://fastcompression.blogspot.fr/2015/08/accessing-unaligned-memory.html for details.
* Prefer these methods in priority order (0 > 1 > 2)
* Default : method 1 if supported, else method 0
*/
#ifndef MEM_FORCE_MEMORY_ACCESS /* can be defined externally, on command line for example */
# if defined(__INTEL_COMPILER) || defined(__GNUC__) || defined(__ICCARM__)
# ifdef __GNUC__
# define MEM_FORCE_MEMORY_ACCESS 1
# endif
#endif
@@ -190,30 +184,19 @@ MEM_STATIC void MEM_write64(void* memPtr, U64 value) { *(U64*)memPtr = value; }
#elif defined(MEM_FORCE_MEMORY_ACCESS) && (MEM_FORCE_MEMORY_ACCESS==1)
/* __pack instructions are safer, but compiler specific, hence potentially problematic for some compilers */
/* currently only defined for gcc and icc */
#if defined(_MSC_VER) || (defined(__INTEL_COMPILER) && defined(WIN32))
__pragma( pack(push, 1) )
typedef struct { U16 v; } unalign16;
typedef struct { U32 v; } unalign32;
typedef struct { U64 v; } unalign64;
typedef struct { size_t v; } unalignArch;
__pragma( pack(pop) )
#else
typedef struct { U16 v; } __attribute__((packed)) unalign16;
typedef struct { U32 v; } __attribute__((packed)) unalign32;
typedef struct { U64 v; } __attribute__((packed)) unalign64;
typedef struct { size_t v; } __attribute__((packed)) unalignArch;
#endif
typedef __attribute__((aligned(1))) U16 unalign16;
typedef __attribute__((aligned(1))) U32 unalign32;
typedef __attribute__((aligned(1))) U64 unalign64;
typedef __attribute__((aligned(1))) size_t unalignArch;
MEM_STATIC U16 MEM_read16(const void* ptr) { return ((const unalign16*)ptr)->v; }
MEM_STATIC U32 MEM_read32(const void* ptr) { return ((const unalign32*)ptr)->v; }
MEM_STATIC U64 MEM_read64(const void* ptr) { return ((const unalign64*)ptr)->v; }
MEM_STATIC size_t MEM_readST(const void* ptr) { return ((const unalignArch*)ptr)->v; }
MEM_STATIC U16 MEM_read16(const void* ptr) { return *(const unalign16*)ptr; }
MEM_STATIC U32 MEM_read32(const void* ptr) { return *(const unalign32*)ptr; }
MEM_STATIC U64 MEM_read64(const void* ptr) { return *(const unalign64*)ptr; }
MEM_STATIC size_t MEM_readST(const void* ptr) { return *(const unalignArch*)ptr; }
MEM_STATIC void MEM_write16(void* memPtr, U16 value) { ((unalign16*)memPtr)->v = value; }
MEM_STATIC void MEM_write32(void* memPtr, U32 value) { ((unalign32*)memPtr)->v = value; }
MEM_STATIC void MEM_write64(void* memPtr, U64 value) { ((unalign64*)memPtr)->v = value; }
MEM_STATIC void MEM_write16(void* memPtr, U16 value) { *(unalign16*)memPtr = value; }
MEM_STATIC void MEM_write32(void* memPtr, U32 value) { *(unalign32*)memPtr = value; }
MEM_STATIC void MEM_write64(void* memPtr, U64 value) { *(unalign64*)memPtr = value; }
#else
lib/compress/zstd_compress.c
+2 -2
View File
@@ -1370,8 +1370,8 @@ ZSTD_adjustCParams_internal(ZSTD_compressionParameters cPar,
}
/* resize windowLog if input is small enough, to use less memory */
if ( (srcSize < maxWindowResize)
&& (dictSize < maxWindowResize) ) {
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 :
lib/decompress/zstd_decompress.c
+8 -3
View File
@@ -1549,7 +1549,7 @@ unsigned ZSTD_getDictID_fromDict(const void* dict, size_t dictSize)
* ZSTD_getFrameHeader(), which will provide a more precise error code. */
unsigned ZSTD_getDictID_fromFrame(const void* src, size_t srcSize)
{
ZSTD_frameHeader zfp = { 0, 0, 0, ZSTD_frame, 0, 0, 0 };
ZSTD_frameHeader zfp = { 0, 0, 0, ZSTD_frame, 0, 0, 0, 0, 0 };
size_t const hError = ZSTD_getFrameHeader(&zfp, src, srcSize);
if (ZSTD_isError(hError)) return 0;
return zfp.dictID;
@@ -2058,6 +2058,7 @@ size_t ZSTD_decompressStream(ZSTD_DStream* zds, ZSTD_outBuffer* output, ZSTD_inB
size_t const decompressedSize = ZSTD_decompress_usingDDict(zds, op, (size_t)(oend-op), istart, cSize, ZSTD_getDDict(zds));
if (ZSTD_isError(decompressedSize)) return decompressedSize;
DEBUGLOG(4, "shortcut to single-pass ZSTD_decompress_usingDDict()")
assert(istart != NULL);
ip = istart + cSize;
op = op ? op + decompressedSize : op; /* can occur if frameContentSize = 0 (empty frame) */
zds->expected = 0;
@@ -2143,6 +2144,7 @@ size_t ZSTD_decompressStream(ZSTD_DStream* zds, ZSTD_outBuffer* output, ZSTD_inB
}
if ((size_t)(iend-ip) >= neededInSize) { /* decode directly from src */
FORWARD_IF_ERROR(ZSTD_decompressContinueStream(zds, &op, oend, ip, neededInSize), "");
assert(ip != NULL);
ip += neededInSize;
/* Function modifies the stage so we must break */
break;
@@ -2166,8 +2168,11 @@ size_t ZSTD_decompressStream(ZSTD_DStream* zds, ZSTD_outBuffer* output, ZSTD_inB
"should never happen");
loadedSize = ZSTD_limitCopy(zds->inBuff + zds->inPos, toLoad, ip, (size_t)(iend-ip));
}
ip += loadedSize;
zds->inPos += loadedSize;
if (loadedSize != 0) {
/* ip may be NULL */
ip += loadedSize;
zds->inPos += loadedSize;
}
if (loadedSize < toLoad) { someMoreWork = 0; break; } /* not enough input, wait for more */
/* decode loaded input */
lib/legacy/zstd_v01.c
-39
View File
@@ -190,25 +190,6 @@ typedef signed long long S64;
/****************************************************************
* Memory I/O
*****************************************************************/
/* FSE_FORCE_MEMORY_ACCESS
* By default, access to unaligned memory is controlled by `memcpy()`, which is safe and portable.
* Unfortunately, on some target/compiler combinations, the generated assembly is sub-optimal.
* The below switch allow to select different access method for improved performance.
* Method 0 (default) : use `memcpy()`. Safe and portable.
* Method 1 : `__packed` statement. It depends on compiler extension (ie, not portable).
* This method is safe if your compiler supports it, and *generally* as fast or faster than `memcpy`.
* Method 2 : direct access. This method is portable but violate C standard.
* It can generate buggy code on targets generating assembly depending on alignment.
* But in some circumstances, it's the only known way to get the most performance (ie GCC + ARMv6)
* See https://fastcompression.blogspot.fr/2015/08/accessing-unaligned-memory.html for details.
* Prefer these methods in priority order (0 > 1 > 2)
*/
#ifndef FSE_FORCE_MEMORY_ACCESS /* can be defined externally, on command line for example */
# if defined(__INTEL_COMPILER) || defined(__GNUC__) || defined(__ICCARM__)
# define FSE_FORCE_MEMORY_ACCESS 1
# endif
#endif
static unsigned FSE_32bits(void)
{
@@ -221,24 +202,6 @@ static unsigned FSE_isLittleEndian(void)
return one.c[0];
}
#if defined(FSE_FORCE_MEMORY_ACCESS) && (FSE_FORCE_MEMORY_ACCESS==2)
static U16 FSE_read16(const void* memPtr) { return *(const U16*) memPtr; }
static U32 FSE_read32(const void* memPtr) { return *(const U32*) memPtr; }
static U64 FSE_read64(const void* memPtr) { return *(const U64*) memPtr; }
#elif defined(FSE_FORCE_MEMORY_ACCESS) && (FSE_FORCE_MEMORY_ACCESS==1)
/* __pack instructions are safer, but compiler specific, hence potentially problematic for some compilers */
/* currently only defined for gcc and icc */
typedef union { U16 u16; U32 u32; U64 u64; } __attribute__((packed)) unalign;
static U16 FSE_read16(const void* ptr) { return ((const unalign*)ptr)->u16; }
static U32 FSE_read32(const void* ptr) { return ((const unalign*)ptr)->u32; }
static U64 FSE_read64(const void* ptr) { return ((const unalign*)ptr)->u64; }
#else
static U16 FSE_read16(const void* memPtr)
{
U16 val; memcpy(&val, memPtr, sizeof(val)); return val;
@@ -254,8 +217,6 @@ static U64 FSE_read64(const void* memPtr)
U64 val; memcpy(&val, memPtr, sizeof(val)); return val;
}
#endif /* FSE_FORCE_MEMORY_ACCESS */
static U16 FSE_readLE16(const void* memPtr)
{
if (FSE_isLittleEndian())
lib/legacy/zstd_v02.c
-48
View File
@@ -115,24 +115,6 @@ extern "C" {
/****************************************************************
* Memory I/O
*****************************************************************/
/* MEM_FORCE_MEMORY_ACCESS
* By default, access to unaligned memory is controlled by `memcpy()`, which is safe and portable.
* Unfortunately, on some target/compiler combinations, the generated assembly is sub-optimal.
* The below switch allow to select different access method for improved performance.
* Method 0 (default) : use `memcpy()`. Safe and portable.
* Method 1 : `__packed` statement. It depends on compiler extension (ie, not portable).
* This method is safe if your compiler supports it, and *generally* as fast or faster than `memcpy`.
* Method 2 : direct access. This method is portable but violate C standard.
* It can generate buggy code on targets generating assembly depending on alignment.
* But in some circumstances, it's the only known way to get the most performance (ie GCC + ARMv6)
* See https://fastcompression.blogspot.fr/2015/08/accessing-unaligned-memory.html for details.
* Prefer these methods in priority order (0 > 1 > 2)
*/
#ifndef MEM_FORCE_MEMORY_ACCESS /* can be defined externally, on command line for example */
# if defined(__INTEL_COMPILER) || defined(__GNUC__) || defined(__ICCARM__)
# define MEM_FORCE_MEMORY_ACCESS 1
# endif
#endif
MEM_STATIC unsigned MEM_32bits(void) { return sizeof(void*)==4; }
MEM_STATIC unsigned MEM_64bits(void) { return sizeof(void*)==8; }
@@ -143,33 +125,6 @@ MEM_STATIC unsigned MEM_isLittleEndian(void)
return one.c[0];
}
#if defined(MEM_FORCE_MEMORY_ACCESS) && (MEM_FORCE_MEMORY_ACCESS==2)
/* violates C standard on structure alignment.
Only use if no other choice to achieve best performance on target platform */
MEM_STATIC U16 MEM_read16(const void* memPtr) { return *(const U16*) memPtr; }
MEM_STATIC U32 MEM_read32(const void* memPtr) { return *(const U32*) memPtr; }
MEM_STATIC U64 MEM_read64(const void* memPtr) { return *(const U64*) memPtr; }
MEM_STATIC void MEM_write16(void* memPtr, U16 value) { *(U16*)memPtr = value; }
#elif defined(MEM_FORCE_MEMORY_ACCESS) && (MEM_FORCE_MEMORY_ACCESS==1)
/* __pack instructions are safer, but compiler specific, hence potentially problematic for some compilers */
/* currently only defined for gcc and icc */
typedef union { U16 u16; U32 u32; U64 u64; } __attribute__((packed)) unalign;
MEM_STATIC U16 MEM_read16(const void* ptr) { return ((const unalign*)ptr)->u16; }
MEM_STATIC U32 MEM_read32(const void* ptr) { return ((const unalign*)ptr)->u32; }
MEM_STATIC U64 MEM_read64(const void* ptr) { return ((const unalign*)ptr)->u64; }
MEM_STATIC void MEM_write16(void* memPtr, U16 value) { ((unalign*)memPtr)->u16 = value; }
#else
/* default method, safe and standard.
can sometimes prove slower */
MEM_STATIC U16 MEM_read16(const void* memPtr)
{
U16 val; memcpy(&val, memPtr, sizeof(val)); return val;
@@ -190,9 +145,6 @@ MEM_STATIC void MEM_write16(void* memPtr, U16 value)
memcpy(memPtr, &value, sizeof(value));
}
#endif /* MEM_FORCE_MEMORY_ACCESS */
MEM_STATIC U16 MEM_readLE16(const void* memPtr)
{
if (MEM_isLittleEndian())
lib/legacy/zstd_v03.c
-49
View File
@@ -116,24 +116,6 @@ extern "C" {
/****************************************************************
* Memory I/O
*****************************************************************/
/* MEM_FORCE_MEMORY_ACCESS
* By default, access to unaligned memory is controlled by `memcpy()`, which is safe and portable.
* Unfortunately, on some target/compiler combinations, the generated assembly is sub-optimal.
* The below switch allow to select different access method for improved performance.
* Method 0 (default) : use `memcpy()`. Safe and portable.
* Method 1 : `__packed` statement. It depends on compiler extension (ie, not portable).
* This method is safe if your compiler supports it, and *generally* as fast or faster than `memcpy`.
* Method 2 : direct access. This method is portable but violate C standard.
* It can generate buggy code on targets generating assembly depending on alignment.
* But in some circumstances, it's the only known way to get the most performance (ie GCC + ARMv6)
* See https://fastcompression.blogspot.fr/2015/08/accessing-unaligned-memory.html for details.
* Prefer these methods in priority order (0 > 1 > 2)
*/
#ifndef MEM_FORCE_MEMORY_ACCESS /* can be defined externally, on command line for example */
# if defined(__INTEL_COMPILER) || defined(__GNUC__) || defined(__ICCARM__)
# define MEM_FORCE_MEMORY_ACCESS 1
# endif
#endif
MEM_STATIC unsigned MEM_32bits(void) { return sizeof(void*)==4; }
MEM_STATIC unsigned MEM_64bits(void) { return sizeof(void*)==8; }
@@ -144,33 +126,6 @@ MEM_STATIC unsigned MEM_isLittleEndian(void)
return one.c[0];
}
#if defined(MEM_FORCE_MEMORY_ACCESS) && (MEM_FORCE_MEMORY_ACCESS==2)
/* violates C standard on structure alignment.
Only use if no other choice to achieve best performance on target platform */
MEM_STATIC U16 MEM_read16(const void* memPtr) { return *(const U16*) memPtr; }
MEM_STATIC U32 MEM_read32(const void* memPtr) { return *(const U32*) memPtr; }
MEM_STATIC U64 MEM_read64(const void* memPtr) { return *(const U64*) memPtr; }
MEM_STATIC void MEM_write16(void* memPtr, U16 value) { *(U16*)memPtr = value; }
#elif defined(MEM_FORCE_MEMORY_ACCESS) && (MEM_FORCE_MEMORY_ACCESS==1)
/* __pack instructions are safer, but compiler specific, hence potentially problematic for some compilers */
/* currently only defined for gcc and icc */
typedef union { U16 u16; U32 u32; U64 u64; } __attribute__((packed)) unalign;
MEM_STATIC U16 MEM_read16(const void* ptr) { return ((const unalign*)ptr)->u16; }
MEM_STATIC U32 MEM_read32(const void* ptr) { return ((const unalign*)ptr)->u32; }
MEM_STATIC U64 MEM_read64(const void* ptr) { return ((const unalign*)ptr)->u64; }
MEM_STATIC void MEM_write16(void* memPtr, U16 value) { ((unalign*)memPtr)->u16 = value; }
#else
/* default method, safe and standard.
can sometimes prove slower */
MEM_STATIC U16 MEM_read16(const void* memPtr)
{
U16 val; memcpy(&val, memPtr, sizeof(val)); return val;
@@ -191,10 +146,6 @@ MEM_STATIC void MEM_write16(void* memPtr, U16 value)
memcpy(memPtr, &value, sizeof(value));
}
#endif /* MEM_FORCE_MEMORY_ACCESS */
MEM_STATIC U16 MEM_readLE16(const void* memPtr)
{
if (MEM_isLittleEndian())
lib/legacy/zstd_v04.c
-48
View File
@@ -87,24 +87,6 @@ extern "C" {
/****************************************************************
* Memory I/O
*****************************************************************/
/* MEM_FORCE_MEMORY_ACCESS
* By default, access to unaligned memory is controlled by `memcpy()`, which is safe and portable.
* Unfortunately, on some target/compiler combinations, the generated assembly is sub-optimal.
* The below switch allow to select different access method for improved performance.
* Method 0 (default) : use `memcpy()`. Safe and portable.
* Method 1 : `__packed` statement. It depends on compiler extension (ie, not portable).
* This method is safe if your compiler supports it, and *generally* as fast or faster than `memcpy`.
* Method 2 : direct access. This method is portable but violate C standard.
* It can generate buggy code on targets generating assembly depending on alignment.
* But in some circumstances, it's the only known way to get the most performance (ie GCC + ARMv6)
* See https://fastcompression.blogspot.fr/2015/08/accessing-unaligned-memory.html for details.
* Prefer these methods in priority order (0 > 1 > 2)
*/
#ifndef MEM_FORCE_MEMORY_ACCESS /* can be defined externally, on command line for example */
# if defined(__INTEL_COMPILER) || defined(__GNUC__) || defined(__ICCARM__)
# define MEM_FORCE_MEMORY_ACCESS 1
# endif
#endif
MEM_STATIC unsigned MEM_32bits(void) { return sizeof(void*)==4; }
MEM_STATIC unsigned MEM_64bits(void) { return sizeof(void*)==8; }
@@ -115,33 +97,6 @@ MEM_STATIC unsigned MEM_isLittleEndian(void)
return one.c[0];
}
#if defined(MEM_FORCE_MEMORY_ACCESS) && (MEM_FORCE_MEMORY_ACCESS==2)
/* violates C standard on structure alignment.
Only use if no other choice to achieve best performance on target platform */
MEM_STATIC U16 MEM_read16(const void* memPtr) { return *(const U16*) memPtr; }
MEM_STATIC U32 MEM_read32(const void* memPtr) { return *(const U32*) memPtr; }
MEM_STATIC U64 MEM_read64(const void* memPtr) { return *(const U64*) memPtr; }
MEM_STATIC void MEM_write16(void* memPtr, U16 value) { *(U16*)memPtr = value; }
#elif defined(MEM_FORCE_MEMORY_ACCESS) && (MEM_FORCE_MEMORY_ACCESS==1)
/* __pack instructions are safer, but compiler specific, hence potentially problematic for some compilers */
/* currently only defined for gcc and icc */
typedef union { U16 u16; U32 u32; U64 u64; } __attribute__((packed)) unalign;
MEM_STATIC U16 MEM_read16(const void* ptr) { return ((const unalign*)ptr)->u16; }
MEM_STATIC U32 MEM_read32(const void* ptr) { return ((const unalign*)ptr)->u32; }
MEM_STATIC U64 MEM_read64(const void* ptr) { return ((const unalign*)ptr)->u64; }
MEM_STATIC void MEM_write16(void* memPtr, U16 value) { ((unalign*)memPtr)->u16 = value; }
#else
/* default method, safe and standard.
can sometimes prove slower */
MEM_STATIC U16 MEM_read16(const void* memPtr)
{
U16 val; memcpy(&val, memPtr, sizeof(val)); return val;
@@ -162,9 +117,6 @@ MEM_STATIC void MEM_write16(void* memPtr, U16 value)
memcpy(memPtr, &value, sizeof(value));
}
#endif /* MEM_FORCE_MEMORY_ACCESS */
MEM_STATIC U16 MEM_readLE16(const void* memPtr)
{
if (MEM_isLittleEndian())
lib/legacy/zstd_v05.c
-52
View File
@@ -106,24 +106,6 @@ extern "C" {
/*-**************************************************************
* Memory I/O
*****************************************************************/
/* MEM_FORCE_MEMORY_ACCESS :
* By default, access to unaligned memory is controlled by `memcpy()`, which is safe and portable.
* Unfortunately, on some target/compiler combinations, the generated assembly is sub-optimal.
* The below switch allow to select different access method for improved performance.
* Method 0 (default) : use `memcpy()`. Safe and portable.
* Method 1 : `__packed` statement. It depends on compiler extension (ie, not portable).
* This method is safe if your compiler supports it, and *generally* as fast or faster than `memcpy`.
* Method 2 : direct access. This method is portable but violate C standard.
* It can generate buggy code on targets depending on alignment.
* In some circumstances, it's the only known way to get the most performance (ie GCC + ARMv6)
* See https://fastcompression.blogspot.fr/2015/08/accessing-unaligned-memory.html for details.
* Prefer these methods in priority order (0 > 1 > 2)
*/
#ifndef MEM_FORCE_MEMORY_ACCESS /* can be defined externally, on command line for example */
# if defined(__INTEL_COMPILER) || defined(__GNUC__) || defined(__ICCARM__)
# define MEM_FORCE_MEMORY_ACCESS 1
# endif
#endif
MEM_STATIC unsigned MEM_32bits(void) { return sizeof(void*)==4; }
MEM_STATIC unsigned MEM_64bits(void) { return sizeof(void*)==8; }
@@ -134,37 +116,6 @@ MEM_STATIC unsigned MEM_isLittleEndian(void)
return one.c[0];
}
#if defined(MEM_FORCE_MEMORY_ACCESS) && (MEM_FORCE_MEMORY_ACCESS==2)
/* violates C standard, by lying on structure alignment.
Only use if no other choice to achieve best performance on target platform */
MEM_STATIC U16 MEM_read16(const void* memPtr) { return *(const U16*) memPtr; }
MEM_STATIC U32 MEM_read32(const void* memPtr) { return *(const U32*) memPtr; }
MEM_STATIC U64 MEM_read64(const void* memPtr) { return *(const U64*) memPtr; }
MEM_STATIC void MEM_write16(void* memPtr, U16 value) { *(U16*)memPtr = value; }
MEM_STATIC void MEM_write32(void* memPtr, U32 value) { *(U32*)memPtr = value; }
MEM_STATIC void MEM_write64(void* memPtr, U64 value) { *(U64*)memPtr = value; }
#elif defined(MEM_FORCE_MEMORY_ACCESS) && (MEM_FORCE_MEMORY_ACCESS==1)
/* __pack instructions are safer, but compiler specific, hence potentially problematic for some compilers */
/* currently only defined for gcc and icc */
typedef union { U16 u16; U32 u32; U64 u64; size_t st; } __attribute__((packed)) unalign;
MEM_STATIC U16 MEM_read16(const void* ptr) { return ((const unalign*)ptr)->u16; }
MEM_STATIC U32 MEM_read32(const void* ptr) { return ((const unalign*)ptr)->u32; }
MEM_STATIC U64 MEM_read64(const void* ptr) { return ((const unalign*)ptr)->u64; }
MEM_STATIC void MEM_write16(void* memPtr, U16 value) { ((unalign*)memPtr)->u16 = value; }
MEM_STATIC void MEM_write32(void* memPtr, U32 value) { ((unalign*)memPtr)->u32 = value; }
MEM_STATIC void MEM_write64(void* memPtr, U64 value) { ((unalign*)memPtr)->u64 = value; }
#else
/* default method, safe and standard.
can sometimes prove slower */
MEM_STATIC U16 MEM_read16(const void* memPtr)
{
U16 val; memcpy(&val, memPtr, sizeof(val)); return val;
@@ -195,9 +146,6 @@ MEM_STATIC void MEM_write64(void* memPtr, U64 value)
memcpy(memPtr, &value, sizeof(value));
}
#endif /* MEM_FORCE_MEMORY_ACCESS */
MEM_STATIC U16 MEM_readLE16(const void* memPtr)
{
if (MEM_isLittleEndian())
lib/legacy/zstd_v06.c
+2 -49
View File
@@ -108,24 +108,6 @@ extern "C" {
/*-**************************************************************
* Memory I/O
*****************************************************************/
/* MEM_FORCE_MEMORY_ACCESS :
* By default, access to unaligned memory is controlled by `memcpy()`, which is safe and portable.
* Unfortunately, on some target/compiler combinations, the generated assembly is sub-optimal.
* The below switch allow to select different access method for improved performance.
* Method 0 (default) : use `memcpy()`. Safe and portable.
* Method 1 : `__packed` statement. It depends on compiler extension (ie, not portable).
* This method is safe if your compiler supports it, and *generally* as fast or faster than `memcpy`.
* Method 2 : direct access. This method is portable but violate C standard.
* It can generate buggy code on targets depending on alignment.
* In some circumstances, it's the only known way to get the most performance (ie GCC + ARMv6)
* See https://fastcompression.blogspot.fr/2015/08/accessing-unaligned-memory.html for details.
* Prefer these methods in priority order (0 > 1 > 2)
*/
#ifndef MEM_FORCE_MEMORY_ACCESS /* can be defined externally, on command line for example */
# if defined(__INTEL_COMPILER) || defined(__GNUC__) || defined(__ICCARM__)
# define MEM_FORCE_MEMORY_ACCESS 1
# endif
#endif
MEM_STATIC unsigned MEM_32bits(void) { return sizeof(size_t)==4; }
MEM_STATIC unsigned MEM_64bits(void) { return sizeof(size_t)==8; }
@@ -136,33 +118,6 @@ MEM_STATIC unsigned MEM_isLittleEndian(void)
return one.c[0];
}
#if defined(MEM_FORCE_MEMORY_ACCESS) && (MEM_FORCE_MEMORY_ACCESS==2)
/* violates C standard, by lying on structure alignment.
Only use if no other choice to achieve best performance on target platform */
MEM_STATIC U16 MEM_read16(const void* memPtr) { return *(const U16*) memPtr; }
MEM_STATIC U32 MEM_read32(const void* memPtr) { return *(const U32*) memPtr; }
MEM_STATIC U64 MEM_read64(const void* memPtr) { return *(const U64*) memPtr; }
MEM_STATIC void MEM_write16(void* memPtr, U16 value) { *(U16*)memPtr = value; }
#elif defined(MEM_FORCE_MEMORY_ACCESS) && (MEM_FORCE_MEMORY_ACCESS==1)
/* __pack instructions are safer, but compiler specific, hence potentially problematic for some compilers */
/* currently only defined for gcc and icc */
typedef union { U16 u16; U32 u32; U64 u64; size_t st; } __attribute__((packed)) unalign;
MEM_STATIC U16 MEM_read16(const void* ptr) { return ((const unalign*)ptr)->u16; }
MEM_STATIC U32 MEM_read32(const void* ptr) { return ((const unalign*)ptr)->u32; }
MEM_STATIC U64 MEM_read64(const void* ptr) { return ((const unalign*)ptr)->u64; }
MEM_STATIC void MEM_write16(void* memPtr, U16 value) { ((unalign*)memPtr)->u16 = value; }
#else
/* default method, safe and standard.
can sometimes prove slower */
MEM_STATIC U16 MEM_read16(const void* memPtr)
{
U16 val; memcpy(&val, memPtr, sizeof(val)); return val;
@@ -183,9 +138,6 @@ MEM_STATIC void MEM_write16(void* memPtr, U16 value)
memcpy(memPtr, &value, sizeof(value));
}
#endif /* MEM_FORCE_MEMORY_ACCESS */
MEM_STATIC U32 MEM_swap32(U32 in)
{
#if defined(_MSC_VER) /* Visual Studio */
@@ -4035,7 +3987,8 @@ size_t ZBUFFv06_decompressContinue(ZBUFFv06_DCtx* zbd,
size_t const toLoad = hSize - zbd->lhSize; /* if hSize!=0, hSize > zbd->lhSize */
if (ZSTDv06_isError(hSize)) return hSize;
if (toLoad > (size_t)(iend-ip)) { /* not enough input to load full header */
memcpy(zbd->headerBuffer + zbd->lhSize, ip, iend-ip);
if (ip != NULL)
memcpy(zbd->headerBuffer + zbd->lhSize, ip, iend-ip);
zbd->lhSize += iend-ip;
*dstCapacityPtr = 0;
return (hSize - zbd->lhSize) + ZSTDv06_blockHeaderSize; /* remaining header bytes + next block header */
lib/legacy/zstd_v07.c
+2 -48
View File
@@ -268,24 +268,6 @@ extern "C" {
/*-**************************************************************
* Memory I/O
*****************************************************************/
/* MEM_FORCE_MEMORY_ACCESS :
* By default, access to unaligned memory is controlled by `memcpy()`, which is safe and portable.
* Unfortunately, on some target/compiler combinations, the generated assembly is sub-optimal.
* The below switch allow to select different access method for improved performance.
* Method 0 (default) : use `memcpy()`. Safe and portable.
* Method 1 : `__packed` statement. It depends on compiler extension (ie, not portable).
* This method is safe if your compiler supports it, and *generally* as fast or faster than `memcpy`.
* Method 2 : direct access. This method is portable but violate C standard.
* It can generate buggy code on targets depending on alignment.
* In some circumstances, it's the only known way to get the most performance (ie GCC + ARMv6)
* See https://fastcompression.blogspot.fr/2015/08/accessing-unaligned-memory.html for details.
* Prefer these methods in priority order (0 > 1 > 2)
*/
#ifndef MEM_FORCE_MEMORY_ACCESS /* can be defined externally, on command line for example */
# if defined(__INTEL_COMPILER) || defined(__GNUC__) || defined(__ICCARM__)
# define MEM_FORCE_MEMORY_ACCESS 1
# endif
#endif
MEM_STATIC unsigned MEM_32bits(void) { return sizeof(size_t)==4; }
MEM_STATIC unsigned MEM_64bits(void) { return sizeof(size_t)==8; }
@@ -296,33 +278,6 @@ MEM_STATIC unsigned MEM_isLittleEndian(void)
return one.c[0];
}
#if defined(MEM_FORCE_MEMORY_ACCESS) && (MEM_FORCE_MEMORY_ACCESS==2)
/* violates C standard, by lying on structure alignment.
Only use if no other choice to achieve best performance on target platform */
MEM_STATIC U16 MEM_read16(const void* memPtr) { return *(const U16*) memPtr; }
MEM_STATIC U32 MEM_read32(const void* memPtr) { return *(const U32*) memPtr; }
MEM_STATIC U64 MEM_read64(const void* memPtr) { return *(const U64*) memPtr; }
MEM_STATIC void MEM_write16(void* memPtr, U16 value) { *(U16*)memPtr = value; }
#elif defined(MEM_FORCE_MEMORY_ACCESS) && (MEM_FORCE_MEMORY_ACCESS==1)
/* __pack instructions are safer, but compiler specific, hence potentially problematic for some compilers */
/* currently only defined for gcc and icc */
typedef union { U16 u16; U32 u32; U64 u64; size_t st; } __attribute__((packed)) unalign;
MEM_STATIC U16 MEM_read16(const void* ptr) { return ((const unalign*)ptr)->u16; }
MEM_STATIC U32 MEM_read32(const void* ptr) { return ((const unalign*)ptr)->u32; }
MEM_STATIC U64 MEM_read64(const void* ptr) { return ((const unalign*)ptr)->u64; }
MEM_STATIC void MEM_write16(void* memPtr, U16 value) { ((unalign*)memPtr)->u16 = value; }
#else
/* default method, safe and standard.
can sometimes prove slower */
MEM_STATIC U16 MEM_read16(const void* memPtr)
{
U16 val; memcpy(&val, memPtr, sizeof(val)); return val;
@@ -343,8 +298,6 @@ MEM_STATIC void MEM_write16(void* memPtr, U16 value)
memcpy(memPtr, &value, sizeof(value));
}
#endif /* MEM_FORCE_MEMORY_ACCESS */
MEM_STATIC U32 MEM_swap32(U32 in)
{
#if defined(_MSC_VER) /* Visual Studio */
@@ -4417,7 +4370,8 @@ size_t ZBUFFv07_decompressContinue(ZBUFFv07_DCtx* zbd,
if (hSize != 0) {
size_t const toLoad = hSize - zbd->lhSize; /* if hSize!=0, hSize > zbd->lhSize */
if (toLoad > (size_t)(iend-ip)) { /* not enough input to load full header */
memcpy(zbd->headerBuffer + zbd->lhSize, ip, iend-ip);
if (ip != NULL)
memcpy(zbd->headerBuffer + zbd->lhSize, ip, iend-ip);
zbd->lhSize += iend-ip;
*dstCapacityPtr = 0;
return (hSize - zbd->lhSize) + ZSTDv07_blockHeaderSize; /* remaining header bytes + next block header */
lib/zstd.h
+2
View File
@@ -2573,6 +2573,8 @@ typedef struct {
unsigned headerSize;
unsigned dictID;
unsigned checksumFlag;
unsigned _reserved1;
unsigned _reserved2;
} ZSTD_frameHeader;
/*! ZSTD_getFrameHeader() :