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

Flatten ZSTD_row_getMatchMask (#2681)

Browse Source 
* Flatten ZSTD_row_getMatchMask

* Remove the SIMD abstraction layer.
* Add big endian support.
* Align `hashTags` within `tagRow` to a 16-byte boundary. 
* Switch SSE2 to use aligned reads.
* Optimize scalar path using SWAR.
* Optimize neon path for `n == 32`
* Work around minor clang issue for NEON (https://bugs.llvm.org/show_bug.cgi?id=49577)

* replace memcpy with MEM_readST

* silence alignment warnings

* fix neon casts

* Update zstd_lazy.c

* unify simd preprocessor detection (#3)

* remove duplicate asserts

* tweak rotates

* improve endian detection

* add cast

there is a fun little catch-22 with gcc: result from pmovmskb has to be cast to uint32_t to avoid a zero-extension
but must be uint16_t to get gcc to generate a rotate instruction..

* more casts

* fix casts

better work-around for the (bogus) warning: unary minus on unsigned
This commit is contained in:
aqrit
2021-06-09 08:50:25 +03:00
committed by GitHub
parent 8a3bdfaa7b
commit dd4f6aa9e6
5 changed files with 137 additions and 227 deletions
Download Patch File Download Diff File Expand all files Collapse all files
lib/common/compiler.h
+16
View File
@@ -197,6 +197,22 @@
#define STATIC_BMI2 0
#endif
/* compile time determination of SIMD support */
#if !defined(ZSTD_NO_INTRINSICS)
# if defined(__SSE2__) || defined(_M_AMD64) || (defined (_M_IX86) && defined(_M_IX86_FP) && (_M_IX86_FP >= 2))
# define ZSTD_ARCH_X86_SSE2
# endif
# if defined(__ARM_NEON) || defined(_M_ARM64)
# define ZSTD_ARCH_ARM_NEON
# endif
#
# if defined(ZSTD_ARCH_X86_SSE2)
# include <emmintrin.h>
# elif defined(ZSTD_ARCH_ARM_NEON)
# include <arm_neon.h>
# endif
#endif
/* compat. with non-clang compilers */
#ifndef __has_builtin
# define __has_builtin(x) 0
lib/common/mem.h
+14
View File
@@ -153,8 +153,22 @@ MEM_STATIC unsigned MEM_64bits(void) { return sizeof(size_t)==8; }
MEM_STATIC unsigned MEM_isLittleEndian(void)
{
#if defined(__BYTE_ORDER__) && defined(__ORDER_LITTLE_ENDIAN__) && (__BYTE_ORDER__ == __ORDER_LITTLE_ENDIAN__)
return 1;
#elif defined(__BYTE_ORDER__) && defined(__ORDER_BIG_ENDIAN__) && (__BYTE_ORDER__ == __ORDER_BIG_ENDIAN__)
return 0;
#elif defined(__clang__) && __LITTLE_ENDIAN__
return 1;
#elif defined(__clang__) && __BIG_ENDIAN__
return 0;
#elif defined(_MSC_VER) && (_M_AMD64 || _M_IX86)
return 1;
#elif defined(__DMC__) && defined(_M_IX86)
return 1;
#else
const union { U32 u; BYTE c[4]; } one = { 1 }; /* don't use static : performance detrimental */
return one.c[0];
#endif
}
#if defined(MEM_FORCE_MEMORY_ACCESS) && (MEM_FORCE_MEMORY_ACCESS==2)
lib/common/zstd_internal.h
+2 -5
View File
@@ -19,9 +19,6 @@
/*-*************************************
* Dependencies
***************************************/
#if !defined(ZSTD_NO_INTRINSICS) && defined(__ARM_NEON)
#include <arm_neon.h>
#endif
#include "compiler.h"
#include "mem.h"
#include "debug.h" /* assert, DEBUGLOG, RAWLOG, g_debuglevel */
@@ -247,7 +244,7 @@ static UNUSED_ATTR const U32 OF_defaultNormLog = OF_DEFAULTNORMLOG;
* Shared functions to include for inlining
*********************************************/
static void ZSTD_copy8(void* dst, const void* src) {
#if !defined(ZSTD_NO_INTRINSICS) && defined(__ARM_NEON)
#if defined(ZSTD_ARCH_ARM_NEON)
vst1_u8((uint8_t*)dst, vld1_u8((const uint8_t*)src));
#else
ZSTD_memcpy(dst, src, 8);
@@ -256,7 +253,7 @@ static void ZSTD_copy8(void* dst, const void* src) {
#define COPY8(d,s) { ZSTD_copy8(d,s); d+=8; s+=8; }
static void ZSTD_copy16(void* dst, const void* src) {
#if !defined(ZSTD_NO_INTRINSICS) && defined(__ARM_NEON)
#if defined(ZSTD_ARCH_ARM_NEON)
vst1q_u8((uint8_t*)dst, vld1q_u8((const uint8_t*)src));
#else
ZSTD_memcpy(dst, src, 16);