vbatts/llama.cpp
1
0
Fork 0
Code Issues Pull requests Projects Releases Packages Wiki Activity Actions

Enable F16C/CVT16 vector extensions on MSVC

Browse source 
__F16C__ macro does not exist in MSVC, but is implied with AVX2/AVX512
This commit is contained in:
anzz1 2023-03-22 04:05:49 +02:00
parent 50ea0027a4
commit c43d45696f
1 changed files with 11 additions and 8 deletions
Download patch file Download diff file Expand all files Collapse all files

19
ggml.c
View file

@ -79,6 +79,16 @@ static int sched_yield (void) {
typedef void* thread_ret_t; typedef void* thread_ret_t;
#endif #endif
// __FMA__ and __F16C__ are not defined in MSVC, however they are implied with AVX2/AVX512
#if defined(_MSC_VER) && (defined(__AVX2__) || defined(__AVX512F__))
#ifndef __FMA__
#define __FMA__
#endif
#ifndef __F16C__
#define __F16C__
#endif
#endif
#ifdef __HAIKU__ #ifdef __HAIKU__
#define static_assert(cond, msg) _Static_assert(cond, msg) #define static_assert(cond, msg) _Static_assert(cond, msg)
#endif #endif
@ -407,16 +417,9 @@ static const size_t CACHE_LINE_SIZE_F32 = CACHE_LINE_SIZE/sizeof(float);
#define QK 32 #define QK 32
#if __AVX2__ || __AVX512F__
// __FMA__ is not defined in MSVC, however it is implied with AVX2/AVX512
#if defined(_MSC_VER) && !defined(__FMA__)
#define __FMA__
#endif
// AVX routines provided by GH user Const-me // AVX routines provided by GH user Const-me
// ref: https://github.com/ggerganov/ggml/pull/27#issuecomment-1464934600 // ref: https://github.com/ggerganov/ggml/pull/27#issuecomment-1464934600
#if __AVX2__ || __AVX512F__
// Unpack 32 4-bit fields into 32 bytes // Unpack 32 4-bit fields into 32 bytes
// The output vector contains 32 bytes, each one in [ 0 .. 15 ] interval // The output vector contains 32 bytes, each one in [ 0 .. 15 ] interval
static inline __m256i bytesFromNibbles( const uint8_t* rsi ) static inline __m256i bytesFromNibbles( const uint8_t* rsi )