root/llama.cpp
1
0
Fork 0
mirror of https://github.com/ggerganov/llama.cpp.git synced 2024-12-25 02:44:36 +00:00
Code Issues Actions 9 Packages Projects Releases Wiki Activity

cuda : vsubss4 for older versions of ROCm/clang (#2942)

Browse Source
This commit is contained in:
Engininja2 2023-09-01 15:33:19 -06:00 committed by GitHub
parent 69fdbb9abc
commit f04d002844
No known key found for this signature in database
GPG Key ID: 4AEE18F83AFDEB23
1 changed files with 17 additions and 0 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

17
ggml-cuda.cu
View File

@ -81,12 +81,29 @@
#if defined(GGML_USE_HIPBLAS) #if defined(GGML_USE_HIPBLAS)
#define __CUDA_ARCH__ 1300 #define __CUDA_ARCH__ 1300
#ifndef __has_builtin
#define __has_builtin(x) 0
#endif
typedef int8_t int8x4_t __attribute__((ext_vector_type(4))); typedef int8_t int8x4_t __attribute__((ext_vector_type(4)));
static __device__ __forceinline__ int __vsubss4(const int a, const int b) { static __device__ __forceinline__ int __vsubss4(const int a, const int b) {
const int8x4_t va = reinterpret_cast<const int8x4_t&>(a); const int8x4_t va = reinterpret_cast<const int8x4_t&>(a);
const int8x4_t vb = reinterpret_cast<const int8x4_t&>(b); const int8x4_t vb = reinterpret_cast<const int8x4_t&>(b);
#if __has_builtin(__builtin_elementwise_sub_sat)
const int8x4_t c = __builtin_elementwise_sub_sat(va, vb); const int8x4_t c = __builtin_elementwise_sub_sat(va, vb);
return reinterpret_cast<const int&>(c); return reinterpret_cast<const int&>(c);
#else
int8x4_t c;
int16_t tmp;
#pragma unroll
for (int i = 0; i < 4; i++) {
tmp = va[i] - vb[i];
if(tmp > std::numeric_limits<int8_t>::max()) tmp = std::numeric_limits<int8_t>::max();
if(tmp < std::numeric_limits<int8_t>::min()) tmp = std::numeric_limits<int8_t>::min();
c[i] = tmp;
}
return reinterpret_cast<int&>(c);
#endif // __has_builtin(__builtin_elementwise_sub_sat)
} }
static __device__ __forceinline__ int __dp4a(const int a, const int b, int c) { static __device__ __forceinline__ int __dp4a(const int a, const int b, int c) {