Mirrors/llama.cpp
1
0
Fork 0
mirror of https://github.com/ggerganov/llama.cpp.git synced 2024-12-26 14:20:31 +01:00
Code Issues Packages Projects Releases Wiki Activity

ggml : fix loongson compile warnings

Browse Source 
ggml-ci
This commit is contained in:
Georgi Gerganov 2024-05-25 20:24:12 +03:00
parent d5c05821f3
commit fd5de67bb7
No known key found for this signature in database
GPG Key ID: BF970631944C16B7
2 changed files with 13 additions and 6 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

6
ggml-quants.c
View File

@ -7404,6 +7404,9 @@ void ggml_vec_dot_q4_K_q8_K(int n, float * restrict s, size_t bs, const void * r
*s = vec_extract(vsumf0, 0); *s = vec_extract(vsumf0, 0);
#elif defined __loongarch_asx #elif defined __loongarch_asx
GGML_UNUSED(kmask1);
GGML_UNUSED(kmask2);
GGML_UNUSED(kmask3);
const __m256i m4 = __lasx_xvreplgr2vr_b(0xF); const __m256i m4 = __lasx_xvreplgr2vr_b(0xF);
@ -8002,6 +8005,9 @@ void ggml_vec_dot_q5_K_q8_K(int n, float * restrict s, size_t bs, const void * r
*s = vec_extract(vsumf0, 0); *s = vec_extract(vsumf0, 0);
#elif defined __loongarch_asx #elif defined __loongarch_asx
GGML_UNUSED(kmask1);
GGML_UNUSED(kmask2);
GGML_UNUSED(kmask3);
const __m256i m4 = __lasx_xvreplgr2vr_b(0xF); const __m256i m4 = __lasx_xvreplgr2vr_b(0xF);
const __m128i mzero = __lsx_vldi(0); const __m128i mzero = __lsx_vldi(0);

13
ggml.c
View File

@ -1576,11 +1576,11 @@ do { \
// F16 arithmetic is not supported by AVX, so we use F32 instead // F16 arithmetic is not supported by AVX, so we use F32 instead
#define GGML_F32Cx8 __m256 #define GGML_F32Cx8 __m256
#define GGML_F32Cx8_ZERO (__m256)__lasx_xvldi(0) #define GGML_F32Cx8_ZERO (__m256)__lasx_xvldi(0)
#define GGML_F32Cx8_SET1(x) (__m256)__lasx_xvreplgr2vr_w((x)) #define GGML_F32Cx8_SET1(x) (__m256)__lasx_xvreplgr2vr_w((x))
static inline __m256 __lasx_f32cx8_load(const ggml_fp16_t *x) { static inline __m256 __lasx_f32cx8_load(const ggml_fp16_t * x) {
float tmp[8]; float tmp[8];
for (int i = 0; i < 8; i++) { for (int i = 0; i < 8; i++) {
@ -1589,13 +1589,14 @@ static inline __m256 __lasx_f32cx8_load(const ggml_fp16_t *x) {
return (__m256)__lasx_xvld(tmp, 0); return (__m256)__lasx_xvld(tmp, 0);
} }
static inline void __lasx_f32cx8_store(ggml_fp16_t *x, __m256 y) { static inline void __lasx_f32cx8_store(ggml_fp16_t * x, __m256 y) {
float arr[8]; float arr[8];
__lasx_xvst(y, arr, 0); __lasx_xvst(y, arr, 0);
for (int i = 0; i < 8; i++) for (int i = 0; i < 8; i++) {
x[i] = GGML_FP32_TO_FP16(arr[i]); x[i] = GGML_FP32_TO_FP16(arr[i]);
}
} }
#define GGML_F32Cx8_LOAD(x) __lasx_f32cx8_load(x) #define GGML_F32Cx8_LOAD(x) __lasx_f32cx8_load(x)
#define GGML_F32Cx8_STORE(x, y) __lasx_f32cx8_store(x, y) #define GGML_F32Cx8_STORE(x, y) __lasx_f32cx8_store(x, y)
@ -1671,7 +1672,7 @@ static inline void __lasx_f32cx8_store(ggml_fp16_t *x, __m256 y) {
#define GGML_F16_STEP 32 #define GGML_F16_STEP 32
#define GGML_F16_EPR 4 #define GGML_F16_EPR 4
static inline __m128 __lsx_f16x4_load(ggml_fp16_t *x) { static inline __m128 __lsx_f16x4_load(const ggml_fp16_t * x) {
float tmp[4]; float tmp[4];
tmp[0] = GGML_FP16_TO_FP32(x[0]); tmp[0] = GGML_FP16_TO_FP32(x[0]);
@ -1682,7 +1683,7 @@ static inline __m128 __lsx_f16x4_load(ggml_fp16_t *x) {
return __lsx_vld(tmp, 0); return __lsx_vld(tmp, 0);
} }
static inline void __lsx_f16x4_store(ggml_fp16_t *x, __m128 y) { static inline void __lsx_f16x4_store(ggml_fp16_t * x, __m128 y) {
float arr[4]; float arr[4];
__lsx_vst(y, arr, 0); __lsx_vst(y, arr, 0);