CS348Project/llama.cpp
5
0
Fork 0
mirror of https://github.com/ggml-org/llama.cpp.git synced 2025-10-31 08:51:55 +00:00
Code Issues Packages Projects Releases Wiki Activity

ggml : support GGML_TYPE_F32 ".from_float" trait

Browse Source
This commit is contained in:
Georgi Gerganov
2025-06-22 18:44:42 +03:00
parent 630c84a2bd
commit e89709721b
2 changed files with 6 additions and 0 deletions
Download Patch File Download Diff File Expand all files Collapse all files

1
ggml/include/ggml-cpu.h
View File

@@ -133,6 +133,7 @@ extern "C" {
GGML_BACKEND_API ggml_backend_reg_t ggml_backend_cpu_reg(void); GGML_BACKEND_API ggml_backend_reg_t ggml_backend_cpu_reg(void);
GGML_BACKEND_API void ggml_cpu_fp32_to_fp32(const float *, float *, int64_t);
GGML_BACKEND_API void ggml_cpu_fp32_to_fp16(const float *, ggml_fp16_t *, int64_t); GGML_BACKEND_API void ggml_cpu_fp32_to_fp16(const float *, ggml_fp16_t *, int64_t);
GGML_BACKEND_API void ggml_cpu_fp16_to_fp32(const ggml_fp16_t *, float *, int64_t); GGML_BACKEND_API void ggml_cpu_fp16_to_fp32(const ggml_fp16_t *, float *, int64_t);
GGML_BACKEND_API void ggml_cpu_fp32_to_bf16(const float *, ggml_bf16_t *, int64_t); GGML_BACKEND_API void ggml_cpu_fp32_to_bf16(const float *, ggml_bf16_t *, int64_t);

5
ggml/src/ggml-cpu/ggml-cpu.c
View File

@@ -192,6 +192,7 @@ typedef pthread_t ggml_thread_t;
static const struct ggml_type_traits_cpu type_traits_cpu[GGML_TYPE_COUNT] = { static const struct ggml_type_traits_cpu type_traits_cpu[GGML_TYPE_COUNT] = {
[GGML_TYPE_F32] = { [GGML_TYPE_F32] = {
.from_float = (ggml_from_float_t) ggml_cpu_fp32_to_fp32,
.vec_dot = (ggml_vec_dot_t) ggml_vec_dot_f32, .vec_dot = (ggml_vec_dot_t) ggml_vec_dot_f32,
.vec_dot_type = GGML_TYPE_F32, .vec_dot_type = GGML_TYPE_F32,
.nrows = 1, .nrows = 1,
@@ -3126,6 +3127,10 @@ enum ggml_status ggml_graph_compute_with_ctx(struct ggml_context * ctx, struct g
return ggml_graph_compute(cgraph, &cplan); return ggml_graph_compute(cgraph, &cplan);
} }
void ggml_cpu_fp32_to_fp32(const float * x, float * y, int64_t n) {
memcpy(y, x, n * sizeof(float));
}
void ggml_cpu_fp32_to_fp16(const float * x, ggml_fp16_t * y, int64_t n) { void ggml_cpu_fp32_to_fp16(const float * x, ggml_fp16_t * y, int64_t n) {
int64_t i = 0; int64_t i = 0;
#if defined(__F16C__) #if defined(__F16C__)