tqcq/llama.cpp
0
0
Fork 0
mirror of https://github.com/ggml-org/llama.cpp.git synced 2025-08-13 11:57:43 -04:00
Code Issues Packages Projects Releases Wiki Activity

ggml-cpu: add ggml fp16->fp32 and fp32->fp16 scalar simd

Browse Source 
Signed-off-by: Aaron Teo <aaron.teo1@ibm.com>
This commit is contained in:
Aaron Teo
2025-06-20 19:54:38 +08:00
parent 2f58bbcbb8
commit ae9c5f928a
2 changed files with 26 additions and 2 deletions
Download Patch File Download Diff File Expand all files Collapse all files

4
ggml/src/ggml-cpu/simd-mappings.h
View File

@@ -982,8 +982,8 @@ static inline float32x4_t __lzs_f16cx4_load(const ggml_fp16_t * x) {
static inline void __lzs_f16cx4_store(ggml_fp16_t * x, float32x4_t v_y) {
#ifdef __NNPA__
float32x4_t zero = vec_splats(0.0f);
uint16x8_t v_x = vec_round_from_fp32(v_y, zero, 0);
float32x4_t v_zero = vec_splats(0.0f);
uint16x8_t v_x = vec_round_from_fp32(v_y, v_zero, 0);
x[0] = vec_extract(v_x, 0);
x[1] = vec_extract(v_x, 1);
x[2] = vec_extract(v_x, 2);

24
ggml/src/ggml-impl.h
View File

@@ -322,6 +322,7 @@ GGML_API void ggml_aligned_free(void * ptr, size_t size);
// 16-bit float
// on Arm, we use __fp16
// on x86, we use uint16_t
// on s390x, we use ZDNN_DLFLOAT16 with NNPA
//
// for old CUDA compilers (<= 11), we use uint16_t: ref https://github.com/ggml-org/llama.cpp/pull/10616
// for MUSA compilers , we use uint16_t: ref https://github.com/ggml-org/llama.cpp/pull/11843
@@ -417,6 +418,29 @@ GGML_API void ggml_aligned_free(void * ptr, size_t size);
#define GGML_FP16_TO_FP32(x) GGML_COMPUTE_FP16_TO_FP32(x)
#define GGML_FP32_TO_FP16(x) GGML_COMPUTE_FP32_TO_FP16(x)
#elif defined(__NNPA__)
#define GGML_COMPUTE_FP16_TO_FP32(x) ggml_compute_fp16_to_fp32(x)
#define GGML_COMPUTE_FP32_TO_FP16(x) ggml_compute_fp32_to_fp16(x)
#define GGML_FP16_TO_FP32(x) GGML_COMPUTE_FP16_TO_FP32(x)
#define GGML_FP32_TO_FP16(x) GGML_COMPUTE_FP32_TO_FP16(x)
// TODO: Determine if inline assembly is faster
static inline float ggml_compute_fp16_to_fp32(ggml_fp16_t h) {
uint16x8_t v_h = vec_splats(h);
uint16x8_t nnpa_dlf16 = vec_convert_from_fp16(v_h, 0);
return vec_extend_to_fp32_hi(nnpa_dlf16, 0)[0];
}
// TODO: Determine if inline assembly is faster
static inline ggml_fp16_t ggml_compute_fp32_to_fp16(float f) {
float32x4_t v_f = vec_splats(f);
float32x4_t v_zero = vec_splats(0.0f);
uint16x8_t v_h = vec_round_from_fp32(v_f, v_zero, 0);
return vec_extract(v_h, 0);
}
#else
// FP16 <-> FP32