tqcq/llama.cpp
0
0
Fork 0
mirror of https://github.com/ggml-org/llama.cpp.git synced 2025-07-27 11:43:35 -04:00
Code Issues Packages Projects Releases Wiki Activity

ggml-cpu: clarified vector naming

Browse Source 
Signed-off-by: Aaron Teo <aaron.teo1@ibm.com>
This commit is contained in:
Aaron Teo
2025-06-21 17:39:35 +08:00
parent 4621a23c14
commit 987d1690e4
1 changed files with 12 additions and 13 deletions
Download Patch File Download Diff File Expand all files Collapse all files

25
ggml/src/ggml-cpu/ggml-cpu.c
View File

@ -3145,15 +3145,15 @@ void ggml_cpu_fp32_to_fp16(const float * x, ggml_fp16_t * y, int64_t n) {
for (; i + 7 < n; i += 8) { for (; i + 7 < n; i += 8) {
float32x4_t v_xh = vec_xl(0, (const float *)(x + i + 0)); float32x4_t v_xh = vec_xl(0, (const float *)(x + i + 0));
float32x4_t v_xl = vec_xl(0, (const float *)(x + i + 4)); float32x4_t v_xl = vec_xl(0, (const float *)(x + i + 4));
uint16x8_t v_xd = vec_round_from_fp32(v_xh, v_xl, 0); uint16x8_t v_yd = vec_round_from_fp32(v_xh, v_xl, 0);
uint16x8_t v_y = vec_convert_to_fp16(v_xd, 0); uint16x8_t v_y = vec_convert_to_fp16(v_yd, 0);
vec_xst(v_y, 0, (ggml_fp16_t *)(y + i)); vec_xst(v_y, 0, (ggml_fp16_t *)(y + i));
} }
for (; i + 3 < n; i += 4) { for (; i + 3 < n; i += 4) {
float32x4_t v_x = vec_xl(0, (const float *)(x + i)); float32x4_t v_x = vec_xl(0, (const float *)(x + i));
float32x4_t v_zero = vec_splats(0.0f); float32x4_t v_zero = vec_splats(0.0f);
uint16x8_t v_xd = vec_round_from_fp32(v_x, v_zero, 0); uint16x8_t v_yd = vec_round_from_fp32(v_x, v_zero, 0);
uint16x8_t v_y = vec_convert_to_fp16(v_xd, 0); uint16x8_t v_y = vec_convert_to_fp16(v_yd, 0);
vec_xst(v_y, 0, (ggml_fp16_t *)(y + i)); vec_xst(v_y, 0, (ggml_fp16_t *)(y + i));
} }
#endif #endif
@ -3185,18 +3185,17 @@ void ggml_cpu_fp16_to_fp32(const ggml_fp16_t * x, float * y, int64_t n) {
#elif defined(__NNPA__) #elif defined(__NNPA__)
for (; i + 7 < n; i += 8) { for (; i + 7 < n; i += 8) {
uint16x8_t v_x = vec_xl(0, (const ggml_fp16_t *)(x + i)); uint16x8_t v_x = vec_xl(0, (const ggml_fp16_t *)(x + i));
uint16x8_t v_xd = vec_convert_from_fp16(v_x, 0); uint16x8_t v_yd = vec_convert_from_fp16(v_x, 0);
float32x4_t v_xdh = vec_extend_to_fp32_hi(v_xd, 0); float32x4_t v_yh = vec_extend_to_fp32_hi(v_yd, 0);
float32x4_t v_xdl = vec_extend_to_fp32_lo(v_xd, 0); float32x4_t v_yl = vec_extend_to_fp32_lo(v_yd, 0);
vec_xst(v_xdh, 0, (float *)(y + i + 0)); vec_xst(v_yh, 0, (float *)(y + i + 0));
vec_xst(v_xdl, 0, (float *)(y + i + 4)); vec_xst(v_yl, 0, (float *)(y + i + 4));
} }
for (; i + 3 < n; i += 4) { for (; i + 3 < n; i += 4) {
uint16x8_t v_x = vec_xl(0, (const ggml_fp16_t *)(x + i)); uint16x8_t v_x = vec_xl(0, (const ggml_fp16_t *)(x + i));
uint16x8_t v_xd = vec_convert_from_fp16(v_x, 0); uint16x8_t v_yd = vec_convert_from_fp16(v_x, 0);
float32x4_t v_xdh = vec_extend_to_fp32_hi(v_xd, 0); float32x4_t v_yh = vec_extend_to_fp32_hi(v_yd, 0);
vec_xst(v_xdh, 0, (float *)(y + i)); vec_xst(v_yh, 0, (float *)(y + i));
} }
#endif #endif