tqcq/llama.cpp
0
0
Fork 0
mirror of https://github.com/ggml-org/llama.cpp.git synced 2025-06-30 12:55:17 +00:00
Code Issues Packages Projects Releases Wiki Activity

ggml-cpu: add debugging prints to see if dlf16 is correct

Browse Source 
Signed-off-by: Aaron Teo <aaron.teo1@ibm.com>
This commit is contained in:
Aaron Teo
2025-06-21 14:50:33 +08:00
parent 0394a006c5
commit 48b820d05f
1 changed files with 5 additions and 1 deletions
Download Patch File Download Diff File Expand all files Collapse all files

6
ggml/src/ggml-cpu/simd-mappings.h
View File

@ -1,6 +1,7 @@
#pragma once
#include "ggml-cpu-impl.h"
#include <signal.h>
//
// simd mappings
@ -964,9 +965,12 @@ static inline void __lsx_f16x4_store(ggml_fp16_t * x, __m128 y) {
static inline float32x4_t __lzs_f16cx4_load(const ggml_fp16_t * x) {
#ifdef __NNPA__
print("%s: %8d %8d %8d %8d\n", __func__, x[0], x[1], x[2], x[3]);
uint16x8_t v_x = vec_xl(0, (const ggml_fp16_t *)x);
uint16x8_t nnpa_dlf16 = vec_convert_from_fp16(v_x, 0);
return vec_extend_to_fp32_hi(nnpa_dlf16, 0);
float32x4_t result = vec_extend_to_fp32_hi(nnpa_dlf16, 0);
print("%s: %8d %8d %8d %8d\n", __func__, result[0], result[1], result[2], result[3]);
return result;
#else
float tmp[4];