Lumpiasty/llama.cpp
1
0
Fork 0
Code Issues Pull Requests Packages Projects Releases Wiki Activity

ggml-cpu: use UE4M3 LUT in ARM NVFP4 dot product (#25331)

Browse Source
This commit is contained in:
ragz4125
2026-07-06 19:06:40 +08:00
committed by GitHub
parent 3b4fca11ac
commit 20a04b2206
2 changed files with 6 additions and 6 deletions
Download Patch File Download Diff File Expand all files Collapse all files
ggml/src/ggml-cpu/simd-mappings.h
+2 -2
View File
@@ -131,8 +131,8 @@ extern float ggml_table_f32_ue4m3[1 << 8];
#define GGML_CPU_E8M0_TO_FP32_HALF(x) GGML_E8M0_TO_FP32_HALF(x)
#endif
// Use lookup table for UE4M3 on x86 (faster than bit manipulation)
#if defined(__AVX__) || defined(__AVX2__) || defined(__AVX512F__)
// Use lookup table for UE4M3 on x86 and ARM (faster than bit manipulation)
#if defined(__AVX__) || defined(__AVX2__) || defined(__AVX512F__) || defined(__ARM_NEON)
#define GGML_CPU_UE4M3_TO_FP32(x) ggml_table_f32_ue4m3[(uint8_t)(x)]
#else
#define GGML_CPU_UE4M3_TO_FP32(x) ggml_ue4m3_to_fp32(x)