mirrors/llama.cpp
1
0
Fork 0
mirror of https://github.com/ggerganov/llama.cpp.git synced 2025-04-20 21:46:07 +00:00
Code Issues Packages Projects Releases Wiki Activity

ggml: don't include arm_neon.h when using CUDA 12 with ARM Neon (ggml/1187)

Browse Source 
fix
This commit is contained in:
cmdr2 2025-04-10 17:53:08 +05:30 committed by Georgi Gerganov
parent fe92821ea9
commit cb79c2e7fa
1 changed files with 8 additions and 7 deletions

15
ggml/src/ggml-impl.h

@ -16,6 +16,14 @@
#include <arm_sve.h>
#endif // __ARM_FEATURE_SVE
#if defined(__ARM_NEON) && !defined(__CUDACC__) && !defined(__MUSACC__)
// if YCM cannot find <arm_neon.h>, make a symbolic link to it, for example:
//
// $ ln -sfn /Library/Developer/CommandLineTools/usr/lib/clang/13.1.6/include/arm_neon.h ./src/
//
#include <arm_neon.h>
#endif
#if defined(__F16C__)
#include <immintrin.h>
#endif
@ -317,13 +325,6 @@ GGML_API void ggml_aligned_free(void * ptr, size_t size);
// for MUSA compilers , we use uint16_t: ref https://github.com/ggml-org/llama.cpp/pull/11843
//
#if defined(__ARM_NEON) && !(defined(__CUDACC__) && __CUDACC_VER_MAJOR__ <= 11) && !defined(__MUSACC__)
// if YCM cannot find <arm_neon.h>, make a symbolic link to it, for example:
//
// $ ln -sfn /Library/Developer/CommandLineTools/usr/lib/clang/13.1.6/include/arm_neon.h ./src/
//
#include <arm_neon.h>
#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)