ver4a/llama.cpp
1
0
Fork 0
Code Issues Pull requests Projects Releases Packages Wiki Activity

CUDA: rename macros to avoid conflicts with WinAPI (#10736)

Browse source 
* Renames NVIDIA GPU-architecture flags to avoid name clashes with WinAPI. (e.g. CC_PASCAL, GPU architecture or WinAPI pascal compiler flag?)

* Reverts erroneous rename in SYCL-code.

* Renames GGML_CUDA_MIN_CC_DP4A to GGML_CUDA_CC_DP4A.

* Renames the rest of the compute capability macros for consistency.
This commit is contained in:
Andreas Kieslinger 2024-12-10 18:23:24 +01:00 • committed by GitHub
parent a86ad841f1
commit 750cb3e246
No known key found for this signature in database
GPG key ID: B5690EEEBB952194
10 changed files with 69 additions and 71 deletions
Download patch file Download diff file Expand all files Collapse all files

2
ggml/src/ggml-cuda/fattn.cu
View file

@ -304,7 +304,7 @@ void ggml_cuda_flash_attn_ext(ggml_backend_cuda_context & ctx, ggml_tensor * dst
const enum ggml_prec prec = ggml_flash_attn_ext_get_prec(KQV);
// On AMD the tile kernels perform poorly, use the vec kernel instead:
if (cc >= CC_OFFSET_AMD) {
if (cc >= GGML_CUDA_CC_OFFSET_AMD) {
if (prec == GGML_PREC_DEFAULT && fast_fp16_available(cc)) {
ggml_cuda_flash_attn_ext_vec_f16(ctx, dst);
} else {