happyz
/
llama.cpp
mirror of https://github.com/ggerganov/llama.cpp.git
1
0
Fork 0
Code Issues Packages Projects Releases Wiki Activity

fix gramma and empty spaces

This commit is contained in:
zhang hui 2025-12-16 21:27:14 +08:00
parent cffa070b08
commit cad07fa4b5
1 changed files with 5 additions and 5 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

2
ggml/src/ggml-cuda/mma.cuh
View File

@ -612,7 +612,7 @@ namespace ggml_cuda_mma {
ggml_cuda_memcpy_1<sizeof(t.x)>(t.x, xs0 + t.get_i(0) * stride + t.get_j(0)); ggml_cuda_memcpy_1<sizeof(t.x)>(t.x, xs0 + t.get_i(0) * stride + t.get_j(0));
} }
#elif defined(AMD_WMMA_AVAILABLE) #elif defined(AMD_WMMA_AVAILABLE)
// All wmma layout has continues data when i-major. // All wmma layout has contiguous data when i-major.
if constexpr (is_i_major(dl)) { if constexpr (is_i_major(dl)) {
// the data must be aligned to 16 bytes when bigger than ggml_cuda_get_max_cpy_bytes() // the data must be aligned to 16 bytes when bigger than ggml_cuda_get_max_cpy_bytes()
constexpr int aligned_copy_bytes = ggml_cuda_get_max_cpy_bytes(); constexpr int aligned_copy_bytes = ggml_cuda_get_max_cpy_bytes();