CS348Project/llama.cpp
5
0
Fork 0
mirror of https://github.com/ggml-org/llama.cpp.git synced 2025-11-16 11:27:03 +00:00
Code Issues Packages Projects Releases Wiki Activity

CUDA: FA support for Deepseek (Ampere or newer) (#13306)

Browse Source 
* CUDA: FA support for Deepseek (Ampere or newer)

* do loop unrolling via C++ template
This commit is contained in:
Johannes Gäßler
2025-05-09 13:34:58 +02:00
committed by GitHub
parent 27ebfcacba
commit 0cf6725e9f
33 changed files with 852 additions and 547 deletions
Download Patch File Download Diff File Expand all files Collapse all files

11
ggml/src/ggml-cuda/cp-async.cuh
View File

@@ -2,6 +2,17 @@
#include "common.cuh"
static __device__ __forceinline__ unsigned int ggml_cuda_cvta_generic_to_shared(void * generic_ptr) {
#ifdef CP_ASYNC_AVAILABLE
return __cvta_generic_to_shared(generic_ptr);
#else
GGML_UNUSED(generic_ptr);
NO_DEVICE_CODE;
return 0;
#endif // CP_ASYNC_AVAILABLE
}
// Copies data from global to shared memory, cg == cache global.
// Both the src and dst pointers must be aligned to 16 bit.
// Shared memory uses 32 bit addressing, the pointer is passed as unsigned int.