CS348Project/llama.cpp
5
0
Fork 0
mirror of https://github.com/ggml-org/llama.cpp.git synced 2025-11-04 09:32:00 +00:00
Code Issues Packages Projects Releases Wiki Activity

ggml : remove OpenCL (#7735)

Browse Source 
ggml-ci
This commit is contained in:
Georgi Gerganov
2024-06-04 21:23:20 +03:00
committed by GitHub
parent 0cd6bd3483
commit 554c247caf
21 changed files with 29 additions and 2639 deletions
Download Patch File Download Diff File Expand all files Collapse all files

10
llama.cpp
View File

@@ -13,8 +13,6 @@
#ifdef GGML_USE_CUDA
# include "ggml-cuda.h"
#elif defined(GGML_USE_CLBLAST)
# include "ggml-opencl.h"
#elif defined(GGML_USE_VULKAN)
# include "ggml-vulkan.h"
#elif defined(GGML_USE_SYCL)
@@ -2406,8 +2404,6 @@ static ggml_backend_buffer_type_t llama_default_buffer_type_offload(const llama_
buft = ggml_backend_vk_buffer_type(gpu);
#elif defined(GGML_USE_SYCL)
buft = ggml_backend_sycl_buffer_type(gpu);
#elif defined(GGML_USE_CLBLAST)
buft = ggml_backend_opencl_buffer_type();
#elif defined(GGML_USE_KOMPUTE)
buft = ggml_backend_kompute_buffer_type(gpu);
if (buft == nullptr) {
@@ -2530,10 +2526,6 @@ static bool llama_kv_cache_init(
}
}
#ifdef GGML_USE_CLBLAST
offload = false;
#endif
// count used buffer types
std::map<ggml_backend_buffer_type_t, int> buft_layer_count;
if (offload) {
@@ -15921,7 +15913,7 @@ bool llama_supports_mlock(void) {
}
bool llama_supports_gpu_offload(void) {
#if defined(GGML_USE_CUDA) || defined(GGML_USE_CLBLAST) || defined(GGML_USE_METAL) || defined(GGML_USE_VULKAN) || \
#if defined(GGML_USE_CUDA) || defined(GGML_USE_METAL) || defined(GGML_USE_VULKAN) || \
defined(GGML_USE_SYCL) || defined(GGML_USE_KOMPUTE) || defined(GGML_USE_RPC)
// Defined when llama.cpp is compiled with support for offloading model layers to GPU.
return true;