Lumpiasty/llama.cpp
1
0
Fork 0
Code Issues Pull Requests Packages Projects Releases Wiki Activity

SYCL: fix use-after-free bug with async memcpy in MoE prefill (#24676)

Browse Source 
* SYCL: fix a bug with async memcpy

* make mmid_row_mapping_host persistent

* comment on stream->wait

* Apply suggestion from @sanmai

* Apply suggestion from @sanmai

* Apply suggestion from @sanmai
This commit is contained in:
Alexey Kopytko
2026-06-17 08:57:29 +03:00
committed by GitHub
parent 9b260fc9ef
commit ebbc1e51c1
2 changed files with 10 additions and 6 deletions
Download Patch File Download Diff File Expand all files Collapse all files
ggml/src/ggml-sycl/common.hpp
+7
View File
@@ -324,6 +324,11 @@ void ggml_sycl_free_device(void *ptr, sycl::queue &q);
void release_extra_gpu(ggml_tensor_extra_gpu * extra, std::vector<queue_ptr> streams={});
struct mmid_row_mapping {
int32_t i1;
int32_t i2;
};
namespace sycl_ex = sycl::ext::oneapi::experimental;
struct ggml_backend_sycl_context {
int device;
@@ -421,6 +426,8 @@ struct ggml_backend_sycl_context {
std::unique_ptr<ggml_sycl_pool> host_pools[GGML_SYCL_MAX_DEVICES];
std::vector<mmid_row_mapping> mmid_row_mapping_host;
static std::unique_ptr<ggml_sycl_pool> new_pool_for_device(queue_ptr qptr, int device);
static std::unique_ptr<ggml_sycl_pool> new_pool_for_host(queue_ptr qptr, int device);