CS348Project/llama.cpp
5
0
Fork 0
mirror of https://github.com/ggml-org/llama.cpp.git synced 2025-10-30 08:42:00 +00:00
Code Issues Packages Projects Releases Wiki Activity

opencl: add backend_synchronize (#13939)

Browse Source 
* This is not needed by the normal use where the result is read
  using `tensor_get`, but it allows perf mode of `test-backend-ops`
  to properly measure performance.
This commit is contained in:
lhez
2025-06-02 16:54:58 -07:00
committed by GitHub
parent bfb1e012a0
commit 71e74a3ac9
1 changed files with 7 additions and 2 deletions
Download Patch File Download Diff File Expand all files Collapse all files

9
ggml/src/ggml-opencl/ggml-opencl.cpp
View File

@@ -2022,7 +2022,12 @@ static bool ggml_backend_opencl_cpy_tensor_async(ggml_backend_t backend, const g
}
static void ggml_backend_opencl_synchronize(ggml_backend_t backend) {
GGML_UNUSED(backend);
auto * backend_ctx = static_cast<ggml_backend_opencl_context *>(backend->context);
cl_event evt;
CL_CHECK(clEnqueueBarrierWithWaitList(backend_ctx->queue, 0, nullptr, &evt));
CL_CHECK(clWaitForEvents(1, &evt));
CL_CHECK(clReleaseEvent(evt));
}
// Syncronizes the 'backend_ctx's device with others so that commands
@@ -2225,7 +2230,7 @@ static ggml_backend_i ggml_backend_opencl_i = {
/* .set_tensor_async = */ NULL, /* ggml_backend_opencl_set_tensor_async */
/* .get_tensor_async = */ NULL, /* ggml_backend_opencl_get_tensor_async */
/* .cpy_tensor_async = */ NULL, /* ggml_backend_opencl_cpy_tensor_async */
/* .synchronize = */ NULL, /* ggml_backend_opencl_synchronize */
/* .synchronize = */ ggml_backend_opencl_synchronize,
/* .graph_plan_create = */ NULL,
/* .graph_plan_free = */ NULL,
/* .graph_plan_update = */ NULL,