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

SYCL: implement memset ggml backend buffer interface (#12580)

Browse Source 
* SYCL: implement memset ggml backend buffer interface

* use GGML_ABORT macro

* Do not wait for all queues to finish for memset operation
This commit is contained in:
Akarshan Biswas
2025-03-27 07:16:00 +05:30
committed by GitHub
parent bd40678df7
commit f17a3bb4e8
1 changed files with 19 additions and 1 deletions
Download Patch File Download Diff File Expand all files Collapse all files

20
ggml/src/ggml-sycl/ggml-sycl.cpp
View File

@@ -37,6 +37,7 @@
#include "ggml-backend-impl.h" #include "ggml-backend-impl.h"
#include "ggml-sycl/backend.hpp" #include "ggml-sycl/backend.hpp"
#include "ggml-sycl/common.hpp"
#include "ggml-sycl/presets.hpp" #include "ggml-sycl/presets.hpp"
#include "ggml-sycl/gemm.hpp" #include "ggml-sycl/gemm.hpp"
#include "ggml-sycl/sycl_hw.hpp" #include "ggml-sycl/sycl_hw.hpp"
@@ -490,6 +491,23 @@ catch (sycl::exception const &exc) {
std::exit(1); std::exit(1);
} }
static void ggml_backend_sycl_buffer_memset_tensor(ggml_backend_buffer_t buffer, ggml_tensor * tensor, uint8_t value,
size_t offset, size_t size) {
GGML_SYCL_DEBUG(" [SYCL] call %s\n", __func__);
ggml_backend_sycl_buffer_context * ctx = (ggml_backend_sycl_buffer_context *) buffer->context;
SYCL_CHECK(ggml_sycl_set_device(ctx->device));
auto stream = &(dpct::dev_mgr::instance().get_device(ctx->device).default_queue());
if (size == 0) {
return; // Nothing to do
}
if (tensor->data == nullptr) {
GGML_ABORT("Error: Tensor data pointer is null.\n");
}
void * target_ptr = static_cast<char *>(tensor->data) + offset;
SYCL_CHECK(CHECK_TRY_ERROR((*stream).memset(target_ptr, value, size)));
SYCL_CHECK(CHECK_TRY_ERROR((*stream).wait()));
}
static void ggml_backend_sycl_buffer_reset(ggml_backend_buffer_t buffer) { static void ggml_backend_sycl_buffer_reset(ggml_backend_buffer_t buffer) {
GGML_SYCL_DEBUG("[SYCL] call %s\n", __func__); GGML_SYCL_DEBUG("[SYCL] call %s\n", __func__);
if (buffer == nullptr) { if (buffer == nullptr) {
@@ -510,7 +528,7 @@ static const ggml_backend_buffer_i ggml_backend_sycl_buffer_interface = {
/* .free_buffer = */ ggml_backend_sycl_buffer_free_buffer, /* .free_buffer = */ ggml_backend_sycl_buffer_free_buffer,
/* .get_base = */ ggml_backend_sycl_buffer_get_base, /* .get_base = */ ggml_backend_sycl_buffer_get_base,
/* .init_tensor = */ ggml_backend_sycl_buffer_init_tensor, /* .init_tensor = */ ggml_backend_sycl_buffer_init_tensor,
/* .memset_tensor = */ NULL, /* .memset_tensor = */ ggml_backend_sycl_buffer_memset_tensor,
/* .set_tensor = */ ggml_backend_sycl_buffer_set_tensor, /* .set_tensor = */ ggml_backend_sycl_buffer_set_tensor,
/* .get_tensor = */ ggml_backend_sycl_buffer_get_tensor, /* .get_tensor = */ ggml_backend_sycl_buffer_get_tensor,
/* .cpy_tensor = */ ggml_backend_sycl_buffer_cpy_tensor, /* .cpy_tensor = */ ggml_backend_sycl_buffer_cpy_tensor,