CS348Project/llama.cpp
5
0
Fork 0
mirror of https://github.com/ggml-org/llama.cpp.git synced 2025-11-17 11:37:10 +00:00
Code Issues Packages Projects Releases Wiki Activity

SYCL: Remove misleading ggml_sycl_op_flatten function (#12387)

Browse Source 
* SYCL: Remove misleading ggml_sycl_op_flatten function

* remove trailing whitespace

* Fix L2 norm from rebase

* remove try catch block from element_wise.cpp

* remove comment from common.hp

* ggml-sycl.cpp: Add try catch sycl::exception block in compute_forward

* norm.cpp: remove try catch exception block
This commit is contained in:
Akarshan Biswas
2025-03-31 14:55:24 +05:30
committed by GitHub
parent f52d59d771
commit 6c02a032fa
12 changed files with 369 additions and 586 deletions
Download Patch File Download Diff File Expand all files Collapse all files

82
ggml/src/ggml-sycl/norm.cpp
View File

@@ -397,90 +397,78 @@ static void l2_norm_f32_sycl(const float* x, float* dst, const int ncols,
}
}
void ggml_sycl_op_norm(ggml_backend_sycl_context& ctx, const ggml_tensor* src0, const ggml_tensor* src1,
ggml_tensor* dst, const float* src0_dd,
const float* src1_dd, float* dst_dd,
const queue_ptr& main_stream) {
void ggml_sycl_op_norm(ggml_backend_sycl_context& ctx, ggml_tensor* dst) {
GGML_ASSERT(src0->type == GGML_TYPE_F32);
GGML_ASSERT(dst->src[0]->type == GGML_TYPE_F32);
GGML_ASSERT(dst->type == GGML_TYPE_F32);
const int64_t ne00 = src0->ne[0];
const int64_t nrows = ggml_nrows(src0);
const int64_t ne00 = dst->src[0]->ne[0];
const int64_t nrows = ggml_nrows(dst->src[0]);
dpct::queue_ptr main_stream = ctx.stream();
SYCL_CHECK(ggml_sycl_set_device(ctx.device));
const float * src0_dd = static_cast<const float *>(dst->src[0]->data);
float * dst_dd = static_cast<float *>(dst->data);
float eps;
memcpy(&eps, dst->op_params, sizeof(float));
norm_f32_sycl(src0_dd, dst_dd, ne00, nrows, eps, main_stream, ctx.device);
(void)src1;
(void)dst;
(void)src1_dd;
}
void ggml_sycl_op_group_norm(ggml_backend_sycl_context& ctx, const ggml_tensor* src0,
const ggml_tensor* src1, ggml_tensor* dst,
const float* src0_dd, const float* src1_dd,
float* dst_dd,
const queue_ptr& main_stream) {
void ggml_sycl_op_group_norm(ggml_backend_sycl_context& ctx, ggml_tensor* dst) {
GGML_ASSERT(src0->type == GGML_TYPE_F32);
GGML_ASSERT(dst->src[0]->type == GGML_TYPE_F32);
GGML_ASSERT(dst->type == GGML_TYPE_F32);
int num_groups = dst->op_params[0];
dpct::queue_ptr main_stream = ctx.stream();
SYCL_CHECK(ggml_sycl_set_device(ctx.device));
const float * src0_dd = static_cast<const float *>(dst->src[0]->data);
float * dst_dd = static_cast<float *>(dst->data);
float eps;
memcpy(&eps, dst->op_params + 1, sizeof(float));
int group_size = src0->ne[0] * src0->ne[1] * ((src0->ne[2] + num_groups - 1) / num_groups);
group_norm_f32_sycl(src0_dd, dst_dd, num_groups, eps, group_size, src0->ne[0] * src0->ne[1] * src0->ne[2], main_stream, ctx.device);
(void)src1;
(void)dst;
(void)src1_dd;
GGML_UNUSED(ctx);
int group_size = dst->src[0]->ne[0] * dst->src[0]->ne[1] * ((dst->src[0]->ne[2] + num_groups - 1) / num_groups);
group_norm_f32_sycl(src0_dd, dst_dd, num_groups, eps, group_size, dst->src[0]->ne[0] * dst->src[0]->ne[1] * dst->src[0]->ne[2], main_stream, ctx.device);
}
void ggml_sycl_op_rms_norm(ggml_backend_sycl_context& ctx, const ggml_tensor* src0,
const ggml_tensor* src1, ggml_tensor* dst,
const float* src0_dd, const float* src1_dd,
float* dst_dd,
const queue_ptr& main_stream) {
void ggml_sycl_op_rms_norm(ggml_backend_sycl_context & ctx, ggml_tensor * dst) {
GGML_ASSERT(src0->type == GGML_TYPE_F32);
GGML_ASSERT(dst->src[0]->type == GGML_TYPE_F32);
GGML_ASSERT(dst->type == GGML_TYPE_F32);
const int64_t ne00 = src0->ne[0];
const int64_t nrows = ggml_nrows(src0);
const int64_t ne00 = dst->src[0]->ne[0];
const int64_t nrows = ggml_nrows(dst->src[0]);
dpct::queue_ptr main_stream = ctx.stream();
SYCL_CHECK(ggml_sycl_set_device(ctx.device));
const float * src0_dd = static_cast<const float *>(dst->src[0]->data);
float * dst_dd = static_cast<float *>(dst->data);
float eps;
memcpy(&eps, dst->op_params, sizeof(float));
rms_norm_f32_sycl(src0_dd, dst_dd, ne00, nrows, eps, main_stream, ctx.device);
(void)src1;
(void)dst;
(void)src1_dd;
}
void ggml_sycl_op_l2_norm(ggml_backend_sycl_context& ctx, const ggml_tensor* src0,
const ggml_tensor* src1, ggml_tensor* dst,
const float* src0_dd, const float* src1_dd,
float* dst_dd,
const queue_ptr& main_stream) {
void ggml_sycl_op_l2_norm(ggml_backend_sycl_context& ctx, ggml_tensor* dst) {
GGML_ASSERT(src0->type == GGML_TYPE_F32);
GGML_ASSERT(dst->src[0]->type == GGML_TYPE_F32);
GGML_ASSERT(dst->type == GGML_TYPE_F32);
const int64_t ne00 = src0->ne[0];
const int64_t nrows = ggml_nrows(src0);
dpct::queue_ptr main_stream = ctx.stream();
SYCL_CHECK(ggml_sycl_set_device(ctx.device));
const int64_t ne00 = dst->src[0]->ne[0];
const int64_t nrows = ggml_nrows(dst->src[0]);
const float * src0_dd = static_cast<const float *>(dst->src[0]->data);
float * dst_dd = static_cast<float *>(dst->data);
float eps;
memcpy(&eps, dst->op_params, sizeof(float));
l2_norm_f32_sycl(src0_dd, dst_dd, ne00, nrows, eps, main_stream, ctx.device);
(void)src1;
(void)dst;
(void)src1_dd;
}