ExternalVendorCode/whisper.cpp
1
0
Fork 0
mirror of https://github.com/ggerganov/whisper.cpp.git synced 2025-06-22 16:38:58 +00:00
Code Issues Actions Packages Projects Releases Wiki Activity

SYCL: Remove misleading ggml_sycl_op_flatten function (llama/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 Georgi Gerganov
parent 93631b2be6
commit 2e2f0f954b
12 changed files with 369 additions and 586 deletions
Download Patch File Download Diff File Expand all files Collapse all files

28
ggml/src/ggml-sycl/common.hpp
View File

@ -494,12 +494,6 @@ static __dpct_inline__ Tp* get_pointer(sycl::local_accessor<Tp, dim> acc) {
int64_t downsample_sycl_global_range(int64_t accumulate_block_num, int64_t block_size);
typedef void (*ggml_sycl_op_flatten_t)(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);
template<float (*bin_op)(const float, const float), typename src0_t, typename src1_t, typename dst_t>
static void k_bin_bcast(const src0_t * src0, const src1_t * src1, dst_t * dst,
int ne0, int ne1, int ne2, int ne3,
@ -757,24 +751,22 @@ struct bin_bcast_sycl {
template <class op>
inline void ggml_sycl_op_bin_bcast(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) {
const ggml_tensor *src1, ggml_tensor *dst) {
dpct::queue_ptr main_stream = ctx.stream();
if (src0->type == GGML_TYPE_F32 && dst->type == GGML_TYPE_F32) {
op()(ctx, src0, src1, dst, src0_dd, src1_dd, dst_dd, main_stream);
op()(ctx, src0, src1, dst, (const float *)src0->data, (const float *)src1->data, (float *)dst->data, main_stream);
} else if (src0->type == GGML_TYPE_F16 && dst->type == GGML_TYPE_F16) {
op()(ctx, src0, src1, dst, (const sycl::half *)src0_dd, src1_dd,
(sycl::half *)dst_dd, main_stream);
op()(ctx, src0, src1, dst, (const sycl::half *)src0->data, (const float *)src1->data,
(sycl::half *)dst->data, main_stream);
} else if (src0->type == GGML_TYPE_F16 && dst->type == GGML_TYPE_F32) {
op()(ctx, src0, src1, dst, (const sycl::half *)src0_dd, src1_dd, dst_dd,
op()(ctx, src0, src1, dst, (const sycl::half *)src0->data, (const float *)src1->data, (float *)dst->data,
main_stream);
} else if (src0->type == GGML_TYPE_I32 && dst->type == GGML_TYPE_I32) {
op()(ctx, src0, src1, dst, (const int32_t *)src0_dd, (const int32_t *)src1_dd, (int32_t *)dst_dd,
op()(ctx, src0, src1, dst, (const int32_t *)src0->data, (const int32_t *)src1->data, (int32_t *)dst->data,
main_stream);
} else if (src0->type == GGML_TYPE_I16 && dst->type == GGML_TYPE_I16) {
op()(ctx, src0, src1, dst, (const int16_t *)src0_dd, (const int16_t *)src1_dd, (int16_t *)dst_dd,
op()(ctx, src0, src1, dst, (const int16_t *)src0->data, (const int16_t *)src1->data, (int16_t *)dst->data,
main_stream);
} else {
fprintf(stderr, "%s: unsupported types: dst: %s, src0: %s, src1: %s\n", __func__,
@ -784,8 +776,4 @@ inline void ggml_sycl_op_bin_bcast(ggml_backend_sycl_context & ctx, const ggml_t
}
bool gpu_has_xmx(sycl::device &dev);
void ggml_sycl_op_flatten(ggml_backend_sycl_context & ctx, const ggml_tensor *src0,
const ggml_tensor *src1, ggml_tensor *dst,
const ggml_sycl_op_flatten_t op);
#endif // GGML_SYCL_COMMON_HPP