ver4a/llama.cpp
1
0
Fork 0
Code Issues Pull requests Projects Releases Packages Wiki Activity

sycl: cleanup oneDNN related code (#12097)

Browse source
This commit is contained in:
Svetlozar Georgiev 2025-03-21 02:15:56 +00:00 • committed by GitHub
parent e04643063b
commit 9ffcc9e374
No known key found for this signature in database
GPG key ID: B5690EEEBB952194
5 changed files with 90 additions and 66 deletions
Download patch file Download diff file Expand all files Collapse all files

12
ggml/src/ggml-sycl/ggml-sycl.cpp
View file

@ -2058,9 +2058,9 @@ inline void ggml_sycl_op_mul_mat_sycl(
const to_fp32_sycl_t to_fp32_sycl = ggml_get_to_fp32_sycl(GGML_TYPE_F16, dst);
to_fp32_sycl(dst_f16.get(), dst_dd_i, row_diff*src1_ncols, stream);
#else
auto dnnl_stream = ctx.stream_dnnl(stream);
DnnlGemmWrapper::row_gemm(dnnl_stream, false, true, src1_ncols, row_diff, ne10, src1_ptr, DnnlGemmWrapper::to_dt<sycl::half>(),
src0_ptr, DnnlGemmWrapper::to_dt<sycl::half>(), dst_f16.get(), DnnlGemmWrapper::to_dt<sycl::half>());
DnnlGemmWrapper::row_gemm(ctx, false, true, src1_ncols, row_diff, ne10, src1_ptr,
DnnlGemmWrapper::to_dt<sycl::half>(), src0_ptr, DnnlGemmWrapper::to_dt<sycl::half>(),
dst_f16.get(), DnnlGemmWrapper::to_dt<sycl::half>(), stream);
const to_fp32_sycl_t to_fp32_sycl = ggml_get_to_fp32_sycl(GGML_TYPE_F16, dst);
to_fp32_sycl(dst_f16.get(), dst_dd_i, row_diff* src1_ncols, stream);
#endif
@ -2099,9 +2099,9 @@ inline void ggml_sycl_op_mul_mat_sycl(
dst_dd_i, ldc)));
# endif
#else
auto dnnl_stream = ctx.stream_dnnl(stream);
DnnlGemmWrapper::row_gemm(dnnl_stream, false, true, src1_ncols, row_diff, ne10, src1_ddf1_i, DnnlGemmWrapper::to_dt<float>(),
src0_ddf_i, DnnlGemmWrapper::to_dt<float>(), dst_dd_i, DnnlGemmWrapper::to_dt<float>());
DnnlGemmWrapper::row_gemm(ctx, false, true, src1_ncols, row_diff, ne10, src1_ddf1_i,
DnnlGemmWrapper::to_dt<float>(), src0_ddf_i, DnnlGemmWrapper::to_dt<float>(),
dst_dd_i, DnnlGemmWrapper::to_dt<float>(), stream);
#endif
}
GGML_UNUSED(dst);