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

Remove unneccesary waits

Browse source
This commit is contained in:
Aidan 2024-02-15 13:48:33 +00:00
parent 88c272a622
commit c7f0639ae4
1 changed files with 0 additions and 2 deletions
Download patch file Download diff file Expand all files Collapse all files

2
ggml-sycl.cpp
View file

@ -1801,7 +1801,6 @@ namespace dpct
coeff + batch_size * 1 /*beta*/, reinterpret_cast<Tc **>(c), coeff + batch_size * 1 /*beta*/, reinterpret_cast<Tc **>(c),
dims + batch_size * 5 /*ldc*/, batch_size, dims + batch_size * 5 /*ldc*/, batch_size,
dims + batch_size * 6 /*group_size*/); dims + batch_size * 6 /*group_size*/);
q.wait_and_throw();
q.submit([&](sycl::handler &cgh) { q.submit([&](sycl::handler &cgh) {
cgh.depends_on(e); cgh.depends_on(e);
@ -13378,7 +13377,6 @@ static void ggml_sycl_mul_mat_batched_sycl(const ggml_tensor *src0,
src1_f16_alloc.alloc(ne_src1); src1_f16_alloc.alloc(ne_src1);
GGML_ASSERT(to_fp16_sycl != nullptr); GGML_ASSERT(to_fp16_sycl != nullptr);
to_fp16_sycl(src1_ddf, src1_f16_alloc.get(), ne_src1, main_stream); to_fp16_sycl(src1_ddf, src1_f16_alloc.get(), ne_src1, main_stream);
main_stream->wait_and_throw();
} }
sycl::half *src1_f16 = src1->type == GGML_TYPE_F16 ? (sycl::half *)src1_ddf sycl::half *src1_f16 = src1->type == GGML_TYPE_F16 ? (sycl::half *)src1_ddf
: src1_f16_alloc.get(); : src1_f16_alloc.get();