Skip to content
New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Revert "[GPU] Enable fusing dynamic eltwise with gemm in onednn case" #29756

Merged
merged 1 commit into from
Mar 27, 2025
Merged

Revert "[GPU] Enable fusing dynamic eltwise with gemm in onednn case" #29756

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
1 commit
Select commit
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
3 changes: 1 addition & 2 deletions src/plugins/intel_gpu/src/graph/graph_optimizer/prepare_primitive_fusing.cpp
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -1142,8 +1142,7 @@ void prepare_primitive_fusing::fuse_simple_primitives(program &p) {
auto eltw_in_size = peer_node->get_output_layout();
if (eltw_in_size.is_dynamic()
// this whitelist condition is temporarily and to be relaxed soon.
&& !fused_node->is_type<fully_connected>()
&& !fused_node->is_type<gemm>())
&& !fused_node->is_type<fully_connected>())
return;
}
if (parent1.first->is_type<convolution>() && !conv_supports_fusings(parent1.first->as<convolution>()))
Expand Down
16 changes: 13 additions & 3 deletions src/plugins/intel_gpu/tests/unit/fusions/gemm_fusion_test.cpp
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -482,6 +482,13 @@ class gemm_2in_dynamic_add : public gemm_2in_add {};
TEST_P(gemm_2in_dynamic_add, add) {
auto p = GetParam();

if (engine.get_device_info().supports_immad) {
p.expected_fused_primitives++;
if (!p.kernel_name.empty()) {
p.expected_not_fused_primitives++;
}
}

cfg_fused.set_property(ov::intel_gpu::allow_new_shape_infer(true));
cfg_not_fused.set_property(ov::intel_gpu::allow_new_shape_infer(true));

Expand Down Expand Up @@ -518,10 +525,13 @@ TEST_P(gemm_2in_dynamic_add, add) {
}

INSTANTIATE_TEST_SUITE_P(fusings_gpu, gemm_2in_dynamic_add, ::testing::ValuesIn(std::vector<gemm_test_params>{
gemm_test_params{ CASE_GEMM_2IN_FP16_3D_1, 4, 4, "gemm_tiled_opt", broadcast_kinds::batch, eltwise_mode::sum },
gemm_test_params{ CASE_GEMM_2IN_FP16_3D_1, 4, 5, "", broadcast_kinds::batch, eltwise_mode::sum },
gemm_test_params{ CASE_GEMM_2IN_FP16_3D_1, 4, 5, "", broadcast_kinds::feature, eltwise_mode::sum },
gemm_test_params{ CASE_GEMM_2IN_FP16_3D_2, 4, 5, "", broadcast_kinds::feature, eltwise_mode::sum },
gemm_test_params{ CASE_GEMM_2IN_FP16_3D_2, 4, 5, "", broadcast_kinds::batch, eltwise_mode::sum },
// Reference graph can be fused because force_implementation leads to optimize_data = true;
gemm_test_params{ CASE_GEMM_2IN_FP16_3D_1, 4, 4, "gemm_tiled_opt", broadcast_kinds::feature, eltwise_mode::sum },
gemm_test_params{ CASE_GEMM_2IN_FP16_3D_2, 4, 4, "gemm_tiled_opt", broadcast_kinds::feature, eltwise_mode::sum },
gemm_test_params{ CASE_GEMM_2IN_FP16_3D_2, 4, 4, "gemm_tiled_opt", broadcast_kinds::batch, eltwise_mode::sum },
gemm_test_params{ CASE_GEMM_2IN_FP16_3D_1, 4, 4, "gemm_tiled_opt", broadcast_kinds::batch, eltwise_mode::sum }
}));

class gemm_2in_act_scale_quantize_i8 : public GemmFusingTest {};
Expand Down
Loading