Skip to content

Fix OpenCL Build Failure for int16/uint16 Fused Activations #36407

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

Open
peterchen-intel wants to merge 8 commits into
base: master
Choose a base branch
from
Open

Fix OpenCL Build Failure for int16/uint16 Fused Activations #36407

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
8 commits
Select commit Hold shift + click to select a range
67ea06f
Fix OpenCL Build Failure for int16/uint16 Fused Activations
peterchen-intel Jun 15, 2026
81b077b
Merge branch 'master' into CVS-188408/fixing
peterchen-intel Jun 15, 2026
74a7df1
GPU: Fix missing INT16/UINT16 in standalone activation nodes
peterchen-intel Jun 16, 2026
b5f3c13
Merge branch 'master' into CVS-188408/fixing
peterchen-intel Jun 16, 2026
85abc5b
Update per review comments
peterchen-intel Jun 24, 2026
10d1573
Merge branch 'master' into CVS-188408/fixing
peterchen-intel Jun 24, 2026
4aa1f3e
Merge branch 'master' into CVS-188408/fixing
peterchen-intel Jul 5, 2026
676a7fd
Potential fix for pull request finding
peterchen-intel Jul 5, 2026
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
2 changes: 2 additions & 0 deletions src/plugins/intel_gpu/src/graph/impls/ocl/activation.cpp
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -89,6 +89,8 @@ attach_activation_impl::attach_activation_impl() {
auto types = {
data_types::f32,
data_types::f16,
data_types::i16,
data_types::u16,
data_types::i8,
data_types::u8,
data_types::i32
Expand Down
2 changes: 1 addition & 1 deletion src/plugins/intel_gpu/src/kernel_selector/kernel_base.cpp
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -62,7 +62,7 @@ static bool IsTypeUsedIn(Datatype type, const base_params& params) {

Datatype KernelBase::GetUnitType(const base_params& params) const {
Datatype types_prioritized[] =
{Datatype::INT8, Datatype::F16, Datatype::INT32, Datatype::INT64, Datatype::UINT8, Datatype::UINT32};
{Datatype::INT8, Datatype::INT16, Datatype::UINT16, Datatype::F16, Datatype::INT32, Datatype::INT64, Datatype::UINT8, Datatype::UINT32};
Comment thread
peterchen-intel marked this conversation as resolved.

Comment thread
peterchen-intel marked this conversation as resolved.
for (Datatype type : types_prioritized)
if (IsTypeUsedIn(type, params))
Expand Down
4 changes: 4 additions & 0 deletions src/plugins/intel_gpu/src/kernel_selector/kernels/activation/activation_kernel_ref.cpp
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -13,11 +13,15 @@ ParamsKey ActivationKernelRef::GetSupportedKey() const {
k.EnableInputDataType(Datatype::INT8);
k.EnableInputDataType(Datatype::UINT8);
k.EnableInputDataType(Datatype::INT32);
k.EnableInputDataType(Datatype::INT16);
k.EnableInputDataType(Datatype::UINT16);
k.EnableInputDataType(Datatype::F16);
k.EnableInputDataType(Datatype::F32);
k.EnableOutputDataType(Datatype::INT8);
k.EnableOutputDataType(Datatype::INT32);
k.EnableOutputDataType(Datatype::UINT8);
k.EnableOutputDataType(Datatype::INT16);
k.EnableOutputDataType(Datatype::UINT16);
k.EnableOutputDataType(Datatype::F16);
k.EnableOutputDataType(Datatype::F32);
k.EnableDifferentTypes();
Expand Down
42 changes: 42 additions & 0 deletions src/plugins/intel_gpu/tests/unit/fusions/activation_fusion_test.cpp
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -337,3 +337,45 @@ INSTANTIATE_TEST_SUITE_P(DISABLED_fusings_gpu, activation_eltwise_activation, ::
activation_test_params{ CASE_ACTIVATION_3D_F32_4, 2, 4, "activation_ref" }, // FIXME - accuracy bug
activation_test_params{ CASE_ACTIVATION_3D_F32_5, 2, 4, "activation_ref" }, // FIXME - accuracy bug
}));

TEST_F(ActivationFusingTest, reorder_activation_i16) {
// This tests CVS-188408: Ensure int16 Reorder -> Abs activation compiles cleanly
// Prior to fix, this would fail with CL_BUILD_PROGRAM_FAILURE due to "fabs" macro ambiguity.
auto& engine = get_test_engine();
ov::PartialShape input_shape = { 1, 4, 1, 1 };
layout in_layout{ input_shape, data_types::f32, format::bfyx };

topology topology(
input_layout("input", in_layout),
reorder("reorder", input_info("input"), format::bfyx, data_types::i16),
activation("abs", input_info("reorder"), activation_func::abs),
reorder("reorder_out", input_info("abs"), format::bfyx, data_types::f32)
);

ExecutionConfig config = get_test_default_config(engine);
config.set_property(ov::intel_gpu::allow_new_shape_infer(true));
config.set_property(ov::intel_gpu::optimize_data(true));
network network(engine, topology, config);

bool abs_fused = false;
for (const auto& pi : network.get_primitives_info()) {
for (const auto& fused_id : pi.c_fused_ids) {
if (fused_id == "abs") {
abs_fused = true;
break;
}
}
if (abs_fused)
break;
}
EXPECT_TRUE(abs_fused) << "Expected 'abs' to be fused when optimize_data(true)";
Comment on lines +360 to +371

auto input_mem = engine.allocate_memory(in_layout);
set_values(input_mem, { -5.0f, 10.0f, -15.0f, 20.0f });
network.set_input_data("input", input_mem);
auto outputs = network.execute();
auto out_mem = outputs.at("reorder_out").get_memory();
cldnn::mem_lock<float> out_ptr(out_mem, get_test_stream());
std::vector<float> expected_output = { 5.0f, 10.0f, 15.0f, 20.0f };
for (size_t i = 0; i < expected_output.size(); ++i) { EXPECT_EQ(out_ptr[i], expected_output[i]); }
}
Loading