Skip to content

[GPU] Add ellipsis mode support for strided slice #24069

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

Merged
p-durandin merged 1 commit into from
Apr 17, 2024
Merged

[GPU] Add ellipsis mode support for strided slice #24069

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
48 changes: 38 additions & 10 deletions src/plugins/intel_gpu/src/graph/impls/ocl/strided_slice.cpp
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -12,11 +12,36 @@

namespace {
template <typename T, typename DT, typename = typename std::enable_if<std::is_convertible<DT, T>::value>::type>
std::vector<T>& pad_vector_to_size(std::vector<T>& data, size_t size, DT value) {
for (size_t i = data.size(); i < size; ++i) {
data.push_back(static_cast<T>(value));
void pad_vector_to_size(std::vector<T>& data, size_t size, DT value, const std::vector<int64_t>& ellipsis_mask) {
bool apply_ellipsis_mask = std::count(ellipsis_mask.begin(), ellipsis_mask.end(), 1) == 1;
if (apply_ellipsis_mask && data.size() == ellipsis_mask.size()) {
std::vector<T> temp;
size_t ellipsis_pos1 = 0;
for (size_t i = 0; i < ellipsis_mask.size(); i++) {
if (ellipsis_mask[i] == 1) {
ellipsis_pos1 = i;
break;
}
}

size_t dims_after = data.size() - ellipsis_pos1 - 1;
size_t ellipsis_pos2 = size - dims_after - 1;;

for (size_t i = 0; i < ellipsis_pos1; i++)
temp.push_back(data[i]);

for (size_t i = ellipsis_pos1; i < ellipsis_pos2 + 1; i++)
temp.push_back(value);

for (size_t i = 1; i < size - ellipsis_pos2; i++)
temp.push_back(data[i + ellipsis_pos1]);

data = temp;
} else {
for (size_t i = data.size(); i < size; ++i) {
data.push_back(static_cast<T>(value));
}
}
return data;
}

template <typename T, typename MT>
Expand Down Expand Up @@ -74,7 +99,7 @@ struct strided_slice_impl : typed_primitive_impl_ocl<strided_slice> {

// Getting data from constant inputs. There are 3 args: Begin, End, Stride
if (!begin.empty() && !params.has_dynamic_tensors()) {
pad_vector_to_size(begin, dims_num, 0);
pad_vector_to_size(begin, dims_num, 0, prim->ellipsis_mask);
params.begin_type = kernel_selector::base_params::ArgType::Constant;
params.striding_params.push_back(begin);
} else {
Expand All @@ -91,7 +116,7 @@ struct strided_slice_impl : typed_primitive_impl_ocl<strided_slice> {
return offset;
};
if (!end.empty() && !params.has_dynamic_tensors()) {
pad_vector_to_size(end, dims_num, 1);
pad_vector_to_size(end, dims_num, 1, prim->ellipsis_mask);
params.end_type = kernel_selector::base_params::ArgType::Constant;
params.striding_params.push_back(end);
} else {
Expand All @@ -108,7 +133,7 @@ struct strided_slice_impl : typed_primitive_impl_ocl<strided_slice> {
return offset;
};
if (!strides.empty() && !params.has_dynamic_tensors()) {
pad_vector_to_size(strides, dims_num, 1);
pad_vector_to_size(strides, dims_num, 1, prim->ellipsis_mask);
params.stride_type = kernel_selector::base_params::ArgType::Constant;
params.striding_params.push_back(strides);
} else {
Expand All @@ -122,19 +147,22 @@ struct strided_slice_impl : typed_primitive_impl_ocl<strided_slice> {
auto end_mask_ = prim->end_mask;
auto new_axis_mask_ = prim->new_axis_mask;
auto shrink_axis_mask_ = prim->shrink_axis_mask;
auto ellipsis_mask_ = prim->ellipsis_mask;

std::vector<uint8_t> begin_mask(begin_mask_.begin(), begin_mask_.end());
std::vector<uint8_t> end_mask(end_mask_.begin(), end_mask_.end());
std::vector<uint8_t> new_axis_mask(new_axis_mask_.begin(), new_axis_mask_.end());
std::vector<uint8_t> shrink_axis_mask(shrink_axis_mask_.begin(), shrink_axis_mask_.end());
std::vector<uint8_t> ellipsis_mask(ellipsis_mask_.begin(), ellipsis_mask_.end());
params.end_mask = std::move(end_mask);
pad_vector_to_size(params.end_mask, dims_num, 0);
pad_vector_to_size(params.end_mask, dims_num, 0, prim->ellipsis_mask);
params.begin_mask = std::move(begin_mask);
pad_vector_to_size(params.begin_mask, dims_num, 0);
pad_vector_to_size(params.begin_mask, dims_num, 0, prim->ellipsis_mask);

params.new_axis_mask = new_axis_mask;
params.shrink_axis_mask = shrink_axis_mask;
pad_vector_to_size(params.shrink_axis_mask, dims_num, 0);
params.ellipsis_mask = ellipsis_mask;
pad_vector_to_size(params.shrink_axis_mask, dims_num, 0, prim->ellipsis_mask);

std::vector<size_t> logical_dims = params.inputs[0].LogicalDims();
std::reverse(logical_dims.begin(), logical_dims.end()); // get dims in bfyx order
Expand Down
70 changes: 35 additions & 35 deletions src/plugins/intel_gpu/src/kernel_selector/cl_kernels/strided_slice_ref.cl
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -10,20 +10,20 @@ inline void FUNC(get_slice_step)(OPTIONAL_SHAPE_INFO_ARG
int* step_batch, int* step_feature,
int* step_w, int* step_z, int* step_y, int* step_x)
{
const uint batch_index = 0;
const uint feature_index = 1;
const uint batch_index = DIM_IDX_BATCH;
const uint feature_index = DIM_IDX_FEATURE;
#ifdef OUTPUT_LAYOUT_BFYX
const uint y_index = 2;
const uint x_index = 3;
const uint y_index = DIM_IDX_Y;
const uint x_index = DIM_IDX_X;
#elif OUTPUT_LAYOUT_BFZYX
const uint z_index = 2;
const uint y_index = 3;
const uint x_index = 4;
const uint z_index = DIM_IDX_Z;
const uint y_index = DIM_IDX_Y;
const uint x_index = DIM_IDX_X;
#elif OUTPUT_LAYOUT_BFWZYX
const uint w_index = 2;
const uint z_index = 3;
const uint y_index = 4;
const uint x_index = 5;
const uint w_index = DIM_IDX_W;
const uint z_index = DIM_IDX_Z;
const uint y_index = DIM_IDX_Y;
const uint x_index = DIM_IDX_X;
#endif

*step_batch = batch_index < STRIDE_DIMS ? stride[batch_index] : 1;
Expand Down Expand Up @@ -55,20 +55,20 @@ inline void FUNC(get_slice_end)(OPTIONAL_SHAPE_INFO_ARG
const uint out_z_num = INPUT0_SIZE_Z;
const uint out_y_num = INPUT0_SIZE_Y;
const uint out_x_num = INPUT0_SIZE_X;
const uint batch_index = 0;
const uint feature_index = 1;
const uint batch_index = DIM_IDX_BATCH;
const uint feature_index = DIM_IDX_FEATURE;
#ifdef OUTPUT_LAYOUT_BFYX
const uint y_index = 2;
const uint x_index = 3;
const uint y_index = DIM_IDX_Y;
const uint x_index = DIM_IDX_X;
#elif OUTPUT_LAYOUT_BFZYX
const uint z_index = 2;
const uint y_index = 3;
const uint x_index = 4;
const uint z_index = DIM_IDX_Z;
const uint y_index = DIM_IDX_Y;
const uint x_index = DIM_IDX_X;
#elif OUTPUT_LAYOUT_BFWZYX
const uint w_index = 2;
const uint z_index = 3;
const uint y_index = 4;
const uint x_index = 5;
const uint w_index = DIM_IDX_W;
const uint z_index = DIM_IDX_Z;
const uint y_index = DIM_IDX_Y;
const uint x_index = DIM_IDX_X;
#endif
END_TYPE batch = batch_index < END_DIMS ? end[batch_index] : 0;
END_TYPE feature = feature_index < END_DIMS ? end[feature_index] : 0;
Expand Down Expand Up @@ -100,20 +100,20 @@ inline void FUNC(get_slice_begin)(OPTIONAL_SHAPE_INFO_ARG
int* begin_batch, int* begin_feature,
int* begin_w, int* begin_z, int* begin_y, int* begin_x)
{
const uint batch_index = 0;
const uint feature_index = 1;
const uint batch_index = DIM_IDX_BATCH;
const uint feature_index = DIM_IDX_FEATURE;
#ifdef OUTPUT_LAYOUT_BFYX
const uint y_index = 2;
const uint x_index = 3;
const uint y_index = DIM_IDX_Y;
const uint x_index = DIM_IDX_X;
#elif OUTPUT_LAYOUT_BFZYX
const uint z_index = 2;
const uint y_index = 3;
const uint x_index = 4;
const uint z_index = DIM_IDX_Z;
const uint y_index = DIM_IDX_Y;
const uint x_index = DIM_IDX_X;
#elif OUTPUT_LAYOUT_BFWZYX
const uint w_index = 2;
const uint z_index = 3;
const uint y_index = 4;
const uint x_index = 5;
const uint w_index = DIM_IDX_W;
const uint z_index = DIM_IDX_Z;
const uint y_index = DIM_IDX_Y;
const uint x_index = DIM_IDX_X;
#endif

BEGIN_TYPE batch = batch_index < BEGIN_DIMS ? begin[batch_index] : 0;
Expand Down Expand Up @@ -160,7 +160,7 @@ inline void FUNC(calculate_index)(int* step, int* begin_num, int* end_num, const
{
int real_begin = *begin_num < 0 ? *begin_num + out_num : *begin_num;
int real_end = *end_num < 0 ? *end_num + out_num : *end_num;
if (*step < 0) {
if (*step < 0) {
real_begin = max((int)(0), min((int)(out_num - 1), real_begin));
real_end = max((int)(-1), min((int)out_num, real_end));
if (real_begin < real_end) { // for reversing
Expand Down Expand Up @@ -388,7 +388,7 @@ KERNEL(strided_slice_ref)(OPTIONAL_SHAPE_INFO_ARG
const uint input_index = INPUT0_OFFSET +
(slice_begin_batch + batch * slice_steps_batch) * INPUT0_BATCH_PITCH +
(slice_begin_feature + feature * slice_steps_feature) * INPUT0_FEATURE_PITCH +
#if INPUT0_LAYOUT_BFWZYX
#if INPUT0_LAYOUT_BFWZYX
(slice_begin_w + w * slice_steps_w) * INPUT0_W_PITCH +
(slice_begin_z + z * slice_steps_z) * INPUT0_Z_PITCH +
(slice_begin_y + y * slice_steps_y) * INPUT0_Y_PITCH +
Expand Down
32 changes: 32 additions & 0 deletions src/plugins/intel_gpu/src/kernel_selector/kernels/strided_slice/strided_slice_kernel_ref.cpp
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -165,6 +165,38 @@ JitConstants StridedSliceKernelRef::GetJitConstants(const strided_slice_params&
"NEW_AXIS_MODE",
std::find(params.new_axis_mask.begin(), params.new_axis_mask.end(), 1) != params.new_axis_mask.end()));

std::vector<int> dims_indexes;
bool ellipsis_mode = std::find(params.ellipsis_mask.begin(), params.ellipsis_mask.end(), 1) != params.ellipsis_mask.end();
if (ellipsis_mode) {
size_t ellipsis_pos1 = 0;
for (size_t i = 0; i < params.ellipsis_mask.size(); i++) {
if (params.ellipsis_mask[i] == 1) {
ellipsis_pos1 = i;
break;
}
}

const size_t output_rank = params.outputs[0].Dimentions();
const size_t skip_dims_num = output_rank - params.ellipsis_mask.size() + 1;
int dim_counter = 0;

for (size_t i = 0; i < ellipsis_pos1; i++)
dims_indexes.push_back(dim_counter++);

for (size_t i = 0; i < skip_dims_num; i++)
dims_indexes.push_back(-1);

dim_counter++;
for (size_t i = 0; i < params.ellipsis_mask.size() - ellipsis_pos1 - 1; i++)
dims_indexes.push_back(dim_counter++);

OPENVINO_ASSERT(dims_indexes.size() == output_rank, "[GPU] Number of indexes is expected to match with output rank");
} else {
dims_indexes.resize(params.outputs[0].Dimentions());
std::iota(dims_indexes.begin(), dims_indexes.end(), 0);
}
makeJitConstForParam(jit, "DIM_IDX", dims_indexes);

bool shrink_mode = std::find(params.shrink_axis_mask.begin(), params.shrink_axis_mask.end(), 1) != params.shrink_axis_mask.end();
if (shrink_mode) {
jit.AddConstant(MakeJitConstant("SHRINK_MODE", true));
Expand Down
16 changes: 16 additions & 0 deletions ...ns/intel_gpu/tests/functional/shared_tests_instances/single_layer_tests/strided_slice.cpp
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -152,6 +152,22 @@ std::vector<StridedSliceSpecificParams> ss_only_test_cases_fp32 = {
{ 128, 1, 1024 }})),
{ -1, 0, 0 }, { 0, 0, 0 }, { 1, 1, 1 },
{ 0, 1, 1 }, { 0, 1, 1 }, { 0, 0, 0 }, { 1, 0, 0 }, { 0, 0, 0 } },
StridedSliceSpecificParams{ ov::test::static_shapes_to_test_representation(std::vector<ov::Shape>({
{ 10, 10 }})),
{ -4, 1 }, { -8, 0 }, { -1, 1 },
{ 0, 1 }, { 0, 1 }, { 0, 0 }, { 0, 0 }, { 0, 1 } },
StridedSliceSpecificParams{ ov::test::static_shapes_to_test_representation(std::vector<ov::Shape>({
{ 2, 2, 4, 1 }})),
{ 0, 0 }, { 2, 2 }, { -1, 1 },
{ 1, 0 }, { 1, 0 }, { 0, 0 }, { 0, 0 }, { 0, 1 } },
StridedSliceSpecificParams{ ov::test::static_shapes_to_test_representation(std::vector<ov::Shape>({
{ 2, 2, 4, 1 }})),
{ 0, 0 }, { 4, 1 }, { 1, -1 },
{ 0, 1 }, { 0, 1 }, { 0, 0 }, { 0, 0 }, { 1, 0 } },
StridedSliceSpecificParams{ ov::test::static_shapes_to_test_representation(std::vector<ov::Shape>({
{ 1, 5, 30, 30, 30 }})),
{ 0, 0, 0 }, { 0, 29, 29 }, { 1, 1, 1 },
{1, 1, 1}, {1, 0, 0}, {0, 0, 0}, {0, 0, 0}, {0, 1, 0} },
};

std::vector<StridedSliceSpecificParams> ss_only_test_cases_i64 = {
Expand Down
3 changes: 3 additions & 0 deletions src/plugins/intel_gpu/tests/functional/single_layer_tests/dynamic/strided_slice.cpp
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -368,6 +368,7 @@ const std::vector<StridedSliceParams> testCasesCommon4D = {
StridedSliceParams{ { 0, 0, 0, 20 }, { 1, 2, 30, 30 }, { 1, 1, 2, 1 }, { 0, 0, 0, 1 }, { 0, 1, 0, 1 }, { }, { }, { } },
StridedSliceParams{ { 0, 1, 2, 10 }, { 1, 5, 32, 18 }, { 1, 1, 1, 2 }, { 0, 0, 1, 0 }, { 0, 0, 0, 1 }, { }, { }, { } },
StridedSliceParams{ { 0, 0, 2, 10 }, { 1, 8, 32, 18 }, { 1, 2, 1, 2 }, { 0, 0, 1, 0 }, { 0, 0, 0, 1 }, { }, { }, { } },
StridedSliceParams{ { 2, 10 }, { 32, 18 }, { 1, 2 }, { 1, 0 }, { 0, 1 }, { }, { }, { 1, 0 } },
};

const std::vector<InputShape> inputShapesDynamic4D = {
Expand Down Expand Up @@ -396,6 +397,7 @@ const std::vector<StridedSliceParams> testCasesCommon5D = {
StridedSliceParams{ { 0, 0, 0, 20 }, { 1, 2, 30, 30 }, { 1, 1, 2, 1 }, { 0, 0, 0, 1 }, { 0, 1, 0, 1 }, { }, { }, { } },
StridedSliceParams{ { 0, 1, 2, 10 }, { 1, 5, 32, 18 }, { 1, 1, 1, 2 }, { 0, 0, 1, 0 }, { 0, 0, 0, 1 }, { }, { }, { } },
StridedSliceParams{ { 0, 0, 2, 10 }, { 1, 8, 32, 18 }, { 1, 2, 1, 2 }, { 0, 0, 1, 0 }, { 0, 0, 0, 1 }, { }, { }, { } },
StridedSliceParams{ { 0, 0, 2 }, { 1, 8, 32 }, { 1, 2, 1 }, { 1, 0, 1 }, { 0, 0, 0 }, { }, { }, { 0, 1, 0} },
};

const std::vector<InputShape> inputShapesDynamic5D = {
Expand All @@ -421,6 +423,7 @@ INSTANTIATE_TEST_SUITE_P(smoke_CompareWithRefs_Common_Dynamic_5D, StridedSliceLa
const std::vector<StridedSliceParams> testCasesCommon6D = {
StridedSliceParams{ { 0, 2, 5, 4 }, { 1, 4, 28, 27 }, { 1, 1, 1, 1 }, { 0, 0, 0, 0 }, { 0, 0, 0, 0 }, { }, { }, { } },
StridedSliceParams{ { 0, 0, 10, 20 }, { 1, 5, 28, 26 }, { 1, 1, 1, 2 }, { 0, 0, 0, 0 }, { 0, 0, 0, 0 }, { }, { }, { } },
StridedSliceParams{ { 0, 0, 0 }, { 0, 0, 0 }, { 1, 1, 1 }, { 0, 1, 0 }, { 0, 1, 0 }, { }, { }, { 0, 0, 1 } },
};

const std::vector<InputShape> inputShapesDynamic6D = {
Expand Down