diff options
Diffstat (limited to 'src/gpu/cl/kernels/ClPermuteKernel.cpp')
| -rw-r--r-- | src/gpu/cl/kernels/ClPermuteKernel.cpp | 27 |
1 files changed, 16 insertions, 11 deletions
diff --git a/src/gpu/cl/kernels/ClPermuteKernel.cpp b/src/gpu/cl/kernels/ClPermuteKernel.cpp index 8d4655114b..a4755782ed 100644 --- a/src/gpu/cl/kernels/ClPermuteKernel.cpp +++ b/src/gpu/cl/kernels/ClPermuteKernel.cpp @@ -29,8 +29,9 @@ #include "arm_compute/core/Helpers.h" #include "arm_compute/core/TensorInfo.h" #include "arm_compute/core/Utils.h" -#include "arm_compute/core/Validate.h" #include "arm_compute/core/utils/misc/ShapeCalculator.h" +#include "arm_compute/core/Validate.h" + #include "src/core/CL/CLValidate.h" #include "src/core/helpers/AutoConfiguration.h" #include "src/core/helpers/WindowHelpers.h" @@ -60,13 +61,13 @@ Status validate_arguments(const ITensorInfo *src, const ITensorInfo *dst, const "Permutation up to 4-D src tensor is supported"); ARM_COMPUTE_RETURN_ERROR_ON_MSG(perm.num_dimensions() < 1 || perm.num_dimensions() > 4, "Permutation vector size should be less than or equal to 4"); - for(const auto &p : perm) + for (const auto &p : perm) { ARM_COMPUTE_RETURN_ERROR_ON_MSG(p >= perm.num_dimensions(), "Permutation vector has invalid values"); } // Validate configured dst - if(dst->total_size() != 0) + if (dst->total_size() != 0) { const TensorShape dst_shape = misc::shape_calculator::compute_permutation_output_shape(*src, perm); ARM_COMPUTE_RETURN_ERROR_ON_MISMATCHING_DIMENSIONS(dst->tensor_shape(), dst_shape); @@ -82,10 +83,13 @@ ClPermuteKernel::ClPermuteKernel() _type = CLKernelType::ELEMENTWISE; } -void ClPermuteKernel::configure(const CLCompileContext &compile_context, const ITensorInfo *src, ITensorInfo *dst, const PermutationVector &perm) +void ClPermuteKernel::configure(const CLCompileContext &compile_context, + const ITensorInfo *src, + ITensorInfo *dst, + const PermutationVector &perm) { ARM_COMPUTE_ERROR_ON_NULLPTR(src, dst); - auto padding_info = get_padding_info({ src, dst }); + auto padding_info = get_padding_info({src, dst}); const TensorShape dst_shape = get_dst_shape(src, perm); // Output auto initialization if not yet initialized auto_init_if_empty(*dst, src->clone()->set_tensor_shape(dst_shape)); @@ -96,7 +100,8 @@ void ClPermuteKernel::configure(const CLCompileContext &compile_context, const I // Create kernel CLBuildOptions build_opts; - build_opts.add_option("-DDATA_TYPE=" + get_cl_unsigned_type_from_element_size(data_size_from_type(src->data_type()))); + build_opts.add_option("-DDATA_TYPE=" + + get_cl_unsigned_type_from_element_size(data_size_from_type(src->data_type()))); build_opts.add_option("-DDEPTH_IN=" + support::cpp11::to_string(src->dimension(2))); // New positions of width(W), height(H), channel(C) and batch(D) based on permutation vector build_opts.add_option("-DP1=" + support::cpp11::to_string((_perm.num_dimensions() >= 1) ? perm[0] : 0)); @@ -126,8 +131,9 @@ void ClPermuteKernel::run_op(ITensorPack &tensors, const Window &window, cl::Com ARM_COMPUTE_ERROR_ON_UNCONFIGURED_KERNEL(this); ARM_COMPUTE_ERROR_ON_MISMATCHING_WINDOWS(ICLKernel::window(), window); - const auto src = utils::cast::polymorphic_downcast<const ICLTensor *>(tensors.get_const_tensor(TensorType::ACL_SRC)); - auto dst = utils::cast::polymorphic_downcast<ICLTensor *>(tensors.get_tensor(TensorType::ACL_DST)); + const auto src = + utils::cast::polymorphic_downcast<const ICLTensor *>(tensors.get_const_tensor(TensorType::ACL_SRC)); + auto dst = utils::cast::polymorphic_downcast<ICLTensor *>(tensors.get_tensor(TensorType::ACL_DST)); Window slice_in = window.first_slice_window_4D().collapse(ICLKernel::window(), 2, 4); @@ -144,9 +150,8 @@ void ClPermuteKernel::run_op(ITensorPack &tensors, const Window &window, cl::Com add_4D_tensor_argument(idx, src, slice_in); add_4D_tensor_argument(idx, dst, slice_out); enqueue(queue, *this, slice_in, lws_hint()); - } - while(window.slide_window_slice_4D(slice_in) && window.slide_window_slice_4D(slice_out)); + } while (window.slide_window_slice_4D(slice_in) && window.slide_window_slice_4D(slice_out)); } } // namespace kernels } // namespace opencl -} // namespace arm_compute
\ No newline at end of file +} // namespace arm_compute |