diff options
author | Georgios Pinitas <georgios.pinitas@arm.com> | 2018-07-18 19:51:24 +0100 |
---|---|---|
committer | Anthony Barbier <anthony.barbier@arm.com> | 2018-11-02 16:54:54 +0000 |
commit | 8bc745dfd133f46e68edad511e2933a590602a24 (patch) | |
tree | 440126c9bb569e4eda74a580da9b041afe216eef /src/core/CL/kernels/CLCopyKernel.cpp | |
parent | 201cea1b40597c226bf2c8e59d90bebdf9817dd3 (diff) | |
download | ComputeLibrary-8bc745dfd133f46e68edad511e2933a590602a24.tar.gz |
COMPMID-1124: Validate CLLSTM
-Enables cell-to-input weights when !cifg and peephole
-Makes projection bias conditional
Change-Id: Iee866db9f5d8479c2dfd95d74a2d42492bf07a8d
Reviewed-on: https://eu-gerrit-1.euhpc.arm.com/140543
Tested-by: Jenkins <bsgcomp@arm.com>
Reviewed-by: Les Bell <les.bell@arm.com>
Reviewed-by: Anthony Barbier <anthony.barbier@arm.com>
Diffstat (limited to 'src/core/CL/kernels/CLCopyKernel.cpp')
-rw-r--r-- | src/core/CL/kernels/CLCopyKernel.cpp | 74 |
1 files changed, 56 insertions, 18 deletions
diff --git a/src/core/CL/kernels/CLCopyKernel.cpp b/src/core/CL/kernels/CLCopyKernel.cpp index 4f00ef9eef..1fc8b5bfbe 100644 --- a/src/core/CL/kernels/CLCopyKernel.cpp +++ b/src/core/CL/kernels/CLCopyKernel.cpp @@ -33,10 +33,44 @@ #include "arm_compute/core/Validate.h" #include "arm_compute/core/Window.h" -#include <algorithm> - using namespace arm_compute; +namespace +{ +Status validate_arguments(const ITensorInfo *input, const ITensorInfo *output) +{ + ARM_COMPUTE_RETURN_ERROR_ON_NULLPTR(input, output); + + // Validate output if initialized + if(output->total_size() != 0) + { + ARM_COMPUTE_RETURN_ERROR_ON_MISMATCHING_DIMENSIONS(input->tensor_shape(), output->tensor_shape()); + ARM_COMPUTE_RETURN_ERROR_ON_MISMATCHING_DATA_TYPES(input, output); + } + + return Status{}; +} + +std::pair<Status, Window> validate_and_configure_window(ITensorInfo *input, ITensorInfo *output) +{ + // Output auto inizialitation if not yet initialized + auto_init_if_empty(*output, *input); + + // Configure window + const unsigned int num_elems_processed_per_iteration = 16 / input->element_size(); + + Window win = calculate_max_window(*input, Steps(num_elems_processed_per_iteration)); + + AccessWindowHorizontal input_access(input, 0, num_elems_processed_per_iteration); + AccessWindowHorizontal output_access(output, 0, num_elems_processed_per_iteration); + + bool window_changed = update_window_and_padding(win, input_access, output_access); + + Status err = (window_changed) ? ARM_COMPUTE_CREATE_ERROR(ErrorCode::RUNTIME_ERROR, "Insufficient Padding!") : Status{}; + return std::make_pair(err, win); +} +} // namespace + CLCopyKernel::CLCopyKernel() : _input(nullptr), _output(nullptr) { @@ -44,28 +78,32 @@ CLCopyKernel::CLCopyKernel() void CLCopyKernel::configure(const ICLTensor *input, ICLTensor *output) { - ARM_COMPUTE_ERROR_ON_MISMATCHING_DIMENSIONS(input->info()->tensor_shape(), output->info()->tensor_shape()); - ARM_COMPUTE_ERROR_ON_MISMATCHING_DATA_TYPES(input, output); + ARM_COMPUTE_ERROR_ON_NULLPTR(input, output); + ARM_COMPUTE_ERROR_THROW_ON(validate_arguments(input->info(), output->info())); _input = input; _output = output; + const unsigned int num_elems_processed_per_iteration = 16 / input->info()->element_size(); + // Create kernel CLBuildOptions build_opts; build_opts.add_option("-DDATA_TYPE=" + get_cl_type_from_data_type(input->info()->data_type())); + build_opts.add_option("-DVEC_SIZE=" + support::cpp11::to_string(num_elems_processed_per_iteration)); _kernel = static_cast<cl::Kernel>(CLKernelLibrary::get().create_kernel("copy_tensor", build_opts.options())); - // Configure window - constexpr unsigned int num_elems_processed_per_iteration = 16; - - Window win = calculate_max_window(*input->info(), Steps(num_elems_processed_per_iteration)); - - AccessWindowHorizontal input_access(input->info(), 0, num_elems_processed_per_iteration); - AccessWindowHorizontal output_access(output->info(), 0, num_elems_processed_per_iteration); + // Configure kernel window + auto win_config = validate_and_configure_window(input->info(), output->info()); + ARM_COMPUTE_ERROR_THROW_ON(win_config.first); + ICLKernel::configure(win_config.second); +} - update_window_and_padding(win, input_access, output_access); +Status CLCopyKernel::validate(const arm_compute::ITensorInfo *input, const arm_compute::ITensorInfo *output) +{ + ARM_COMPUTE_RETURN_ON_ERROR(validate_arguments(input, output)); + ARM_COMPUTE_RETURN_ON_ERROR(validate_and_configure_window(input->clone().get(), output->clone().get()).first); - ICLKernel::configure(win); + return Status{}; } void CLCopyKernel::run(const Window &window, cl::CommandQueue &queue) @@ -73,15 +111,15 @@ void CLCopyKernel::run(const Window &window, cl::CommandQueue &queue) ARM_COMPUTE_ERROR_ON_UNCONFIGURED_KERNEL(this); ARM_COMPUTE_ERROR_ON_INVALID_SUBWINDOW(ICLKernel::window(), window); - Window collapsed = window.collapse_if_possible(ICLKernel::window(), Window::DimX); - Window slice = collapsed.first_slice_window_1D(); + Window collapsed = window.collapse_if_possible(ICLKernel::window(), Window::DimZ); + Window slice = collapsed.first_slice_window_3D(); do { unsigned int idx = 0; - add_1D_tensor_argument(idx, _input, slice); - add_1D_tensor_argument(idx, _output, slice); + add_3D_tensor_argument(idx, _input, slice); + add_3D_tensor_argument(idx, _output, slice); enqueue(queue, *this, slice); } - while(collapsed.slide_window_slice_1D(slice)); + while(collapsed.slide_window_slice_3D(slice)); } |