commit 8be9148814b88e5b0cabd5a4d2b1f4ff470a8c1c
author Georgios Pinitas <georgios.pinitas@arm.com> Tue Mar 26 17:23:28 2019 +0000
committer Gian Marco Iodice <gianmarco.iodice@arm.com> Thu Apr 11 09:34:26 2019 +0000
tree 760658b8c7b8917379467bd3fc119a5502faa850
parent a50e702289af66944e860eafc7f3b32f6c5f30be

COMPMID-1959: Implements 2D FFT on OpenCL

Change-Id: I73cf3984a5463acc854c8a59dc2bd9a5234cd99c
Signed-off-by: Georgios Pinitas <georgios.pinitas@arm.com>
Reviewed-on: https://review.mlplatform.org/c/936
Comments-Addressed: Arm Jenkins <bsgcomp@arm.com>
Tested-by: Arm Jenkins <bsgcomp@arm.com>
Reviewed-by: Gian Marco Iodice <gianmarco.iodice@arm.com>

src/core/CL/kernels/CLFFTDigitReverseKernel.cpp

diff --git a/src/core/CL/kernels/CLFFTDigitReverseKernel.cpp b/src/core/CL/kernels/CLFFTDigitReverseKernel.cpp
index d72647c..b04293d 100644
--- a/src/core/CL/kernels/CLFFTDigitReverseKernel.cpp
+++ b/src/core/CL/kernels/CLFFTDigitReverseKernel.cpp
@@ -34,16 +34,19 @@
 {
 namespace
 {
-Status validate_arguments(const ITensorInfo *input, const ITensorInfo *output, const ITensorInfo *idx, unsigned int axis)
+Status validate_arguments(const ITensorInfo *input, const ITensorInfo *output, const ITensorInfo *idx, const FFTDigitReverseKernelInfo &config)
 {
     ARM_COMPUTE_RETURN_ERROR_ON_F16_UNSUPPORTED(input);
-    ARM_COMPUTE_RETURN_ERROR_ON_DATA_TYPE_CHANNEL_NOT_IN(input, 2, DataType::F32);
+    ARM_COMPUTE_RETURN_ERROR_ON(input->data_type() != DataType::F32);
+    ARM_COMPUTE_RETURN_ERROR_ON(input->num_channels() != 1 && input->num_channels() != 2);
     ARM_COMPUTE_RETURN_ERROR_ON_DATA_TYPE_CHANNEL_NOT_IN(idx, 1, DataType::U32);
-    ARM_COMPUTE_RETURN_ERROR_ON(axis != 0);
+    ARM_COMPUTE_RETURN_ERROR_ON(std::set<unsigned int>({ 0, 1 }).count(config.axis) == 0);
+    ARM_COMPUTE_RETURN_ERROR_ON(input->tensor_shape()[config.axis] != idx->tensor_shape().x());
 
     // Checks performed when output is configured
     if((output != nullptr) && (output->total_size() != 0))
     {
+        ARM_COMPUTE_RETURN_ERROR_ON(output->num_channels() != 2);
         ARM_COMPUTE_RETURN_ERROR_ON_MISMATCHING_SHAPES(input, output);
         ARM_COMPUTE_RETURN_ERROR_ON_MISMATCHING_DATA_TYPES(input, output);
     }
@@ -51,11 +54,11 @@
     return Status{};
 }
 
-std::pair<Status, Window> validate_and_configure_window(ITensorInfo *input, ITensorInfo *output, ITensorInfo *idx, unsigned int axis)
+std::pair<Status, Window> validate_and_configure_window(ITensorInfo *input, ITensorInfo *output, ITensorInfo *idx, const FFTDigitReverseKernelInfo &config)
 {
-    ARM_COMPUTE_UNUSED(idx, axis);
+    ARM_COMPUTE_UNUSED(idx, config);
 
-    auto_init_if_empty(*output, *input);
+    auto_init_if_empty(*output, input->clone()->set_num_channels(2));
 
     Window win = calculate_max_window(*output, Steps());
     output->set_valid_region(ValidRegion(Coordinates(), output->tensor_shape()));
@@ -69,25 +72,30 @@
 {
 }
 
-void CLFFTDigitReverseKernel::configure(const ICLTensor *input, ICLTensor *output, const ICLTensor *idx, unsigned int axis)
+void CLFFTDigitReverseKernel::configure(const ICLTensor *input, ICLTensor *output, const ICLTensor *idx, const FFTDigitReverseKernelInfo &config)
 {
     ARM_COMPUTE_ERROR_ON_NULLPTR(input, output, idx);
-    ARM_COMPUTE_ERROR_THROW_ON(validate_arguments(input->info(), output->info(), idx->info(), axis));
+    ARM_COMPUTE_ERROR_THROW_ON(validate_arguments(input->info(), output->info(), idx->info(), config));
 
     _input  = input;
     _output = output;
     _idx    = idx;
 
     // Create kernel
-    _kernel = static_cast<cl::Kernel>(CLKernelLibrary::get().create_kernel("digit_reverse"));
+    CLBuildOptions build_opts;
+    build_opts.add_option("-DVEC_SIZE=" + support::cpp11::to_string(input->info()->num_channels()));
+    build_opts.add_option_if(config.conjugate, "-DCONJ");
+    std::string kernel_name = "fft_digit_reverse_axis_" + support::cpp11::to_string(config.axis);
+    _kernel                 = static_cast<cl::Kernel>(CLKernelLibrary::get().create_kernel(kernel_name, build_opts.options()));
 
     // Configure kernel window
-    auto win_config = validate_and_configure_window(input->info(), output->info(), idx->info(), axis);
+    auto win_config = validate_and_configure_window(input->info(), output->info(), idx->info(), config);
     ARM_COMPUTE_ERROR_THROW_ON(win_config.first);
     ICLKernel::configure_internal(win_config.second);
 
     // Set config_id for enabling LWS tuning
-    _config_id = "digit_reverse_";
+    _config_id = kernel_name;
+    _config_id += "_";
     _config_id += lower_string(string_from_data_type(input->info()->data_type()));
     _config_id += "_";
     _config_id += support::cpp11::to_string(input->info()->dimension(0));
@@ -95,10 +103,10 @@
     _config_id += support::cpp11::to_string(input->info()->dimension(1));
 }
 
-Status CLFFTDigitReverseKernel::validate(const ITensorInfo *input, const ITensorInfo *output, const ITensorInfo *idx, unsigned int axis)
+Status CLFFTDigitReverseKernel::validate(const ITensorInfo *input, const ITensorInfo *output, const ITensorInfo *idx, const FFTDigitReverseKernelInfo &config)
 {
-    ARM_COMPUTE_RETURN_ON_ERROR(validate_arguments(input, output, idx, axis));
-    ARM_COMPUTE_RETURN_ON_ERROR(validate_and_configure_window(input->clone().get(), output->clone().get(), idx->clone().get(), axis).first);
+    ARM_COMPUTE_RETURN_ON_ERROR(validate_arguments(input, output, idx, config));
+    ARM_COMPUTE_RETURN_ON_ERROR(validate_and_configure_window(input->clone().get(), output->clone().get(), idx->clone().get(), config).first);
 
     return Status{};
 }