/*=========================================================================
 *
 *  Copyright UMC Utrecht and contributors
 *
 *  Licensed under the Apache License, Version 2.0 (the "License");
 *  you may not use this file except in compliance with the License.
 *  You may obtain a copy of the License at
 *
 *        http://www.apache.org/licenses/LICENSE-2.0.txt
 *
 *  Unless required by applicable law or agreed to in writing, software
 *  distributed under the License is distributed on an "AS IS" BASIS,
 *  WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
 *  See the License for the specific language governing permissions and
 *  limitations under the License.
 *
 *=========================================================================*/
#include "itkTestHelper.h"

// GPU includes
#include "itkGPUImageFactory.h"
#include "itkGPURecursiveGaussianImageFilterFactory.h"
#include "itkGPUCastImageFilterFactory.h" // used internally in smoothing filter

// ITK include files
#include "itkSmoothingRecursiveGaussianImageFilter.h"
#include "itkImageFileReader.h"
#include "itkImageFileWriter.h"
#include "itkImageRegionConstIterator.h"
#include "itkTimeProbe.h"

#include <iomanip> // setprecision, etc.

//------------------------------------------------------------------------------
// This test compares the CPU with the GPU version of the
// SmoothingRecursiveGaussianImageFilter.
// The filter takes an input image and produces an output image.
// We compare the CPU and GPU output image write RMSE and speed.

int
main( int argc, char * argv[] )
{
  // Check arguments for help
  if( argc < 4 )
  {
    std::cerr << "ERROR: insufficient command line arguments.\n"
              << "  inputFileName outputNameCPU outputNameGPU" << std::endl;
    return EXIT_FAILURE;
  }

  // Setup for debugging
  itk::SetupForDebugging();

  // Create and check OpenCL context
  if( !itk::CreateContext() )
  {
    return EXIT_FAILURE;
  }

  /** Get the command line arguments. */
  const std::string  inputFileName     = argv[ 1 ];
  const std::string  outputFileNameCPU = argv[ 2 ];
  const std::string  outputFileNameGPU = argv[ 3 ];
  const double       epsilon           = 1e-3;
  const unsigned int runTimes          = 5;

  std::cout << std::showpoint << std::setprecision( 4 );

  // Typedefs.
  const unsigned int Dimension = 3;
  typedef float                                    InputPixelType;
  typedef float                                    OutputPixelType;
  typedef itk::Image< InputPixelType, Dimension >  InputImageType;
  typedef itk::Image< OutputPixelType, Dimension > OutputImageType;

  // CPU Typedefs
  typedef itk::SmoothingRecursiveGaussianImageFilter
    < InputImageType, OutputImageType >           FilterType;
  typedef itk::ImageFileReader< InputImageType >  ReaderType;
  typedef itk::ImageFileWriter< OutputImageType > WriterType;

  // Reader
  ReaderType::Pointer reader = ReaderType::New();
  reader->SetFileName( inputFileName );
  reader->Update();

  // Construct the filter
  FilterType::Pointer        filter = FilterType::New();
  FilterType::SigmaArrayType sigmaArray;
  for( unsigned int i = 0; i < Dimension; i++ )
  {
    sigmaArray[ i ] = 3.0;
  }
  filter->SetSigmaArray( sigmaArray );

  std::cout << "Testing the SmoothingRecursiveGaussianImageFilter, CPU vs GPU:\n";
  std::cout << "CPU/GPU sigma #threads time speedup RMSE\n";

  // Time the filter, run on the CPU
  itk::TimeProbe cputimer;
  cputimer.Start();
  for( unsigned int i = 0; i < runTimes; i++ )
  {
    filter->SetInput( reader->GetOutput() );
    try
    {
      filter->Update();
    }
    catch( itk::ExceptionObject & e )
    {
      std::cerr << "ERROR: " << e << std::endl;
      itk::ReleaseContext();
      return EXIT_FAILURE;
    }
    filter->Modified();
  }
  cputimer.Stop();

  std::cout << "CPU " << sigmaArray[ 0 ]
            << " " << filter->GetNumberOfWorkUnits()
            << " " << cputimer.GetMean() / runTimes;

  /** Write the CPU result. */
  WriterType::Pointer writer = WriterType::New();
  writer->SetInput( filter->GetOutput() );
  writer->SetFileName( outputFileNameCPU.c_str() );
  try
  {
    writer->Update();
  }
  catch( itk::ExceptionObject & e )
  {
    std::cerr << "ERROR: " << e << std::endl;
    itk::ReleaseContext();
    return EXIT_FAILURE;
  }

  // Register object factory for GPU image and filter
  // All these filters that are constructed after this point are
  // turned into a GPU filter.
  // Note that we are not registering a
  // GPUSmoothingRecursiveGaussianImageFilter,
  // but the recursive one. We are simply using the original ITK implementation,
  // that internally uses the recursive filter. By registering the recursive
  // filter, we now automatically use it, even if it's usage is hidden by a
  // wrapper.
  typedef typelist::MakeTypeList< float >::Type OCLImageTypes;
  itk::GPUImageFactory2< OCLImageTypes, OCLImageDims >
  ::RegisterOneFactory();
  itk::GPURecursiveGaussianImageFilterFactory2< OCLImageTypes, OCLImageTypes, OCLImageDims >
  ::RegisterOneFactory();
  itk::GPUCastImageFilterFactory2< OCLImageTypes, OCLImageTypes, OCLImageDims >
  ::RegisterOneFactory();

  // Construct the filter
  // Use a try/catch, because construction of this filter will trigger
  // OpenCL compilation, which may fail.
  FilterType::Pointer gpuFilter;
  try
  {
    gpuFilter = FilterType::New();
    itk::ITKObjectEnableWarnings( gpuFilter.GetPointer() );
  }
  catch( itk::ExceptionObject & e )
  {
    std::cerr << "ERROR: " << e << std::endl;
    itk::ReleaseContext();
    return EXIT_FAILURE;
  }
  gpuFilter->SetSigmaArray( sigmaArray );

  // Also need to re-construct the image reader, so that it now
  // reads a GPUImage instead of a normal image.
  // Otherwise, you will get an exception when running the GPU filter:
  // "ERROR: The GPU InputImage is NULL. Filter unable to perform."
  ReaderType::Pointer gpuReader = ReaderType::New();
  gpuReader->SetFileName( inputFileName );

  // \todo: If the following line is uncommented something goes wrong with
  // the ITK pipeline synchronization.
  // Something is still read in, but I guess it is not properly copied to
  // the GPU. The output of the shrink filter is then bogus.
  // The following line is however not needed in a pure CPU implementation.
  gpuReader->Update();

  // Time the filter, run on the GPU
  itk::TimeProbe gputimer;
  gputimer.Start();
  for( unsigned int i = 0; i < runTimes; i++ )
  {
    gpuFilter->SetInput( gpuReader->GetOutput() );
    try
    {
      gpuFilter->Update();
    }
    catch( itk::ExceptionObject & e )
    {
      std::cerr << "ERROR: " << e << std::endl;
      itk::ReleaseContext();
      return EXIT_FAILURE;
    }
    gpuFilter->Modified();
  }
  gputimer.Stop();

  std::cout << "GPU " << sigmaArray[ 0 ]
            << " x " << gputimer.GetMean() / runTimes
            << " " << cputimer.GetMean() / gputimer.GetMean();

  /** Write the GPU result. */
  WriterType::Pointer gpuWriter = WriterType::New();
  gpuWriter->SetInput( gpuFilter->GetOutput() );
  gpuWriter->SetFileName( outputFileNameGPU.c_str() );
  try
  {
    gpuWriter->Update();
  }
  catch( itk::ExceptionObject & e )
  {
    std::cerr << "ERROR: " << e << std::endl;
    itk::ReleaseContext();
    return EXIT_FAILURE;
  }

  // Compute RMSE
  double       RMSrelative = 0.0;
  const double RMSerror    = itk::ComputeRMSE< double, OutputImageType, OutputImageType >
      ( filter->GetOutput(), gpuFilter->GetOutput(), RMSrelative );
  std::cout << " " << RMSerror << std::endl;

  // Check
  if( RMSerror > epsilon )
  {
    std::cerr << "ERROR: RMSE between CPU and GPU result larger than expected" << std::endl;
    itk::ReleaseContext();
    return EXIT_FAILURE;
  }

  // End program.
  itk::ReleaseContext();
  return EXIT_SUCCESS;
} // end main()