Function genSwapKernelGeneral

src/library/generator.transpose.cpp:920–1937

swap lines. a more general kernel generator. this function accepts any ratio in theory. But in practice we restrict it to 1:2, 1:3, 1:5 and 1:10 ration

//swap lines. a more general kernel generator.
//this function accepts any ratio in theory. But in practice we restrict it to 1:2, 1:3, 1:5 and 1:10 ration
clfftStatus genSwapKernelGeneral(const FFTGeneratedTransposeNonSquareAction::Signature & params, std::string& strKernel, std::string& KernelFuncName, const size_t& lwSize, const size_t reShapeFactor)
{
	if (params.fft_placeness == CLFFT_OUTOFPLACE)
		return CLFFT_TRANSPOSED_NOTIMPLEMENTED;

	size_t smaller_dim = (params.fft_N[0] < params.fft_N[1]) ? params.fft_N[0] : params.fft_N[1];
	size_t bigger_dim = (params.fft_N[0] >= params.fft_N[1]) ? params.fft_N[0] : params.fft_N[1];
	size_t dim_ratio = bigger_dim / smaller_dim;
	/*
	if ( (params.fft_N[0] != 2 * params.fft_N[1]) && (params.fft_N[1] != 2 * params.fft_N[0]) &&
		 (params.fft_N[0] != 3 * params.fft_N[1]) && (params.fft_N[1] != 3 * params.fft_N[0]) &&
		 (params.fft_N[0] != 5 * params.fft_N[1]) && (params.fft_N[1] != 5 * params.fft_N[0]) &&
		 (params.fft_N[0] != 10 * params.fft_N[1]) && (params.fft_N[1] != 10 * params.fft_N[0]) )
    */
	if(dim_ratio % 2 != 0 && dim_ratio % 3 != 0 && dim_ratio % 5 != 0 && dim_ratio % 10 != 0)
	{
		return CLFFT_TRANSPOSED_NOTIMPLEMENTED;
	}

	strKernel.reserve(4096);
	std::stringstream transKernel(std::stringstream::out);

	// These strings represent the various data types we read or write in the kernel, depending on how the plan
	// is configured
	std::string dtInput;        // The type read as input into kernel
	std::string dtOutput;       // The type written as output from kernel
	std::string dtPlanar;       // Fundamental type for planar arrays
	std::string tmpBuffType;
	std::string dtComplex;      // Fundamental type for complex arrays

								// NOTE:  Enable only for debug
								// clKernWrite( transKernel, 0 ) << "#pragma OPENCL EXTENSION cl_amd_printf : enable\n" << std::endl;

								//if (params.fft_inputLayout != params.fft_outputLayout)
								//	return CLFFT_TRANSPOSED_NOTIMPLEMENTED;

	switch (params.fft_precision)
	{
	case CLFFT_SINGLE:
	case CLFFT_SINGLE_FAST:
		dtPlanar = "float";
		dtComplex = "float2";
		break;
	case CLFFT_DOUBLE:
	case CLFFT_DOUBLE_FAST:
		dtPlanar = "double";
		dtComplex = "double2";

		// Emit code that enables double precision in the kernel
		clKernWrite(transKernel, 0) << "#ifdef cl_khr_fp64" << std::endl;
		clKernWrite(transKernel, 3) << "#pragma OPENCL EXTENSION cl_khr_fp64 : enable" << std::endl;
		clKernWrite(transKernel, 0) << "#else" << std::endl;
		clKernWrite(transKernel, 3) << "#pragma OPENCL EXTENSION cl_amd_fp64 : enable" << std::endl;
		clKernWrite(transKernel, 0) << "#endif\n" << std::endl;

		break;
	default:
		return CLFFT_TRANSPOSED_NOTIMPLEMENTED;