dtfft_config_t Derived Type

type, public, bind(C) :: dtfft_config_t

Type that can be used to set additional configuration parameters to dtFFT

Inherits

Help

Components

Type	Visibility	Attributes		Name		Initial
logical(kind=c_bool),	public		::	enable_log			Should dtFFT print additional information during plan creation or not. Default is false.
logical(kind=c_bool),	public		::	enable_z_slab			Should dtFFT use Z-slab optimization or not. Default is true. One should consider disabling Z-slab optimization in order to resolve `DTFFT_ERROR_VKFFT_R2R_2D_PLAN` error OR when underlying FFT implementation of 2D plan is too slow. In all other cases it is considered that Z-slab is always faster, since it reduces number of data transpositions.
logical(kind=c_bool),	public		::	enable_y_slab			Should dtFFT use Y-slab optimization or not. Default is false. One should consider disabling Y-slab optimization in order to resolve `DTFFT_ERROR_VKFFT_R2R_2D_PLAN` error OR when underlying FFT implementation of 2D plan is too slow. In all other cases it is considered that Y-slab is always faster, since it reduces number of data transpositions.
integer(kind=c_int32_t),	public		::	n_measure_warmup_iters			Number of warmup iterations to execute during backend and kernel autotuning when effort level is `DTFFT_MEASURE` or higher. Default is 2.
integer(kind=c_int32_t),	public		::	n_measure_iters			Number of iterations to execute during backend and kernel autotuning when effort level is `DTFFT_MEASURE` or higher. Default is 5.
type(dtfft_platform_t),	public		::	platform			Selects platform to execute plan. Default is `DTFFT_PLATFORM_HOST`. This option is only available when dtFFT is built with device support. Even when dtFFT is built with device support, it does not necessarily mean that all plans must be device-related. This enables a single library installation to support both host and CUDA plans.
type(dtfft_stream_t),	public		::	stream			Main CUDA stream that will be used in dtFFT. This parameter is a placeholder for user to set custom stream. Stream that is actually used by dtFFT plan is returned by `plan%get_stream` function. When user sets stream he is responsible of destroying it. Stream must not be destroyed before call to `plan%destroy`.
type(dtfft_backend_t),	public		::	backend			Backend that will be used by dtFFT when `effort` is `DTFFT_ESTIMATE` or `DTFFT_MEASURE`. Default for HOST platform is `DTFFT_BACKEND_MPI_DATATYPE`. Default for CUDA platform is `DTFFT_BACKEND_NCCL` if NCCL is enabled, otherwise `DTFFT_BACKEND_MPI_P2P`.
type(dtfft_backend_t),	public		::	reshape_backend			Backend that will be used by dtFFT for data reshaping from bricks to pencils and vice versa when `effort` is `DTFFT_ESTIMATE` or `DTFFT_MEASURE`. Default for HOST platform is `DTFFT_BACKEND_MPI_DATATYPE`. Default for CUDA platform is `DTFFT_BACKEND_NCCL` if NCCL is enabled, otherwise `DTFFT_BACKEND_MPI_P2P`.
logical(kind=c_bool),	public		::	enable_datatype_backend			Should `DTFFT_BACKEND_MPI_DATATYPE` be considered for autotuning when `effort` is `DTFFT_PATIENT` or `DTFFT_EXHAUSTIVE`. Default is true. This option only works when `platform` is `DTFFT_PLATFORM_HOST`. When `platform` is `DTFFT_PLATFORM_CUDA`, `DTFFT_BACKEND_MPI_DATATYPE` is always disabled during autotuning.
logical(kind=c_bool),	public		::	enable_mpi_backends			Should MPI Backends be enabled when `effort` is `DTFFT_PATIENT` or `DTFFT_EXHAUSTIVE`. Default is false. The following applies only to CUDA builds. MPI Backends are disabled by default during autotuning process due to OpenMPI Bug https://github.com/open-mpi/ompi/issues/12849 It was noticed that during plan autotuning GPU memory not being freed completely. For example: 1024x1024x512 C2C, double precision, single GPU, using Z-slab optimization, with MPI backends enabled, plan autotuning will leak 8Gb GPU memory. Without Z-slab optimization, running on 4 GPUs, will leak 24Gb on each of the GPUs. One of the workarounds is to disable MPI Backends by default, which is done here. Other is to pass “–mca btl_smcuda_use_cuda_ipc 0” to `mpiexec`, but it was noticed that disabling CUDA IPC seriously affects overall performance of MPI algorithms
logical(kind=c_bool),	public		::	enable_pipelined_backends			Should pipelined backends be enabled when `effort` is `DTFFT_PATIENT` or `DTFFT_EXHAUSTIVE`. Default is true.
logical(kind=c_bool),	public		::	enable_nccl_backends			Should NCCL Backends be enabled when `effort` is `DTFFT_PATIENT` or `DTFFT_EXHAUSTIVE`. Default is true. This option is only defined when dtFFT is built with CUDA support.
logical(kind=c_bool),	public		::	enable_nvshmem_backends			Should NVSHMEM Backends be enabled when `effort` is `DTFFT_PATIENT` or `DTFFT_EXHAUSTIVE`. Default is true. This option is only defined when dtFFT is built with CUDA support.
logical(kind=c_bool),	public		::	enable_kernel_autotune			Should dtFFT try to optimize kernel launch parameters during plan creation when `effort` is below `DTFFT_EXHAUSTIVE`. Default is false. Kernel optimization is always enabled for `DTFFT_EXHAUSTIVE` effort level. Setting this option to true enables kernel optimization for lower effort levels (`DTFFT_ESTIMATE`, `DTFFT_MEASURE`, `DTFFT_PATIENT`). This may increase plan creation time but can improve runtime performance. Since kernel optimization is performed without data transfers, the time increase is usually minimal.
logical(kind=c_bool),	public		::	enable_fourier_reshape			Should dtFFT execute reshapes from pencils to bricks and vice versa in Fourier space during calls to `execute`. Default is false. When enabled, data will be in brick layout in Fourier space, which may be useful for certain operations between forward and backward transforms. However, this requires additional data transpositions and will reduce overall FFT performance.

Constructor

public interface dtfft_config_t

Interface to create a new configuration

private pure function config_constructor(enable_log, enable_z_slab, enable_y_slab, n_measure_warmup_iters, n_measure_iters, platform, stream, backend, reshape_backend, enable_datatype_backend, enable_mpi_backends, enable_pipelined_backends, enable_nccl_backends, enable_nvshmem_backends, enable_kernel_autotune, enable_fourier_reshape) result(config)

Creates a new configuration

Arguments

Type	Intent	Optional	Attributes		Name
logical,	intent(in),	optional		::	enable_log	Should dtFFT print additional information during plan creation or not.
logical,	intent(in),	optional		::	enable_z_slab	Should dtFFT use Z-slab optimization or not.
logical,	intent(in),	optional		::	enable_y_slab	Should dtFFT use Y-slab optimization or not.
integer(kind=int32),	intent(in),	optional		::	n_measure_warmup_iters	Number of warmup iterations for measurements
integer(kind=int32),	intent(in),	optional		::	n_measure_iters	Number of measurement iterations
type(dtfft_platform_t),	intent(in),	optional		::	platform	Selects platform to execute plan.
type(dtfft_stream_t),	intent(in),	optional		::	stream	Main CUDA stream that will be used in dtFFT.
type(dtfft_backend_t),	intent(in),	optional		::	backend	Backend that will be used by dtFFT when `effort` is `DTFFT_ESTIMATE` or `DTFFT_MEASURE`.
type(dtfft_backend_t),	intent(in),	optional		::	reshape_backend	Backend that will be used by dtFFT for data reshaping from bricks to pencils and vice versa when `effort` is `DTFFT_ESTIMATE` or `DTFFT_MEASURE`.
logical,	intent(in),	optional		::	enable_datatype_backend	Should `DTFFT_BACKEND_MPI_DATATYPE` be enabled when `effort` is `DTFFT_PATIENT` or not.
logical,	intent(in),	optional		::	enable_mpi_backends	Should MPI Backends be enabled when `effort` is `DTFFT_PATIENT` or not.
logical,	intent(in),	optional		::	enable_pipelined_backends	Should pipelined backends be enabled when `effort` is `DTFFT_PATIENT` or not.
logical,	intent(in),	optional		::	enable_nccl_backends	Should NCCL Backends be enabled when `effort` is `DTFFT_PATIENT` or not.
logical,	intent(in),	optional		::	enable_nvshmem_backends	Should NVSHMEM Backends be enabled when `effort` is `DTFFT_PATIENT` or not.
logical,	intent(in),	optional		::	enable_kernel_autotune	Should dtFFT try to autotune transpose/packing/unpacking kernels size during autotune process or not.
logical,	intent(in),	optional		::	enable_fourier_reshape	Should dtFFT execute reshapes from pencils to bricks and vice versa in Fourier space during calls to `execute` or not.

Return Value type(dtfft_config_t)

Constructed dtFFT config ready to be set by call to dtfft_set_config