fftw3: Using MPI Plans
6.12.3 Using MPI Plans
----------------------
Once an MPI plan is created, you can execute and destroy it using
'fftw_execute', 'fftw_destroy_plan', and the other functions in the
serial interface that operate on generic plans (
Using Plans).
The 'fftw_execute' and 'fftw_destroy_plan' functions, applied to MPI
plans, are _collective_ calls: they must be called for all processes in
the communicator that was used to create the plan.
You must _not_ use the serial new-array plan-execution functions
'fftw_execute_dft' and so on (New-array Execute Functions) with
MPI plans. Such functions are specialized to the problem type, and
there are specific new-array execute functions for MPI plans:
void fftw_mpi_execute_dft(fftw_plan p, fftw_complex *in, fftw_complex *out);
void fftw_mpi_execute_dft_r2c(fftw_plan p, double *in, fftw_complex *out);
void fftw_mpi_execute_dft_c2r(fftw_plan p, fftw_complex *in, double *out);
void fftw_mpi_execute_r2r(fftw_plan p, double *in, double *out);
These functions have the same restrictions as those of the serial
new-array execute functions. They are _always_ safe to apply to the
_same_ 'in' and 'out' arrays that were used to create the plan. They
can only be applied to new arrarys if those arrays have the same types,
dimensions, in-placeness, and alignment as the original arrays, where
the best way to ensure the same alignment is to use FFTW's 'fftw_malloc'
DONTPRINTYET and related allocation functions for all arrays (Memory
Allocation). Note that distributed transposes (*noteFFTW MPI
DONTPRINTYET and related allocation functions for all arrays (Memory
Allocation). Note that distributed transposes (FFTW MPI
Transposes) use 'fftw_mpi_execute_r2r', since they count as rank-zero
r2r plans from FFTW's perspective.
Info Catalog
fftw3: MPI Initialization
fftw3: FFTW MPI Reference
fftw3: MPI Data Distribution Functions