Chris@10: Next: Load balancing, Chris@10: Previous: MPI Data Distribution, Chris@10: Up: MPI Data Distribution Chris@10:
As with the planner interface, the ‘fftw_mpi_local_size’ Chris@10: distribution interface is broken into basic and advanced Chris@10: (‘_many’) interfaces, where the latter allows you to specify the Chris@10: block size manually and also to request block sizes when computing Chris@10: multiple transforms simultaneously. These functions are documented Chris@10: more exhaustively by the FFTW MPI Reference, but we summarize the Chris@10: basic ideas here using a couple of two-dimensional examples. Chris@10: Chris@10:
For the 100 × 200 complex-DFT example, above, we would find Chris@10: the distribution by calling the following function in the basic Chris@10: interface: Chris@10: Chris@10:
ptrdiff_t fftw_mpi_local_size_2d(ptrdiff_t n0, ptrdiff_t n1, MPI_Comm comm, Chris@10: ptrdiff_t *local_n0, ptrdiff_t *local_0_start); Chris@10:Chris@10:
Chris@10: Given the total size of the data to be transformed (here, n0 =
Chris@10: 100 and n1 = 200) and an MPI communicator (comm), this
Chris@10: function provides three numbers.
Chris@10:
Chris@10:
First, it describes the shape of the local data: the current process
Chris@10: should store a local_n0 by n1 slice of the overall
Chris@10: dataset, in row-major order (n1 dimension contiguous), starting
Chris@10: at index local_0_start. That is, if the total dataset is
Chris@10: viewed as a n0 by n1 matrix, the current process should
Chris@10: store the rows local_0_start to
Chris@10: local_0_start+local_n0-1. Obviously, if you are running with
Chris@10: only a single MPI process, that process will store the entire array:
Chris@10: local_0_start will be zero and local_n0 will be
Chris@10: n0. See Row-major Format.
Chris@10:
Chris@10:
Chris@10:
Second, the return value is the total number of data elements (e.g.,
Chris@10: complex numbers for a complex DFT) that should be allocated for the
Chris@10: input and output arrays on the current process (ideally with
Chris@10: fftw_malloc or an ‘fftw_alloc’ function, to ensure optimal
Chris@10: alignment). It might seem that this should always be equal to
Chris@10: local_n0 * n1, but this is not the case. FFTW's
Chris@10: distributed FFT algorithms require data redistributions at
Chris@10: intermediate stages of the transform, and in some circumstances this
Chris@10: may require slightly larger local storage. This is discussed in more
Chris@10: detail below, under Load balancing.
Chris@10:
Chris@10:
Chris@10:
The advanced-interface ‘local_size’ function for multidimensional
Chris@10: transforms returns the same three things (local_n0,
Chris@10: local_0_start, and the total number of elements to allocate),
Chris@10: but takes more inputs:
Chris@10:
Chris@10:
ptrdiff_t fftw_mpi_local_size_many(int rnk, const ptrdiff_t *n, Chris@10: ptrdiff_t howmany, Chris@10: ptrdiff_t block0, Chris@10: MPI_Comm comm, Chris@10: ptrdiff_t *local_n0, Chris@10: ptrdiff_t *local_0_start); Chris@10:Chris@10:
Chris@10: The two-dimensional case above corresponds to rnk = 2 and an
Chris@10: array n of length 2 with n[0] = n0 and n[1] = n1.
Chris@10: This routine is for any rnk > 1; one-dimensional transforms
Chris@10: have their own interface because they work slightly differently, as
Chris@10: discussed below.
Chris@10:
Chris@10:
First, the advanced interface allows you to perform multiple
Chris@10: transforms at once, of interleaved data, as specified by the
Chris@10: howmany parameter. (hoamany is 1 for a single
Chris@10: transform.)
Chris@10:
Chris@10:
Second, here you can specify your desired block size in the n0
Chris@10: dimension, block0. To use FFTW's default block size, pass
Chris@10: FFTW_MPI_DEFAULT_BLOCK (0) for block0. Otherwise, on
Chris@10: P processes, FFTW will return local_n0 equal to
Chris@10: block0 on the first P / block0 processes (rounded down),
Chris@10: return local_n0 equal to n0 - block0 * (P / block0) on
Chris@10: the next process, and local_n0 equal to zero on any remaining
Chris@10: processes. In general, we recommend using the default block size
Chris@10: (which corresponds to n0 / P, rounded up).
Chris@10:
Chris@10:
Chris@10:
For example, suppose you have P = 4 processes and n0 =
Chris@10: 21. The default will be a block size of 6, which will give
Chris@10: local_n0 = 6 on the first three processes and local_n0 =
Chris@10: 3 on the last process. Instead, however, you could specify
Chris@10: block0 = 5 if you wanted, which would give local_n0 = 5
Chris@10: on processes 0 to 2, local_n0 = 6 on process 3. (This choice,
Chris@10: while it may look superficially more “balanced,” has the same
Chris@10: critical path as FFTW's default but requires more communications.)
Chris@10:
Chris@10:
Chris@10: