1 /* 2 * Copyright (c) 2003, 2007-14 Matteo Frigo 3 * Copyright (c) 2003, 2007-14 Massachusetts Institute of Technology 4 * 5 * The following statement of license applies *only* to this header file, 6 * and *not* to the other files distributed with FFTW or derived therefrom: 7 * 8 * Redistribution and use in source and binary forms, with or without 9 * modification, are permitted provided that the following conditions 10 * are met: 11 * 12 * 1. Redistributions of source code must retain the above copyright 13 * notice, this list of conditions and the following disclaimer. 14 * 15 * 2. Redistributions in binary form must reproduce the above copyright 16 * notice, this list of conditions and the following disclaimer in the 17 * documentation and/or other materials provided with the distribution. 18 * 19 * THIS SOFTWARE IS PROVIDED BY THE AUTHOR ``AS IS'' AND ANY EXPRESS 20 * OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED 21 * WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE 22 * ARE DISCLAIMED. IN NO EVENT SHALL THE AUTHOR BE LIABLE FOR ANY 23 * DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL 24 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE 25 * GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS 26 * INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, 27 * WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING 28 * NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF THIS 29 * SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE. 30 */ 31 32 /***************************** NOTE TO USERS ********************************* 33 * 34 * THIS IS A HEADER FILE, NOT A MANUAL 35 * 36 * If you want to know how to use FFTW, please read the manual, 37 * online at http://www.fftw.org/doc/ and also included with FFTW. 38 * For a quick start, see the manual's tutorial section. 39 * 40 * (Reading header files to learn how to use a library is a habit 41 * stemming from code lacking a proper manual. Arguably, it's a 42 * *bad* habit in most cases, because header files can contain 43 * interfaces that are not part of the public, stable API.) 44 * 45 ****************************************************************************/ 46 47 #ifndef FFTW3_MPI_H 48 #define FFTW3_MPI_H 49 50 #include <fftw3.h> 51 #include <mpi.h> 52 53 #ifdef __cplusplus 54 extern "C" 55 { 56 #endif /* __cplusplus */ 57 58 struct fftw_mpi_ddim_do_not_use_me { 59 ptrdiff_t n; /* dimension size */ 60 ptrdiff_t ib; /* input block */ 61 ptrdiff_t ob; /* output block */ 62 }; 63 64 /* 65 huge second-order macro that defines prototypes for all API 66 functions. We expand this macro for each supported precision 67 68 XM: name-mangling macro (MPI) 69 X: name-mangling macro (serial) 70 R: real data type 71 C: complex data type 72 */ 73 74 #define FFTW_MPI_DEFINE_API(XM, X, R, C) \ 75 \ 76 typedef struct fftw_mpi_ddim_do_not_use_me XM(ddim); \ 77 \ 78 FFTW_EXTERN void XM(init)(void); \ 79 FFTW_EXTERN void XM(cleanup)(void); \ 80 \ 81 FFTW_EXTERN ptrdiff_t XM(local_size_many_transposed) \ 82 (int rnk, const ptrdiff_t *n, ptrdiff_t howmany, \ 83 ptrdiff_t block0, ptrdiff_t block1, MPI_Comm comm, \ 84 ptrdiff_t *local_n0, ptrdiff_t *local_0_start, \ 85 ptrdiff_t *local_n1, ptrdiff_t *local_1_start); \ 86 FFTW_EXTERN ptrdiff_t XM(local_size_many) \ 87 (int rnk, const ptrdiff_t *n, ptrdiff_t howmany, \ 88 ptrdiff_t block0, MPI_Comm comm, \ 89 ptrdiff_t *local_n0, ptrdiff_t *local_0_start); \ 90 FFTW_EXTERN ptrdiff_t XM(local_size_transposed) \ 91 (int rnk, const ptrdiff_t *n, MPI_Comm comm, \ 92 ptrdiff_t *local_n0, ptrdiff_t *local_0_start, \ 93 ptrdiff_t *local_n1, ptrdiff_t *local_1_start); \ 94 FFTW_EXTERN ptrdiff_t XM(local_size) \ 95 (int rnk, const ptrdiff_t *n, MPI_Comm comm, \ 96 ptrdiff_t *local_n0, ptrdiff_t *local_0_start); \ 97 FFTW_EXTERN ptrdiff_t XM(local_size_many_1d)( \ 98 ptrdiff_t n0, ptrdiff_t howmany, \ 99 MPI_Comm comm, int sign, unsigned flags, \ 100 ptrdiff_t *local_ni, ptrdiff_t *local_i_start, \ 101 ptrdiff_t *local_no, ptrdiff_t *local_o_start); \ 102 FFTW_EXTERN ptrdiff_t XM(local_size_1d)( \ 103 ptrdiff_t n0, MPI_Comm comm, int sign, unsigned flags, \ 104 ptrdiff_t *local_ni, ptrdiff_t *local_i_start, \ 105 ptrdiff_t *local_no, ptrdiff_t *local_o_start); \ 106 FFTW_EXTERN ptrdiff_t XM(local_size_2d)( \ 107 ptrdiff_t n0, ptrdiff_t n1, MPI_Comm comm, \ 108 ptrdiff_t *local_n0, ptrdiff_t *local_0_start); \ 109 FFTW_EXTERN ptrdiff_t XM(local_size_2d_transposed)( \ 110 ptrdiff_t n0, ptrdiff_t n1, MPI_Comm comm, \ 111 ptrdiff_t *local_n0, ptrdiff_t *local_0_start, \ 112 ptrdiff_t *local_n1, ptrdiff_t *local_1_start); \ 113 FFTW_EXTERN ptrdiff_t XM(local_size_3d)( \ 114 ptrdiff_t n0, ptrdiff_t n1, ptrdiff_t n2, MPI_Comm comm, \ 115 ptrdiff_t *local_n0, ptrdiff_t *local_0_start); \ 116 FFTW_EXTERN ptrdiff_t XM(local_size_3d_transposed)( \ 117 ptrdiff_t n0, ptrdiff_t n1, ptrdiff_t n2, MPI_Comm comm, \ 118 ptrdiff_t *local_n0, ptrdiff_t *local_0_start, \ 119 ptrdiff_t *local_n1, ptrdiff_t *local_1_start); \ 120 \ 121 FFTW_EXTERN X(plan) XM(plan_many_transpose) \ 122 (ptrdiff_t n0, ptrdiff_t n1, \ 123 ptrdiff_t howmany, ptrdiff_t block0, ptrdiff_t block1, \ 124 R *in, R *out, MPI_Comm comm, unsigned flags); \ 125 FFTW_EXTERN X(plan) XM(plan_transpose) \ 126 (ptrdiff_t n0, ptrdiff_t n1, \ 127 R *in, R *out, MPI_Comm comm, unsigned flags); \ 128 \ 129 FFTW_EXTERN X(plan) XM(plan_many_dft) \ 130 (int rnk, const ptrdiff_t *n, ptrdiff_t howmany, \ 131 ptrdiff_t block, ptrdiff_t tblock, C *in, C *out, \ 132 MPI_Comm comm, int sign, unsigned flags); \ 133 FFTW_EXTERN X(plan) XM(plan_dft) \ 134 (int rnk, const ptrdiff_t *n, C *in, C *out, \ 135 MPI_Comm comm, int sign, unsigned flags); \ 136 FFTW_EXTERN X(plan) XM(plan_dft_1d) \ 137 (ptrdiff_t n0, C *in, C *out, \ 138 MPI_Comm comm, int sign, unsigned flags); \ 139 FFTW_EXTERN X(plan) XM(plan_dft_2d) \ 140 (ptrdiff_t n0, ptrdiff_t n1, C *in, C *out, \ 141 MPI_Comm comm, int sign, unsigned flags); \ 142 FFTW_EXTERN X(plan) XM(plan_dft_3d) \ 143 (ptrdiff_t n0, ptrdiff_t n1, ptrdiff_t n2, C *in, C *out, \ 144 MPI_Comm comm, int sign, unsigned flags); \ 145 \ 146 FFTW_EXTERN X(plan) XM(plan_many_r2r) \ 147 (int rnk, const ptrdiff_t *n, ptrdiff_t howmany, \ 148 ptrdiff_t iblock, ptrdiff_t oblock, R *in, R *out, \ 149 MPI_Comm comm, const X(r2r_kind) *kind, unsigned flags); \ 150 FFTW_EXTERN X(plan) XM(plan_r2r) \ 151 (int rnk, const ptrdiff_t *n, R *in, R *out, \ 152 MPI_Comm comm, const X(r2r_kind) *kind, unsigned flags); \ 153 FFTW_EXTERN X(plan) XM(plan_r2r_2d) \ 154 (ptrdiff_t n0, ptrdiff_t n1, R *in, R *out, MPI_Comm comm, \ 155 X(r2r_kind) kind0, X(r2r_kind) kind1, unsigned flags); \ 156 FFTW_EXTERN X(plan) XM(plan_r2r_3d) \ 157 (ptrdiff_t n0, ptrdiff_t n1, ptrdiff_t n2, \ 158 R *in, R *out, MPI_Comm comm, X(r2r_kind) kind0, \ 159 X(r2r_kind) kind1, X(r2r_kind) kind2, unsigned flags); \ 160 \ 161 FFTW_EXTERN X(plan) XM(plan_many_dft_r2c) \ 162 (int rnk, const ptrdiff_t *n, ptrdiff_t howmany, \ 163 ptrdiff_t iblock, ptrdiff_t oblock, R *in, C *out, \ 164 MPI_Comm comm, unsigned flags); \ 165 FFTW_EXTERN X(plan) XM(plan_dft_r2c) \ 166 (int rnk, const ptrdiff_t *n, R *in, C *out, \ 167 MPI_Comm comm, unsigned flags); \ 168 FFTW_EXTERN X(plan) XM(plan_dft_r2c_2d) \ 169 (ptrdiff_t n0, ptrdiff_t n1, R *in, C *out, \ 170 MPI_Comm comm, unsigned flags); \ 171 FFTW_EXTERN X(plan) XM(plan_dft_r2c_3d) \ 172 (ptrdiff_t n0, ptrdiff_t n1, ptrdiff_t n2, R *in, C *out, \ 173 MPI_Comm comm, unsigned flags); \ 174 \ 175 FFTW_EXTERN X(plan) XM(plan_many_dft_c2r) \ 176 (int rnk, const ptrdiff_t *n, ptrdiff_t howmany, \ 177 ptrdiff_t iblock, ptrdiff_t oblock, C *in, R *out, \ 178 MPI_Comm comm, unsigned flags); \ 179 FFTW_EXTERN X(plan) XM(plan_dft_c2r) \ 180 (int rnk, const ptrdiff_t *n, C *in, R *out, \ 181 MPI_Comm comm, unsigned flags); \ 182 FFTW_EXTERN X(plan) XM(plan_dft_c2r_2d) \ 183 (ptrdiff_t n0, ptrdiff_t n1, C *in, R *out, \ 184 MPI_Comm comm, unsigned flags); \ 185 FFTW_EXTERN X(plan) XM(plan_dft_c2r_3d) \ 186 (ptrdiff_t n0, ptrdiff_t n1, ptrdiff_t n2, C *in, R *out, \ 187 MPI_Comm comm, unsigned flags); \ 188 \ 189 FFTW_EXTERN void XM(gather_wisdom)(MPI_Comm comm_); \ 190 FFTW_EXTERN void XM(broadcast_wisdom)(MPI_Comm comm_); \ 191 \ 192 FFTW_EXTERN void XM(execute_dft)(X(plan) p, C *in, C *out); \ 193 FFTW_EXTERN void XM(execute_dft_r2c)(X(plan) p, R *in, C *out); \ 194 FFTW_EXTERN void XM(execute_dft_c2r)(X(plan) p, C *in, R *out); \ 195 FFTW_EXTERN void XM(execute_r2r)(X(plan) p, R *in, R *out); 196 197 198 199 /* end of FFTW_MPI_DEFINE_API macro */ 200 201 #define FFTW_MPI_MANGLE_DOUBLE(name) FFTW_MANGLE_DOUBLE(FFTW_CONCAT(mpi_,name)) 202 #define FFTW_MPI_MANGLE_FLOAT(name) FFTW_MANGLE_FLOAT(FFTW_CONCAT(mpi_,name)) 203 #define FFTW_MPI_MANGLE_LONG_DOUBLE(name) FFTW_MANGLE_LONG_DOUBLE(FFTW_CONCAT(mpi_,name)) 204 205 FFTW_MPI_DEFINE_API(FFTW_MPI_MANGLE_DOUBLE, FFTW_MANGLE_DOUBLE, double, fftw_complex) 206 FFTW_MPI_DEFINE_API(FFTW_MPI_MANGLE_FLOAT, FFTW_MANGLE_FLOAT, float, fftwf_complex) 207 FFTW_MPI_DEFINE_API(FFTW_MPI_MANGLE_LONG_DOUBLE, FFTW_MANGLE_LONG_DOUBLE, long double, fftwl_complex) 208 209 #define FFTW_MPI_DEFAULT_BLOCK (0) 210 211 /* MPI-specific flags */ 212 #define FFTW_MPI_SCRAMBLED_IN (1U << 27) 213 #define FFTW_MPI_SCRAMBLED_OUT (1U << 28) 214 #define FFTW_MPI_TRANSPOSED_IN (1U << 29) 215 #define FFTW_MPI_TRANSPOSED_OUT (1U << 30) 216 217 #ifdef __cplusplus 218 } /* extern "C" */ 219 #endif /* __cplusplus */ 220 221 #endif /* FFTW3_MPI_H */ 222