annotate fft/fftw/fftw-3.3.4/dft/scalar/codelets/n1_15.c @ 40:223f770b5341 kissfft-double tip

Try a double-precision kissfft
author Chris Cannam
date Wed, 07 Sep 2016 10:40:32 +0100
parents 26056e866c29
children
rev   line source
Chris@19 1 /*
Chris@19 2 * Copyright (c) 2003, 2007-14 Matteo Frigo
Chris@19 3 * Copyright (c) 2003, 2007-14 Massachusetts Institute of Technology
Chris@19 4 *
Chris@19 5 * This program is free software; you can redistribute it and/or modify
Chris@19 6 * it under the terms of the GNU General Public License as published by
Chris@19 7 * the Free Software Foundation; either version 2 of the License, or
Chris@19 8 * (at your option) any later version.
Chris@19 9 *
Chris@19 10 * This program is distributed in the hope that it will be useful,
Chris@19 11 * but WITHOUT ANY WARRANTY; without even the implied warranty of
Chris@19 12 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
Chris@19 13 * GNU General Public License for more details.
Chris@19 14 *
Chris@19 15 * You should have received a copy of the GNU General Public License
Chris@19 16 * along with this program; if not, write to the Free Software
Chris@19 17 * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
Chris@19 18 *
Chris@19 19 */
Chris@19 20
Chris@19 21 /* This file was automatically generated --- DO NOT EDIT */
Chris@19 22 /* Generated on Tue Mar 4 13:45:49 EST 2014 */
Chris@19 23
Chris@19 24 #include "codelet-dft.h"
Chris@19 25
Chris@19 26 #ifdef HAVE_FMA
Chris@19 27
Chris@19 28 /* Generated by: ../../../genfft/gen_notw.native -fma -reorder-insns -schedule-for-pipeline -compact -variables 4 -pipeline-latency 4 -n 15 -name n1_15 -include n.h */
Chris@19 29
Chris@19 30 /*
Chris@19 31 * This function contains 156 FP additions, 84 FP multiplications,
Chris@19 32 * (or, 72 additions, 0 multiplications, 84 fused multiply/add),
Chris@19 33 * 75 stack variables, 6 constants, and 60 memory accesses
Chris@19 34 */
Chris@19 35 #include "n.h"
Chris@19 36
Chris@19 37 static void n1_15(const R *ri, const R *ii, R *ro, R *io, stride is, stride os, INT v, INT ivs, INT ovs)
Chris@19 38 {
Chris@19 39 DK(KP951056516, +0.951056516295153572116439333379382143405698634);
Chris@19 40 DK(KP559016994, +0.559016994374947424102293417182819058860154590);
Chris@19 41 DK(KP618033988, +0.618033988749894848204586834365638117720309180);
Chris@19 42 DK(KP250000000, +0.250000000000000000000000000000000000000000000);
Chris@19 43 DK(KP866025403, +0.866025403784438646763723170752936183471402627);
Chris@19 44 DK(KP500000000, +0.500000000000000000000000000000000000000000000);
Chris@19 45 {
Chris@19 46 INT i;
Chris@19 47 for (i = v; i > 0; i = i - 1, ri = ri + ivs, ii = ii + ivs, ro = ro + ovs, io = io + ovs, MAKE_VOLATILE_STRIDE(60, is), MAKE_VOLATILE_STRIDE(60, os)) {
Chris@19 48 E T1r, T1g, T14, T13;
Chris@19 49 {
Chris@19 50 E T5, T2l, Tx, TV, T1z, T1X, T2s, Tr, T24, TT, T2e, T2n, T1Z, T1Q, T1B;
Chris@19 51 E T11, T1H, TW, T2t, Tg, TX, T25, TI, T2h, T2m, T1Y, T1T, T1A;
Chris@19 52 {
Chris@19 53 E T1, T1v, T2, T3, Tu, Tv, TZ, T10;
Chris@19 54 T1 = ri[0];
Chris@19 55 T1v = ii[0];
Chris@19 56 T2 = ri[WS(is, 5)];
Chris@19 57 T3 = ri[WS(is, 10)];
Chris@19 58 Tu = ii[WS(is, 5)];
Chris@19 59 Tv = ii[WS(is, 10)];
Chris@19 60 {
Chris@19 61 E T1k, Tm, TM, TJ, Tl, T2c, T1j, T1m, TP, T1p, Tp, TQ;
Chris@19 62 {
Chris@19 63 E Th, T1h, TK, TL, Tk, Tn, To, T1i;
Chris@19 64 {
Chris@19 65 E Ti, Tj, T1y, T4;
Chris@19 66 Th = ri[WS(is, 6)];
Chris@19 67 T1y = T3 - T2;
Chris@19 68 T4 = T2 + T3;
Chris@19 69 {
Chris@19 70 E T1w, Tw, Tt, T1x;
Chris@19 71 T1w = Tu + Tv;
Chris@19 72 Tw = Tu - Tv;
Chris@19 73 Ti = ri[WS(is, 11)];
Chris@19 74 T5 = T1 + T4;
Chris@19 75 Tt = FNMS(KP500000000, T4, T1);
Chris@19 76 T2l = T1v + T1w;
Chris@19 77 T1x = FNMS(KP500000000, T1w, T1v);
Chris@19 78 Tx = FNMS(KP866025403, Tw, Tt);
Chris@19 79 TV = FMA(KP866025403, Tw, Tt);
Chris@19 80 T1z = FMA(KP866025403, T1y, T1x);
Chris@19 81 T1X = FNMS(KP866025403, T1y, T1x);
Chris@19 82 Tj = ri[WS(is, 1)];
Chris@19 83 }
Chris@19 84 T1h = ii[WS(is, 6)];
Chris@19 85 TK = ii[WS(is, 11)];
Chris@19 86 TL = ii[WS(is, 1)];
Chris@19 87 Tk = Ti + Tj;
Chris@19 88 T1k = Tj - Ti;
Chris@19 89 }
Chris@19 90 Tm = ri[WS(is, 9)];
Chris@19 91 TM = TK - TL;
Chris@19 92 T1i = TK + TL;
Chris@19 93 TJ = FNMS(KP500000000, Tk, Th);
Chris@19 94 Tl = Th + Tk;
Chris@19 95 Tn = ri[WS(is, 14)];
Chris@19 96 To = ri[WS(is, 4)];
Chris@19 97 T2c = T1h + T1i;
Chris@19 98 T1j = FNMS(KP500000000, T1i, T1h);
Chris@19 99 T1m = ii[WS(is, 9)];
Chris@19 100 TP = ii[WS(is, 14)];
Chris@19 101 T1p = To - Tn;
Chris@19 102 Tp = Tn + To;
Chris@19 103 TQ = ii[WS(is, 4)];
Chris@19 104 }
Chris@19 105 {
Chris@19 106 E TN, TS, T1o, T2d;
Chris@19 107 {
Chris@19 108 E TO, T1n, TR, Tq;
Chris@19 109 TN = FNMS(KP866025403, TM, TJ);
Chris@19 110 TZ = FMA(KP866025403, TM, TJ);
Chris@19 111 TO = FNMS(KP500000000, Tp, Tm);
Chris@19 112 Tq = Tm + Tp;
Chris@19 113 T1n = TP + TQ;
Chris@19 114 TR = TP - TQ;
Chris@19 115 T2s = Tl - Tq;
Chris@19 116 Tr = Tl + Tq;
Chris@19 117 T10 = FMA(KP866025403, TR, TO);
Chris@19 118 TS = FNMS(KP866025403, TR, TO);
Chris@19 119 T1o = FNMS(KP500000000, T1n, T1m);
Chris@19 120 T2d = T1m + T1n;
Chris@19 121 }
Chris@19 122 {
Chris@19 123 E T1O, T1l, T1P, T1q;
Chris@19 124 T1O = FNMS(KP866025403, T1k, T1j);
Chris@19 125 T1l = FMA(KP866025403, T1k, T1j);
Chris@19 126 T24 = TN - TS;
Chris@19 127 TT = TN + TS;
Chris@19 128 T1P = FNMS(KP866025403, T1p, T1o);
Chris@19 129 T1q = FMA(KP866025403, T1p, T1o);
Chris@19 130 T2e = T2c - T2d;
Chris@19 131 T2n = T2c + T2d;
Chris@19 132 T1Z = T1O + T1P;
Chris@19 133 T1Q = T1O - T1P;
Chris@19 134 T1r = T1l - T1q;
Chris@19 135 T1B = T1l + T1q;
Chris@19 136 }
Chris@19 137 }
Chris@19 138 }
Chris@19 139 {
Chris@19 140 E T19, Tb, TB, Ty, Ta, T2f, T18, T1b, TE, T1e, Te, TF;
Chris@19 141 {
Chris@19 142 E T6, T16, Tz, TA, T9, T7, T8, Tc, Td, T17;
Chris@19 143 T6 = ri[WS(is, 3)];
Chris@19 144 T7 = ri[WS(is, 8)];
Chris@19 145 T11 = TZ + T10;
Chris@19 146 T1H = TZ - T10;
Chris@19 147 T8 = ri[WS(is, 13)];
Chris@19 148 T16 = ii[WS(is, 3)];
Chris@19 149 Tz = ii[WS(is, 8)];
Chris@19 150 TA = ii[WS(is, 13)];
Chris@19 151 T9 = T7 + T8;
Chris@19 152 T19 = T8 - T7;
Chris@19 153 Tb = ri[WS(is, 12)];
Chris@19 154 TB = Tz - TA;
Chris@19 155 T17 = Tz + TA;
Chris@19 156 Ty = FNMS(KP500000000, T9, T6);
Chris@19 157 Ta = T6 + T9;
Chris@19 158 Tc = ri[WS(is, 2)];
Chris@19 159 Td = ri[WS(is, 7)];
Chris@19 160 T2f = T16 + T17;
Chris@19 161 T18 = FNMS(KP500000000, T17, T16);
Chris@19 162 T1b = ii[WS(is, 12)];
Chris@19 163 TE = ii[WS(is, 2)];
Chris@19 164 T1e = Td - Tc;
Chris@19 165 Te = Tc + Td;
Chris@19 166 TF = ii[WS(is, 7)];
Chris@19 167 }
Chris@19 168 {
Chris@19 169 E TC, TH, T1d, T2g;
Chris@19 170 {
Chris@19 171 E TD, T1c, TG, Tf;
Chris@19 172 TC = FNMS(KP866025403, TB, Ty);
Chris@19 173 TW = FMA(KP866025403, TB, Ty);
Chris@19 174 TD = FNMS(KP500000000, Te, Tb);
Chris@19 175 Tf = Tb + Te;
Chris@19 176 T1c = TE + TF;
Chris@19 177 TG = TE - TF;
Chris@19 178 T2t = Ta - Tf;
Chris@19 179 Tg = Ta + Tf;
Chris@19 180 TX = FMA(KP866025403, TG, TD);
Chris@19 181 TH = FNMS(KP866025403, TG, TD);
Chris@19 182 T1d = FNMS(KP500000000, T1c, T1b);
Chris@19 183 T2g = T1b + T1c;
Chris@19 184 }
Chris@19 185 {
Chris@19 186 E T1R, T1a, T1S, T1f;
Chris@19 187 T1R = FNMS(KP866025403, T19, T18);
Chris@19 188 T1a = FMA(KP866025403, T19, T18);
Chris@19 189 T25 = TC - TH;
Chris@19 190 TI = TC + TH;
Chris@19 191 T1S = FNMS(KP866025403, T1e, T1d);
Chris@19 192 T1f = FMA(KP866025403, T1e, T1d);
Chris@19 193 T2h = T2f - T2g;
Chris@19 194 T2m = T2f + T2g;
Chris@19 195 T1Y = T1R + T1S;
Chris@19 196 T1T = T1R - T1S;
Chris@19 197 T1g = T1a - T1f;
Chris@19 198 T1A = T1a + T1f;
Chris@19 199 }
Chris@19 200 }
Chris@19 201 }
Chris@19 202 }
Chris@19 203 {
Chris@19 204 E TY, T1G, T1M, T1L, T2a, T29, Ts, T22, T21, T20;
Chris@19 205 T2a = Tg - Tr;
Chris@19 206 Ts = Tg + Tr;
Chris@19 207 TY = TW + TX;
Chris@19 208 T1G = TW - TX;
Chris@19 209 T29 = FNMS(KP250000000, Ts, T5);
Chris@19 210 ro[0] = T5 + Ts;
Chris@19 211 {
Chris@19 212 E T2q, T2p, T2o, TU;
Chris@19 213 T2o = T2m + T2n;
Chris@19 214 T2q = T2m - T2n;
Chris@19 215 {
Chris@19 216 E T2k, T2i, T2b, T2j;
Chris@19 217 T2k = FMA(KP618033988, T2e, T2h);
Chris@19 218 T2i = FNMS(KP618033988, T2h, T2e);
Chris@19 219 T2b = FNMS(KP559016994, T2a, T29);
Chris@19 220 T2j = FMA(KP559016994, T2a, T29);
Chris@19 221 ro[WS(os, 3)] = FMA(KP951056516, T2i, T2b);
Chris@19 222 ro[WS(os, 12)] = FNMS(KP951056516, T2i, T2b);
Chris@19 223 ro[WS(os, 6)] = FMA(KP951056516, T2k, T2j);
Chris@19 224 ro[WS(os, 9)] = FNMS(KP951056516, T2k, T2j);
Chris@19 225 T2p = FNMS(KP250000000, T2o, T2l);
Chris@19 226 }
Chris@19 227 io[0] = T2l + T2o;
Chris@19 228 TU = TI + TT;
Chris@19 229 T1M = TI - TT;
Chris@19 230 {
Chris@19 231 E T2r, T2v, T2w, T2u;
Chris@19 232 T2r = FNMS(KP559016994, T2q, T2p);
Chris@19 233 T2v = FMA(KP559016994, T2q, T2p);
Chris@19 234 T2w = FMA(KP618033988, T2s, T2t);
Chris@19 235 T2u = FNMS(KP618033988, T2t, T2s);
Chris@19 236 io[WS(os, 9)] = FMA(KP951056516, T2w, T2v);
Chris@19 237 io[WS(os, 6)] = FNMS(KP951056516, T2w, T2v);
Chris@19 238 io[WS(os, 12)] = FMA(KP951056516, T2u, T2r);
Chris@19 239 io[WS(os, 3)] = FNMS(KP951056516, T2u, T2r);
Chris@19 240 T1L = FNMS(KP250000000, TU, Tx);
Chris@19 241 }
Chris@19 242 ro[WS(os, 5)] = Tx + TU;
Chris@19 243 }
Chris@19 244 T20 = T1Y + T1Z;
Chris@19 245 T22 = T1Y - T1Z;
Chris@19 246 {
Chris@19 247 E T1N, T1V, T1W, T1U;
Chris@19 248 T1N = FNMS(KP559016994, T1M, T1L);
Chris@19 249 T1V = FMA(KP559016994, T1M, T1L);
Chris@19 250 T1W = FMA(KP618033988, T1Q, T1T);
Chris@19 251 T1U = FNMS(KP618033988, T1T, T1Q);
Chris@19 252 ro[WS(os, 11)] = FMA(KP951056516, T1W, T1V);
Chris@19 253 ro[WS(os, 14)] = FNMS(KP951056516, T1W, T1V);
Chris@19 254 ro[WS(os, 8)] = FMA(KP951056516, T1U, T1N);
Chris@19 255 ro[WS(os, 2)] = FNMS(KP951056516, T1U, T1N);
Chris@19 256 T21 = FNMS(KP250000000, T20, T1X);
Chris@19 257 }
Chris@19 258 io[WS(os, 5)] = T1X + T20;
Chris@19 259 {
Chris@19 260 E T1E, T1D, T1C, T12;
Chris@19 261 T1C = T1A + T1B;
Chris@19 262 T1E = T1A - T1B;
Chris@19 263 {
Chris@19 264 E T23, T27, T28, T26;
Chris@19 265 T23 = FNMS(KP559016994, T22, T21);
Chris@19 266 T27 = FMA(KP559016994, T22, T21);
Chris@19 267 T28 = FMA(KP618033988, T24, T25);
Chris@19 268 T26 = FNMS(KP618033988, T25, T24);
Chris@19 269 io[WS(os, 14)] = FMA(KP951056516, T28, T27);
Chris@19 270 io[WS(os, 11)] = FNMS(KP951056516, T28, T27);
Chris@19 271 io[WS(os, 8)] = FNMS(KP951056516, T26, T23);
Chris@19 272 io[WS(os, 2)] = FMA(KP951056516, T26, T23);
Chris@19 273 T1D = FNMS(KP250000000, T1C, T1z);
Chris@19 274 }
Chris@19 275 io[WS(os, 10)] = T1z + T1C;
Chris@19 276 T12 = TY + T11;
Chris@19 277 T14 = TY - T11;
Chris@19 278 {
Chris@19 279 E T1F, T1J, T1K, T1I;
Chris@19 280 T1F = FMA(KP559016994, T1E, T1D);
Chris@19 281 T1J = FNMS(KP559016994, T1E, T1D);
Chris@19 282 T1K = FNMS(KP618033988, T1G, T1H);
Chris@19 283 T1I = FMA(KP618033988, T1H, T1G);
Chris@19 284 io[WS(os, 13)] = FNMS(KP951056516, T1K, T1J);
Chris@19 285 io[WS(os, 7)] = FMA(KP951056516, T1K, T1J);
Chris@19 286 io[WS(os, 4)] = FMA(KP951056516, T1I, T1F);
Chris@19 287 io[WS(os, 1)] = FNMS(KP951056516, T1I, T1F);
Chris@19 288 T13 = FNMS(KP250000000, T12, TV);
Chris@19 289 }
Chris@19 290 ro[WS(os, 10)] = TV + T12;
Chris@19 291 }
Chris@19 292 }
Chris@19 293 }
Chris@19 294 {
Chris@19 295 E T1t, T15, T1s, T1u;
Chris@19 296 T1t = FNMS(KP559016994, T14, T13);
Chris@19 297 T15 = FMA(KP559016994, T14, T13);
Chris@19 298 T1s = FMA(KP618033988, T1r, T1g);
Chris@19 299 T1u = FNMS(KP618033988, T1g, T1r);
Chris@19 300 ro[WS(os, 13)] = FMA(KP951056516, T1u, T1t);
Chris@19 301 ro[WS(os, 7)] = FNMS(KP951056516, T1u, T1t);
Chris@19 302 ro[WS(os, 1)] = FMA(KP951056516, T1s, T15);
Chris@19 303 ro[WS(os, 4)] = FNMS(KP951056516, T1s, T15);
Chris@19 304 }
Chris@19 305 }
Chris@19 306 }
Chris@19 307 }
Chris@19 308
Chris@19 309 static const kdft_desc desc = { 15, "n1_15", {72, 0, 84, 0}, &GENUS, 0, 0, 0, 0 };
Chris@19 310
Chris@19 311 void X(codelet_n1_15) (planner *p) {
Chris@19 312 X(kdft_register) (p, n1_15, &desc);
Chris@19 313 }
Chris@19 314
Chris@19 315 #else /* HAVE_FMA */
Chris@19 316
Chris@19 317 /* Generated by: ../../../genfft/gen_notw.native -compact -variables 4 -pipeline-latency 4 -n 15 -name n1_15 -include n.h */
Chris@19 318
Chris@19 319 /*
Chris@19 320 * This function contains 156 FP additions, 56 FP multiplications,
Chris@19 321 * (or, 128 additions, 28 multiplications, 28 fused multiply/add),
Chris@19 322 * 69 stack variables, 6 constants, and 60 memory accesses
Chris@19 323 */
Chris@19 324 #include "n.h"
Chris@19 325
Chris@19 326 static void n1_15(const R *ri, const R *ii, R *ro, R *io, stride is, stride os, INT v, INT ivs, INT ovs)
Chris@19 327 {
Chris@19 328 DK(KP587785252, +0.587785252292473129168705954639072768597652438);
Chris@19 329 DK(KP951056516, +0.951056516295153572116439333379382143405698634);
Chris@19 330 DK(KP250000000, +0.250000000000000000000000000000000000000000000);
Chris@19 331 DK(KP559016994, +0.559016994374947424102293417182819058860154590);
Chris@19 332 DK(KP500000000, +0.500000000000000000000000000000000000000000000);
Chris@19 333 DK(KP866025403, +0.866025403784438646763723170752936183471402627);
Chris@19 334 {
Chris@19 335 INT i;
Chris@19 336 for (i = v; i > 0; i = i - 1, ri = ri + ivs, ii = ii + ivs, ro = ro + ovs, io = io + ovs, MAKE_VOLATILE_STRIDE(60, is), MAKE_VOLATILE_STRIDE(60, os)) {
Chris@19 337 E T5, T2l, Tx, TV, T1C, T20, Tl, Tq, Tr, TN, TS, TT, T2c, T2d, T2n;
Chris@19 338 E T1O, T1P, T22, T1l, T1q, T1w, TZ, T10, T11, Ta, Tf, Tg, TC, TH, TI;
Chris@19 339 E T2f, T2g, T2m, T1R, T1S, T21, T1a, T1f, T1v, TW, TX, TY;
Chris@19 340 {
Chris@19 341 E T1, T1z, T4, T1y, Tw, T1A, Tt, T1B;
Chris@19 342 T1 = ri[0];
Chris@19 343 T1z = ii[0];
Chris@19 344 {
Chris@19 345 E T2, T3, Tu, Tv;
Chris@19 346 T2 = ri[WS(is, 5)];
Chris@19 347 T3 = ri[WS(is, 10)];
Chris@19 348 T4 = T2 + T3;
Chris@19 349 T1y = KP866025403 * (T3 - T2);
Chris@19 350 Tu = ii[WS(is, 5)];
Chris@19 351 Tv = ii[WS(is, 10)];
Chris@19 352 Tw = KP866025403 * (Tu - Tv);
Chris@19 353 T1A = Tu + Tv;
Chris@19 354 }
Chris@19 355 T5 = T1 + T4;
Chris@19 356 T2l = T1z + T1A;
Chris@19 357 Tt = FNMS(KP500000000, T4, T1);
Chris@19 358 Tx = Tt - Tw;
Chris@19 359 TV = Tt + Tw;
Chris@19 360 T1B = FNMS(KP500000000, T1A, T1z);
Chris@19 361 T1C = T1y + T1B;
Chris@19 362 T20 = T1B - T1y;
Chris@19 363 }
Chris@19 364 {
Chris@19 365 E Th, Tk, TJ, T1h, T1i, T1j, TM, T1k, Tm, Tp, TO, T1m, T1n, T1o, TR;
Chris@19 366 E T1p;
Chris@19 367 {
Chris@19 368 E Ti, Tj, TK, TL;
Chris@19 369 Th = ri[WS(is, 6)];
Chris@19 370 Ti = ri[WS(is, 11)];
Chris@19 371 Tj = ri[WS(is, 1)];
Chris@19 372 Tk = Ti + Tj;
Chris@19 373 TJ = FNMS(KP500000000, Tk, Th);
Chris@19 374 T1h = KP866025403 * (Tj - Ti);
Chris@19 375 T1i = ii[WS(is, 6)];
Chris@19 376 TK = ii[WS(is, 11)];
Chris@19 377 TL = ii[WS(is, 1)];
Chris@19 378 T1j = TK + TL;
Chris@19 379 TM = KP866025403 * (TK - TL);
Chris@19 380 T1k = FNMS(KP500000000, T1j, T1i);
Chris@19 381 }
Chris@19 382 {
Chris@19 383 E Tn, To, TP, TQ;
Chris@19 384 Tm = ri[WS(is, 9)];
Chris@19 385 Tn = ri[WS(is, 14)];
Chris@19 386 To = ri[WS(is, 4)];
Chris@19 387 Tp = Tn + To;
Chris@19 388 TO = FNMS(KP500000000, Tp, Tm);
Chris@19 389 T1m = KP866025403 * (To - Tn);
Chris@19 390 T1n = ii[WS(is, 9)];
Chris@19 391 TP = ii[WS(is, 14)];
Chris@19 392 TQ = ii[WS(is, 4)];
Chris@19 393 T1o = TP + TQ;
Chris@19 394 TR = KP866025403 * (TP - TQ);
Chris@19 395 T1p = FNMS(KP500000000, T1o, T1n);
Chris@19 396 }
Chris@19 397 Tl = Th + Tk;
Chris@19 398 Tq = Tm + Tp;
Chris@19 399 Tr = Tl + Tq;
Chris@19 400 TN = TJ - TM;
Chris@19 401 TS = TO - TR;
Chris@19 402 TT = TN + TS;
Chris@19 403 T2c = T1i + T1j;
Chris@19 404 T2d = T1n + T1o;
Chris@19 405 T2n = T2c + T2d;
Chris@19 406 T1O = T1k - T1h;
Chris@19 407 T1P = T1p - T1m;
Chris@19 408 T22 = T1O + T1P;
Chris@19 409 T1l = T1h + T1k;
Chris@19 410 T1q = T1m + T1p;
Chris@19 411 T1w = T1l + T1q;
Chris@19 412 TZ = TJ + TM;
Chris@19 413 T10 = TO + TR;
Chris@19 414 T11 = TZ + T10;
Chris@19 415 }
Chris@19 416 {
Chris@19 417 E T6, T9, Ty, T16, T17, T18, TB, T19, Tb, Te, TD, T1b, T1c, T1d, TG;
Chris@19 418 E T1e;
Chris@19 419 {
Chris@19 420 E T7, T8, Tz, TA;
Chris@19 421 T6 = ri[WS(is, 3)];
Chris@19 422 T7 = ri[WS(is, 8)];
Chris@19 423 T8 = ri[WS(is, 13)];
Chris@19 424 T9 = T7 + T8;
Chris@19 425 Ty = FNMS(KP500000000, T9, T6);
Chris@19 426 T16 = KP866025403 * (T8 - T7);
Chris@19 427 T17 = ii[WS(is, 3)];
Chris@19 428 Tz = ii[WS(is, 8)];
Chris@19 429 TA = ii[WS(is, 13)];
Chris@19 430 T18 = Tz + TA;
Chris@19 431 TB = KP866025403 * (Tz - TA);
Chris@19 432 T19 = FNMS(KP500000000, T18, T17);
Chris@19 433 }
Chris@19 434 {
Chris@19 435 E Tc, Td, TE, TF;
Chris@19 436 Tb = ri[WS(is, 12)];
Chris@19 437 Tc = ri[WS(is, 2)];
Chris@19 438 Td = ri[WS(is, 7)];
Chris@19 439 Te = Tc + Td;
Chris@19 440 TD = FNMS(KP500000000, Te, Tb);
Chris@19 441 T1b = KP866025403 * (Td - Tc);
Chris@19 442 T1c = ii[WS(is, 12)];
Chris@19 443 TE = ii[WS(is, 2)];
Chris@19 444 TF = ii[WS(is, 7)];
Chris@19 445 T1d = TE + TF;
Chris@19 446 TG = KP866025403 * (TE - TF);
Chris@19 447 T1e = FNMS(KP500000000, T1d, T1c);
Chris@19 448 }
Chris@19 449 Ta = T6 + T9;
Chris@19 450 Tf = Tb + Te;
Chris@19 451 Tg = Ta + Tf;
Chris@19 452 TC = Ty - TB;
Chris@19 453 TH = TD - TG;
Chris@19 454 TI = TC + TH;
Chris@19 455 T2f = T17 + T18;
Chris@19 456 T2g = T1c + T1d;
Chris@19 457 T2m = T2f + T2g;
Chris@19 458 T1R = T19 - T16;
Chris@19 459 T1S = T1e - T1b;
Chris@19 460 T21 = T1R + T1S;
Chris@19 461 T1a = T16 + T19;
Chris@19 462 T1f = T1b + T1e;
Chris@19 463 T1v = T1a + T1f;
Chris@19 464 TW = Ty + TB;
Chris@19 465 TX = TD + TG;
Chris@19 466 TY = TW + TX;
Chris@19 467 }
Chris@19 468 {
Chris@19 469 E T2a, Ts, T29, T2i, T2k, T2e, T2h, T2j, T2b;
Chris@19 470 T2a = KP559016994 * (Tg - Tr);
Chris@19 471 Ts = Tg + Tr;
Chris@19 472 T29 = FNMS(KP250000000, Ts, T5);
Chris@19 473 T2e = T2c - T2d;
Chris@19 474 T2h = T2f - T2g;
Chris@19 475 T2i = FNMS(KP587785252, T2h, KP951056516 * T2e);
Chris@19 476 T2k = FMA(KP951056516, T2h, KP587785252 * T2e);
Chris@19 477 ro[0] = T5 + Ts;
Chris@19 478 T2j = T2a + T29;
Chris@19 479 ro[WS(os, 9)] = T2j - T2k;
Chris@19 480 ro[WS(os, 6)] = T2j + T2k;
Chris@19 481 T2b = T29 - T2a;
Chris@19 482 ro[WS(os, 12)] = T2b - T2i;
Chris@19 483 ro[WS(os, 3)] = T2b + T2i;
Chris@19 484 }
Chris@19 485 {
Chris@19 486 E T2q, T2o, T2p, T2u, T2w, T2s, T2t, T2v, T2r;
Chris@19 487 T2q = KP559016994 * (T2m - T2n);
Chris@19 488 T2o = T2m + T2n;
Chris@19 489 T2p = FNMS(KP250000000, T2o, T2l);
Chris@19 490 T2s = Tl - Tq;
Chris@19 491 T2t = Ta - Tf;
Chris@19 492 T2u = FNMS(KP587785252, T2t, KP951056516 * T2s);
Chris@19 493 T2w = FMA(KP951056516, T2t, KP587785252 * T2s);
Chris@19 494 io[0] = T2l + T2o;
Chris@19 495 T2v = T2q + T2p;
Chris@19 496 io[WS(os, 6)] = T2v - T2w;
Chris@19 497 io[WS(os, 9)] = T2w + T2v;
Chris@19 498 T2r = T2p - T2q;
Chris@19 499 io[WS(os, 3)] = T2r - T2u;
Chris@19 500 io[WS(os, 12)] = T2u + T2r;
Chris@19 501 }
Chris@19 502 {
Chris@19 503 E T1M, TU, T1L, T1U, T1W, T1Q, T1T, T1V, T1N;
Chris@19 504 T1M = KP559016994 * (TI - TT);
Chris@19 505 TU = TI + TT;
Chris@19 506 T1L = FNMS(KP250000000, TU, Tx);
Chris@19 507 T1Q = T1O - T1P;
Chris@19 508 T1T = T1R - T1S;
Chris@19 509 T1U = FNMS(KP587785252, T1T, KP951056516 * T1Q);
Chris@19 510 T1W = FMA(KP951056516, T1T, KP587785252 * T1Q);
Chris@19 511 ro[WS(os, 5)] = Tx + TU;
Chris@19 512 T1V = T1M + T1L;
Chris@19 513 ro[WS(os, 14)] = T1V - T1W;
Chris@19 514 ro[WS(os, 11)] = T1V + T1W;
Chris@19 515 T1N = T1L - T1M;
Chris@19 516 ro[WS(os, 2)] = T1N - T1U;
Chris@19 517 ro[WS(os, 8)] = T1N + T1U;
Chris@19 518 }
Chris@19 519 {
Chris@19 520 E T25, T23, T24, T1Z, T28, T1X, T1Y, T27, T26;
Chris@19 521 T25 = KP559016994 * (T21 - T22);
Chris@19 522 T23 = T21 + T22;
Chris@19 523 T24 = FNMS(KP250000000, T23, T20);
Chris@19 524 T1X = TN - TS;
Chris@19 525 T1Y = TC - TH;
Chris@19 526 T1Z = FNMS(KP587785252, T1Y, KP951056516 * T1X);
Chris@19 527 T28 = FMA(KP951056516, T1Y, KP587785252 * T1X);
Chris@19 528 io[WS(os, 5)] = T20 + T23;
Chris@19 529 T27 = T25 + T24;
Chris@19 530 io[WS(os, 11)] = T27 - T28;
Chris@19 531 io[WS(os, 14)] = T28 + T27;
Chris@19 532 T26 = T24 - T25;
Chris@19 533 io[WS(os, 2)] = T1Z + T26;
Chris@19 534 io[WS(os, 8)] = T26 - T1Z;
Chris@19 535 }
Chris@19 536 {
Chris@19 537 E T1x, T1D, T1E, T1I, T1J, T1G, T1H, T1K, T1F;
Chris@19 538 T1x = KP559016994 * (T1v - T1w);
Chris@19 539 T1D = T1v + T1w;
Chris@19 540 T1E = FNMS(KP250000000, T1D, T1C);
Chris@19 541 T1G = TW - TX;
Chris@19 542 T1H = TZ - T10;
Chris@19 543 T1I = FMA(KP951056516, T1G, KP587785252 * T1H);
Chris@19 544 T1J = FNMS(KP587785252, T1G, KP951056516 * T1H);
Chris@19 545 io[WS(os, 10)] = T1C + T1D;
Chris@19 546 T1K = T1E - T1x;
Chris@19 547 io[WS(os, 7)] = T1J + T1K;
Chris@19 548 io[WS(os, 13)] = T1K - T1J;
Chris@19 549 T1F = T1x + T1E;
Chris@19 550 io[WS(os, 1)] = T1F - T1I;
Chris@19 551 io[WS(os, 4)] = T1I + T1F;
Chris@19 552 }
Chris@19 553 {
Chris@19 554 E T13, T12, T14, T1s, T1u, T1g, T1r, T1t, T15;
Chris@19 555 T13 = KP559016994 * (TY - T11);
Chris@19 556 T12 = TY + T11;
Chris@19 557 T14 = FNMS(KP250000000, T12, TV);
Chris@19 558 T1g = T1a - T1f;
Chris@19 559 T1r = T1l - T1q;
Chris@19 560 T1s = FMA(KP951056516, T1g, KP587785252 * T1r);
Chris@19 561 T1u = FNMS(KP587785252, T1g, KP951056516 * T1r);
Chris@19 562 ro[WS(os, 10)] = TV + T12;
Chris@19 563 T1t = T14 - T13;
Chris@19 564 ro[WS(os, 7)] = T1t - T1u;
Chris@19 565 ro[WS(os, 13)] = T1t + T1u;
Chris@19 566 T15 = T13 + T14;
Chris@19 567 ro[WS(os, 4)] = T15 - T1s;
Chris@19 568 ro[WS(os, 1)] = T15 + T1s;
Chris@19 569 }
Chris@19 570 }
Chris@19 571 }
Chris@19 572 }
Chris@19 573
Chris@19 574 static const kdft_desc desc = { 15, "n1_15", {128, 28, 28, 0}, &GENUS, 0, 0, 0, 0 };
Chris@19 575
Chris@19 576 void X(codelet_n1_15) (planner *p) {
Chris@19 577 X(kdft_register) (p, n1_15, &desc);
Chris@19 578 }
Chris@19 579
Chris@19 580 #endif /* HAVE_FMA */