Mercurial > hg > sv-dependency-builds

annotate src/fftw-3.3.5/dft/scalar/codelets/t1_10.c @ 42:2cd0e3b3e1fd

Find changesets by keywords (author, files, the commit message), revision number or hash, or revset expression.
Current fftw source
author Chris Cannam
date Tue, 18 Oct 2016 13:40:26 +0100
parents
children
rev   line source
Chris@42 1 /*
Chris@42 2 * Copyright (c) 2003, 2007-14 Matteo Frigo
Chris@42 3 * Copyright (c) 2003, 2007-14 Massachusetts Institute of Technology
Chris@42 4 *
Chris@42 5 * This program is free software; you can redistribute it and/or modify
Chris@42 6 * it under the terms of the GNU General Public License as published by
Chris@42 7 * the Free Software Foundation; either version 2 of the License, or
Chris@42 8 * (at your option) any later version.
Chris@42 9 *
Chris@42 10 * This program is distributed in the hope that it will be useful,
Chris@42 11 * but WITHOUT ANY WARRANTY; without even the implied warranty of
Chris@42 12 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
Chris@42 13 * GNU General Public License for more details.
Chris@42 14 *
Chris@42 15 * You should have received a copy of the GNU General Public License
Chris@42 16 * along with this program; if not, write to the Free Software
Chris@42 17 * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
Chris@42 18 *
Chris@42 19 */
Chris@42 20
Chris@42 21 /* This file was automatically generated --- DO NOT EDIT */
Chris@42 22 /* Generated on Sat Jul 30 16:36:13 EDT 2016 */
Chris@42 23
Chris@42 24 #include "codelet-dft.h"
Chris@42 25
Chris@42 26 #ifdef HAVE_FMA
Chris@42 27
Chris@42 28 /* Generated by: ../../../genfft/gen_twiddle.native -fma -reorder-insns -schedule-for-pipeline -compact -variables 4 -pipeline-latency 4 -n 10 -name t1_10 -include t.h */
Chris@42 29
Chris@42 30 /*
Chris@42 31 * This function contains 102 FP additions, 72 FP multiplications,
Chris@42 32 * (or, 48 additions, 18 multiplications, 54 fused multiply/add),
Chris@42 33 * 70 stack variables, 4 constants, and 40 memory accesses
Chris@42 34 */
Chris@42 35 #include "t.h"
Chris@42 36
Chris@42 37 static void t1_10(R *ri, R *ii, const R *W, stride rs, INT mb, INT me, INT ms)
Chris@42 38 {
Chris@42 39 DK(KP951056516, +0.951056516295153572116439333379382143405698634);
Chris@42 40 DK(KP559016994, +0.559016994374947424102293417182819058860154590);
Chris@42 41 DK(KP250000000, +0.250000000000000000000000000000000000000000000);
Chris@42 42 DK(KP618033988, +0.618033988749894848204586834365638117720309180);
Chris@42 43 {
Chris@42 44 INT m;
Chris@42 45 for (m = mb, W = W + (mb * 18); m < me; m = m + 1, ri = ri + ms, ii = ii + ms, W = W + 18, MAKE_VOLATILE_STRIDE(20, rs)) {
Chris@42 46 E T1X, T21, T20, T22;
Chris@42 47 {
Chris@42 48 E T23, T1U, T8, T12, T1y, T25, T1P, T1H, T1Y, T18, T10, T2b, T1K, T1O, T15;
Chris@42 49 E T1Z, T2a, Tz, T24, T1n;
Chris@42 50 {
Chris@42 51 E T1, T1T, T3, T6, T2, T5;
Chris@42 52 T1 = ri[0];
Chris@42 53 T1T = ii[0];
Chris@42 54 T3 = ri[WS(rs, 5)];
Chris@42 55 T6 = ii[WS(rs, 5)];
Chris@42 56 T2 = W[8];
Chris@42 57 T5 = W[9];
Chris@42 58 {
Chris@42 59 E T1w, TY, T1s, T1F, TM, T16, T1u, TS;
Chris@42 60 {
Chris@42 61 E TF, T1p, TO, TR, T1r, TL, TN, TQ, T1t, TP;
Chris@42 62 {
Chris@42 63 E TU, TX, TT, TW;
Chris@42 64 {
Chris@42 65 E TB, TE, T1R, T4, TA, TD;
Chris@42 66 TB = ri[WS(rs, 4)];
Chris@42 67 TE = ii[WS(rs, 4)];
Chris@42 68 T1R = T2 * T6;
Chris@42 69 T4 = T2 * T3;
Chris@42 70 TA = W[6];
Chris@42 71 TD = W[7];
Chris@42 72 {
Chris@42 73 E T1S, T7, T1o, TC;
Chris@42 74 T1S = FNMS(T5, T3, T1R);
Chris@42 75 T7 = FMA(T5, T6, T4);
Chris@42 76 T1o = TA * TE;
Chris@42 77 TC = TA * TB;
Chris@42 78 T23 = T1T - T1S;
Chris@42 79 T1U = T1S + T1T;
Chris@42 80 T8 = T1 - T7;
Chris@42 81 T12 = T1 + T7;
Chris@42 82 TF = FMA(TD, TE, TC);
Chris@42 83 T1p = FNMS(TD, TB, T1o);
Chris@42 84 }
Chris@42 85 }
Chris@42 86 TU = ri[WS(rs, 1)];
Chris@42 87 TX = ii[WS(rs, 1)];
Chris@42 88 TT = W[0];
Chris@42 89 TW = W[1];
Chris@42 90 {
Chris@42 91 E TH, TK, TJ, T1q, TI, T1v, TV, TG;
Chris@42 92 TH = ri[WS(rs, 9)];
Chris@42 93 TK = ii[WS(rs, 9)];
Chris@42 94 T1v = TT * TX;
Chris@42 95 TV = TT * TU;
Chris@42 96 TG = W[16];
Chris@42 97 TJ = W[17];
Chris@42 98 T1w = FNMS(TW, TU, T1v);
Chris@42 99 TY = FMA(TW, TX, TV);
Chris@42 100 T1q = TG * TK;
Chris@42 101 TI = TG * TH;
Chris@42 102 TO = ri[WS(rs, 6)];
Chris@42 103 TR = ii[WS(rs, 6)];
Chris@42 104 T1r = FNMS(TJ, TH, T1q);
Chris@42 105 TL = FMA(TJ, TK, TI);
Chris@42 106 TN = W[10];
Chris@42 107 TQ = W[11];
Chris@42 108 }
Chris@42 109 }
Chris@42 110 T1s = T1p - T1r;
Chris@42 111 T1F = T1p + T1r;
Chris@42 112 TM = TF - TL;
Chris@42 113 T16 = TF + TL;
Chris@42 114 T1t = TN * TR;
Chris@42 115 TP = TN * TO;
Chris@42 116 T1u = FNMS(TQ, TO, T1t);
Chris@42 117 TS = FMA(TQ, TR, TP);
Chris@42 118 }
Chris@42 119 {
Chris@42 120 E T1e, Te, T1l, Tx, Tn, Tq, Tp, T1g, Tk, T1i, To;
Chris@42 121 {
Chris@42 122 E Tt, Tw, Tv, T1k, Tu;
Chris@42 123 {
Chris@42 124 E Ta, Td, T9, Tc, T1d, Tb, Ts;
Chris@42 125 Ta = ri[WS(rs, 2)];
Chris@42 126 Td = ii[WS(rs, 2)];
Chris@42 127 {
Chris@42 128 E T1G, T1x, TZ, T17;
Chris@42 129 T1G = T1u + T1w;
Chris@42 130 T1x = T1u - T1w;
Chris@42 131 TZ = TS - TY;
Chris@42 132 T17 = TS + TY;
Chris@42 133 T1y = T1s - T1x;
Chris@42 134 T25 = T1s + T1x;
Chris@42 135 T1P = T1F + T1G;
Chris@42 136 T1H = T1F - T1G;
Chris@42 137 T1Y = T16 - T17;
Chris@42 138 T18 = T16 + T17;
Chris@42 139 T10 = TM + TZ;
Chris@42 140 T2b = TM - TZ;
Chris@42 141 T9 = W[2];
Chris@42 142 }
Chris@42 143 Tc = W[3];
Chris@42 144 Tt = ri[WS(rs, 3)];
Chris@42 145 Tw = ii[WS(rs, 3)];
Chris@42 146 T1d = T9 * Td;
Chris@42 147 Tb = T9 * Ta;
Chris@42 148 Ts = W[4];
Chris@42 149 Tv = W[5];
Chris@42 150 T1e = FNMS(Tc, Ta, T1d);
Chris@42 151 Te = FMA(Tc, Td, Tb);
Chris@42 152 T1k = Ts * Tw;
Chris@42 153 Tu = Ts * Tt;
Chris@42 154 }
Chris@42 155 {
Chris@42 156 E Tg, Tj, Tf, Ti, T1f, Th, Tm;
Chris@42 157 Tg = ri[WS(rs, 7)];
Chris@42 158 Tj = ii[WS(rs, 7)];
Chris@42 159 T1l = FNMS(Tv, Tt, T1k);
Chris@42 160 Tx = FMA(Tv, Tw, Tu);
Chris@42 161 Tf = W[12];
Chris@42 162 Ti = W[13];
Chris@42 163 Tn = ri[WS(rs, 8)];
Chris@42 164 Tq = ii[WS(rs, 8)];
Chris@42 165 T1f = Tf * Tj;
Chris@42 166 Th = Tf * Tg;
Chris@42 167 Tm = W[14];
Chris@42 168 Tp = W[15];
Chris@42 169 T1g = FNMS(Ti, Tg, T1f);
Chris@42 170 Tk = FMA(Ti, Tj, Th);
Chris@42 171 T1i = Tm * Tq;
Chris@42 172 To = Tm * Tn;
Chris@42 173 }
Chris@42 174 }
Chris@42 175 {
Chris@42 176 E T1h, T1I, Tl, T13, T1j, Tr;
Chris@42 177 T1h = T1e - T1g;
Chris@42 178 T1I = T1e + T1g;
Chris@42 179 Tl = Te - Tk;
Chris@42 180 T13 = Te + Tk;
Chris@42 181 T1j = FNMS(Tp, Tn, T1i);
Chris@42 182 Tr = FMA(Tp, Tq, To);
Chris@42 183 {
Chris@42 184 E T1m, T1J, T14, Ty;
Chris@42 185 T1m = T1j - T1l;
Chris@42 186 T1J = T1j + T1l;
Chris@42 187 T14 = Tr + Tx;
Chris@42 188 Ty = Tr - Tx;
Chris@42 189 T1K = T1I - T1J;
Chris@42 190 T1O = T1I + T1J;
Chris@42 191 T15 = T13 + T14;
Chris@42 192 T1Z = T13 - T14;
Chris@42 193 T2a = Tl - Ty;
Chris@42 194 Tz = Tl + Ty;
Chris@42 195 T24 = T1h + T1m;
Chris@42 196 T1n = T1h - T1m;
Chris@42 197 }
Chris@42 198 }
Chris@42 199 }
Chris@42 200 }
Chris@42 201 }
Chris@42 202 {
Chris@42 203 E T2c, T2e, T29, T2d;
Chris@42 204 {
Chris@42 205 E T1b, T11, T26, T28, T27;
Chris@42 206 T1b = Tz - T10;
Chris@42 207 T11 = Tz + T10;
Chris@42 208 T26 = T24 + T25;
Chris@42 209 T28 = T24 - T25;
Chris@42 210 {
Chris@42 211 E T1B, T1z, T1a, T1A, T1c;
Chris@42 212 T1B = FNMS(KP618033988, T1n, T1y);
Chris@42 213 T1z = FMA(KP618033988, T1y, T1n);
Chris@42 214 ri[WS(rs, 5)] = T8 + T11;
Chris@42 215 T1a = FNMS(KP250000000, T11, T8);
Chris@42 216 T1A = FNMS(KP559016994, T1b, T1a);
Chris@42 217 T1c = FMA(KP559016994, T1b, T1a);
Chris@42 218 T27 = FNMS(KP250000000, T26, T23);
Chris@42 219 T2c = FMA(KP618033988, T2b, T2a);
Chris@42 220 T2e = FNMS(KP618033988, T2a, T2b);
Chris@42 221 ri[WS(rs, 1)] = FMA(KP951056516, T1z, T1c);
Chris@42 222 ri[WS(rs, 9)] = FNMS(KP951056516, T1z, T1c);
Chris@42 223 ri[WS(rs, 3)] = FMA(KP951056516, T1B, T1A);
Chris@42 224 ri[WS(rs, 7)] = FNMS(KP951056516, T1B, T1A);
Chris@42 225 }
Chris@42 226 ii[WS(rs, 5)] = T26 + T23;
Chris@42 227 T29 = FMA(KP559016994, T28, T27);
Chris@42 228 T2d = FNMS(KP559016994, T28, T27);
Chris@42 229 }
Chris@42 230 {
Chris@42 231 E T1E, T1M, T1L, T1N, T19, T1D, T1C, T1Q, T1W, T1V;
Chris@42 232 T19 = T15 + T18;
Chris@42 233 T1D = T15 - T18;
Chris@42 234 ii[WS(rs, 7)] = FMA(KP951056516, T2e, T2d);
Chris@42 235 ii[WS(rs, 3)] = FNMS(KP951056516, T2e, T2d);
Chris@42 236 ii[WS(rs, 9)] = FMA(KP951056516, T2c, T29);
Chris@42 237 ii[WS(rs, 1)] = FNMS(KP951056516, T2c, T29);
Chris@42 238 T1C = FNMS(KP250000000, T19, T12);
Chris@42 239 ri[0] = T12 + T19;
Chris@42 240 T1E = FNMS(KP559016994, T1D, T1C);
Chris@42 241 T1M = FMA(KP559016994, T1D, T1C);
Chris@42 242 T1L = FNMS(KP618033988, T1K, T1H);
Chris@42 243 T1N = FMA(KP618033988, T1H, T1K);
Chris@42 244 T1Q = T1O + T1P;
Chris@42 245 T1W = T1O - T1P;
Chris@42 246 ri[WS(rs, 6)] = FMA(KP951056516, T1N, T1M);
Chris@42 247 ri[WS(rs, 4)] = FNMS(KP951056516, T1N, T1M);
Chris@42 248 ri[WS(rs, 8)] = FMA(KP951056516, T1L, T1E);
Chris@42 249 ri[WS(rs, 2)] = FNMS(KP951056516, T1L, T1E);
Chris@42 250 T1V = FNMS(KP250000000, T1Q, T1U);
Chris@42 251 ii[0] = T1Q + T1U;
Chris@42 252 T1X = FNMS(KP559016994, T1W, T1V);
Chris@42 253 T21 = FMA(KP559016994, T1W, T1V);
Chris@42 254 T20 = FNMS(KP618033988, T1Z, T1Y);
Chris@42 255 T22 = FMA(KP618033988, T1Y, T1Z);
Chris@42 256 }
Chris@42 257 }
Chris@42 258 }
Chris@42 259 ii[WS(rs, 6)] = FNMS(KP951056516, T22, T21);
Chris@42 260 ii[WS(rs, 4)] = FMA(KP951056516, T22, T21);
Chris@42 261 ii[WS(rs, 8)] = FNMS(KP951056516, T20, T1X);
Chris@42 262 ii[WS(rs, 2)] = FMA(KP951056516, T20, T1X);
Chris@42 263 }
Chris@42 264 }
Chris@42 265 }
Chris@42 266
Chris@42 267 static const tw_instr twinstr[] = {
Chris@42 268 {TW_FULL, 0, 10},
Chris@42 269 {TW_NEXT, 1, 0}
Chris@42 270 };
Chris@42 271
Chris@42 272 static const ct_desc desc = { 10, "t1_10", twinstr, &GENUS, {48, 18, 54, 0}, 0, 0, 0 };
Chris@42 273
Chris@42 274 void X(codelet_t1_10) (planner *p) {
Chris@42 275 X(kdft_dit_register) (p, t1_10, &desc);
Chris@42 276 }
Chris@42 277 #else /* HAVE_FMA */
Chris@42 278
Chris@42 279 /* Generated by: ../../../genfft/gen_twiddle.native -compact -variables 4 -pipeline-latency 4 -n 10 -name t1_10 -include t.h */
Chris@42 280
Chris@42 281 /*
Chris@42 282 * This function contains 102 FP additions, 60 FP multiplications,
Chris@42 283 * (or, 72 additions, 30 multiplications, 30 fused multiply/add),
Chris@42 284 * 45 stack variables, 4 constants, and 40 memory accesses
Chris@42 285 */
Chris@42 286 #include "t.h"
Chris@42 287
Chris@42 288 static void t1_10(R *ri, R *ii, const R *W, stride rs, INT mb, INT me, INT ms)
Chris@42 289 {
Chris@42 290 DK(KP587785252, +0.587785252292473129168705954639072768597652438);
Chris@42 291 DK(KP951056516, +0.951056516295153572116439333379382143405698634);
Chris@42 292 DK(KP250000000, +0.250000000000000000000000000000000000000000000);
Chris@42 293 DK(KP559016994, +0.559016994374947424102293417182819058860154590);
Chris@42 294 {
Chris@42 295 INT m;
Chris@42 296 for (m = mb, W = W + (mb * 18); m < me; m = m + 1, ri = ri + ms, ii = ii + ms, W = W + 18, MAKE_VOLATILE_STRIDE(20, rs)) {
Chris@42 297 E T7, T1O, TT, T1C, TF, TQ, TR, T1o, T1p, T1y, TX, TY, TZ, T1d, T1g;
Chris@42 298 E T1M, Ti, Tt, Tu, T1r, T1s, T1x, TU, TV, TW, T16, T19, T1L;
Chris@42 299 {
Chris@42 300 E T1, T1B, T6, T1A;
Chris@42 301 T1 = ri[0];
Chris@42 302 T1B = ii[0];
Chris@42 303 {
Chris@42 304 E T3, T5, T2, T4;
Chris@42 305 T3 = ri[WS(rs, 5)];
Chris@42 306 T5 = ii[WS(rs, 5)];
Chris@42 307 T2 = W[8];
Chris@42 308 T4 = W[9];
Chris@42 309 T6 = FMA(T2, T3, T4 * T5);
Chris@42 310 T1A = FNMS(T4, T3, T2 * T5);
Chris@42 311 }
Chris@42 312 T7 = T1 - T6;
Chris@42 313 T1O = T1B - T1A;
Chris@42 314 TT = T1 + T6;
Chris@42 315 T1C = T1A + T1B;
Chris@42 316 }
Chris@42 317 {
Chris@42 318 E Tz, T1b, TP, T1f, TE, T1c, TK, T1e;
Chris@42 319 {
Chris@42 320 E Tw, Ty, Tv, Tx;
Chris@42 321 Tw = ri[WS(rs, 4)];
Chris@42 322 Ty = ii[WS(rs, 4)];
Chris@42 323 Tv = W[6];
Chris@42 324 Tx = W[7];
Chris@42 325 Tz = FMA(Tv, Tw, Tx * Ty);
Chris@42 326 T1b = FNMS(Tx, Tw, Tv * Ty);
Chris@42 327 }
Chris@42 328 {
Chris@42 329 E TM, TO, TL, TN;
Chris@42 330 TM = ri[WS(rs, 1)];
Chris@42 331 TO = ii[WS(rs, 1)];
Chris@42 332 TL = W[0];
Chris@42 333 TN = W[1];
Chris@42 334 TP = FMA(TL, TM, TN * TO);
Chris@42 335 T1f = FNMS(TN, TM, TL * TO);
Chris@42 336 }
Chris@42 337 {
Chris@42 338 E TB, TD, TA, TC;
Chris@42 339 TB = ri[WS(rs, 9)];
Chris@42 340 TD = ii[WS(rs, 9)];
Chris@42 341 TA = W[16];
Chris@42 342 TC = W[17];
Chris@42 343 TE = FMA(TA, TB, TC * TD);
Chris@42 344 T1c = FNMS(TC, TB, TA * TD);
Chris@42 345 }
Chris@42 346 {
Chris@42 347 E TH, TJ, TG, TI;
Chris@42 348 TH = ri[WS(rs, 6)];
Chris@42 349 TJ = ii[WS(rs, 6)];
Chris@42 350 TG = W[10];
Chris@42 351 TI = W[11];
Chris@42 352 TK = FMA(TG, TH, TI * TJ);
Chris@42 353 T1e = FNMS(TI, TH, TG * TJ);
Chris@42 354 }
Chris@42 355 TF = Tz - TE;
Chris@42 356 TQ = TK - TP;
Chris@42 357 TR = TF + TQ;
Chris@42 358 T1o = T1b + T1c;
Chris@42 359 T1p = T1e + T1f;
Chris@42 360 T1y = T1o + T1p;
Chris@42 361 TX = Tz + TE;
Chris@42 362 TY = TK + TP;
Chris@42 363 TZ = TX + TY;
Chris@42 364 T1d = T1b - T1c;
Chris@42 365 T1g = T1e - T1f;
Chris@42 366 T1M = T1d + T1g;
Chris@42 367 }
Chris@42 368 {
Chris@42 369 E Tc, T14, Ts, T18, Th, T15, Tn, T17;
Chris@42 370 {
Chris@42 371 E T9, Tb, T8, Ta;
Chris@42 372 T9 = ri[WS(rs, 2)];
Chris@42 373 Tb = ii[WS(rs, 2)];
Chris@42 374 T8 = W[2];
Chris@42 375 Ta = W[3];
Chris@42 376 Tc = FMA(T8, T9, Ta * Tb);
Chris@42 377 T14 = FNMS(Ta, T9, T8 * Tb);
Chris@42 378 }
Chris@42 379 {
Chris@42 380 E Tp, Tr, To, Tq;
Chris@42 381 Tp = ri[WS(rs, 3)];
Chris@42 382 Tr = ii[WS(rs, 3)];
Chris@42 383 To = W[4];
Chris@42 384 Tq = W[5];
Chris@42 385 Ts = FMA(To, Tp, Tq * Tr);
Chris@42 386 T18 = FNMS(Tq, Tp, To * Tr);
Chris@42 387 }
Chris@42 388 {
Chris@42 389 E Te, Tg, Td, Tf;
Chris@42 390 Te = ri[WS(rs, 7)];
Chris@42 391 Tg = ii[WS(rs, 7)];
Chris@42 392 Td = W[12];
Chris@42 393 Tf = W[13];
Chris@42 394 Th = FMA(Td, Te, Tf * Tg);
Chris@42 395 T15 = FNMS(Tf, Te, Td * Tg);
Chris@42 396 }
Chris@42 397 {
Chris@42 398 E Tk, Tm, Tj, Tl;
Chris@42 399 Tk = ri[WS(rs, 8)];
Chris@42 400 Tm = ii[WS(rs, 8)];
Chris@42 401 Tj = W[14];
Chris@42 402 Tl = W[15];
Chris@42 403 Tn = FMA(Tj, Tk, Tl * Tm);
Chris@42 404 T17 = FNMS(Tl, Tk, Tj * Tm);
Chris@42 405 }
Chris@42 406 Ti = Tc - Th;
Chris@42 407 Tt = Tn - Ts;
Chris@42 408 Tu = Ti + Tt;
Chris@42 409 T1r = T14 + T15;
Chris@42 410 T1s = T17 + T18;
Chris@42 411 T1x = T1r + T1s;
Chris@42 412 TU = Tc + Th;
Chris@42 413 TV = Tn + Ts;
Chris@42 414 TW = TU + TV;
Chris@42 415 T16 = T14 - T15;
Chris@42 416 T19 = T17 - T18;
Chris@42 417 T1L = T16 + T19;
Chris@42 418 }
Chris@42 419 {
Chris@42 420 E T11, TS, T12, T1i, T1k, T1a, T1h, T1j, T13;
Chris@42 421 T11 = KP559016994 * (Tu - TR);
Chris@42 422 TS = Tu + TR;
Chris@42 423 T12 = FNMS(KP250000000, TS, T7);
Chris@42 424 T1a = T16 - T19;
Chris@42 425 T1h = T1d - T1g;
Chris@42 426 T1i = FMA(KP951056516, T1a, KP587785252 * T1h);
Chris@42 427 T1k = FNMS(KP587785252, T1a, KP951056516 * T1h);
Chris@42 428 ri[WS(rs, 5)] = T7 + TS;
Chris@42 429 T1j = T12 - T11;
Chris@42 430 ri[WS(rs, 7)] = T1j - T1k;
Chris@42 431 ri[WS(rs, 3)] = T1j + T1k;
Chris@42 432 T13 = T11 + T12;
Chris@42 433 ri[WS(rs, 9)] = T13 - T1i;
Chris@42 434 ri[WS(rs, 1)] = T13 + T1i;
Chris@42 435 }
Chris@42 436 {
Chris@42 437 E T1N, T1P, T1Q, T1U, T1W, T1S, T1T, T1V, T1R;
Chris@42 438 T1N = KP559016994 * (T1L - T1M);
Chris@42 439 T1P = T1L + T1M;
Chris@42 440 T1Q = FNMS(KP250000000, T1P, T1O);
Chris@42 441 T1S = Ti - Tt;
Chris@42 442 T1T = TF - TQ;
Chris@42 443 T1U = FMA(KP951056516, T1S, KP587785252 * T1T);
Chris@42 444 T1W = FNMS(KP587785252, T1S, KP951056516 * T1T);
Chris@42 445 ii[WS(rs, 5)] = T1P + T1O;
Chris@42 446 T1V = T1Q - T1N;
Chris@42 447 ii[WS(rs, 3)] = T1V - T1W;
Chris@42 448 ii[WS(rs, 7)] = T1W + T1V;
Chris@42 449 T1R = T1N + T1Q;
Chris@42 450 ii[WS(rs, 1)] = T1R - T1U;
Chris@42 451 ii[WS(rs, 9)] = T1U + T1R;
Chris@42 452 }
Chris@42 453 {
Chris@42 454 E T1m, T10, T1l, T1u, T1w, T1q, T1t, T1v, T1n;
Chris@42 455 T1m = KP559016994 * (TW - TZ);
Chris@42 456 T10 = TW + TZ;
Chris@42 457 T1l = FNMS(KP250000000, T10, TT);
Chris@42 458 T1q = T1o - T1p;
Chris@42 459 T1t = T1r - T1s;
Chris@42 460 T1u = FNMS(KP587785252, T1t, KP951056516 * T1q);
Chris@42 461 T1w = FMA(KP951056516, T1t, KP587785252 * T1q);
Chris@42 462 ri[0] = TT + T10;
Chris@42 463 T1v = T1m + T1l;
Chris@42 464 ri[WS(rs, 4)] = T1v - T1w;
Chris@42 465 ri[WS(rs, 6)] = T1v + T1w;
Chris@42 466 T1n = T1l - T1m;
Chris@42 467 ri[WS(rs, 2)] = T1n - T1u;
Chris@42 468 ri[WS(rs, 8)] = T1n + T1u;
Chris@42 469 }
Chris@42 470 {
Chris@42 471 E T1H, T1z, T1G, T1F, T1J, T1D, T1E, T1K, T1I;
Chris@42 472 T1H = KP559016994 * (T1x - T1y);
Chris@42 473 T1z = T1x + T1y;
Chris@42 474 T1G = FNMS(KP250000000, T1z, T1C);
Chris@42 475 T1D = TX - TY;
Chris@42 476 T1E = TU - TV;
Chris@42 477 T1F = FNMS(KP587785252, T1E, KP951056516 * T1D);
Chris@42 478 T1J = FMA(KP951056516, T1E, KP587785252 * T1D);
Chris@42 479 ii[0] = T1z + T1C;
Chris@42 480 T1K = T1H + T1G;
Chris@42 481 ii[WS(rs, 4)] = T1J + T1K;
Chris@42 482 ii[WS(rs, 6)] = T1K - T1J;
Chris@42 483 T1I = T1G - T1H;
Chris@42 484 ii[WS(rs, 2)] = T1F + T1I;
Chris@42 485 ii[WS(rs, 8)] = T1I - T1F;
Chris@42 486 }
Chris@42 487 }
Chris@42 488 }
Chris@42 489 }
Chris@42 490
Chris@42 491 static const tw_instr twinstr[] = {
Chris@42 492 {TW_FULL, 0, 10},
Chris@42 493 {TW_NEXT, 1, 0}
Chris@42 494 };
Chris@42 495
Chris@42 496 static const ct_desc desc = { 10, "t1_10", twinstr, &GENUS, {72, 30, 30, 0}, 0, 0, 0 };
Chris@42 497
Chris@42 498 void X(codelet_t1_10) (planner *p) {
Chris@42 499 X(kdft_dit_register) (p, t1_10, &desc);
Chris@42 500 }
Chris@42 501 #endif /* HAVE_FMA */