annotate src/fftw-3.3.3/rdft/scalar/r2cf/hc2cfdft_32.c @ 169:223a55898ab9 tip default

Add null config files
author Chris Cannam <cannam@all-day-breakfast.com>
date Mon, 02 Mar 2020 14:03:47 +0000
parents 89f5e221ed7b
children
rev   line source
cannam@95 1 /*
cannam@95 2 * Copyright (c) 2003, 2007-11 Matteo Frigo
cannam@95 3 * Copyright (c) 2003, 2007-11 Massachusetts Institute of Technology
cannam@95 4 *
cannam@95 5 * This program is free software; you can redistribute it and/or modify
cannam@95 6 * it under the terms of the GNU General Public License as published by
cannam@95 7 * the Free Software Foundation; either version 2 of the License, or
cannam@95 8 * (at your option) any later version.
cannam@95 9 *
cannam@95 10 * This program is distributed in the hope that it will be useful,
cannam@95 11 * but WITHOUT ANY WARRANTY; without even the implied warranty of
cannam@95 12 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
cannam@95 13 * GNU General Public License for more details.
cannam@95 14 *
cannam@95 15 * You should have received a copy of the GNU General Public License
cannam@95 16 * along with this program; if not, write to the Free Software
cannam@95 17 * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
cannam@95 18 *
cannam@95 19 */
cannam@95 20
cannam@95 21 /* This file was automatically generated --- DO NOT EDIT */
cannam@95 22 /* Generated on Sun Nov 25 07:40:46 EST 2012 */
cannam@95 23
cannam@95 24 #include "codelet-rdft.h"
cannam@95 25
cannam@95 26 #ifdef HAVE_FMA
cannam@95 27
cannam@95 28 /* Generated by: ../../../genfft/gen_hc2cdft.native -fma -reorder-insns -schedule-for-pipeline -compact -variables 4 -pipeline-latency 4 -n 32 -dit -name hc2cfdft_32 -include hc2cf.h */
cannam@95 29
cannam@95 30 /*
cannam@95 31 * This function contains 498 FP additions, 324 FP multiplications,
cannam@95 32 * (or, 300 additions, 126 multiplications, 198 fused multiply/add),
cannam@95 33 * 172 stack variables, 8 constants, and 128 memory accesses
cannam@95 34 */
cannam@95 35 #include "hc2cf.h"
cannam@95 36
cannam@95 37 static void hc2cfdft_32(R *Rp, R *Ip, R *Rm, R *Im, const R *W, stride rs, INT mb, INT me, INT ms)
cannam@95 38 {
cannam@95 39 DK(KP980785280, +0.980785280403230449126182236134239036973933731);
cannam@95 40 DK(KP831469612, +0.831469612302545237078788377617905756738560812);
cannam@95 41 DK(KP668178637, +0.668178637919298919997757686523080761552472251);
cannam@95 42 DK(KP198912367, +0.198912367379658006911597622644676228597850501);
cannam@95 43 DK(KP923879532, +0.923879532511286756128183189396788286822416626);
cannam@95 44 DK(KP414213562, +0.414213562373095048801688724209698078569671875);
cannam@95 45 DK(KP707106781, +0.707106781186547524400844362104849039284835938);
cannam@95 46 DK(KP500000000, +0.500000000000000000000000000000000000000000000);
cannam@95 47 {
cannam@95 48 INT m;
cannam@95 49 for (m = mb, W = W + ((mb - 1) * 62); m < me; m = m + 1, Rp = Rp + ms, Ip = Ip + ms, Rm = Rm - ms, Im = Im - ms, W = W + 62, MAKE_VOLATILE_STRIDE(128, rs)) {
cannam@95 50 E T9X, Ta0;
cannam@95 51 {
cannam@95 52 E T3B, T89, T61, T8l, T2F, T7p, T8t, T4B, T7I, T5e, T7L, T1n, T7R, T5E, T82;
cannam@95 53 E T4u, T3m, T8k, T5W, T8a, T2r, T8u, T4G, T7q, T59, T7K, T7H, T12, T5z, T81;
cannam@95 54 E T7Q, T4h, T4Y, T7D, T7A, Tl, T5o, T3Q, T84, T7V, T2V, T4M, T7t, T7s, T1K;
cannam@95 55 E T5L, T8e, T8n, T38, T7v, T4R, T7w, T25, T5Q, T8h, T8o, T3V, T3S, T5p, T3T;
cannam@95 56 E T41, Tz, T3Y, TE, TA, T51, T5r, T3Z, Tv, T50, TB, T3U, T40;
cannam@95 57 {
cannam@95 58 E T49, T46, T5v, T47, T4f, TV, T4c, T10, TW, T57, T5x, T4d, TR, T56, TX;
cannam@95 59 E T48, T4e;
cannam@95 60 {
cannam@95 61 E T4m, T4j, T5A, T4k, T4s, T1g, T4p, T1l, T1h, T5c, T5C, T4q, T1c, T5b, T1i;
cannam@95 62 E T4l, T4r;
cannam@95 63 {
cannam@95 64 E T2E, T4y, T2B, T4A;
cannam@95 65 {
cannam@95 66 E T3y, T3z, T3t, T5Z, T3x, T2v, T3r, T3q, T3n, T2A, T3o, T2s;
cannam@95 67 {
cannam@95 68 E T2C, T2D, T3w, T3u, T3v;
cannam@95 69 T2C = Ip[0];
cannam@95 70 T2D = Im[0];
cannam@95 71 T3u = Rm[0];
cannam@95 72 T3v = Rp[0];
cannam@95 73 T3y = W[1];
cannam@95 74 T3z = T2C + T2D;
cannam@95 75 T2E = T2C - T2D;
cannam@95 76 T4y = T3v + T3u;
cannam@95 77 T3w = T3u - T3v;
cannam@95 78 T3t = W[0];
cannam@95 79 {
cannam@95 80 E T2y, T2z, T2t, T2u;
cannam@95 81 T2t = Ip[WS(rs, 8)];
cannam@95 82 T2u = Im[WS(rs, 8)];
cannam@95 83 T5Z = T3y * T3w;
cannam@95 84 T3x = T3t * T3w;
cannam@95 85 T2y = Rp[WS(rs, 8)];
cannam@95 86 T2v = T2t - T2u;
cannam@95 87 T3r = T2t + T2u;
cannam@95 88 T2z = Rm[WS(rs, 8)];
cannam@95 89 T3q = W[33];
cannam@95 90 T3n = W[32];
cannam@95 91 T2A = T2y + T2z;
cannam@95 92 T3o = T2z - T2y;
cannam@95 93 T2s = W[30];
cannam@95 94 }
cannam@95 95 }
cannam@95 96 {
cannam@95 97 E T3A, T5X, T4z, T2w, T3s, T3p, T5Y, T60, T2x;
cannam@95 98 T3A = FNMS(T3y, T3z, T3x);
cannam@95 99 T3p = T3n * T3o;
cannam@95 100 T5X = T3q * T3o;
cannam@95 101 T4z = T2s * T2A;
cannam@95 102 T2w = T2s * T2v;
cannam@95 103 T3s = FNMS(T3q, T3r, T3p);
cannam@95 104 T5Y = FMA(T3n, T3r, T5X);
cannam@95 105 T60 = FMA(T3t, T3z, T5Z);
cannam@95 106 T2x = W[31];
cannam@95 107 T3B = T3s + T3A;
cannam@95 108 T89 = T3A - T3s;
cannam@95 109 T61 = T5Y + T60;
cannam@95 110 T8l = T60 - T5Y;
cannam@95 111 T2B = FNMS(T2x, T2A, T2w);
cannam@95 112 T4A = FMA(T2x, T2v, T4z);
cannam@95 113 }
cannam@95 114 }
cannam@95 115 {
cannam@95 116 E T16, T1b, T17, T5a, T1d, T4o, T18;
cannam@95 117 {
cannam@95 118 E T19, T1a, T13, T4i, T14, T15;
cannam@95 119 T14 = Ip[WS(rs, 3)];
cannam@95 120 T15 = Im[WS(rs, 3)];
cannam@95 121 T2F = T2B + T2E;
cannam@95 122 T7p = T2E - T2B;
cannam@95 123 T8t = T4y - T4A;
cannam@95 124 T4B = T4y + T4A;
cannam@95 125 T4m = T14 + T15;
cannam@95 126 T16 = T14 - T15;
cannam@95 127 T19 = Rp[WS(rs, 3)];
cannam@95 128 T1a = Rm[WS(rs, 3)];
cannam@95 129 T13 = W[10];
cannam@95 130 T4i = W[12];
cannam@95 131 {
cannam@95 132 E T1e, T1f, T1j, T1k;
cannam@95 133 T1e = Ip[WS(rs, 11)];
cannam@95 134 T4j = T19 - T1a;
cannam@95 135 T1b = T19 + T1a;
cannam@95 136 T17 = T13 * T16;
cannam@95 137 T5A = T4i * T4m;
cannam@95 138 T4k = T4i * T4j;
cannam@95 139 T5a = T13 * T1b;
cannam@95 140 T1f = Im[WS(rs, 11)];
cannam@95 141 T1j = Rp[WS(rs, 11)];
cannam@95 142 T1k = Rm[WS(rs, 11)];
cannam@95 143 T1d = W[42];
cannam@95 144 T4s = T1e + T1f;
cannam@95 145 T1g = T1e - T1f;
cannam@95 146 T4p = T1j - T1k;
cannam@95 147 T1l = T1j + T1k;
cannam@95 148 T4o = W[44];
cannam@95 149 T1h = T1d * T1g;
cannam@95 150 }
cannam@95 151 }
cannam@95 152 T18 = W[11];
cannam@95 153 T5c = T1d * T1l;
cannam@95 154 T5C = T4o * T4s;
cannam@95 155 T4q = T4o * T4p;
cannam@95 156 T1c = FNMS(T18, T1b, T17);
cannam@95 157 T5b = FMA(T18, T16, T5a);
cannam@95 158 T1i = W[43];
cannam@95 159 T4l = W[13];
cannam@95 160 T4r = W[45];
cannam@95 161 }
cannam@95 162 }
cannam@95 163 {
cannam@95 164 E T4D, T2g, T2q, T4F;
cannam@95 165 {
cannam@95 166 E T3d, T3e, T2a, T2f, T3a, T5S, T3c, T4C, T2b, T3j, T2k, T3k, T2p, T3h, T3g;
cannam@95 167 E T2h, T5U, T3b, T27;
cannam@95 168 {
cannam@95 169 E T28, T29, T2d, T2e, T5d, T1m;
cannam@95 170 T28 = Ip[WS(rs, 4)];
cannam@95 171 T5d = FMA(T1i, T1g, T5c);
cannam@95 172 T1m = FNMS(T1i, T1l, T1h);
cannam@95 173 {
cannam@95 174 E T5B, T4n, T5D, T4t;
cannam@95 175 T5B = FNMS(T4l, T4j, T5A);
cannam@95 176 T4n = FMA(T4l, T4m, T4k);
cannam@95 177 T5D = FNMS(T4r, T4p, T5C);
cannam@95 178 T4t = FMA(T4r, T4s, T4q);
cannam@95 179 T7I = T5b - T5d;
cannam@95 180 T5e = T5b + T5d;
cannam@95 181 T7L = T1c - T1m;
cannam@95 182 T1n = T1c + T1m;
cannam@95 183 T7R = T5D - T5B;
cannam@95 184 T5E = T5B + T5D;
cannam@95 185 T82 = T4t - T4n;
cannam@95 186 T4u = T4n + T4t;
cannam@95 187 T29 = Im[WS(rs, 4)];
cannam@95 188 }
cannam@95 189 T2d = Rp[WS(rs, 4)];
cannam@95 190 T2e = Rm[WS(rs, 4)];
cannam@95 191 T3d = W[17];
cannam@95 192 T3e = T28 + T29;
cannam@95 193 T2a = T28 - T29;
cannam@95 194 T3b = T2e - T2d;
cannam@95 195 T2f = T2d + T2e;
cannam@95 196 T3a = W[16];
cannam@95 197 T27 = W[14];
cannam@95 198 T5S = T3d * T3b;
cannam@95 199 }
cannam@95 200 {
cannam@95 201 E T2i, T2j, T2n, T2o;
cannam@95 202 T2i = Ip[WS(rs, 12)];
cannam@95 203 T3c = T3a * T3b;
cannam@95 204 T4C = T27 * T2f;
cannam@95 205 T2b = T27 * T2a;
cannam@95 206 T2j = Im[WS(rs, 12)];
cannam@95 207 T2n = Rp[WS(rs, 12)];
cannam@95 208 T2o = Rm[WS(rs, 12)];
cannam@95 209 T3j = W[49];
cannam@95 210 T2k = T2i - T2j;
cannam@95 211 T3k = T2i + T2j;
cannam@95 212 T2p = T2n + T2o;
cannam@95 213 T3h = T2o - T2n;
cannam@95 214 T3g = W[48];
cannam@95 215 T2h = W[46];
cannam@95 216 T5U = T3j * T3h;
cannam@95 217 }
cannam@95 218 {
cannam@95 219 E T3f, T3i, T4E, T2l;
cannam@95 220 T3f = FNMS(T3d, T3e, T3c);
cannam@95 221 T3i = T3g * T3h;
cannam@95 222 T4E = T2h * T2p;
cannam@95 223 T2l = T2h * T2k;
cannam@95 224 {
cannam@95 225 E T5T, T3l, T5V, T2c, T2m;
cannam@95 226 T5T = FMA(T3a, T3e, T5S);
cannam@95 227 T3l = FNMS(T3j, T3k, T3i);
cannam@95 228 T5V = FMA(T3g, T3k, T5U);
cannam@95 229 T2c = W[15];
cannam@95 230 T2m = W[47];
cannam@95 231 T3m = T3f + T3l;
cannam@95 232 T8k = T3f - T3l;
cannam@95 233 T5W = T5T + T5V;
cannam@95 234 T8a = T5T - T5V;
cannam@95 235 T4D = FMA(T2c, T2a, T4C);
cannam@95 236 T2g = FNMS(T2c, T2f, T2b);
cannam@95 237 T2q = FNMS(T2m, T2p, T2l);
cannam@95 238 T4F = FMA(T2m, T2k, T4E);
cannam@95 239 }
cannam@95 240 }
cannam@95 241 }
cannam@95 242 {
cannam@95 243 E TL, TQ, TM, T55, TS, T4b, TN;
cannam@95 244 {
cannam@95 245 E TO, TP, TI, T45, TJ, TK;
cannam@95 246 TJ = Ip[WS(rs, 15)];
cannam@95 247 TK = Im[WS(rs, 15)];
cannam@95 248 T2r = T2g + T2q;
cannam@95 249 T8u = T2g - T2q;
cannam@95 250 T4G = T4D + T4F;
cannam@95 251 T7q = T4D - T4F;
cannam@95 252 T49 = TJ + TK;
cannam@95 253 TL = TJ - TK;
cannam@95 254 TO = Rp[WS(rs, 15)];
cannam@95 255 TP = Rm[WS(rs, 15)];
cannam@95 256 TI = W[58];
cannam@95 257 T45 = W[60];
cannam@95 258 {
cannam@95 259 E TT, TU, TY, TZ;
cannam@95 260 TT = Ip[WS(rs, 7)];
cannam@95 261 T46 = TO - TP;
cannam@95 262 TQ = TO + TP;
cannam@95 263 TM = TI * TL;
cannam@95 264 T5v = T45 * T49;
cannam@95 265 T47 = T45 * T46;
cannam@95 266 T55 = TI * TQ;
cannam@95 267 TU = Im[WS(rs, 7)];
cannam@95 268 TY = Rp[WS(rs, 7)];
cannam@95 269 TZ = Rm[WS(rs, 7)];
cannam@95 270 TS = W[26];
cannam@95 271 T4f = TT + TU;
cannam@95 272 TV = TT - TU;
cannam@95 273 T4c = TY - TZ;
cannam@95 274 T10 = TY + TZ;
cannam@95 275 T4b = W[28];
cannam@95 276 TW = TS * TV;
cannam@95 277 }
cannam@95 278 }
cannam@95 279 TN = W[59];
cannam@95 280 T57 = TS * T10;
cannam@95 281 T5x = T4b * T4f;
cannam@95 282 T4d = T4b * T4c;
cannam@95 283 TR = FNMS(TN, TQ, TM);
cannam@95 284 T56 = FMA(TN, TL, T55);
cannam@95 285 TX = W[27];
cannam@95 286 T48 = W[61];
cannam@95 287 T4e = W[29];
cannam@95 288 }
cannam@95 289 }
cannam@95 290 }
cannam@95 291 {
cannam@95 292 E T8c, T8d, T8f, T8g;
cannam@95 293 {
cannam@95 294 E T3I, T3F, T5k, T3G, T3O, Te, T3L, Tj, Tf, T4W, T5m, T3M, Ta, T4V, Tg;
cannam@95 295 E T3H, T3N;
cannam@95 296 {
cannam@95 297 E T4, T9, T5, T4U, Tb, T3K, T1, T3E, T6;
cannam@95 298 {
cannam@95 299 E T2, T3, T7, T8, T58, T11;
cannam@95 300 T2 = Ip[WS(rs, 1)];
cannam@95 301 T58 = FMA(TX, TV, T57);
cannam@95 302 T11 = FNMS(TX, T10, TW);
cannam@95 303 {
cannam@95 304 E T5w, T4a, T5y, T4g;
cannam@95 305 T5w = FNMS(T48, T46, T5v);
cannam@95 306 T4a = FMA(T48, T49, T47);
cannam@95 307 T5y = FNMS(T4e, T4c, T5x);
cannam@95 308 T4g = FMA(T4e, T4f, T4d);
cannam@95 309 T59 = T56 + T58;
cannam@95 310 T7K = T56 - T58;
cannam@95 311 T7H = TR - T11;
cannam@95 312 T12 = TR + T11;
cannam@95 313 T5z = T5w + T5y;
cannam@95 314 T81 = T5w - T5y;
cannam@95 315 T7Q = T4g - T4a;
cannam@95 316 T4h = T4a + T4g;
cannam@95 317 T3 = Im[WS(rs, 1)];
cannam@95 318 }
cannam@95 319 T7 = Rp[WS(rs, 1)];
cannam@95 320 T8 = Rm[WS(rs, 1)];
cannam@95 321 T1 = W[2];
cannam@95 322 T3I = T2 + T3;
cannam@95 323 T4 = T2 - T3;
cannam@95 324 T3F = T7 - T8;
cannam@95 325 T9 = T7 + T8;
cannam@95 326 T3E = W[4];
cannam@95 327 T5 = T1 * T4;
cannam@95 328 }
cannam@95 329 {
cannam@95 330 E Tc, Td, Th, Ti;
cannam@95 331 Tc = Ip[WS(rs, 9)];
cannam@95 332 T4U = T1 * T9;
cannam@95 333 T5k = T3E * T3I;
cannam@95 334 T3G = T3E * T3F;
cannam@95 335 Td = Im[WS(rs, 9)];
cannam@95 336 Th = Rp[WS(rs, 9)];
cannam@95 337 Ti = Rm[WS(rs, 9)];
cannam@95 338 Tb = W[34];
cannam@95 339 T3O = Tc + Td;
cannam@95 340 Te = Tc - Td;
cannam@95 341 T3L = Th - Ti;
cannam@95 342 Tj = Th + Ti;
cannam@95 343 T3K = W[36];
cannam@95 344 Tf = Tb * Te;
cannam@95 345 }
cannam@95 346 T6 = W[3];
cannam@95 347 T4W = Tb * Tj;
cannam@95 348 T5m = T3K * T3O;
cannam@95 349 T3M = T3K * T3L;
cannam@95 350 Ta = FNMS(T6, T9, T5);
cannam@95 351 T4V = FMA(T6, T4, T4U);
cannam@95 352 Tg = W[35];
cannam@95 353 T3H = W[5];
cannam@95 354 T3N = W[37];
cannam@95 355 }
cannam@95 356 {
cannam@95 357 E T1t, T2N, T2M, T2J, T1y, T2L, T5H, T4I, T1u, T2S, T1D, T2T, T1I, T2Q, T2P;
cannam@95 358 E T1A, T5J;
cannam@95 359 {
cannam@95 360 E T2K, T1q, T1w, T1x;
cannam@95 361 {
cannam@95 362 E T1r, T7U, T7T, T1s, T4X, Tk;
cannam@95 363 T1r = Ip[WS(rs, 2)];
cannam@95 364 T4X = FMA(Tg, Te, T4W);
cannam@95 365 Tk = FNMS(Tg, Tj, Tf);
cannam@95 366 {
cannam@95 367 E T5l, T3J, T5n, T3P;
cannam@95 368 T5l = FNMS(T3H, T3F, T5k);
cannam@95 369 T3J = FMA(T3H, T3I, T3G);
cannam@95 370 T5n = FNMS(T3N, T3L, T5m);
cannam@95 371 T3P = FMA(T3N, T3O, T3M);
cannam@95 372 T4Y = T4V + T4X;
cannam@95 373 T7D = T4V - T4X;
cannam@95 374 T7A = Ta - Tk;
cannam@95 375 Tl = Ta + Tk;
cannam@95 376 T7U = T5l - T5n;
cannam@95 377 T5o = T5l + T5n;
cannam@95 378 T7T = T3P - T3J;
cannam@95 379 T3Q = T3J + T3P;
cannam@95 380 T1s = Im[WS(rs, 2)];
cannam@95 381 }
cannam@95 382 T1w = Rp[WS(rs, 2)];
cannam@95 383 T84 = T7U + T7T;
cannam@95 384 T7V = T7T - T7U;
cannam@95 385 T1t = T1r - T1s;
cannam@95 386 T2N = T1r + T1s;
cannam@95 387 T1x = Rm[WS(rs, 2)];
cannam@95 388 }
cannam@95 389 T2M = W[9];
cannam@95 390 T2J = W[8];
cannam@95 391 T1y = T1w + T1x;
cannam@95 392 T2K = T1x - T1w;
cannam@95 393 T1q = W[6];
cannam@95 394 {
cannam@95 395 E T1B, T1C, T1G, T1H;
cannam@95 396 T1B = Ip[WS(rs, 10)];
cannam@95 397 T2L = T2J * T2K;
cannam@95 398 T5H = T2M * T2K;
cannam@95 399 T4I = T1q * T1y;
cannam@95 400 T1u = T1q * T1t;
cannam@95 401 T1C = Im[WS(rs, 10)];
cannam@95 402 T1G = Rp[WS(rs, 10)];
cannam@95 403 T1H = Rm[WS(rs, 10)];
cannam@95 404 T2S = W[41];
cannam@95 405 T1D = T1B - T1C;
cannam@95 406 T2T = T1B + T1C;
cannam@95 407 T1I = T1G + T1H;
cannam@95 408 T2Q = T1H - T1G;
cannam@95 409 T2P = W[40];
cannam@95 410 T1A = W[38];
cannam@95 411 T5J = T2S * T2Q;
cannam@95 412 }
cannam@95 413 }
cannam@95 414 {
cannam@95 415 E T2R, T4K, T1E, T1z, T4J, T1F, T1v, T2O, T2U;
cannam@95 416 T1v = W[7];
cannam@95 417 T2R = T2P * T2Q;
cannam@95 418 T4K = T1A * T1I;
cannam@95 419 T1E = T1A * T1D;
cannam@95 420 T1z = FNMS(T1v, T1y, T1u);
cannam@95 421 T4J = FMA(T1v, T1t, T4I);
cannam@95 422 T1F = W[39];
cannam@95 423 T2O = FNMS(T2M, T2N, T2L);
cannam@95 424 T2U = FNMS(T2S, T2T, T2R);
cannam@95 425 {
cannam@95 426 E T5I, T4L, T1J, T5K;
cannam@95 427 T5I = FMA(T2J, T2N, T5H);
cannam@95 428 T4L = FMA(T1F, T1D, T4K);
cannam@95 429 T1J = FNMS(T1F, T1I, T1E);
cannam@95 430 T8c = T2O - T2U;
cannam@95 431 T2V = T2O + T2U;
cannam@95 432 T5K = FMA(T2P, T2T, T5J);
cannam@95 433 T4M = T4J + T4L;
cannam@95 434 T7t = T4J - T4L;
cannam@95 435 T7s = T1z - T1J;
cannam@95 436 T1K = T1z + T1J;
cannam@95 437 T8d = T5I - T5K;
cannam@95 438 T5L = T5I + T5K;
cannam@95 439 }
cannam@95 440 }
cannam@95 441 }
cannam@95 442 }
cannam@95 443 {
cannam@95 444 E T2Z, T30, T1O, T1T, T2W, T5M, T2Y, T4N, T1P, T35, T1Y, T36, T23, T33, T32;
cannam@95 445 E T1V, T5O, T2X, T1L;
cannam@95 446 {
cannam@95 447 E T1M, T1N, T1R, T1S;
cannam@95 448 T1M = Ip[WS(rs, 14)];
cannam@95 449 T8e = T8c - T8d;
cannam@95 450 T8n = T8c + T8d;
cannam@95 451 T1N = Im[WS(rs, 14)];
cannam@95 452 T1R = Rp[WS(rs, 14)];
cannam@95 453 T1S = Rm[WS(rs, 14)];
cannam@95 454 T2Z = W[57];
cannam@95 455 T30 = T1M + T1N;
cannam@95 456 T1O = T1M - T1N;
cannam@95 457 T2X = T1S - T1R;
cannam@95 458 T1T = T1R + T1S;
cannam@95 459 T2W = W[56];
cannam@95 460 T1L = W[54];
cannam@95 461 T5M = T2Z * T2X;
cannam@95 462 }
cannam@95 463 {
cannam@95 464 E T1W, T1X, T21, T22;
cannam@95 465 T1W = Ip[WS(rs, 6)];
cannam@95 466 T2Y = T2W * T2X;
cannam@95 467 T4N = T1L * T1T;
cannam@95 468 T1P = T1L * T1O;
cannam@95 469 T1X = Im[WS(rs, 6)];
cannam@95 470 T21 = Rp[WS(rs, 6)];
cannam@95 471 T22 = Rm[WS(rs, 6)];
cannam@95 472 T35 = W[25];
cannam@95 473 T1Y = T1W - T1X;
cannam@95 474 T36 = T1W + T1X;
cannam@95 475 T23 = T21 + T22;
cannam@95 476 T33 = T22 - T21;
cannam@95 477 T32 = W[24];
cannam@95 478 T1V = W[22];
cannam@95 479 T5O = T35 * T33;
cannam@95 480 }
cannam@95 481 {
cannam@95 482 E T34, T4P, T1Z, T1U, T4O, T20, T1Q, T31, T37;
cannam@95 483 T1Q = W[55];
cannam@95 484 T34 = T32 * T33;
cannam@95 485 T4P = T1V * T23;
cannam@95 486 T1Z = T1V * T1Y;
cannam@95 487 T1U = FNMS(T1Q, T1T, T1P);
cannam@95 488 T4O = FMA(T1Q, T1O, T4N);
cannam@95 489 T20 = W[23];
cannam@95 490 T31 = FNMS(T2Z, T30, T2Y);
cannam@95 491 T37 = FNMS(T35, T36, T34);
cannam@95 492 {
cannam@95 493 E T5N, T4Q, T24, T5P;
cannam@95 494 T5N = FMA(T2W, T30, T5M);
cannam@95 495 T4Q = FMA(T20, T1Y, T4P);
cannam@95 496 T24 = FNMS(T20, T23, T1Z);
cannam@95 497 T8f = T31 - T37;
cannam@95 498 T38 = T31 + T37;
cannam@95 499 T5P = FMA(T32, T36, T5O);
cannam@95 500 T7v = T4O - T4Q;
cannam@95 501 T4R = T4O + T4Q;
cannam@95 502 T7w = T1U - T24;
cannam@95 503 T25 = T1U + T24;
cannam@95 504 T8g = T5N - T5P;
cannam@95 505 T5Q = T5N + T5P;
cannam@95 506 }
cannam@95 507 }
cannam@95 508 }
cannam@95 509 {
cannam@95 510 E Tp, Tu, Tq, T4Z, Tw, T3X, Tm, T3R, Tr;
cannam@95 511 {
cannam@95 512 E Tn, To, Ts, Tt;
cannam@95 513 Tn = Ip[WS(rs, 5)];
cannam@95 514 T8h = T8f + T8g;
cannam@95 515 T8o = T8g - T8f;
cannam@95 516 To = Im[WS(rs, 5)];
cannam@95 517 Ts = Rp[WS(rs, 5)];
cannam@95 518 Tt = Rm[WS(rs, 5)];
cannam@95 519 Tm = W[18];
cannam@95 520 T3V = Tn + To;
cannam@95 521 Tp = Tn - To;
cannam@95 522 T3S = Ts - Tt;
cannam@95 523 Tu = Ts + Tt;
cannam@95 524 T3R = W[20];
cannam@95 525 Tq = Tm * Tp;
cannam@95 526 }
cannam@95 527 {
cannam@95 528 E Tx, Ty, TC, TD;
cannam@95 529 Tx = Ip[WS(rs, 13)];
cannam@95 530 T4Z = Tm * Tu;
cannam@95 531 T5p = T3R * T3V;
cannam@95 532 T3T = T3R * T3S;
cannam@95 533 Ty = Im[WS(rs, 13)];
cannam@95 534 TC = Rp[WS(rs, 13)];
cannam@95 535 TD = Rm[WS(rs, 13)];
cannam@95 536 Tw = W[50];
cannam@95 537 T41 = Tx + Ty;
cannam@95 538 Tz = Tx - Ty;
cannam@95 539 T3Y = TC - TD;
cannam@95 540 TE = TC + TD;
cannam@95 541 T3X = W[52];
cannam@95 542 TA = Tw * Tz;
cannam@95 543 }
cannam@95 544 Tr = W[19];
cannam@95 545 T51 = Tw * TE;
cannam@95 546 T5r = T3X * T41;
cannam@95 547 T3Z = T3X * T3Y;
cannam@95 548 Tv = FNMS(Tr, Tu, Tq);
cannam@95 549 T50 = FMA(Tr, Tp, T4Z);
cannam@95 550 TB = W[51];
cannam@95 551 T3U = W[21];
cannam@95 552 T40 = W[53];
cannam@95 553 }
cannam@95 554 }
cannam@95 555 }
cannam@95 556 {
cannam@95 557 E T6y, T7B, T7E, T6u, T6S, T85, T7Y, T6s, T6v, T6x, T6R, T6r, T6F, T6D, T6C;
cannam@95 558 E T6G, T6M, T6K, T6J, T6N, T6l, T6o, T7j, T7m;
cannam@95 559 {
cannam@95 560 E T6i, T1p, T68, T2H, T67, T5g, T6h, T4T, T4w, T5G, T6d, T3D, T6c, T6m, T63;
cannam@95 561 E T6e;
cannam@95 562 {
cannam@95 563 E T5t, T43, T26, T2G, T54, T5f, T4H, T4S;
cannam@95 564 {
cannam@95 565 E T1o, T53, T7W, T7X, TH, T52, TF, T5q;
cannam@95 566 T6y = T12 - T1n;
cannam@95 567 T1o = T12 + T1n;
cannam@95 568 T52 = FMA(TB, Tz, T51);
cannam@95 569 TF = FNMS(TB, TE, TA);
cannam@95 570 T5q = FNMS(T3U, T3S, T5p);
cannam@95 571 {
cannam@95 572 E T3W, T5s, T42, TG;
cannam@95 573 T3W = FMA(T3U, T3V, T3T);
cannam@95 574 T5s = FNMS(T40, T3Y, T5r);
cannam@95 575 T42 = FMA(T40, T41, T3Z);
cannam@95 576 T7B = T50 - T52;
cannam@95 577 T53 = T50 + T52;
cannam@95 578 T7E = Tv - TF;
cannam@95 579 TG = Tv + TF;
cannam@95 580 T7W = T5s - T5q;
cannam@95 581 T5t = T5q + T5s;
cannam@95 582 T7X = T3W - T42;
cannam@95 583 T43 = T3W + T42;
cannam@95 584 TH = Tl + TG;
cannam@95 585 T6u = Tl - TG;
cannam@95 586 }
cannam@95 587 T6S = T1K - T25;
cannam@95 588 T26 = T1K + T25;
cannam@95 589 T85 = T7W - T7X;
cannam@95 590 T7Y = T7W + T7X;
cannam@95 591 T6i = TH - T1o;
cannam@95 592 T1p = TH + T1o;
cannam@95 593 T2G = T2r + T2F;
cannam@95 594 T6s = T2F - T2r;
cannam@95 595 T6v = T4Y - T53;
cannam@95 596 T54 = T4Y + T53;
cannam@95 597 T5f = T59 + T5e;
cannam@95 598 T6x = T59 - T5e;
cannam@95 599 }
cannam@95 600 T6R = T4B - T4G;
cannam@95 601 T4H = T4B + T4G;
cannam@95 602 T68 = T2G - T26;
cannam@95 603 T2H = T26 + T2G;
cannam@95 604 T67 = T5f - T54;
cannam@95 605 T5g = T54 + T5f;
cannam@95 606 T4S = T4M + T4R;
cannam@95 607 T6r = T4R - T4M;
cannam@95 608 {
cannam@95 609 E T5u, T6b, T5F, T44, T4v;
cannam@95 610 T6F = T43 - T3Q;
cannam@95 611 T44 = T3Q + T43;
cannam@95 612 T4v = T4h + T4u;
cannam@95 613 T6D = T4u - T4h;
cannam@95 614 T6C = T5t - T5o;
cannam@95 615 T5u = T5o + T5t;
cannam@95 616 T6h = T4H - T4S;
cannam@95 617 T4T = T4H + T4S;
cannam@95 618 T6b = T44 - T4v;
cannam@95 619 T4w = T44 + T4v;
cannam@95 620 T6G = T5z - T5E;
cannam@95 621 T5F = T5z + T5E;
cannam@95 622 {
cannam@95 623 E T5R, T62, T39, T3C, T6a;
cannam@95 624 T6M = T2V - T38;
cannam@95 625 T39 = T2V + T38;
cannam@95 626 T3C = T3m + T3B;
cannam@95 627 T6K = T3B - T3m;
cannam@95 628 T6a = T5F - T5u;
cannam@95 629 T5G = T5u + T5F;
cannam@95 630 T6J = T5Q - T5L;
cannam@95 631 T5R = T5L + T5Q;
cannam@95 632 T6d = T3C - T39;
cannam@95 633 T3D = T39 + T3C;
cannam@95 634 T6N = T61 - T5W;
cannam@95 635 T62 = T5W + T61;
cannam@95 636 T6c = T6a + T6b;
cannam@95 637 T6m = T6a - T6b;
cannam@95 638 T63 = T5R + T62;
cannam@95 639 T6e = T62 - T5R;
cannam@95 640 }
cannam@95 641 }
cannam@95 642 }
cannam@95 643 {
cannam@95 644 E T5j, T6n, T6f, T64;
cannam@95 645 {
cannam@95 646 E T5i, T5h, T65, T66, T2I, T4x;
cannam@95 647 T5j = T2H - T1p;
cannam@95 648 T2I = T1p + T2H;
cannam@95 649 T4x = T3D - T4w;
cannam@95 650 T5i = T4w + T3D;
cannam@95 651 T6n = T6d + T6e;
cannam@95 652 T6f = T6d - T6e;
cannam@95 653 T5h = T4T - T5g;
cannam@95 654 T65 = T4T + T5g;
cannam@95 655 Im[WS(rs, 15)] = KP500000000 * (T4x - T2I);
cannam@95 656 Ip[0] = KP500000000 * (T2I + T4x);
cannam@95 657 T66 = T5G + T63;
cannam@95 658 T64 = T5G - T63;
cannam@95 659 Rp[0] = KP500000000 * (T65 + T66);
cannam@95 660 Rm[WS(rs, 15)] = KP500000000 * (T65 - T66);
cannam@95 661 Rp[WS(rs, 8)] = KP500000000 * (T5h + T5i);
cannam@95 662 Rm[WS(rs, 7)] = KP500000000 * (T5h - T5i);
cannam@95 663 }
cannam@95 664 {
cannam@95 665 E T6k, T6j, T6p, T6q, T69, T6g;
cannam@95 666 T6l = T68 - T67;
cannam@95 667 T69 = T67 + T68;
cannam@95 668 T6g = T6c + T6f;
cannam@95 669 T6k = T6f - T6c;
cannam@95 670 T6j = T6h - T6i;
cannam@95 671 T6p = T6h + T6i;
cannam@95 672 Im[WS(rs, 7)] = KP500000000 * (T64 - T5j);
cannam@95 673 Ip[WS(rs, 8)] = KP500000000 * (T5j + T64);
cannam@95 674 Im[WS(rs, 11)] = -(KP500000000 * (FNMS(KP707106781, T6g, T69)));
cannam@95 675 Ip[WS(rs, 4)] = KP500000000 * (FMA(KP707106781, T6g, T69));
cannam@95 676 T6q = T6m + T6n;
cannam@95 677 T6o = T6m - T6n;
cannam@95 678 Rp[WS(rs, 4)] = KP500000000 * (FMA(KP707106781, T6q, T6p));
cannam@95 679 Rm[WS(rs, 11)] = KP500000000 * (FNMS(KP707106781, T6q, T6p));
cannam@95 680 Rp[WS(rs, 12)] = KP500000000 * (FMA(KP707106781, T6k, T6j));
cannam@95 681 Rm[WS(rs, 3)] = KP500000000 * (FNMS(KP707106781, T6k, T6j));
cannam@95 682 }
cannam@95 683 }
cannam@95 684 }
cannam@95 685 {
cannam@95 686 E T75, T6t, T7f, T6T, T76, T6W, T7g, T6A, T7b, T6L, T7a, T7k, T70, T6I, T6U;
cannam@95 687 E T6w;
cannam@95 688 Im[WS(rs, 3)] = -(KP500000000 * (FNMS(KP707106781, T6o, T6l)));
cannam@95 689 Ip[WS(rs, 12)] = KP500000000 * (FMA(KP707106781, T6o, T6l));
cannam@95 690 T75 = T6s - T6r;
cannam@95 691 T6t = T6r + T6s;
cannam@95 692 T7f = T6R - T6S;
cannam@95 693 T6T = T6R + T6S;
cannam@95 694 T6U = T6v + T6u;
cannam@95 695 T6w = T6u - T6v;
cannam@95 696 {
cannam@95 697 E T78, T6E, T6V, T6z, T79, T6H;
cannam@95 698 T6V = T6x - T6y;
cannam@95 699 T6z = T6x + T6y;
cannam@95 700 T78 = T6C - T6D;
cannam@95 701 T6E = T6C + T6D;
cannam@95 702 T76 = T6V - T6U;
cannam@95 703 T6W = T6U + T6V;
cannam@95 704 T7g = T6w - T6z;
cannam@95 705 T6A = T6w + T6z;
cannam@95 706 T79 = T6G - T6F;
cannam@95 707 T6H = T6F + T6G;
cannam@95 708 T7b = T6K - T6J;
cannam@95 709 T6L = T6J + T6K;
cannam@95 710 T7a = FMA(KP414213562, T79, T78);
cannam@95 711 T7k = FNMS(KP414213562, T78, T79);
cannam@95 712 T70 = FNMS(KP414213562, T6E, T6H);
cannam@95 713 T6I = FMA(KP414213562, T6H, T6E);
cannam@95 714 }
cannam@95 715 {
cannam@95 716 E T6Z, T6B, T73, T6X, T7c, T6O;
cannam@95 717 T6Z = FNMS(KP707106781, T6A, T6t);
cannam@95 718 T6B = FMA(KP707106781, T6A, T6t);
cannam@95 719 T73 = FMA(KP707106781, T6W, T6T);
cannam@95 720 T6X = FNMS(KP707106781, T6W, T6T);
cannam@95 721 T7c = T6N - T6M;
cannam@95 722 T6O = T6M + T6N;
cannam@95 723 {
cannam@95 724 E T7i, T7h, T7n, T7o;
cannam@95 725 {
cannam@95 726 E T77, T7l, T71, T6P, T7e, T7d;
cannam@95 727 T7j = FMA(KP707106781, T76, T75);
cannam@95 728 T77 = FNMS(KP707106781, T76, T75);
cannam@95 729 T7d = FMA(KP414213562, T7c, T7b);
cannam@95 730 T7l = FNMS(KP414213562, T7b, T7c);
cannam@95 731 T71 = FMA(KP414213562, T6L, T6O);
cannam@95 732 T6P = FNMS(KP414213562, T6O, T6L);
cannam@95 733 T7e = T7a - T7d;
cannam@95 734 T7i = T7a + T7d;
cannam@95 735 T7h = FMA(KP707106781, T7g, T7f);
cannam@95 736 T7n = FNMS(KP707106781, T7g, T7f);
cannam@95 737 {
cannam@95 738 E T72, T74, T6Y, T6Q;
cannam@95 739 T72 = T70 - T71;
cannam@95 740 T74 = T70 + T71;
cannam@95 741 T6Y = T6P - T6I;
cannam@95 742 T6Q = T6I + T6P;
cannam@95 743 Im[WS(rs, 1)] = -(KP500000000 * (FNMS(KP923879532, T7e, T77)));
cannam@95 744 Ip[WS(rs, 14)] = KP500000000 * (FMA(KP923879532, T7e, T77));
cannam@95 745 Im[WS(rs, 5)] = -(KP500000000 * (FNMS(KP923879532, T72, T6Z)));
cannam@95 746 Ip[WS(rs, 10)] = KP500000000 * (FMA(KP923879532, T72, T6Z));
cannam@95 747 Rp[WS(rs, 2)] = KP500000000 * (FMA(KP923879532, T74, T73));
cannam@95 748 Rm[WS(rs, 13)] = KP500000000 * (FNMS(KP923879532, T74, T73));
cannam@95 749 Rp[WS(rs, 10)] = KP500000000 * (FMA(KP923879532, T6Y, T6X));
cannam@95 750 Rm[WS(rs, 5)] = KP500000000 * (FNMS(KP923879532, T6Y, T6X));
cannam@95 751 Im[WS(rs, 13)] = -(KP500000000 * (FNMS(KP923879532, T6Q, T6B)));
cannam@95 752 Ip[WS(rs, 2)] = KP500000000 * (FMA(KP923879532, T6Q, T6B));
cannam@95 753 T7o = T7k + T7l;
cannam@95 754 T7m = T7k - T7l;
cannam@95 755 }
cannam@95 756 }
cannam@95 757 Rm[WS(rs, 1)] = KP500000000 * (FMA(KP923879532, T7o, T7n));
cannam@95 758 Rp[WS(rs, 14)] = KP500000000 * (FNMS(KP923879532, T7o, T7n));
cannam@95 759 Rp[WS(rs, 6)] = KP500000000 * (FMA(KP923879532, T7i, T7h));
cannam@95 760 Rm[WS(rs, 9)] = KP500000000 * (FNMS(KP923879532, T7i, T7h));
cannam@95 761 }
cannam@95 762 }
cannam@95 763 }
cannam@95 764 {
cannam@95 765 E T9x, T9T, T8L, T7z, T97, T9J, T8V, T8z, T8M, T8C, T8W, T7O, T9O, T9Y, T9E;
cannam@95 766 E T9t, T8Q, T90, T8G, T88, T8p, T8m, T9K, T9A, T9U, T9e, T8R, T8j, T9R, T9Z;
cannam@95 767 E T9F, T9m;
cannam@95 768 {
cannam@95 769 E T9c, T9b, T99, T98, T7S, T86, T83, T9q, T9M, T9p, T9r, T7Z, T9z, T9a;
cannam@95 770 {
cannam@95 771 E T95, T7r, T9v, T8v, T8w, T8x, T9w, T7y, T7u, T7x;
cannam@95 772 T95 = T7q + T7p;
cannam@95 773 T7r = T7p - T7q;
cannam@95 774 T9v = T8t - T8u;
cannam@95 775 T8v = T8t + T8u;
cannam@95 776 T8w = T7t + T7s;
cannam@95 777 T7u = T7s - T7t;
cannam@95 778 Im[WS(rs, 9)] = -(KP500000000 * (FNMS(KP923879532, T7m, T7j)));
cannam@95 779 Ip[WS(rs, 6)] = KP500000000 * (FMA(KP923879532, T7m, T7j));
cannam@95 780 T7x = T7v + T7w;
cannam@95 781 T8x = T7v - T7w;
cannam@95 782 T9w = T7u - T7x;
cannam@95 783 T7y = T7u + T7x;
cannam@95 784 {
cannam@95 785 E T7J, T8A, T7G, T7M;
cannam@95 786 {
cannam@95 787 E T7C, T96, T8y, T7F;
cannam@95 788 T9c = T7A + T7B;
cannam@95 789 T7C = T7A - T7B;
cannam@95 790 T9x = FMA(KP707106781, T9w, T9v);
cannam@95 791 T9T = FNMS(KP707106781, T9w, T9v);
cannam@95 792 T8L = FNMS(KP707106781, T7y, T7r);
cannam@95 793 T7z = FMA(KP707106781, T7y, T7r);
cannam@95 794 T96 = T8x - T8w;
cannam@95 795 T8y = T8w + T8x;
cannam@95 796 T7F = T7D + T7E;
cannam@95 797 T9b = T7D - T7E;
cannam@95 798 T99 = T7H + T7I;
cannam@95 799 T7J = T7H - T7I;
cannam@95 800 T97 = FMA(KP707106781, T96, T95);
cannam@95 801 T9J = FNMS(KP707106781, T96, T95);
cannam@95 802 T8V = FNMS(KP707106781, T8y, T8v);
cannam@95 803 T8z = FMA(KP707106781, T8y, T8v);
cannam@95 804 T8A = FMA(KP414213562, T7C, T7F);
cannam@95 805 T7G = FNMS(KP414213562, T7F, T7C);
cannam@95 806 T7M = T7K + T7L;
cannam@95 807 T98 = T7K - T7L;
cannam@95 808 }
cannam@95 809 {
cannam@95 810 E T9n, T9o, T8B, T7N;
cannam@95 811 T7S = T7Q + T7R;
cannam@95 812 T9n = T7R - T7Q;
cannam@95 813 T9o = T85 - T84;
cannam@95 814 T86 = T84 + T85;
cannam@95 815 T83 = T81 + T82;
cannam@95 816 T9q = T81 - T82;
cannam@95 817 T8B = FNMS(KP414213562, T7J, T7M);
cannam@95 818 T7N = FMA(KP414213562, T7M, T7J);
cannam@95 819 T9M = FMA(KP707106781, T9o, T9n);
cannam@95 820 T9p = FNMS(KP707106781, T9o, T9n);
cannam@95 821 T8M = T8B - T8A;
cannam@95 822 T8C = T8A + T8B;
cannam@95 823 T8W = T7G - T7N;
cannam@95 824 T7O = T7G + T7N;
cannam@95 825 T9r = T7Y - T7V;
cannam@95 826 T7Z = T7V + T7Y;
cannam@95 827 }
cannam@95 828 }
cannam@95 829 }
cannam@95 830 {
cannam@95 831 E T8O, T80, T9N, T9s, T8P, T87;
cannam@95 832 T9N = FMA(KP707106781, T9r, T9q);
cannam@95 833 T9s = FNMS(KP707106781, T9r, T9q);
cannam@95 834 T8O = FNMS(KP707106781, T7Z, T7S);
cannam@95 835 T80 = FMA(KP707106781, T7Z, T7S);
cannam@95 836 T9O = FMA(KP198912367, T9N, T9M);
cannam@95 837 T9Y = FNMS(KP198912367, T9M, T9N);
cannam@95 838 T9E = FMA(KP668178637, T9p, T9s);
cannam@95 839 T9t = FNMS(KP668178637, T9s, T9p);
cannam@95 840 T8P = FNMS(KP707106781, T86, T83);
cannam@95 841 T87 = FMA(KP707106781, T86, T83);
cannam@95 842 T9z = FNMS(KP414213562, T98, T99);
cannam@95 843 T9a = FMA(KP414213562, T99, T98);
cannam@95 844 T8Q = FNMS(KP668178637, T8P, T8O);
cannam@95 845 T90 = FMA(KP668178637, T8O, T8P);
cannam@95 846 T8G = FNMS(KP198912367, T80, T87);
cannam@95 847 T88 = FMA(KP198912367, T87, T80);
cannam@95 848 }
cannam@95 849 {
cannam@95 850 E T8b, T9j, T9P, T9i, T9k, T8i, T9Q, T9l;
cannam@95 851 {
cannam@95 852 E T9g, T9h, T9y, T9d;
cannam@95 853 T8b = T89 - T8a;
cannam@95 854 T9g = T8a + T89;
cannam@95 855 T9h = T8n - T8o;
cannam@95 856 T8p = T8n + T8o;
cannam@95 857 T8m = T8k + T8l;
cannam@95 858 T9j = T8l - T8k;
cannam@95 859 T9y = FMA(KP414213562, T9b, T9c);
cannam@95 860 T9d = FNMS(KP414213562, T9c, T9b);
cannam@95 861 T9P = FMA(KP707106781, T9h, T9g);
cannam@95 862 T9i = FNMS(KP707106781, T9h, T9g);
cannam@95 863 T9K = T9y + T9z;
cannam@95 864 T9A = T9y - T9z;
cannam@95 865 T9U = T9d + T9a;
cannam@95 866 T9e = T9a - T9d;
cannam@95 867 T9k = T8h - T8e;
cannam@95 868 T8i = T8e + T8h;
cannam@95 869 }
cannam@95 870 T9Q = FMA(KP707106781, T9k, T9j);
cannam@95 871 T9l = FNMS(KP707106781, T9k, T9j);
cannam@95 872 T8R = FNMS(KP707106781, T8i, T8b);
cannam@95 873 T8j = FMA(KP707106781, T8i, T8b);
cannam@95 874 T9R = FMA(KP198912367, T9Q, T9P);
cannam@95 875 T9Z = FNMS(KP198912367, T9P, T9Q);
cannam@95 876 T9F = FMA(KP668178637, T9i, T9l);
cannam@95 877 T9m = FNMS(KP668178637, T9l, T9i);
cannam@95 878 }
cannam@95 879 }
cannam@95 880 {
cannam@95 881 E T8Z, T92, T9D, T9G;
cannam@95 882 {
cannam@95 883 E T8F, T7P, T8J, T8D, T8S, T8q;
cannam@95 884 T8F = FNMS(KP923879532, T7O, T7z);
cannam@95 885 T7P = FMA(KP923879532, T7O, T7z);
cannam@95 886 T8J = FMA(KP923879532, T8C, T8z);
cannam@95 887 T8D = FNMS(KP923879532, T8C, T8z);
cannam@95 888 T8S = FNMS(KP707106781, T8p, T8m);
cannam@95 889 T8q = FMA(KP707106781, T8p, T8m);
cannam@95 890 {
cannam@95 891 E T8Y, T8X, T93, T94;
cannam@95 892 {
cannam@95 893 E T8N, T91, T8H, T8r, T8U, T8T;
cannam@95 894 T8Z = FMA(KP923879532, T8M, T8L);
cannam@95 895 T8N = FNMS(KP923879532, T8M, T8L);
cannam@95 896 T8T = FMA(KP668178637, T8S, T8R);
cannam@95 897 T91 = FNMS(KP668178637, T8R, T8S);
cannam@95 898 T8H = FMA(KP198912367, T8j, T8q);
cannam@95 899 T8r = FNMS(KP198912367, T8q, T8j);
cannam@95 900 T8U = T8Q + T8T;
cannam@95 901 T8Y = T8T - T8Q;
cannam@95 902 T8X = FMA(KP923879532, T8W, T8V);
cannam@95 903 T93 = FNMS(KP923879532, T8W, T8V);
cannam@95 904 {
cannam@95 905 E T8I, T8K, T8E, T8s;
cannam@95 906 T8I = T8G - T8H;
cannam@95 907 T8K = T8G + T8H;
cannam@95 908 T8E = T8r - T88;
cannam@95 909 T8s = T88 + T8r;
cannam@95 910 Im[WS(rs, 2)] = -(KP500000000 * (FMA(KP831469612, T8U, T8N)));
cannam@95 911 Ip[WS(rs, 13)] = KP500000000 * (FNMS(KP831469612, T8U, T8N));
cannam@95 912 Im[WS(rs, 6)] = -(KP500000000 * (FNMS(KP980785280, T8I, T8F)));
cannam@95 913 Ip[WS(rs, 9)] = KP500000000 * (FMA(KP980785280, T8I, T8F));
cannam@95 914 Rp[WS(rs, 1)] = KP500000000 * (FMA(KP980785280, T8K, T8J));
cannam@95 915 Rm[WS(rs, 14)] = KP500000000 * (FNMS(KP980785280, T8K, T8J));
cannam@95 916 Rp[WS(rs, 9)] = KP500000000 * (FMA(KP980785280, T8E, T8D));
cannam@95 917 Rm[WS(rs, 6)] = KP500000000 * (FNMS(KP980785280, T8E, T8D));
cannam@95 918 Im[WS(rs, 14)] = -(KP500000000 * (FNMS(KP980785280, T8s, T7P)));
cannam@95 919 Ip[WS(rs, 1)] = KP500000000 * (FMA(KP980785280, T8s, T7P));
cannam@95 920 T94 = T90 + T91;
cannam@95 921 T92 = T90 - T91;
cannam@95 922 }
cannam@95 923 }
cannam@95 924 Rm[WS(rs, 2)] = KP500000000 * (FMA(KP831469612, T94, T93));
cannam@95 925 Rp[WS(rs, 13)] = KP500000000 * (FNMS(KP831469612, T94, T93));
cannam@95 926 Rp[WS(rs, 5)] = KP500000000 * (FMA(KP831469612, T8Y, T8X));
cannam@95 927 Rm[WS(rs, 10)] = KP500000000 * (FNMS(KP831469612, T8Y, T8X));
cannam@95 928 }
cannam@95 929 }
cannam@95 930 {
cannam@95 931 E T9C, T9B, T9H, T9I, T9f, T9u;
cannam@95 932 T9D = FNMS(KP923879532, T9e, T97);
cannam@95 933 T9f = FMA(KP923879532, T9e, T97);
cannam@95 934 T9u = T9m - T9t;
cannam@95 935 T9C = T9t + T9m;
cannam@95 936 T9B = FNMS(KP923879532, T9A, T9x);
cannam@95 937 T9H = FMA(KP923879532, T9A, T9x);
cannam@95 938 Im[WS(rs, 10)] = -(KP500000000 * (FNMS(KP831469612, T92, T8Z)));
cannam@95 939 Ip[WS(rs, 5)] = KP500000000 * (FMA(KP831469612, T92, T8Z));
cannam@95 940 Im[WS(rs, 12)] = -(KP500000000 * (FNMS(KP831469612, T9u, T9f)));
cannam@95 941 Ip[WS(rs, 3)] = KP500000000 * (FMA(KP831469612, T9u, T9f));
cannam@95 942 T9I = T9E + T9F;
cannam@95 943 T9G = T9E - T9F;
cannam@95 944 Rp[WS(rs, 3)] = KP500000000 * (FMA(KP831469612, T9I, T9H));
cannam@95 945 Rm[WS(rs, 12)] = KP500000000 * (FNMS(KP831469612, T9I, T9H));
cannam@95 946 Rp[WS(rs, 11)] = KP500000000 * (FMA(KP831469612, T9C, T9B));
cannam@95 947 Rm[WS(rs, 4)] = KP500000000 * (FNMS(KP831469612, T9C, T9B));
cannam@95 948 }
cannam@95 949 {
cannam@95 950 E T9W, T9V, Ta1, Ta2, T9L, T9S;
cannam@95 951 T9X = FNMS(KP923879532, T9K, T9J);
cannam@95 952 T9L = FMA(KP923879532, T9K, T9J);
cannam@95 953 T9S = T9O - T9R;
cannam@95 954 T9W = T9O + T9R;
cannam@95 955 T9V = FNMS(KP923879532, T9U, T9T);
cannam@95 956 Ta1 = FMA(KP923879532, T9U, T9T);
cannam@95 957 Im[WS(rs, 4)] = -(KP500000000 * (FNMS(KP831469612, T9G, T9D)));
cannam@95 958 Ip[WS(rs, 11)] = KP500000000 * (FMA(KP831469612, T9G, T9D));
cannam@95 959 Im[0] = -(KP500000000 * (FNMS(KP980785280, T9S, T9L)));
cannam@95 960 Ip[WS(rs, 15)] = KP500000000 * (FMA(KP980785280, T9S, T9L));
cannam@95 961 Ta2 = T9Y + T9Z;
cannam@95 962 Ta0 = T9Y - T9Z;
cannam@95 963 Rm[0] = KP500000000 * (FMA(KP980785280, Ta2, Ta1));
cannam@95 964 Rp[WS(rs, 15)] = KP500000000 * (FNMS(KP980785280, Ta2, Ta1));
cannam@95 965 Rp[WS(rs, 7)] = KP500000000 * (FMA(KP980785280, T9W, T9V));
cannam@95 966 Rm[WS(rs, 8)] = KP500000000 * (FNMS(KP980785280, T9W, T9V));
cannam@95 967 }
cannam@95 968 }
cannam@95 969 }
cannam@95 970 }
cannam@95 971 }
cannam@95 972 Im[WS(rs, 8)] = -(KP500000000 * (FNMS(KP980785280, Ta0, T9X)));
cannam@95 973 Ip[WS(rs, 7)] = KP500000000 * (FMA(KP980785280, Ta0, T9X));
cannam@95 974 }
cannam@95 975 }
cannam@95 976 }
cannam@95 977
cannam@95 978 static const tw_instr twinstr[] = {
cannam@95 979 {TW_FULL, 1, 32},
cannam@95 980 {TW_NEXT, 1, 0}
cannam@95 981 };
cannam@95 982
cannam@95 983 static const hc2c_desc desc = { 32, "hc2cfdft_32", twinstr, &GENUS, {300, 126, 198, 0} };
cannam@95 984
cannam@95 985 void X(codelet_hc2cfdft_32) (planner *p) {
cannam@95 986 X(khc2c_register) (p, hc2cfdft_32, &desc, HC2C_VIA_DFT);
cannam@95 987 }
cannam@95 988 #else /* HAVE_FMA */
cannam@95 989
cannam@95 990 /* Generated by: ../../../genfft/gen_hc2cdft.native -compact -variables 4 -pipeline-latency 4 -n 32 -dit -name hc2cfdft_32 -include hc2cf.h */
cannam@95 991
cannam@95 992 /*
cannam@95 993 * This function contains 498 FP additions, 228 FP multiplications,
cannam@95 994 * (or, 404 additions, 134 multiplications, 94 fused multiply/add),
cannam@95 995 * 106 stack variables, 9 constants, and 128 memory accesses
cannam@95 996 */
cannam@95 997 #include "hc2cf.h"
cannam@95 998
cannam@95 999 static void hc2cfdft_32(R *Rp, R *Ip, R *Rm, R *Im, const R *W, stride rs, INT mb, INT me, INT ms)
cannam@95 1000 {
cannam@95 1001 DK(KP277785116, +0.277785116509801112371415406974266437187468595);
cannam@95 1002 DK(KP415734806, +0.415734806151272618539394188808952878369280406);
cannam@95 1003 DK(KP097545161, +0.097545161008064133924142434238511120463845809);
cannam@95 1004 DK(KP490392640, +0.490392640201615224563091118067119518486966865);
cannam@95 1005 DK(KP707106781, +0.707106781186547524400844362104849039284835938);
cannam@95 1006 DK(KP191341716, +0.191341716182544885864229992015199433380672281);
cannam@95 1007 DK(KP461939766, +0.461939766255643378064091594698394143411208313);
cannam@95 1008 DK(KP353553390, +0.353553390593273762200422181052424519642417969);
cannam@95 1009 DK(KP500000000, +0.500000000000000000000000000000000000000000000);
cannam@95 1010 {
cannam@95 1011 INT m;
cannam@95 1012 for (m = mb, W = W + ((mb - 1) * 62); m < me; m = m + 1, Rp = Rp + ms, Ip = Ip + ms, Rm = Rm - ms, Im = Im - ms, W = W + 62, MAKE_VOLATILE_STRIDE(128, rs)) {
cannam@95 1013 E T2S, T5K, T52, T5N, T7p, T8r, T7i, T8o, T2q, T7t, T45, T6L, T2d, T7u, T48;
cannam@95 1014 E T6M, T1A, T4c, T4f, T1T, T3f, T5M, T7e, T7l, T6J, T7x, T4V, T5J, T7b, T7k;
cannam@95 1015 E T6G, T7w, Tj, TC, T5r, T4k, T4n, T5s, T3D, T5C, T6V, T72, T4G, T5F, T6u;
cannam@95 1016 E T86, T6S, T71, T6r, T85, TW, T1f, T5v, T4r, T4u, T5u, T40, T5G, T76, T8k;
cannam@95 1017 E T4N, T5D, T6B, T89, T6Z, T8h, T6y, T88;
cannam@95 1018 {
cannam@95 1019 E T1Y, T22, T2L, T4W, T2p, T43, T2A, T50, T27, T2b, T2Q, T4X, T2h, T2l, T2F;
cannam@95 1020 E T4Z;
cannam@95 1021 {
cannam@95 1022 E T1W, T1X, T2K, T20, T21, T2I, T2H, T2J;
cannam@95 1023 T1W = Ip[WS(rs, 4)];
cannam@95 1024 T1X = Im[WS(rs, 4)];
cannam@95 1025 T2K = T1W + T1X;
cannam@95 1026 T20 = Rp[WS(rs, 4)];
cannam@95 1027 T21 = Rm[WS(rs, 4)];
cannam@95 1028 T2I = T20 - T21;
cannam@95 1029 T1Y = T1W - T1X;
cannam@95 1030 T22 = T20 + T21;
cannam@95 1031 T2H = W[16];
cannam@95 1032 T2J = W[17];
cannam@95 1033 T2L = FMA(T2H, T2I, T2J * T2K);
cannam@95 1034 T4W = FNMS(T2J, T2I, T2H * T2K);
cannam@95 1035 }
cannam@95 1036 {
cannam@95 1037 E T2n, T2o, T2z, T2v, T2w, T2x, T2u, T2y;
cannam@95 1038 T2n = Ip[0];
cannam@95 1039 T2o = Im[0];
cannam@95 1040 T2z = T2n + T2o;
cannam@95 1041 T2v = Rm[0];
cannam@95 1042 T2w = Rp[0];
cannam@95 1043 T2x = T2v - T2w;
cannam@95 1044 T2p = T2n - T2o;
cannam@95 1045 T43 = T2w + T2v;
cannam@95 1046 T2u = W[0];
cannam@95 1047 T2y = W[1];
cannam@95 1048 T2A = FNMS(T2y, T2z, T2u * T2x);
cannam@95 1049 T50 = FMA(T2y, T2x, T2u * T2z);
cannam@95 1050 }
cannam@95 1051 {
cannam@95 1052 E T25, T26, T2P, T29, T2a, T2N, T2M, T2O;
cannam@95 1053 T25 = Ip[WS(rs, 12)];
cannam@95 1054 T26 = Im[WS(rs, 12)];
cannam@95 1055 T2P = T25 + T26;
cannam@95 1056 T29 = Rp[WS(rs, 12)];
cannam@95 1057 T2a = Rm[WS(rs, 12)];
cannam@95 1058 T2N = T29 - T2a;
cannam@95 1059 T27 = T25 - T26;
cannam@95 1060 T2b = T29 + T2a;
cannam@95 1061 T2M = W[48];
cannam@95 1062 T2O = W[49];
cannam@95 1063 T2Q = FMA(T2M, T2N, T2O * T2P);
cannam@95 1064 T4X = FNMS(T2O, T2N, T2M * T2P);
cannam@95 1065 }
cannam@95 1066 {
cannam@95 1067 E T2f, T2g, T2E, T2j, T2k, T2C, T2B, T2D;
cannam@95 1068 T2f = Ip[WS(rs, 8)];
cannam@95 1069 T2g = Im[WS(rs, 8)];
cannam@95 1070 T2E = T2f + T2g;
cannam@95 1071 T2j = Rp[WS(rs, 8)];
cannam@95 1072 T2k = Rm[WS(rs, 8)];
cannam@95 1073 T2C = T2j - T2k;
cannam@95 1074 T2h = T2f - T2g;
cannam@95 1075 T2l = T2j + T2k;
cannam@95 1076 T2B = W[32];
cannam@95 1077 T2D = W[33];
cannam@95 1078 T2F = FMA(T2B, T2C, T2D * T2E);
cannam@95 1079 T4Z = FNMS(T2D, T2C, T2B * T2E);
cannam@95 1080 }
cannam@95 1081 {
cannam@95 1082 E T2G, T2R, T7g, T7h;
cannam@95 1083 T2G = T2A - T2F;
cannam@95 1084 T2R = T2L + T2Q;
cannam@95 1085 T2S = T2G - T2R;
cannam@95 1086 T5K = T2R + T2G;
cannam@95 1087 {
cannam@95 1088 E T4Y, T51, T7n, T7o;
cannam@95 1089 T4Y = T4W + T4X;
cannam@95 1090 T51 = T4Z + T50;
cannam@95 1091 T52 = T4Y + T51;
cannam@95 1092 T5N = T51 - T4Y;
cannam@95 1093 T7n = T2Q - T2L;
cannam@95 1094 T7o = T50 - T4Z;
cannam@95 1095 T7p = T7n + T7o;
cannam@95 1096 T8r = T7o - T7n;
cannam@95 1097 }
cannam@95 1098 T7g = T2F + T2A;
cannam@95 1099 T7h = T4W - T4X;
cannam@95 1100 T7i = T7g - T7h;
cannam@95 1101 T8o = T7h + T7g;
cannam@95 1102 {
cannam@95 1103 E T2m, T44, T2e, T2i;
cannam@95 1104 T2e = W[30];
cannam@95 1105 T2i = W[31];
cannam@95 1106 T2m = FNMS(T2i, T2l, T2e * T2h);
cannam@95 1107 T44 = FMA(T2e, T2l, T2i * T2h);
cannam@95 1108 T2q = T2m + T2p;
cannam@95 1109 T7t = T43 - T44;
cannam@95 1110 T45 = T43 + T44;
cannam@95 1111 T6L = T2p - T2m;
cannam@95 1112 }
cannam@95 1113 {
cannam@95 1114 E T23, T46, T2c, T47;
cannam@95 1115 {
cannam@95 1116 E T1V, T1Z, T24, T28;
cannam@95 1117 T1V = W[14];
cannam@95 1118 T1Z = W[15];
cannam@95 1119 T23 = FNMS(T1Z, T22, T1V * T1Y);
cannam@95 1120 T46 = FMA(T1V, T22, T1Z * T1Y);
cannam@95 1121 T24 = W[46];
cannam@95 1122 T28 = W[47];
cannam@95 1123 T2c = FNMS(T28, T2b, T24 * T27);
cannam@95 1124 T47 = FMA(T24, T2b, T28 * T27);
cannam@95 1125 }
cannam@95 1126 T2d = T23 + T2c;
cannam@95 1127 T7u = T23 - T2c;
cannam@95 1128 T48 = T46 + T47;
cannam@95 1129 T6M = T46 - T47;
cannam@95 1130 }
cannam@95 1131 }
cannam@95 1132 }
cannam@95 1133 {
cannam@95 1134 E T1q, T4a, T2X, T4P, T1S, T4e, T3d, T4T, T1z, T4b, T32, T4Q, T1J, T4d, T38;
cannam@95 1135 E T4S;
cannam@95 1136 {
cannam@95 1137 E T1l, T2W, T1p, T2U;
cannam@95 1138 {
cannam@95 1139 E T1j, T1k, T1n, T1o;
cannam@95 1140 T1j = Ip[WS(rs, 2)];
cannam@95 1141 T1k = Im[WS(rs, 2)];
cannam@95 1142 T1l = T1j - T1k;
cannam@95 1143 T2W = T1j + T1k;
cannam@95 1144 T1n = Rp[WS(rs, 2)];
cannam@95 1145 T1o = Rm[WS(rs, 2)];
cannam@95 1146 T1p = T1n + T1o;
cannam@95 1147 T2U = T1n - T1o;
cannam@95 1148 }
cannam@95 1149 {
cannam@95 1150 E T1i, T1m, T2T, T2V;
cannam@95 1151 T1i = W[6];
cannam@95 1152 T1m = W[7];
cannam@95 1153 T1q = FNMS(T1m, T1p, T1i * T1l);
cannam@95 1154 T4a = FMA(T1i, T1p, T1m * T1l);
cannam@95 1155 T2T = W[8];
cannam@95 1156 T2V = W[9];
cannam@95 1157 T2X = FMA(T2T, T2U, T2V * T2W);
cannam@95 1158 T4P = FNMS(T2V, T2U, T2T * T2W);
cannam@95 1159 }
cannam@95 1160 }
cannam@95 1161 {
cannam@95 1162 E T1N, T3c, T1R, T3a;
cannam@95 1163 {
cannam@95 1164 E T1L, T1M, T1P, T1Q;
cannam@95 1165 T1L = Ip[WS(rs, 6)];
cannam@95 1166 T1M = Im[WS(rs, 6)];
cannam@95 1167 T1N = T1L - T1M;
cannam@95 1168 T3c = T1L + T1M;
cannam@95 1169 T1P = Rp[WS(rs, 6)];
cannam@95 1170 T1Q = Rm[WS(rs, 6)];
cannam@95 1171 T1R = T1P + T1Q;
cannam@95 1172 T3a = T1P - T1Q;
cannam@95 1173 }
cannam@95 1174 {
cannam@95 1175 E T1K, T1O, T39, T3b;
cannam@95 1176 T1K = W[22];
cannam@95 1177 T1O = W[23];
cannam@95 1178 T1S = FNMS(T1O, T1R, T1K * T1N);
cannam@95 1179 T4e = FMA(T1K, T1R, T1O * T1N);
cannam@95 1180 T39 = W[24];
cannam@95 1181 T3b = W[25];
cannam@95 1182 T3d = FMA(T39, T3a, T3b * T3c);
cannam@95 1183 T4T = FNMS(T3b, T3a, T39 * T3c);
cannam@95 1184 }
cannam@95 1185 }
cannam@95 1186 {
cannam@95 1187 E T1u, T31, T1y, T2Z;
cannam@95 1188 {
cannam@95 1189 E T1s, T1t, T1w, T1x;
cannam@95 1190 T1s = Ip[WS(rs, 10)];
cannam@95 1191 T1t = Im[WS(rs, 10)];
cannam@95 1192 T1u = T1s - T1t;
cannam@95 1193 T31 = T1s + T1t;
cannam@95 1194 T1w = Rp[WS(rs, 10)];
cannam@95 1195 T1x = Rm[WS(rs, 10)];
cannam@95 1196 T1y = T1w + T1x;
cannam@95 1197 T2Z = T1w - T1x;
cannam@95 1198 }
cannam@95 1199 {
cannam@95 1200 E T1r, T1v, T2Y, T30;
cannam@95 1201 T1r = W[38];
cannam@95 1202 T1v = W[39];
cannam@95 1203 T1z = FNMS(T1v, T1y, T1r * T1u);
cannam@95 1204 T4b = FMA(T1r, T1y, T1v * T1u);
cannam@95 1205 T2Y = W[40];
cannam@95 1206 T30 = W[41];
cannam@95 1207 T32 = FMA(T2Y, T2Z, T30 * T31);
cannam@95 1208 T4Q = FNMS(T30, T2Z, T2Y * T31);
cannam@95 1209 }
cannam@95 1210 }
cannam@95 1211 {
cannam@95 1212 E T1E, T37, T1I, T35;
cannam@95 1213 {
cannam@95 1214 E T1C, T1D, T1G, T1H;
cannam@95 1215 T1C = Ip[WS(rs, 14)];
cannam@95 1216 T1D = Im[WS(rs, 14)];
cannam@95 1217 T1E = T1C - T1D;
cannam@95 1218 T37 = T1C + T1D;
cannam@95 1219 T1G = Rp[WS(rs, 14)];
cannam@95 1220 T1H = Rm[WS(rs, 14)];
cannam@95 1221 T1I = T1G + T1H;
cannam@95 1222 T35 = T1G - T1H;
cannam@95 1223 }
cannam@95 1224 {
cannam@95 1225 E T1B, T1F, T34, T36;
cannam@95 1226 T1B = W[54];
cannam@95 1227 T1F = W[55];
cannam@95 1228 T1J = FNMS(T1F, T1I, T1B * T1E);
cannam@95 1229 T4d = FMA(T1B, T1I, T1F * T1E);
cannam@95 1230 T34 = W[56];
cannam@95 1231 T36 = W[57];
cannam@95 1232 T38 = FMA(T34, T35, T36 * T37);
cannam@95 1233 T4S = FNMS(T36, T35, T34 * T37);
cannam@95 1234 }
cannam@95 1235 }
cannam@95 1236 {
cannam@95 1237 E T33, T3e, T4R, T4U;
cannam@95 1238 T1A = T1q + T1z;
cannam@95 1239 T4c = T4a + T4b;
cannam@95 1240 T4f = T4d + T4e;
cannam@95 1241 T1T = T1J + T1S;
cannam@95 1242 T33 = T2X + T32;
cannam@95 1243 T3e = T38 + T3d;
cannam@95 1244 T3f = T33 + T3e;
cannam@95 1245 T5M = T3e - T33;
cannam@95 1246 {
cannam@95 1247 E T7c, T7d, T6H, T6I;
cannam@95 1248 T7c = T4S - T4T;
cannam@95 1249 T7d = T3d - T38;
cannam@95 1250 T7e = T7c + T7d;
cannam@95 1251 T7l = T7c - T7d;
cannam@95 1252 T6H = T4d - T4e;
cannam@95 1253 T6I = T1J - T1S;
cannam@95 1254 T6J = T6H + T6I;
cannam@95 1255 T7x = T6H - T6I;
cannam@95 1256 }
cannam@95 1257 T4R = T4P + T4Q;
cannam@95 1258 T4U = T4S + T4T;
cannam@95 1259 T4V = T4R + T4U;
cannam@95 1260 T5J = T4U - T4R;
cannam@95 1261 {
cannam@95 1262 E T79, T7a, T6E, T6F;
cannam@95 1263 T79 = T32 - T2X;
cannam@95 1264 T7a = T4P - T4Q;
cannam@95 1265 T7b = T79 - T7a;
cannam@95 1266 T7k = T7a + T79;
cannam@95 1267 T6E = T1q - T1z;
cannam@95 1268 T6F = T4a - T4b;
cannam@95 1269 T6G = T6E - T6F;
cannam@95 1270 T7w = T6F + T6E;
cannam@95 1271 }
cannam@95 1272 }
cannam@95 1273 }
cannam@95 1274 {
cannam@95 1275 E T9, T4i, T3l, T4A, TB, T4m, T3B, T4E, Ti, T4j, T3q, T4B, Ts, T4l, T3w;
cannam@95 1276 E T4D;
cannam@95 1277 {
cannam@95 1278 E T4, T3k, T8, T3i;
cannam@95 1279 {
cannam@95 1280 E T2, T3, T6, T7;
cannam@95 1281 T2 = Ip[WS(rs, 1)];
cannam@95 1282 T3 = Im[WS(rs, 1)];
cannam@95 1283 T4 = T2 - T3;
cannam@95 1284 T3k = T2 + T3;
cannam@95 1285 T6 = Rp[WS(rs, 1)];
cannam@95 1286 T7 = Rm[WS(rs, 1)];
cannam@95 1287 T8 = T6 + T7;
cannam@95 1288 T3i = T6 - T7;
cannam@95 1289 }
cannam@95 1290 {
cannam@95 1291 E T1, T5, T3h, T3j;
cannam@95 1292 T1 = W[2];
cannam@95 1293 T5 = W[3];
cannam@95 1294 T9 = FNMS(T5, T8, T1 * T4);
cannam@95 1295 T4i = FMA(T1, T8, T5 * T4);
cannam@95 1296 T3h = W[4];
cannam@95 1297 T3j = W[5];
cannam@95 1298 T3l = FMA(T3h, T3i, T3j * T3k);
cannam@95 1299 T4A = FNMS(T3j, T3i, T3h * T3k);
cannam@95 1300 }
cannam@95 1301 }
cannam@95 1302 {
cannam@95 1303 E Tw, T3A, TA, T3y;
cannam@95 1304 {
cannam@95 1305 E Tu, Tv, Ty, Tz;
cannam@95 1306 Tu = Ip[WS(rs, 13)];
cannam@95 1307 Tv = Im[WS(rs, 13)];
cannam@95 1308 Tw = Tu - Tv;
cannam@95 1309 T3A = Tu + Tv;
cannam@95 1310 Ty = Rp[WS(rs, 13)];
cannam@95 1311 Tz = Rm[WS(rs, 13)];
cannam@95 1312 TA = Ty + Tz;
cannam@95 1313 T3y = Ty - Tz;
cannam@95 1314 }
cannam@95 1315 {
cannam@95 1316 E Tt, Tx, T3x, T3z;
cannam@95 1317 Tt = W[50];
cannam@95 1318 Tx = W[51];
cannam@95 1319 TB = FNMS(Tx, TA, Tt * Tw);
cannam@95 1320 T4m = FMA(Tt, TA, Tx * Tw);
cannam@95 1321 T3x = W[52];
cannam@95 1322 T3z = W[53];
cannam@95 1323 T3B = FMA(T3x, T3y, T3z * T3A);
cannam@95 1324 T4E = FNMS(T3z, T3y, T3x * T3A);
cannam@95 1325 }
cannam@95 1326 }
cannam@95 1327 {
cannam@95 1328 E Td, T3p, Th, T3n;
cannam@95 1329 {
cannam@95 1330 E Tb, Tc, Tf, Tg;
cannam@95 1331 Tb = Ip[WS(rs, 9)];
cannam@95 1332 Tc = Im[WS(rs, 9)];
cannam@95 1333 Td = Tb - Tc;
cannam@95 1334 T3p = Tb + Tc;
cannam@95 1335 Tf = Rp[WS(rs, 9)];
cannam@95 1336 Tg = Rm[WS(rs, 9)];
cannam@95 1337 Th = Tf + Tg;
cannam@95 1338 T3n = Tf - Tg;
cannam@95 1339 }
cannam@95 1340 {
cannam@95 1341 E Ta, Te, T3m, T3o;
cannam@95 1342 Ta = W[34];
cannam@95 1343 Te = W[35];
cannam@95 1344 Ti = FNMS(Te, Th, Ta * Td);
cannam@95 1345 T4j = FMA(Ta, Th, Te * Td);
cannam@95 1346 T3m = W[36];
cannam@95 1347 T3o = W[37];
cannam@95 1348 T3q = FMA(T3m, T3n, T3o * T3p);
cannam@95 1349 T4B = FNMS(T3o, T3n, T3m * T3p);
cannam@95 1350 }
cannam@95 1351 }
cannam@95 1352 {
cannam@95 1353 E Tn, T3v, Tr, T3t;
cannam@95 1354 {
cannam@95 1355 E Tl, Tm, Tp, Tq;
cannam@95 1356 Tl = Ip[WS(rs, 5)];
cannam@95 1357 Tm = Im[WS(rs, 5)];
cannam@95 1358 Tn = Tl - Tm;
cannam@95 1359 T3v = Tl + Tm;
cannam@95 1360 Tp = Rp[WS(rs, 5)];
cannam@95 1361 Tq = Rm[WS(rs, 5)];
cannam@95 1362 Tr = Tp + Tq;
cannam@95 1363 T3t = Tp - Tq;
cannam@95 1364 }
cannam@95 1365 {
cannam@95 1366 E Tk, To, T3s, T3u;
cannam@95 1367 Tk = W[18];
cannam@95 1368 To = W[19];
cannam@95 1369 Ts = FNMS(To, Tr, Tk * Tn);
cannam@95 1370 T4l = FMA(Tk, Tr, To * Tn);
cannam@95 1371 T3s = W[20];
cannam@95 1372 T3u = W[21];
cannam@95 1373 T3w = FMA(T3s, T3t, T3u * T3v);
cannam@95 1374 T4D = FNMS(T3u, T3t, T3s * T3v);
cannam@95 1375 }
cannam@95 1376 }
cannam@95 1377 Tj = T9 + Ti;
cannam@95 1378 TC = Ts + TB;
cannam@95 1379 T5r = Tj - TC;
cannam@95 1380 T4k = T4i + T4j;
cannam@95 1381 T4n = T4l + T4m;
cannam@95 1382 T5s = T4k - T4n;
cannam@95 1383 {
cannam@95 1384 E T3r, T3C, T6T, T6U;
cannam@95 1385 T3r = T3l + T3q;
cannam@95 1386 T3C = T3w + T3B;
cannam@95 1387 T3D = T3r + T3C;
cannam@95 1388 T5C = T3C - T3r;
cannam@95 1389 T6T = T4E - T4D;
cannam@95 1390 T6U = T3w - T3B;
cannam@95 1391 T6V = T6T + T6U;
cannam@95 1392 T72 = T6T - T6U;
cannam@95 1393 }
cannam@95 1394 {
cannam@95 1395 E T4C, T4F, T6s, T6t;
cannam@95 1396 T4C = T4A + T4B;
cannam@95 1397 T4F = T4D + T4E;
cannam@95 1398 T4G = T4C + T4F;
cannam@95 1399 T5F = T4F - T4C;
cannam@95 1400 T6s = T4i - T4j;
cannam@95 1401 T6t = Ts - TB;
cannam@95 1402 T6u = T6s + T6t;
cannam@95 1403 T86 = T6s - T6t;
cannam@95 1404 }
cannam@95 1405 {
cannam@95 1406 E T6Q, T6R, T6p, T6q;
cannam@95 1407 T6Q = T3q - T3l;
cannam@95 1408 T6R = T4A - T4B;
cannam@95 1409 T6S = T6Q - T6R;
cannam@95 1410 T71 = T6R + T6Q;
cannam@95 1411 T6p = T9 - Ti;
cannam@95 1412 T6q = T4l - T4m;
cannam@95 1413 T6r = T6p - T6q;
cannam@95 1414 T85 = T6p + T6q;
cannam@95 1415 }
cannam@95 1416 }
cannam@95 1417 {
cannam@95 1418 E TM, T4p, T3I, T4H, T1e, T4t, T3Y, T4L, TV, T4q, T3N, T4I, T15, T4s, T3T;
cannam@95 1419 E T4K;
cannam@95 1420 {
cannam@95 1421 E TH, T3H, TL, T3F;
cannam@95 1422 {
cannam@95 1423 E TF, TG, TJ, TK;
cannam@95 1424 TF = Ip[WS(rs, 15)];
cannam@95 1425 TG = Im[WS(rs, 15)];
cannam@95 1426 TH = TF - TG;
cannam@95 1427 T3H = TF + TG;
cannam@95 1428 TJ = Rp[WS(rs, 15)];
cannam@95 1429 TK = Rm[WS(rs, 15)];
cannam@95 1430 TL = TJ + TK;
cannam@95 1431 T3F = TJ - TK;
cannam@95 1432 }
cannam@95 1433 {
cannam@95 1434 E TE, TI, T3E, T3G;
cannam@95 1435 TE = W[58];
cannam@95 1436 TI = W[59];
cannam@95 1437 TM = FNMS(TI, TL, TE * TH);
cannam@95 1438 T4p = FMA(TE, TL, TI * TH);
cannam@95 1439 T3E = W[60];
cannam@95 1440 T3G = W[61];
cannam@95 1441 T3I = FMA(T3E, T3F, T3G * T3H);
cannam@95 1442 T4H = FNMS(T3G, T3F, T3E * T3H);
cannam@95 1443 }
cannam@95 1444 }
cannam@95 1445 {
cannam@95 1446 E T19, T3X, T1d, T3V;
cannam@95 1447 {
cannam@95 1448 E T17, T18, T1b, T1c;
cannam@95 1449 T17 = Ip[WS(rs, 11)];
cannam@95 1450 T18 = Im[WS(rs, 11)];
cannam@95 1451 T19 = T17 - T18;
cannam@95 1452 T3X = T17 + T18;
cannam@95 1453 T1b = Rp[WS(rs, 11)];
cannam@95 1454 T1c = Rm[WS(rs, 11)];
cannam@95 1455 T1d = T1b + T1c;
cannam@95 1456 T3V = T1b - T1c;
cannam@95 1457 }
cannam@95 1458 {
cannam@95 1459 E T16, T1a, T3U, T3W;
cannam@95 1460 T16 = W[42];
cannam@95 1461 T1a = W[43];
cannam@95 1462 T1e = FNMS(T1a, T1d, T16 * T19);
cannam@95 1463 T4t = FMA(T16, T1d, T1a * T19);
cannam@95 1464 T3U = W[44];
cannam@95 1465 T3W = W[45];
cannam@95 1466 T3Y = FMA(T3U, T3V, T3W * T3X);
cannam@95 1467 T4L = FNMS(T3W, T3V, T3U * T3X);
cannam@95 1468 }
cannam@95 1469 }
cannam@95 1470 {
cannam@95 1471 E TQ, T3M, TU, T3K;
cannam@95 1472 {
cannam@95 1473 E TO, TP, TS, TT;
cannam@95 1474 TO = Ip[WS(rs, 7)];
cannam@95 1475 TP = Im[WS(rs, 7)];
cannam@95 1476 TQ = TO - TP;
cannam@95 1477 T3M = TO + TP;
cannam@95 1478 TS = Rp[WS(rs, 7)];
cannam@95 1479 TT = Rm[WS(rs, 7)];
cannam@95 1480 TU = TS + TT;
cannam@95 1481 T3K = TS - TT;
cannam@95 1482 }
cannam@95 1483 {
cannam@95 1484 E TN, TR, T3J, T3L;
cannam@95 1485 TN = W[26];
cannam@95 1486 TR = W[27];
cannam@95 1487 TV = FNMS(TR, TU, TN * TQ);
cannam@95 1488 T4q = FMA(TN, TU, TR * TQ);
cannam@95 1489 T3J = W[28];
cannam@95 1490 T3L = W[29];
cannam@95 1491 T3N = FMA(T3J, T3K, T3L * T3M);
cannam@95 1492 T4I = FNMS(T3L, T3K, T3J * T3M);
cannam@95 1493 }
cannam@95 1494 }
cannam@95 1495 {
cannam@95 1496 E T10, T3S, T14, T3Q;
cannam@95 1497 {
cannam@95 1498 E TY, TZ, T12, T13;
cannam@95 1499 TY = Ip[WS(rs, 3)];
cannam@95 1500 TZ = Im[WS(rs, 3)];
cannam@95 1501 T10 = TY - TZ;
cannam@95 1502 T3S = TY + TZ;
cannam@95 1503 T12 = Rp[WS(rs, 3)];
cannam@95 1504 T13 = Rm[WS(rs, 3)];
cannam@95 1505 T14 = T12 + T13;
cannam@95 1506 T3Q = T12 - T13;
cannam@95 1507 }
cannam@95 1508 {
cannam@95 1509 E TX, T11, T3P, T3R;
cannam@95 1510 TX = W[10];
cannam@95 1511 T11 = W[11];
cannam@95 1512 T15 = FNMS(T11, T14, TX * T10);
cannam@95 1513 T4s = FMA(TX, T14, T11 * T10);
cannam@95 1514 T3P = W[12];
cannam@95 1515 T3R = W[13];
cannam@95 1516 T3T = FMA(T3P, T3Q, T3R * T3S);
cannam@95 1517 T4K = FNMS(T3R, T3Q, T3P * T3S);
cannam@95 1518 }
cannam@95 1519 }
cannam@95 1520 TW = TM + TV;
cannam@95 1521 T1f = T15 + T1e;
cannam@95 1522 T5v = TW - T1f;
cannam@95 1523 T4r = T4p + T4q;
cannam@95 1524 T4u = T4s + T4t;
cannam@95 1525 T5u = T4r - T4u;
cannam@95 1526 {
cannam@95 1527 E T3O, T3Z, T74, T75;
cannam@95 1528 T3O = T3I + T3N;
cannam@95 1529 T3Z = T3T + T3Y;
cannam@95 1530 T40 = T3O + T3Z;
cannam@95 1531 T5G = T3Z - T3O;
cannam@95 1532 T74 = T4H - T4I;
cannam@95 1533 T75 = T3Y - T3T;
cannam@95 1534 T76 = T74 + T75;
cannam@95 1535 T8k = T74 - T75;
cannam@95 1536 }
cannam@95 1537 {
cannam@95 1538 E T4J, T4M, T6z, T6A;
cannam@95 1539 T4J = T4H + T4I;
cannam@95 1540 T4M = T4K + T4L;
cannam@95 1541 T4N = T4J + T4M;
cannam@95 1542 T5D = T4J - T4M;
cannam@95 1543 T6z = T4p - T4q;
cannam@95 1544 T6A = T15 - T1e;
cannam@95 1545 T6B = T6z + T6A;
cannam@95 1546 T89 = T6z - T6A;
cannam@95 1547 }
cannam@95 1548 {
cannam@95 1549 E T6X, T6Y, T6w, T6x;
cannam@95 1550 T6X = T3N - T3I;
cannam@95 1551 T6Y = T4K - T4L;
cannam@95 1552 T6Z = T6X - T6Y;
cannam@95 1553 T8h = T6X + T6Y;
cannam@95 1554 T6w = TM - TV;
cannam@95 1555 T6x = T4s - T4t;
cannam@95 1556 T6y = T6w - T6x;
cannam@95 1557 T88 = T6w + T6x;
cannam@95 1558 }
cannam@95 1559 }
cannam@95 1560 {
cannam@95 1561 E T1h, T5i, T5c, T5m, T5f, T5n, T2s, T58, T42, T4y, T4w, T57, T54, T56, T4h;
cannam@95 1562 E T5h;
cannam@95 1563 {
cannam@95 1564 E TD, T1g, T5a, T5b;
cannam@95 1565 TD = Tj + TC;
cannam@95 1566 T1g = TW + T1f;
cannam@95 1567 T1h = TD + T1g;
cannam@95 1568 T5i = TD - T1g;
cannam@95 1569 T5a = T4N - T4G;
cannam@95 1570 T5b = T3D - T40;
cannam@95 1571 T5c = T5a + T5b;
cannam@95 1572 T5m = T5a - T5b;
cannam@95 1573 }
cannam@95 1574 {
cannam@95 1575 E T5d, T5e, T1U, T2r;
cannam@95 1576 T5d = T3f + T2S;
cannam@95 1577 T5e = T52 - T4V;
cannam@95 1578 T5f = T5d - T5e;
cannam@95 1579 T5n = T5d + T5e;
cannam@95 1580 T1U = T1A + T1T;
cannam@95 1581 T2r = T2d + T2q;
cannam@95 1582 T2s = T1U + T2r;
cannam@95 1583 T58 = T2r - T1U;
cannam@95 1584 }
cannam@95 1585 {
cannam@95 1586 E T3g, T41, T4o, T4v;
cannam@95 1587 T3g = T2S - T3f;
cannam@95 1588 T41 = T3D + T40;
cannam@95 1589 T42 = T3g - T41;
cannam@95 1590 T4y = T41 + T3g;
cannam@95 1591 T4o = T4k + T4n;
cannam@95 1592 T4v = T4r + T4u;
cannam@95 1593 T4w = T4o + T4v;
cannam@95 1594 T57 = T4v - T4o;
cannam@95 1595 }
cannam@95 1596 {
cannam@95 1597 E T4O, T53, T49, T4g;
cannam@95 1598 T4O = T4G + T4N;
cannam@95 1599 T53 = T4V + T52;
cannam@95 1600 T54 = T4O - T53;
cannam@95 1601 T56 = T4O + T53;
cannam@95 1602 T49 = T45 + T48;
cannam@95 1603 T4g = T4c + T4f;
cannam@95 1604 T4h = T49 + T4g;
cannam@95 1605 T5h = T49 - T4g;
cannam@95 1606 }
cannam@95 1607 {
cannam@95 1608 E T2t, T55, T4x, T4z;
cannam@95 1609 T2t = T1h + T2s;
cannam@95 1610 Ip[0] = KP500000000 * (T2t + T42);
cannam@95 1611 Im[WS(rs, 15)] = KP500000000 * (T42 - T2t);
cannam@95 1612 T55 = T4h + T4w;
cannam@95 1613 Rm[WS(rs, 15)] = KP500000000 * (T55 - T56);
cannam@95 1614 Rp[0] = KP500000000 * (T55 + T56);
cannam@95 1615 T4x = T4h - T4w;
cannam@95 1616 Rm[WS(rs, 7)] = KP500000000 * (T4x - T4y);
cannam@95 1617 Rp[WS(rs, 8)] = KP500000000 * (T4x + T4y);
cannam@95 1618 T4z = T2s - T1h;
cannam@95 1619 Ip[WS(rs, 8)] = KP500000000 * (T4z + T54);
cannam@95 1620 Im[WS(rs, 7)] = KP500000000 * (T54 - T4z);
cannam@95 1621 }
cannam@95 1622 {
cannam@95 1623 E T59, T5g, T5p, T5q;
cannam@95 1624 T59 = KP500000000 * (T57 + T58);
cannam@95 1625 T5g = KP353553390 * (T5c + T5f);
cannam@95 1626 Ip[WS(rs, 4)] = T59 + T5g;
cannam@95 1627 Im[WS(rs, 11)] = T5g - T59;
cannam@95 1628 T5p = KP500000000 * (T5h + T5i);
cannam@95 1629 T5q = KP353553390 * (T5m + T5n);
cannam@95 1630 Rm[WS(rs, 11)] = T5p - T5q;
cannam@95 1631 Rp[WS(rs, 4)] = T5p + T5q;
cannam@95 1632 }
cannam@95 1633 {
cannam@95 1634 E T5j, T5k, T5l, T5o;
cannam@95 1635 T5j = KP500000000 * (T5h - T5i);
cannam@95 1636 T5k = KP353553390 * (T5f - T5c);
cannam@95 1637 Rm[WS(rs, 3)] = T5j - T5k;
cannam@95 1638 Rp[WS(rs, 12)] = T5j + T5k;
cannam@95 1639 T5l = KP500000000 * (T58 - T57);
cannam@95 1640 T5o = KP353553390 * (T5m - T5n);
cannam@95 1641 Ip[WS(rs, 12)] = T5l + T5o;
cannam@95 1642 Im[WS(rs, 3)] = T5o - T5l;
cannam@95 1643 }
cannam@95 1644 }
cannam@95 1645 {
cannam@95 1646 E T5x, T6g, T6a, T6k, T6d, T6l, T5A, T66, T5I, T60, T5T, T6f, T5W, T65, T5P;
cannam@95 1647 E T61;
cannam@95 1648 {
cannam@95 1649 E T5t, T5w, T68, T69;
cannam@95 1650 T5t = T5r - T5s;
cannam@95 1651 T5w = T5u + T5v;
cannam@95 1652 T5x = KP353553390 * (T5t + T5w);
cannam@95 1653 T6g = KP353553390 * (T5t - T5w);
cannam@95 1654 T68 = T5D - T5C;
cannam@95 1655 T69 = T5G - T5F;
cannam@95 1656 T6a = FMA(KP461939766, T68, KP191341716 * T69);
cannam@95 1657 T6k = FNMS(KP461939766, T69, KP191341716 * T68);
cannam@95 1658 }
cannam@95 1659 {
cannam@95 1660 E T6b, T6c, T5y, T5z;
cannam@95 1661 T6b = T5K - T5J;
cannam@95 1662 T6c = T5N - T5M;
cannam@95 1663 T6d = FNMS(KP461939766, T6c, KP191341716 * T6b);
cannam@95 1664 T6l = FMA(KP461939766, T6b, KP191341716 * T6c);
cannam@95 1665 T5y = T4f - T4c;
cannam@95 1666 T5z = T2q - T2d;
cannam@95 1667 T5A = KP500000000 * (T5y + T5z);
cannam@95 1668 T66 = KP500000000 * (T5z - T5y);
cannam@95 1669 }
cannam@95 1670 {
cannam@95 1671 E T5E, T5H, T5R, T5S;
cannam@95 1672 T5E = T5C + T5D;
cannam@95 1673 T5H = T5F + T5G;
cannam@95 1674 T5I = FMA(KP191341716, T5E, KP461939766 * T5H);
cannam@95 1675 T60 = FNMS(KP191341716, T5H, KP461939766 * T5E);
cannam@95 1676 T5R = T45 - T48;
cannam@95 1677 T5S = T1A - T1T;
cannam@95 1678 T5T = KP500000000 * (T5R + T5S);
cannam@95 1679 T6f = KP500000000 * (T5R - T5S);
cannam@95 1680 }
cannam@95 1681 {
cannam@95 1682 E T5U, T5V, T5L, T5O;
cannam@95 1683 T5U = T5s + T5r;
cannam@95 1684 T5V = T5u - T5v;
cannam@95 1685 T5W = KP353553390 * (T5U + T5V);
cannam@95 1686 T65 = KP353553390 * (T5V - T5U);
cannam@95 1687 T5L = T5J + T5K;
cannam@95 1688 T5O = T5M + T5N;
cannam@95 1689 T5P = FNMS(KP191341716, T5O, KP461939766 * T5L);
cannam@95 1690 T61 = FMA(KP191341716, T5L, KP461939766 * T5O);
cannam@95 1691 }
cannam@95 1692 {
cannam@95 1693 E T5B, T5Q, T63, T64;
cannam@95 1694 T5B = T5x + T5A;
cannam@95 1695 T5Q = T5I + T5P;
cannam@95 1696 Ip[WS(rs, 2)] = T5B + T5Q;
cannam@95 1697 Im[WS(rs, 13)] = T5Q - T5B;
cannam@95 1698 T63 = T5T + T5W;
cannam@95 1699 T64 = T60 + T61;
cannam@95 1700 Rm[WS(rs, 13)] = T63 - T64;
cannam@95 1701 Rp[WS(rs, 2)] = T63 + T64;
cannam@95 1702 }
cannam@95 1703 {
cannam@95 1704 E T5X, T5Y, T5Z, T62;
cannam@95 1705 T5X = T5T - T5W;
cannam@95 1706 T5Y = T5P - T5I;
cannam@95 1707 Rm[WS(rs, 5)] = T5X - T5Y;
cannam@95 1708 Rp[WS(rs, 10)] = T5X + T5Y;
cannam@95 1709 T5Z = T5A - T5x;
cannam@95 1710 T62 = T60 - T61;
cannam@95 1711 Ip[WS(rs, 10)] = T5Z + T62;
cannam@95 1712 Im[WS(rs, 5)] = T62 - T5Z;
cannam@95 1713 }
cannam@95 1714 {
cannam@95 1715 E T67, T6e, T6n, T6o;
cannam@95 1716 T67 = T65 + T66;
cannam@95 1717 T6e = T6a + T6d;
cannam@95 1718 Ip[WS(rs, 6)] = T67 + T6e;
cannam@95 1719 Im[WS(rs, 9)] = T6e - T67;
cannam@95 1720 T6n = T6f + T6g;
cannam@95 1721 T6o = T6k + T6l;
cannam@95 1722 Rm[WS(rs, 9)] = T6n - T6o;
cannam@95 1723 Rp[WS(rs, 6)] = T6n + T6o;
cannam@95 1724 }
cannam@95 1725 {
cannam@95 1726 E T6h, T6i, T6j, T6m;
cannam@95 1727 T6h = T6f - T6g;
cannam@95 1728 T6i = T6d - T6a;
cannam@95 1729 Rm[WS(rs, 1)] = T6h - T6i;
cannam@95 1730 Rp[WS(rs, 14)] = T6h + T6i;
cannam@95 1731 T6j = T66 - T65;
cannam@95 1732 T6m = T6k - T6l;
cannam@95 1733 Ip[WS(rs, 14)] = T6j + T6m;
cannam@95 1734 Im[WS(rs, 1)] = T6m - T6j;
cannam@95 1735 }
cannam@95 1736 }
cannam@95 1737 {
cannam@95 1738 E T6D, T7W, T6O, T7M, T7C, T7L, T7z, T7V, T7r, T81, T7H, T7T, T78, T80, T7G;
cannam@95 1739 E T7Q;
cannam@95 1740 {
cannam@95 1741 E T6v, T6C, T7v, T7y;
cannam@95 1742 T6v = FNMS(KP191341716, T6u, KP461939766 * T6r);
cannam@95 1743 T6C = FMA(KP461939766, T6y, KP191341716 * T6B);
cannam@95 1744 T6D = T6v + T6C;
cannam@95 1745 T7W = T6v - T6C;
cannam@95 1746 {
cannam@95 1747 E T6K, T6N, T7A, T7B;
cannam@95 1748 T6K = KP353553390 * (T6G + T6J);
cannam@95 1749 T6N = KP500000000 * (T6L - T6M);
cannam@95 1750 T6O = T6K + T6N;
cannam@95 1751 T7M = T6N - T6K;
cannam@95 1752 T7A = FMA(KP191341716, T6r, KP461939766 * T6u);
cannam@95 1753 T7B = FNMS(KP191341716, T6y, KP461939766 * T6B);
cannam@95 1754 T7C = T7A + T7B;
cannam@95 1755 T7L = T7B - T7A;
cannam@95 1756 }
cannam@95 1757 T7v = KP500000000 * (T7t + T7u);
cannam@95 1758 T7y = KP353553390 * (T7w + T7x);
cannam@95 1759 T7z = T7v + T7y;
cannam@95 1760 T7V = T7v - T7y;
cannam@95 1761 {
cannam@95 1762 E T7j, T7R, T7q, T7S, T7f, T7m;
cannam@95 1763 T7f = KP707106781 * (T7b + T7e);
cannam@95 1764 T7j = T7f + T7i;
cannam@95 1765 T7R = T7i - T7f;
cannam@95 1766 T7m = KP707106781 * (T7k + T7l);
cannam@95 1767 T7q = T7m + T7p;
cannam@95 1768 T7S = T7p - T7m;
cannam@95 1769 T7r = FNMS(KP097545161, T7q, KP490392640 * T7j);
cannam@95 1770 T81 = FMA(KP415734806, T7R, KP277785116 * T7S);
cannam@95 1771 T7H = FMA(KP097545161, T7j, KP490392640 * T7q);
cannam@95 1772 T7T = FNMS(KP415734806, T7S, KP277785116 * T7R);
cannam@95 1773 }
cannam@95 1774 {
cannam@95 1775 E T70, T7O, T77, T7P, T6W, T73;
cannam@95 1776 T6W = KP707106781 * (T6S + T6V);
cannam@95 1777 T70 = T6W + T6Z;
cannam@95 1778 T7O = T6Z - T6W;
cannam@95 1779 T73 = KP707106781 * (T71 + T72);
cannam@95 1780 T77 = T73 + T76;
cannam@95 1781 T7P = T76 - T73;
cannam@95 1782 T78 = FMA(KP490392640, T70, KP097545161 * T77);
cannam@95 1783 T80 = FNMS(KP415734806, T7O, KP277785116 * T7P);
cannam@95 1784 T7G = FNMS(KP097545161, T70, KP490392640 * T77);
cannam@95 1785 T7Q = FMA(KP277785116, T7O, KP415734806 * T7P);
cannam@95 1786 }
cannam@95 1787 }
cannam@95 1788 {
cannam@95 1789 E T6P, T7s, T7J, T7K;
cannam@95 1790 T6P = T6D + T6O;
cannam@95 1791 T7s = T78 + T7r;
cannam@95 1792 Ip[WS(rs, 1)] = T6P + T7s;
cannam@95 1793 Im[WS(rs, 14)] = T7s - T6P;
cannam@95 1794 T7J = T7z + T7C;
cannam@95 1795 T7K = T7G + T7H;
cannam@95 1796 Rm[WS(rs, 14)] = T7J - T7K;
cannam@95 1797 Rp[WS(rs, 1)] = T7J + T7K;
cannam@95 1798 }
cannam@95 1799 {
cannam@95 1800 E T7D, T7E, T7F, T7I;
cannam@95 1801 T7D = T7z - T7C;
cannam@95 1802 T7E = T7r - T78;
cannam@95 1803 Rm[WS(rs, 6)] = T7D - T7E;
cannam@95 1804 Rp[WS(rs, 9)] = T7D + T7E;
cannam@95 1805 T7F = T6O - T6D;
cannam@95 1806 T7I = T7G - T7H;
cannam@95 1807 Ip[WS(rs, 9)] = T7F + T7I;
cannam@95 1808 Im[WS(rs, 6)] = T7I - T7F;
cannam@95 1809 }
cannam@95 1810 {
cannam@95 1811 E T7N, T7U, T83, T84;
cannam@95 1812 T7N = T7L + T7M;
cannam@95 1813 T7U = T7Q + T7T;
cannam@95 1814 Ip[WS(rs, 5)] = T7N + T7U;
cannam@95 1815 Im[WS(rs, 10)] = T7U - T7N;
cannam@95 1816 T83 = T7V + T7W;
cannam@95 1817 T84 = T80 + T81;
cannam@95 1818 Rm[WS(rs, 10)] = T83 - T84;
cannam@95 1819 Rp[WS(rs, 5)] = T83 + T84;
cannam@95 1820 }
cannam@95 1821 {
cannam@95 1822 E T7X, T7Y, T7Z, T82;
cannam@95 1823 T7X = T7V - T7W;
cannam@95 1824 T7Y = T7T - T7Q;
cannam@95 1825 Rm[WS(rs, 2)] = T7X - T7Y;
cannam@95 1826 Rp[WS(rs, 13)] = T7X + T7Y;
cannam@95 1827 T7Z = T7M - T7L;
cannam@95 1828 T82 = T80 - T81;
cannam@95 1829 Ip[WS(rs, 13)] = T7Z + T82;
cannam@95 1830 Im[WS(rs, 2)] = T82 - T7Z;
cannam@95 1831 }
cannam@95 1832 }
cannam@95 1833 {
cannam@95 1834 E T8b, T8U, T8e, T8K, T8A, T8J, T8x, T8T, T8t, T8Z, T8F, T8R, T8m, T8Y, T8E;
cannam@95 1835 E T8O;
cannam@95 1836 {
cannam@95 1837 E T87, T8a, T8v, T8w;
cannam@95 1838 T87 = FNMS(KP461939766, T86, KP191341716 * T85);
cannam@95 1839 T8a = FMA(KP191341716, T88, KP461939766 * T89);
cannam@95 1840 T8b = T87 + T8a;
cannam@95 1841 T8U = T87 - T8a;
cannam@95 1842 {
cannam@95 1843 E T8c, T8d, T8y, T8z;
cannam@95 1844 T8c = KP353553390 * (T7x - T7w);
cannam@95 1845 T8d = KP500000000 * (T6M + T6L);
cannam@95 1846 T8e = T8c + T8d;
cannam@95 1847 T8K = T8d - T8c;
cannam@95 1848 T8y = FMA(KP461939766, T85, KP191341716 * T86);
cannam@95 1849 T8z = FNMS(KP461939766, T88, KP191341716 * T89);
cannam@95 1850 T8A = T8y + T8z;
cannam@95 1851 T8J = T8z - T8y;
cannam@95 1852 }
cannam@95 1853 T8v = KP500000000 * (T7t - T7u);
cannam@95 1854 T8w = KP353553390 * (T6G - T6J);
cannam@95 1855 T8x = T8v + T8w;
cannam@95 1856 T8T = T8v - T8w;
cannam@95 1857 {
cannam@95 1858 E T8p, T8P, T8s, T8Q, T8n, T8q;
cannam@95 1859 T8n = KP707106781 * (T7l - T7k);
cannam@95 1860 T8p = T8n + T8o;
cannam@95 1861 T8P = T8o - T8n;
cannam@95 1862 T8q = KP707106781 * (T7b - T7e);
cannam@95 1863 T8s = T8q + T8r;
cannam@95 1864 T8Q = T8r - T8q;
cannam@95 1865 T8t = FNMS(KP277785116, T8s, KP415734806 * T8p);
cannam@95 1866 T8Z = FMA(KP490392640, T8P, KP097545161 * T8Q);
cannam@95 1867 T8F = FMA(KP277785116, T8p, KP415734806 * T8s);
cannam@95 1868 T8R = FNMS(KP490392640, T8Q, KP097545161 * T8P);
cannam@95 1869 }
cannam@95 1870 {
cannam@95 1871 E T8i, T8M, T8l, T8N, T8g, T8j;
cannam@95 1872 T8g = KP707106781 * (T72 - T71);
cannam@95 1873 T8i = T8g + T8h;
cannam@95 1874 T8M = T8h - T8g;
cannam@95 1875 T8j = KP707106781 * (T6S - T6V);
cannam@95 1876 T8l = T8j + T8k;
cannam@95 1877 T8N = T8k - T8j;
cannam@95 1878 T8m = FMA(KP415734806, T8i, KP277785116 * T8l);
cannam@95 1879 T8Y = FNMS(KP490392640, T8M, KP097545161 * T8N);
cannam@95 1880 T8E = FNMS(KP277785116, T8i, KP415734806 * T8l);
cannam@95 1881 T8O = FMA(KP097545161, T8M, KP490392640 * T8N);
cannam@95 1882 }
cannam@95 1883 }
cannam@95 1884 {
cannam@95 1885 E T8f, T8u, T8H, T8I;
cannam@95 1886 T8f = T8b + T8e;
cannam@95 1887 T8u = T8m + T8t;
cannam@95 1888 Ip[WS(rs, 3)] = T8f + T8u;
cannam@95 1889 Im[WS(rs, 12)] = T8u - T8f;
cannam@95 1890 T8H = T8x + T8A;
cannam@95 1891 T8I = T8E + T8F;
cannam@95 1892 Rm[WS(rs, 12)] = T8H - T8I;
cannam@95 1893 Rp[WS(rs, 3)] = T8H + T8I;
cannam@95 1894 }
cannam@95 1895 {
cannam@95 1896 E T8B, T8C, T8D, T8G;
cannam@95 1897 T8B = T8x - T8A;
cannam@95 1898 T8C = T8t - T8m;
cannam@95 1899 Rm[WS(rs, 4)] = T8B - T8C;
cannam@95 1900 Rp[WS(rs, 11)] = T8B + T8C;
cannam@95 1901 T8D = T8e - T8b;
cannam@95 1902 T8G = T8E - T8F;
cannam@95 1903 Ip[WS(rs, 11)] = T8D + T8G;
cannam@95 1904 Im[WS(rs, 4)] = T8G - T8D;
cannam@95 1905 }
cannam@95 1906 {
cannam@95 1907 E T8L, T8S, T91, T92;
cannam@95 1908 T8L = T8J + T8K;
cannam@95 1909 T8S = T8O + T8R;
cannam@95 1910 Ip[WS(rs, 7)] = T8L + T8S;
cannam@95 1911 Im[WS(rs, 8)] = T8S - T8L;
cannam@95 1912 T91 = T8T + T8U;
cannam@95 1913 T92 = T8Y + T8Z;
cannam@95 1914 Rm[WS(rs, 8)] = T91 - T92;
cannam@95 1915 Rp[WS(rs, 7)] = T91 + T92;
cannam@95 1916 }
cannam@95 1917 {
cannam@95 1918 E T8V, T8W, T8X, T90;
cannam@95 1919 T8V = T8T - T8U;
cannam@95 1920 T8W = T8R - T8O;
cannam@95 1921 Rm[0] = T8V - T8W;
cannam@95 1922 Rp[WS(rs, 15)] = T8V + T8W;
cannam@95 1923 T8X = T8K - T8J;
cannam@95 1924 T90 = T8Y - T8Z;
cannam@95 1925 Ip[WS(rs, 15)] = T8X + T90;
cannam@95 1926 Im[0] = T90 - T8X;
cannam@95 1927 }
cannam@95 1928 }
cannam@95 1929 }
cannam@95 1930 }
cannam@95 1931 }
cannam@95 1932
cannam@95 1933 static const tw_instr twinstr[] = {
cannam@95 1934 {TW_FULL, 1, 32},
cannam@95 1935 {TW_NEXT, 1, 0}
cannam@95 1936 };
cannam@95 1937
cannam@95 1938 static const hc2c_desc desc = { 32, "hc2cfdft_32", twinstr, &GENUS, {404, 134, 94, 0} };
cannam@95 1939
cannam@95 1940 void X(codelet_hc2cfdft_32) (planner *p) {
cannam@95 1941 X(khc2c_register) (p, hc2cfdft_32, &desc, HC2C_VIA_DFT);
cannam@95 1942 }
cannam@95 1943 #endif /* HAVE_FMA */