Mercurial > hg > sv-dependency-builds

annotate src/fftw-3.3.8/rdft/scalar/r2cb/hc2cbdft_16.c @ 82:d0c2a83c1364

Find changesets by keywords (author, files, the commit message), revision number or hash, or revset expression.
Add FFTW 3.3.8 source, and a Linux build
author Chris Cannam
date Tue, 19 Nov 2019 14:52:55 +0000
parents
children
rev   line source
Chris@82 1 /*
Chris@82 2 * Copyright (c) 2003, 2007-14 Matteo Frigo
Chris@82 3 * Copyright (c) 2003, 2007-14 Massachusetts Institute of Technology
Chris@82 4 *
Chris@82 5 * This program is free software; you can redistribute it and/or modify
Chris@82 6 * it under the terms of the GNU General Public License as published by
Chris@82 7 * the Free Software Foundation; either version 2 of the License, or
Chris@82 8 * (at your option) any later version.
Chris@82 9 *
Chris@82 10 * This program is distributed in the hope that it will be useful,
Chris@82 11 * but WITHOUT ANY WARRANTY; without even the implied warranty of
Chris@82 12 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
Chris@82 13 * GNU General Public License for more details.
Chris@82 14 *
Chris@82 15 * You should have received a copy of the GNU General Public License
Chris@82 16 * along with this program; if not, write to the Free Software
Chris@82 17 * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
Chris@82 18 *
Chris@82 19 */
Chris@82 20
Chris@82 21 /* This file was automatically generated --- DO NOT EDIT */
Chris@82 22 /* Generated on Thu May 24 08:07:58 EDT 2018 */
Chris@82 23
Chris@82 24 #include "rdft/codelet-rdft.h"
Chris@82 25
Chris@82 26 #if defined(ARCH_PREFERS_FMA) || defined(ISA_EXTENSION_PREFERS_FMA)
Chris@82 27
Chris@82 28 /* Generated by: ../../../genfft/gen_hc2cdft.native -fma -compact -variables 4 -pipeline-latency 4 -sign 1 -n 16 -dif -name hc2cbdft_16 -include rdft/scalar/hc2cb.h */
Chris@82 29
Chris@82 30 /*
Chris@82 31 * This function contains 206 FP additions, 100 FP multiplications,
Chris@82 32 * (or, 136 additions, 30 multiplications, 70 fused multiply/add),
Chris@82 33 * 66 stack variables, 3 constants, and 64 memory accesses
Chris@82 34 */
Chris@82 35 #include "rdft/scalar/hc2cb.h"
Chris@82 36
Chris@82 37 static void hc2cbdft_16(R *Rp, R *Ip, R *Rm, R *Im, const R *W, stride rs, INT mb, INT me, INT ms)
Chris@82 38 {
Chris@82 39 DK(KP923879532, +0.923879532511286756128183189396788286822416626);
Chris@82 40 DK(KP414213562, +0.414213562373095048801688724209698078569671875);
Chris@82 41 DK(KP707106781, +0.707106781186547524400844362104849039284835938);
Chris@82 42 {
Chris@82 43 INT m;
Chris@82 44 for (m = mb, W = W + ((mb - 1) * 30); m < me; m = m + 1, Rp = Rp + ms, Ip = Ip + ms, Rm = Rm - ms, Im = Im - ms, W = W + 30, MAKE_VOLATILE_STRIDE(64, rs)) {
Chris@82 45 E Tf, T20, T32, T3Q, T3f, T3V, TN, T2a, T1m, T2f, T2G, T3G, T2T, T3L, T1F;
Chris@82 46 E T26, T2J, T2M, T2N, T2U, T2V, T3H, Tu, T25, T3i, T3R, T1a, T2g, T1y, T21;
Chris@82 47 E T39, T3W, T1p, T2b;
Chris@82 48 {
Chris@82 49 E T3, T1e, TA, T1C, T6, Tx, T1h, T1D, Td, T1A, TL, T1k, Ta, T1z, TG;
Chris@82 50 E T1j;
Chris@82 51 {
Chris@82 52 E T1, T2, T1f, T1g;
Chris@82 53 T1 = Rp[0];
Chris@82 54 T2 = Rm[WS(rs, 7)];
Chris@82 55 T3 = T1 + T2;
Chris@82 56 T1e = T1 - T2;
Chris@82 57 {
Chris@82 58 E Ty, Tz, T4, T5;
Chris@82 59 Ty = Ip[0];
Chris@82 60 Tz = Im[WS(rs, 7)];
Chris@82 61 TA = Ty + Tz;
Chris@82 62 T1C = Ty - Tz;
Chris@82 63 T4 = Rp[WS(rs, 4)];
Chris@82 64 T5 = Rm[WS(rs, 3)];
Chris@82 65 T6 = T4 + T5;
Chris@82 66 Tx = T4 - T5;
Chris@82 67 }
Chris@82 68 T1f = Ip[WS(rs, 4)];
Chris@82 69 T1g = Im[WS(rs, 3)];
Chris@82 70 T1h = T1f + T1g;
Chris@82 71 T1D = T1f - T1g;
Chris@82 72 {
Chris@82 73 E Tb, Tc, TH, TI, TJ, TK;
Chris@82 74 Tb = Rm[WS(rs, 1)];
Chris@82 75 Tc = Rp[WS(rs, 6)];
Chris@82 76 TH = Tb - Tc;
Chris@82 77 TI = Im[WS(rs, 1)];
Chris@82 78 TJ = Ip[WS(rs, 6)];
Chris@82 79 TK = TI + TJ;
Chris@82 80 Td = Tb + Tc;
Chris@82 81 T1A = TJ - TI;
Chris@82 82 TL = TH + TK;
Chris@82 83 T1k = TH - TK;
Chris@82 84 }
Chris@82 85 {
Chris@82 86 E T8, T9, TC, TD, TE, TF;
Chris@82 87 T8 = Rp[WS(rs, 2)];
Chris@82 88 T9 = Rm[WS(rs, 5)];
Chris@82 89 TC = T8 - T9;
Chris@82 90 TD = Ip[WS(rs, 2)];
Chris@82 91 TE = Im[WS(rs, 5)];
Chris@82 92 TF = TD + TE;
Chris@82 93 Ta = T8 + T9;
Chris@82 94 T1z = TD - TE;
Chris@82 95 TG = TC + TF;
Chris@82 96 T1j = TC - TF;
Chris@82 97 }
Chris@82 98 }
Chris@82 99 {
Chris@82 100 E T7, Te, T30, T31;
Chris@82 101 T7 = T3 + T6;
Chris@82 102 Te = Ta + Td;
Chris@82 103 Tf = T7 + Te;
Chris@82 104 T20 = T7 - Te;
Chris@82 105 T30 = TA - Tx;
Chris@82 106 T31 = T1j - T1k;
Chris@82 107 T32 = FMA(KP707106781, T31, T30);
Chris@82 108 T3Q = FNMS(KP707106781, T31, T30);
Chris@82 109 }
Chris@82 110 {
Chris@82 111 E T3d, T3e, TB, TM;
Chris@82 112 T3d = T1e + T1h;
Chris@82 113 T3e = TG + TL;
Chris@82 114 T3f = FNMS(KP707106781, T3e, T3d);
Chris@82 115 T3V = FMA(KP707106781, T3e, T3d);
Chris@82 116 TB = Tx + TA;
Chris@82 117 TM = TG - TL;
Chris@82 118 TN = FMA(KP707106781, TM, TB);
Chris@82 119 T2a = FNMS(KP707106781, TM, TB);
Chris@82 120 }
Chris@82 121 {
Chris@82 122 E T1i, T1l, T2E, T2F;
Chris@82 123 T1i = T1e - T1h;
Chris@82 124 T1l = T1j + T1k;
Chris@82 125 T1m = FMA(KP707106781, T1l, T1i);
Chris@82 126 T2f = FNMS(KP707106781, T1l, T1i);
Chris@82 127 T2E = T3 - T6;
Chris@82 128 T2F = T1A - T1z;
Chris@82 129 T2G = T2E + T2F;
Chris@82 130 T3G = T2E - T2F;
Chris@82 131 }
Chris@82 132 {
Chris@82 133 E T2R, T2S, T1B, T1E;
Chris@82 134 T2R = Ta - Td;
Chris@82 135 T2S = T1C - T1D;
Chris@82 136 T2T = T2R + T2S;
Chris@82 137 T3L = T2S - T2R;
Chris@82 138 T1B = T1z + T1A;
Chris@82 139 T1E = T1C + T1D;
Chris@82 140 T1F = T1B + T1E;
Chris@82 141 T26 = T1E - T1B;
Chris@82 142 }
Chris@82 143 }
Chris@82 144 {
Chris@82 145 E Ti, T1s, Tl, T1t, TS, TX, T34, T33, T2I, T2H, Tp, T1v, Ts, T1w, T13;
Chris@82 146 E T18, T37, T36, T2L, T2K;
Chris@82 147 {
Chris@82 148 E TT, TR, TO, TW;
Chris@82 149 {
Chris@82 150 E Tg, Th, TP, TQ;
Chris@82 151 Tg = Rp[WS(rs, 1)];
Chris@82 152 Th = Rm[WS(rs, 6)];
Chris@82 153 Ti = Tg + Th;
Chris@82 154 TT = Tg - Th;
Chris@82 155 TP = Ip[WS(rs, 1)];
Chris@82 156 TQ = Im[WS(rs, 6)];
Chris@82 157 TR = TP + TQ;
Chris@82 158 T1s = TP - TQ;
Chris@82 159 }
Chris@82 160 {
Chris@82 161 E Tj, Tk, TU, TV;
Chris@82 162 Tj = Rp[WS(rs, 5)];
Chris@82 163 Tk = Rm[WS(rs, 2)];
Chris@82 164 Tl = Tj + Tk;
Chris@82 165 TO = Tj - Tk;
Chris@82 166 TU = Ip[WS(rs, 5)];
Chris@82 167 TV = Im[WS(rs, 2)];
Chris@82 168 TW = TU + TV;
Chris@82 169 T1t = TU - TV;
Chris@82 170 }
Chris@82 171 TS = TO + TR;
Chris@82 172 TX = TT - TW;
Chris@82 173 T34 = TR - TO;
Chris@82 174 T33 = TT + TW;
Chris@82 175 T2I = T1s - T1t;
Chris@82 176 T2H = Ti - Tl;
Chris@82 177 }
Chris@82 178 {
Chris@82 179 E T14, T12, TZ, T17;
Chris@82 180 {
Chris@82 181 E Tn, To, T10, T11;
Chris@82 182 Tn = Rm[0];
Chris@82 183 To = Rp[WS(rs, 7)];
Chris@82 184 Tp = Tn + To;
Chris@82 185 T14 = Tn - To;
Chris@82 186 T10 = Im[0];
Chris@82 187 T11 = Ip[WS(rs, 7)];
Chris@82 188 T12 = T10 + T11;
Chris@82 189 T1v = T11 - T10;
Chris@82 190 }
Chris@82 191 {
Chris@82 192 E Tq, Tr, T15, T16;
Chris@82 193 Tq = Rp[WS(rs, 3)];
Chris@82 194 Tr = Rm[WS(rs, 4)];
Chris@82 195 Ts = Tq + Tr;
Chris@82 196 TZ = Tq - Tr;
Chris@82 197 T15 = Ip[WS(rs, 3)];
Chris@82 198 T16 = Im[WS(rs, 4)];
Chris@82 199 T17 = T15 + T16;
Chris@82 200 T1w = T15 - T16;
Chris@82 201 }
Chris@82 202 T13 = TZ - T12;
Chris@82 203 T18 = T14 - T17;
Chris@82 204 T37 = TZ + T12;
Chris@82 205 T36 = T14 + T17;
Chris@82 206 T2L = T1v - T1w;
Chris@82 207 T2K = Tp - Ts;
Chris@82 208 }
Chris@82 209 T2J = T2H - T2I;
Chris@82 210 T2M = T2K + T2L;
Chris@82 211 T2N = T2J + T2M;
Chris@82 212 T2U = T2H + T2I;
Chris@82 213 T2V = T2L - T2K;
Chris@82 214 T3H = T2V - T2U;
Chris@82 215 {
Chris@82 216 E Tm, Tt, T3g, T3h;
Chris@82 217 Tm = Ti + Tl;
Chris@82 218 Tt = Tp + Ts;
Chris@82 219 Tu = Tm + Tt;
Chris@82 220 T25 = Tm - Tt;
Chris@82 221 T3g = FNMS(KP414213562, T33, T34);
Chris@82 222 T3h = FNMS(KP414213562, T36, T37);
Chris@82 223 T3i = T3g + T3h;
Chris@82 224 T3R = T3h - T3g;
Chris@82 225 }
Chris@82 226 {
Chris@82 227 E TY, T19, T1u, T1x;
Chris@82 228 TY = FMA(KP414213562, TX, TS);
Chris@82 229 T19 = FNMS(KP414213562, T18, T13);
Chris@82 230 T1a = TY + T19;
Chris@82 231 T2g = T19 - TY;
Chris@82 232 T1u = T1s + T1t;
Chris@82 233 T1x = T1v + T1w;
Chris@82 234 T1y = T1u + T1x;
Chris@82 235 T21 = T1x - T1u;
Chris@82 236 }
Chris@82 237 {
Chris@82 238 E T35, T38, T1n, T1o;
Chris@82 239 T35 = FMA(KP414213562, T34, T33);
Chris@82 240 T38 = FMA(KP414213562, T37, T36);
Chris@82 241 T39 = T35 - T38;
Chris@82 242 T3W = T35 + T38;
Chris@82 243 T1n = FNMS(KP414213562, TS, TX);
Chris@82 244 T1o = FMA(KP414213562, T13, T18);
Chris@82 245 T1p = T1n + T1o;
Chris@82 246 T2b = T1n - T1o;
Chris@82 247 }
Chris@82 248 }
Chris@82 249 {
Chris@82 250 E Tv, T1G, T1b, T1q, T1c, T1H, Tw, T1r, T1I, T1d;
Chris@82 251 Tv = Tf + Tu;
Chris@82 252 T1G = T1y + T1F;
Chris@82 253 T1b = FMA(KP923879532, T1a, TN);
Chris@82 254 T1q = FMA(KP923879532, T1p, T1m);
Chris@82 255 Tw = W[0];
Chris@82 256 T1c = Tw * T1b;
Chris@82 257 T1H = Tw * T1q;
Chris@82 258 T1d = W[1];
Chris@82 259 T1r = FMA(T1d, T1q, T1c);
Chris@82 260 T1I = FNMS(T1d, T1b, T1H);
Chris@82 261 Rp[0] = Tv - T1r;
Chris@82 262 Ip[0] = T1G + T1I;
Chris@82 263 Rm[0] = Tv + T1r;
Chris@82 264 Im[0] = T1I - T1G;
Chris@82 265 }
Chris@82 266 {
Chris@82 267 E T1N, T1J, T1L, T1M, T1V, T1Q, T1T, T1R, T1X, T1K, T1P;
Chris@82 268 T1N = T1F - T1y;
Chris@82 269 T1K = Tf - Tu;
Chris@82 270 T1J = W[14];
Chris@82 271 T1L = T1J * T1K;
Chris@82 272 T1M = W[15];
Chris@82 273 T1V = T1M * T1K;
Chris@82 274 T1Q = FNMS(KP923879532, T1a, TN);
Chris@82 275 T1T = FNMS(KP923879532, T1p, T1m);
Chris@82 276 T1P = W[16];
Chris@82 277 T1R = T1P * T1Q;
Chris@82 278 T1X = T1P * T1T;
Chris@82 279 {
Chris@82 280 E T1O, T1W, T1U, T1Y, T1S;
Chris@82 281 T1O = FNMS(T1M, T1N, T1L);
Chris@82 282 T1W = FMA(T1J, T1N, T1V);
Chris@82 283 T1S = W[17];
Chris@82 284 T1U = FMA(T1S, T1T, T1R);
Chris@82 285 T1Y = FNMS(T1S, T1Q, T1X);
Chris@82 286 Rp[WS(rs, 4)] = T1O - T1U;
Chris@82 287 Ip[WS(rs, 4)] = T1W + T1Y;
Chris@82 288 Rm[WS(rs, 4)] = T1O + T1U;
Chris@82 289 Im[WS(rs, 4)] = T1Y - T1W;
Chris@82 290 }
Chris@82 291 }
Chris@82 292 {
Chris@82 293 E T2r, T2n, T2p, T2q, T2z, T2u, T2x, T2v, T2B, T2o, T2t;
Chris@82 294 T2r = T26 - T25;
Chris@82 295 T2o = T20 - T21;
Chris@82 296 T2n = W[22];
Chris@82 297 T2p = T2n * T2o;
Chris@82 298 T2q = W[23];
Chris@82 299 T2z = T2q * T2o;
Chris@82 300 T2u = FNMS(KP923879532, T2b, T2a);
Chris@82 301 T2x = FNMS(KP923879532, T2g, T2f);
Chris@82 302 T2t = W[24];
Chris@82 303 T2v = T2t * T2u;
Chris@82 304 T2B = T2t * T2x;
Chris@82 305 {
Chris@82 306 E T2s, T2A, T2y, T2C, T2w;
Chris@82 307 T2s = FNMS(T2q, T2r, T2p);
Chris@82 308 T2A = FMA(T2n, T2r, T2z);
Chris@82 309 T2w = W[25];
Chris@82 310 T2y = FMA(T2w, T2x, T2v);
Chris@82 311 T2C = FNMS(T2w, T2u, T2B);
Chris@82 312 Rp[WS(rs, 6)] = T2s - T2y;
Chris@82 313 Ip[WS(rs, 6)] = T2A + T2C;
Chris@82 314 Rm[WS(rs, 6)] = T2s + T2y;
Chris@82 315 Im[WS(rs, 6)] = T2C - T2A;
Chris@82 316 }
Chris@82 317 }
Chris@82 318 {
Chris@82 319 E T27, T1Z, T23, T24, T2j, T2c, T2h, T2d, T2l, T22, T29;
Chris@82 320 T27 = T25 + T26;
Chris@82 321 T22 = T20 + T21;
Chris@82 322 T1Z = W[6];
Chris@82 323 T23 = T1Z * T22;
Chris@82 324 T24 = W[7];
Chris@82 325 T2j = T24 * T22;
Chris@82 326 T2c = FMA(KP923879532, T2b, T2a);
Chris@82 327 T2h = FMA(KP923879532, T2g, T2f);
Chris@82 328 T29 = W[8];
Chris@82 329 T2d = T29 * T2c;
Chris@82 330 T2l = T29 * T2h;
Chris@82 331 {
Chris@82 332 E T28, T2k, T2i, T2m, T2e;
Chris@82 333 T28 = FNMS(T24, T27, T23);
Chris@82 334 T2k = FMA(T1Z, T27, T2j);
Chris@82 335 T2e = W[9];
Chris@82 336 T2i = FMA(T2e, T2h, T2d);
Chris@82 337 T2m = FNMS(T2e, T2c, T2l);
Chris@82 338 Rp[WS(rs, 2)] = T28 - T2i;
Chris@82 339 Ip[WS(rs, 2)] = T2k + T2m;
Chris@82 340 Rm[WS(rs, 2)] = T28 + T2i;
Chris@82 341 Im[WS(rs, 2)] = T2m - T2k;
Chris@82 342 }
Chris@82 343 }
Chris@82 344 {
Chris@82 345 E T3N, T47, T43, T45, T46, T4f, T3F, T3J, T3K, T3Z, T3S, T3X, T3T, T41, T4a;
Chris@82 346 E T4d, T4b, T4h;
Chris@82 347 {
Chris@82 348 E T3M, T44, T3I, T3P, T49;
Chris@82 349 T3M = T2J - T2M;
Chris@82 350 T3N = FMA(KP707106781, T3M, T3L);
Chris@82 351 T47 = FNMS(KP707106781, T3M, T3L);
Chris@82 352 T44 = FNMS(KP707106781, T3H, T3G);
Chris@82 353 T43 = W[26];
Chris@82 354 T45 = T43 * T44;
Chris@82 355 T46 = W[27];
Chris@82 356 T4f = T46 * T44;
Chris@82 357 T3I = FMA(KP707106781, T3H, T3G);
Chris@82 358 T3F = W[10];
Chris@82 359 T3J = T3F * T3I;
Chris@82 360 T3K = W[11];
Chris@82 361 T3Z = T3K * T3I;
Chris@82 362 T3S = FMA(KP923879532, T3R, T3Q);
Chris@82 363 T3X = FNMS(KP923879532, T3W, T3V);
Chris@82 364 T3P = W[12];
Chris@82 365 T3T = T3P * T3S;
Chris@82 366 T41 = T3P * T3X;
Chris@82 367 T4a = FNMS(KP923879532, T3R, T3Q);
Chris@82 368 T4d = FMA(KP923879532, T3W, T3V);
Chris@82 369 T49 = W[28];
Chris@82 370 T4b = T49 * T4a;
Chris@82 371 T4h = T49 * T4d;
Chris@82 372 }
Chris@82 373 {
Chris@82 374 E T3O, T40, T3Y, T42, T3U;
Chris@82 375 T3O = FNMS(T3K, T3N, T3J);
Chris@82 376 T40 = FMA(T3F, T3N, T3Z);
Chris@82 377 T3U = W[13];
Chris@82 378 T3Y = FMA(T3U, T3X, T3T);
Chris@82 379 T42 = FNMS(T3U, T3S, T41);
Chris@82 380 Rp[WS(rs, 3)] = T3O - T3Y;
Chris@82 381 Ip[WS(rs, 3)] = T40 + T42;
Chris@82 382 Rm[WS(rs, 3)] = T3O + T3Y;
Chris@82 383 Im[WS(rs, 3)] = T42 - T40;
Chris@82 384 }
Chris@82 385 {
Chris@82 386 E T48, T4g, T4e, T4i, T4c;
Chris@82 387 T48 = FNMS(T46, T47, T45);
Chris@82 388 T4g = FMA(T43, T47, T4f);
Chris@82 389 T4c = W[29];
Chris@82 390 T4e = FMA(T4c, T4d, T4b);
Chris@82 391 T4i = FNMS(T4c, T4a, T4h);
Chris@82 392 Rp[WS(rs, 7)] = T48 - T4e;
Chris@82 393 Ip[WS(rs, 7)] = T4g + T4i;
Chris@82 394 Rm[WS(rs, 7)] = T48 + T4e;
Chris@82 395 Im[WS(rs, 7)] = T4i - T4g;
Chris@82 396 }
Chris@82 397 }
Chris@82 398 {
Chris@82 399 E T2X, T3t, T3p, T3r, T3s, T3B, T2D, T2P, T2Q, T3l, T3a, T3j, T3b, T3n, T3w;
Chris@82 400 E T3z, T3x, T3D;
Chris@82 401 {
Chris@82 402 E T2W, T3q, T2O, T2Z, T3v;
Chris@82 403 T2W = T2U + T2V;
Chris@82 404 T2X = FMA(KP707106781, T2W, T2T);
Chris@82 405 T3t = FNMS(KP707106781, T2W, T2T);
Chris@82 406 T3q = FNMS(KP707106781, T2N, T2G);
Chris@82 407 T3p = W[18];
Chris@82 408 T3r = T3p * T3q;
Chris@82 409 T3s = W[19];
Chris@82 410 T3B = T3s * T3q;
Chris@82 411 T2O = FMA(KP707106781, T2N, T2G);
Chris@82 412 T2D = W[2];
Chris@82 413 T2P = T2D * T2O;
Chris@82 414 T2Q = W[3];
Chris@82 415 T3l = T2Q * T2O;
Chris@82 416 T3a = FMA(KP923879532, T39, T32);
Chris@82 417 T3j = FNMS(KP923879532, T3i, T3f);
Chris@82 418 T2Z = W[4];
Chris@82 419 T3b = T2Z * T3a;
Chris@82 420 T3n = T2Z * T3j;
Chris@82 421 T3w = FNMS(KP923879532, T39, T32);
Chris@82 422 T3z = FMA(KP923879532, T3i, T3f);
Chris@82 423 T3v = W[20];
Chris@82 424 T3x = T3v * T3w;
Chris@82 425 T3D = T3v * T3z;
Chris@82 426 }
Chris@82 427 {
Chris@82 428 E T2Y, T3m, T3k, T3o, T3c;
Chris@82 429 T2Y = FNMS(T2Q, T2X, T2P);
Chris@82 430 T3m = FMA(T2D, T2X, T3l);
Chris@82 431 T3c = W[5];
Chris@82 432 T3k = FMA(T3c, T3j, T3b);
Chris@82 433 T3o = FNMS(T3c, T3a, T3n);
Chris@82 434 Rp[WS(rs, 1)] = T2Y - T3k;
Chris@82 435 Ip[WS(rs, 1)] = T3m + T3o;
Chris@82 436 Rm[WS(rs, 1)] = T2Y + T3k;
Chris@82 437 Im[WS(rs, 1)] = T3o - T3m;
Chris@82 438 }
Chris@82 439 {
Chris@82 440 E T3u, T3C, T3A, T3E, T3y;
Chris@82 441 T3u = FNMS(T3s, T3t, T3r);
Chris@82 442 T3C = FMA(T3p, T3t, T3B);
Chris@82 443 T3y = W[21];
Chris@82 444 T3A = FMA(T3y, T3z, T3x);
Chris@82 445 T3E = FNMS(T3y, T3w, T3D);
Chris@82 446 Rp[WS(rs, 5)] = T3u - T3A;
Chris@82 447 Ip[WS(rs, 5)] = T3C + T3E;
Chris@82 448 Rm[WS(rs, 5)] = T3u + T3A;
Chris@82 449 Im[WS(rs, 5)] = T3E - T3C;
Chris@82 450 }
Chris@82 451 }
Chris@82 452 }
Chris@82 453 }
Chris@82 454 }
Chris@82 455
Chris@82 456 static const tw_instr twinstr[] = {
Chris@82 457 {TW_FULL, 1, 16},
Chris@82 458 {TW_NEXT, 1, 0}
Chris@82 459 };
Chris@82 460
Chris@82 461 static const hc2c_desc desc = { 16, "hc2cbdft_16", twinstr, &GENUS, {136, 30, 70, 0} };
Chris@82 462
Chris@82 463 void X(codelet_hc2cbdft_16) (planner *p) {
Chris@82 464 X(khc2c_register) (p, hc2cbdft_16, &desc, HC2C_VIA_DFT);
Chris@82 465 }
Chris@82 466 #else
Chris@82 467
Chris@82 468 /* Generated by: ../../../genfft/gen_hc2cdft.native -compact -variables 4 -pipeline-latency 4 -sign 1 -n 16 -dif -name hc2cbdft_16 -include rdft/scalar/hc2cb.h */
Chris@82 469
Chris@82 470 /*
Chris@82 471 * This function contains 206 FP additions, 84 FP multiplications,
Chris@82 472 * (or, 168 additions, 46 multiplications, 38 fused multiply/add),
Chris@82 473 * 60 stack variables, 3 constants, and 64 memory accesses
Chris@82 474 */
Chris@82 475 #include "rdft/scalar/hc2cb.h"
Chris@82 476
Chris@82 477 static void hc2cbdft_16(R *Rp, R *Ip, R *Rm, R *Im, const R *W, stride rs, INT mb, INT me, INT ms)
Chris@82 478 {
Chris@82 479 DK(KP923879532, +0.923879532511286756128183189396788286822416626);
Chris@82 480 DK(KP382683432, +0.382683432365089771728459984030398866761344562);
Chris@82 481 DK(KP707106781, +0.707106781186547524400844362104849039284835938);
Chris@82 482 {
Chris@82 483 INT m;
Chris@82 484 for (m = mb, W = W + ((mb - 1) * 30); m < me; m = m + 1, Rp = Rp + ms, Ip = Ip + ms, Rm = Rm - ms, Im = Im - ms, W = W + 30, MAKE_VOLATILE_STRIDE(64, rs)) {
Chris@82 485 E TB, T2L, T30, T1n, Tf, T1U, T2H, T3p, T1E, T1Z, TM, T31, T2s, T3k, T1i;
Chris@82 486 E T2M, Tu, T1Y, T2Q, T2X, T2T, T2Y, TY, T1d, T19, T1e, T2v, T2C, T2y, T2D;
Chris@82 487 E T1x, T1V;
Chris@82 488 {
Chris@82 489 E T3, T1j, TA, T1B, T6, Tx, T1m, T1C, Ta, TC, TF, T1y, Td, TH, TK;
Chris@82 490 E T1z;
Chris@82 491 {
Chris@82 492 E T1, T2, Ty, Tz;
Chris@82 493 T1 = Rp[0];
Chris@82 494 T2 = Rm[WS(rs, 7)];
Chris@82 495 T3 = T1 + T2;
Chris@82 496 T1j = T1 - T2;
Chris@82 497 Ty = Ip[0];
Chris@82 498 Tz = Im[WS(rs, 7)];
Chris@82 499 TA = Ty + Tz;
Chris@82 500 T1B = Ty - Tz;
Chris@82 501 }
Chris@82 502 {
Chris@82 503 E T4, T5, T1k, T1l;
Chris@82 504 T4 = Rp[WS(rs, 4)];
Chris@82 505 T5 = Rm[WS(rs, 3)];
Chris@82 506 T6 = T4 + T5;
Chris@82 507 Tx = T4 - T5;
Chris@82 508 T1k = Ip[WS(rs, 4)];
Chris@82 509 T1l = Im[WS(rs, 3)];
Chris@82 510 T1m = T1k + T1l;
Chris@82 511 T1C = T1k - T1l;
Chris@82 512 }
Chris@82 513 {
Chris@82 514 E T8, T9, TD, TE;
Chris@82 515 T8 = Rp[WS(rs, 2)];
Chris@82 516 T9 = Rm[WS(rs, 5)];
Chris@82 517 Ta = T8 + T9;
Chris@82 518 TC = T8 - T9;
Chris@82 519 TD = Ip[WS(rs, 2)];
Chris@82 520 TE = Im[WS(rs, 5)];
Chris@82 521 TF = TD + TE;
Chris@82 522 T1y = TD - TE;
Chris@82 523 }
Chris@82 524 {
Chris@82 525 E Tb, Tc, TI, TJ;
Chris@82 526 Tb = Rm[WS(rs, 1)];
Chris@82 527 Tc = Rp[WS(rs, 6)];
Chris@82 528 Td = Tb + Tc;
Chris@82 529 TH = Tb - Tc;
Chris@82 530 TI = Im[WS(rs, 1)];
Chris@82 531 TJ = Ip[WS(rs, 6)];
Chris@82 532 TK = TI + TJ;
Chris@82 533 T1z = TJ - TI;
Chris@82 534 }
Chris@82 535 {
Chris@82 536 E T7, Te, TG, TL;
Chris@82 537 TB = Tx + TA;
Chris@82 538 T2L = TA - Tx;
Chris@82 539 T30 = T1j + T1m;
Chris@82 540 T1n = T1j - T1m;
Chris@82 541 T7 = T3 + T6;
Chris@82 542 Te = Ta + Td;
Chris@82 543 Tf = T7 + Te;
Chris@82 544 T1U = T7 - Te;
Chris@82 545 {
Chris@82 546 E T2F, T2G, T1A, T1D;
Chris@82 547 T2F = Ta - Td;
Chris@82 548 T2G = T1B - T1C;
Chris@82 549 T2H = T2F + T2G;
Chris@82 550 T3p = T2G - T2F;
Chris@82 551 T1A = T1y + T1z;
Chris@82 552 T1D = T1B + T1C;
Chris@82 553 T1E = T1A + T1D;
Chris@82 554 T1Z = T1D - T1A;
Chris@82 555 }
Chris@82 556 TG = TC + TF;
Chris@82 557 TL = TH + TK;
Chris@82 558 TM = KP707106781 * (TG - TL);
Chris@82 559 T31 = KP707106781 * (TG + TL);
Chris@82 560 {
Chris@82 561 E T2q, T2r, T1g, T1h;
Chris@82 562 T2q = T3 - T6;
Chris@82 563 T2r = T1z - T1y;
Chris@82 564 T2s = T2q + T2r;
Chris@82 565 T3k = T2q - T2r;
Chris@82 566 T1g = TC - TF;
Chris@82 567 T1h = TH - TK;
Chris@82 568 T1i = KP707106781 * (T1g + T1h);
Chris@82 569 T2M = KP707106781 * (T1g - T1h);
Chris@82 570 }
Chris@82 571 }
Chris@82 572 }
Chris@82 573 {
Chris@82 574 E Ti, TT, TR, T1r, Tl, TO, TW, T1s, Tp, T14, T12, T1u, Ts, TZ, T17;
Chris@82 575 E T1v;
Chris@82 576 {
Chris@82 577 E Tg, Th, TP, TQ;
Chris@82 578 Tg = Rp[WS(rs, 1)];
Chris@82 579 Th = Rm[WS(rs, 6)];
Chris@82 580 Ti = Tg + Th;
Chris@82 581 TT = Tg - Th;
Chris@82 582 TP = Ip[WS(rs, 1)];
Chris@82 583 TQ = Im[WS(rs, 6)];
Chris@82 584 TR = TP + TQ;
Chris@82 585 T1r = TP - TQ;
Chris@82 586 }
Chris@82 587 {
Chris@82 588 E Tj, Tk, TU, TV;
Chris@82 589 Tj = Rp[WS(rs, 5)];
Chris@82 590 Tk = Rm[WS(rs, 2)];
Chris@82 591 Tl = Tj + Tk;
Chris@82 592 TO = Tj - Tk;
Chris@82 593 TU = Ip[WS(rs, 5)];
Chris@82 594 TV = Im[WS(rs, 2)];
Chris@82 595 TW = TU + TV;
Chris@82 596 T1s = TU - TV;
Chris@82 597 }
Chris@82 598 {
Chris@82 599 E Tn, To, T10, T11;
Chris@82 600 Tn = Rm[0];
Chris@82 601 To = Rp[WS(rs, 7)];
Chris@82 602 Tp = Tn + To;
Chris@82 603 T14 = Tn - To;
Chris@82 604 T10 = Im[0];
Chris@82 605 T11 = Ip[WS(rs, 7)];
Chris@82 606 T12 = T10 + T11;
Chris@82 607 T1u = T11 - T10;
Chris@82 608 }
Chris@82 609 {
Chris@82 610 E Tq, Tr, T15, T16;
Chris@82 611 Tq = Rp[WS(rs, 3)];
Chris@82 612 Tr = Rm[WS(rs, 4)];
Chris@82 613 Ts = Tq + Tr;
Chris@82 614 TZ = Tq - Tr;
Chris@82 615 T15 = Ip[WS(rs, 3)];
Chris@82 616 T16 = Im[WS(rs, 4)];
Chris@82 617 T17 = T15 + T16;
Chris@82 618 T1v = T15 - T16;
Chris@82 619 }
Chris@82 620 {
Chris@82 621 E Tm, Tt, T2O, T2P;
Chris@82 622 Tm = Ti + Tl;
Chris@82 623 Tt = Tp + Ts;
Chris@82 624 Tu = Tm + Tt;
Chris@82 625 T1Y = Tm - Tt;
Chris@82 626 T2O = TR - TO;
Chris@82 627 T2P = TT + TW;
Chris@82 628 T2Q = FMA(KP382683432, T2O, KP923879532 * T2P);
Chris@82 629 T2X = FNMS(KP923879532, T2O, KP382683432 * T2P);
Chris@82 630 }
Chris@82 631 {
Chris@82 632 E T2R, T2S, TS, TX;
Chris@82 633 T2R = TZ + T12;
Chris@82 634 T2S = T14 + T17;
Chris@82 635 T2T = FMA(KP382683432, T2R, KP923879532 * T2S);
Chris@82 636 T2Y = FNMS(KP923879532, T2R, KP382683432 * T2S);
Chris@82 637 TS = TO + TR;
Chris@82 638 TX = TT - TW;
Chris@82 639 TY = FMA(KP923879532, TS, KP382683432 * TX);
Chris@82 640 T1d = FNMS(KP382683432, TS, KP923879532 * TX);
Chris@82 641 }
Chris@82 642 {
Chris@82 643 E T13, T18, T2t, T2u;
Chris@82 644 T13 = TZ - T12;
Chris@82 645 T18 = T14 - T17;
Chris@82 646 T19 = FNMS(KP382683432, T18, KP923879532 * T13);
Chris@82 647 T1e = FMA(KP382683432, T13, KP923879532 * T18);
Chris@82 648 T2t = Ti - Tl;
Chris@82 649 T2u = T1r - T1s;
Chris@82 650 T2v = T2t - T2u;
Chris@82 651 T2C = T2t + T2u;
Chris@82 652 }
Chris@82 653 {
Chris@82 654 E T2w, T2x, T1t, T1w;
Chris@82 655 T2w = Tp - Ts;
Chris@82 656 T2x = T1u - T1v;
Chris@82 657 T2y = T2w + T2x;
Chris@82 658 T2D = T2x - T2w;
Chris@82 659 T1t = T1r + T1s;
Chris@82 660 T1w = T1u + T1v;
Chris@82 661 T1x = T1t + T1w;
Chris@82 662 T1V = T1w - T1t;
Chris@82 663 }
Chris@82 664 }
Chris@82 665 {
Chris@82 666 E Tv, T1F, T1b, T1N, T1p, T1P, T1L, T1R;
Chris@82 667 Tv = Tf + Tu;
Chris@82 668 T1F = T1x + T1E;
Chris@82 669 {
Chris@82 670 E TN, T1a, T1f, T1o;
Chris@82 671 TN = TB + TM;
Chris@82 672 T1a = TY + T19;
Chris@82 673 T1b = TN + T1a;
Chris@82 674 T1N = TN - T1a;
Chris@82 675 T1f = T1d + T1e;
Chris@82 676 T1o = T1i + T1n;
Chris@82 677 T1p = T1f + T1o;
Chris@82 678 T1P = T1o - T1f;
Chris@82 679 {
Chris@82 680 E T1I, T1K, T1H, T1J;
Chris@82 681 T1I = Tf - Tu;
Chris@82 682 T1K = T1E - T1x;
Chris@82 683 T1H = W[14];
Chris@82 684 T1J = W[15];
Chris@82 685 T1L = FNMS(T1J, T1K, T1H * T1I);
Chris@82 686 T1R = FMA(T1J, T1I, T1H * T1K);
Chris@82 687 }
Chris@82 688 }
Chris@82 689 {
Chris@82 690 E T1q, T1G, Tw, T1c;
Chris@82 691 Tw = W[0];
Chris@82 692 T1c = W[1];
Chris@82 693 T1q = FMA(Tw, T1b, T1c * T1p);
Chris@82 694 T1G = FNMS(T1c, T1b, Tw * T1p);
Chris@82 695 Rp[0] = Tv - T1q;
Chris@82 696 Ip[0] = T1F + T1G;
Chris@82 697 Rm[0] = Tv + T1q;
Chris@82 698 Im[0] = T1G - T1F;
Chris@82 699 }
Chris@82 700 {
Chris@82 701 E T1Q, T1S, T1M, T1O;
Chris@82 702 T1M = W[16];
Chris@82 703 T1O = W[17];
Chris@82 704 T1Q = FMA(T1M, T1N, T1O * T1P);
Chris@82 705 T1S = FNMS(T1O, T1N, T1M * T1P);
Chris@82 706 Rp[WS(rs, 4)] = T1L - T1Q;
Chris@82 707 Ip[WS(rs, 4)] = T1R + T1S;
Chris@82 708 Rm[WS(rs, 4)] = T1L + T1Q;
Chris@82 709 Im[WS(rs, 4)] = T1S - T1R;
Chris@82 710 }
Chris@82 711 }
Chris@82 712 {
Chris@82 713 E T25, T2j, T29, T2l, T21, T2b, T2h, T2n;
Chris@82 714 {
Chris@82 715 E T23, T24, T27, T28;
Chris@82 716 T23 = TB - TM;
Chris@82 717 T24 = T1d - T1e;
Chris@82 718 T25 = T23 + T24;
Chris@82 719 T2j = T23 - T24;
Chris@82 720 T27 = T19 - TY;
Chris@82 721 T28 = T1n - T1i;
Chris@82 722 T29 = T27 + T28;
Chris@82 723 T2l = T28 - T27;
Chris@82 724 }
Chris@82 725 {
Chris@82 726 E T1W, T20, T1T, T1X;
Chris@82 727 T1W = T1U + T1V;
Chris@82 728 T20 = T1Y + T1Z;
Chris@82 729 T1T = W[6];
Chris@82 730 T1X = W[7];
Chris@82 731 T21 = FNMS(T1X, T20, T1T * T1W);
Chris@82 732 T2b = FMA(T1X, T1W, T1T * T20);
Chris@82 733 }
Chris@82 734 {
Chris@82 735 E T2e, T2g, T2d, T2f;
Chris@82 736 T2e = T1U - T1V;
Chris@82 737 T2g = T1Z - T1Y;
Chris@82 738 T2d = W[22];
Chris@82 739 T2f = W[23];
Chris@82 740 T2h = FNMS(T2f, T2g, T2d * T2e);
Chris@82 741 T2n = FMA(T2f, T2e, T2d * T2g);
Chris@82 742 }
Chris@82 743 {
Chris@82 744 E T2a, T2c, T22, T26;
Chris@82 745 T22 = W[8];
Chris@82 746 T26 = W[9];
Chris@82 747 T2a = FMA(T22, T25, T26 * T29);
Chris@82 748 T2c = FNMS(T26, T25, T22 * T29);
Chris@82 749 Rp[WS(rs, 2)] = T21 - T2a;
Chris@82 750 Ip[WS(rs, 2)] = T2b + T2c;
Chris@82 751 Rm[WS(rs, 2)] = T21 + T2a;
Chris@82 752 Im[WS(rs, 2)] = T2c - T2b;
Chris@82 753 }
Chris@82 754 {
Chris@82 755 E T2m, T2o, T2i, T2k;
Chris@82 756 T2i = W[24];
Chris@82 757 T2k = W[25];
Chris@82 758 T2m = FMA(T2i, T2j, T2k * T2l);
Chris@82 759 T2o = FNMS(T2k, T2j, T2i * T2l);
Chris@82 760 Rp[WS(rs, 6)] = T2h - T2m;
Chris@82 761 Ip[WS(rs, 6)] = T2n + T2o;
Chris@82 762 Rm[WS(rs, 6)] = T2h + T2m;
Chris@82 763 Im[WS(rs, 6)] = T2o - T2n;
Chris@82 764 }
Chris@82 765 }
Chris@82 766 {
Chris@82 767 E T2A, T38, T2I, T3a, T2V, T3d, T33, T3f, T2z, T2E;
Chris@82 768 T2z = KP707106781 * (T2v + T2y);
Chris@82 769 T2A = T2s + T2z;
Chris@82 770 T38 = T2s - T2z;
Chris@82 771 T2E = KP707106781 * (T2C + T2D);
Chris@82 772 T2I = T2E + T2H;
Chris@82 773 T3a = T2H - T2E;
Chris@82 774 {
Chris@82 775 E T2N, T2U, T2Z, T32;
Chris@82 776 T2N = T2L + T2M;
Chris@82 777 T2U = T2Q - T2T;
Chris@82 778 T2V = T2N + T2U;
Chris@82 779 T3d = T2N - T2U;
Chris@82 780 T2Z = T2X + T2Y;
Chris@82 781 T32 = T30 - T31;
Chris@82 782 T33 = T2Z + T32;
Chris@82 783 T3f = T32 - T2Z;
Chris@82 784 }
Chris@82 785 {
Chris@82 786 E T2J, T35, T34, T36;
Chris@82 787 {
Chris@82 788 E T2p, T2B, T2K, T2W;
Chris@82 789 T2p = W[2];
Chris@82 790 T2B = W[3];
Chris@82 791 T2J = FNMS(T2B, T2I, T2p * T2A);
Chris@82 792 T35 = FMA(T2B, T2A, T2p * T2I);
Chris@82 793 T2K = W[4];
Chris@82 794 T2W = W[5];
Chris@82 795 T34 = FMA(T2K, T2V, T2W * T33);
Chris@82 796 T36 = FNMS(T2W, T2V, T2K * T33);
Chris@82 797 }
Chris@82 798 Rp[WS(rs, 1)] = T2J - T34;
Chris@82 799 Ip[WS(rs, 1)] = T35 + T36;
Chris@82 800 Rm[WS(rs, 1)] = T2J + T34;
Chris@82 801 Im[WS(rs, 1)] = T36 - T35;
Chris@82 802 }
Chris@82 803 {
Chris@82 804 E T3b, T3h, T3g, T3i;
Chris@82 805 {
Chris@82 806 E T37, T39, T3c, T3e;
Chris@82 807 T37 = W[18];
Chris@82 808 T39 = W[19];
Chris@82 809 T3b = FNMS(T39, T3a, T37 * T38);
Chris@82 810 T3h = FMA(T39, T38, T37 * T3a);
Chris@82 811 T3c = W[20];
Chris@82 812 T3e = W[21];
Chris@82 813 T3g = FMA(T3c, T3d, T3e * T3f);
Chris@82 814 T3i = FNMS(T3e, T3d, T3c * T3f);
Chris@82 815 }
Chris@82 816 Rp[WS(rs, 5)] = T3b - T3g;
Chris@82 817 Ip[WS(rs, 5)] = T3h + T3i;
Chris@82 818 Rm[WS(rs, 5)] = T3b + T3g;
Chris@82 819 Im[WS(rs, 5)] = T3i - T3h;
Chris@82 820 }
Chris@82 821 }
Chris@82 822 {
Chris@82 823 E T3m, T3E, T3q, T3G, T3v, T3J, T3z, T3L, T3l, T3o;
Chris@82 824 T3l = KP707106781 * (T2D - T2C);
Chris@82 825 T3m = T3k + T3l;
Chris@82 826 T3E = T3k - T3l;
Chris@82 827 T3o = KP707106781 * (T2v - T2y);
Chris@82 828 T3q = T3o + T3p;
Chris@82 829 T3G = T3p - T3o;
Chris@82 830 {
Chris@82 831 E T3t, T3u, T3x, T3y;
Chris@82 832 T3t = T2L - T2M;
Chris@82 833 T3u = T2X - T2Y;
Chris@82 834 T3v = T3t + T3u;
Chris@82 835 T3J = T3t - T3u;
Chris@82 836 T3x = T31 + T30;
Chris@82 837 T3y = T2Q + T2T;
Chris@82 838 T3z = T3x - T3y;
Chris@82 839 T3L = T3y + T3x;
Chris@82 840 }
Chris@82 841 {
Chris@82 842 E T3r, T3B, T3A, T3C;
Chris@82 843 {
Chris@82 844 E T3j, T3n, T3s, T3w;
Chris@82 845 T3j = W[10];
Chris@82 846 T3n = W[11];
Chris@82 847 T3r = FNMS(T3n, T3q, T3j * T3m);
Chris@82 848 T3B = FMA(T3n, T3m, T3j * T3q);
Chris@82 849 T3s = W[12];
Chris@82 850 T3w = W[13];
Chris@82 851 T3A = FMA(T3s, T3v, T3w * T3z);
Chris@82 852 T3C = FNMS(T3w, T3v, T3s * T3z);
Chris@82 853 }
Chris@82 854 Rp[WS(rs, 3)] = T3r - T3A;
Chris@82 855 Ip[WS(rs, 3)] = T3B + T3C;
Chris@82 856 Rm[WS(rs, 3)] = T3r + T3A;
Chris@82 857 Im[WS(rs, 3)] = T3C - T3B;
Chris@82 858 }
Chris@82 859 {
Chris@82 860 E T3H, T3N, T3M, T3O;
Chris@82 861 {
Chris@82 862 E T3D, T3F, T3I, T3K;
Chris@82 863 T3D = W[26];
Chris@82 864 T3F = W[27];
Chris@82 865 T3H = FNMS(T3F, T3G, T3D * T3E);
Chris@82 866 T3N = FMA(T3F, T3E, T3D * T3G);
Chris@82 867 T3I = W[28];
Chris@82 868 T3K = W[29];
Chris@82 869 T3M = FMA(T3I, T3J, T3K * T3L);
Chris@82 870 T3O = FNMS(T3K, T3J, T3I * T3L);
Chris@82 871 }
Chris@82 872 Rp[WS(rs, 7)] = T3H - T3M;
Chris@82 873 Ip[WS(rs, 7)] = T3N + T3O;
Chris@82 874 Rm[WS(rs, 7)] = T3H + T3M;
Chris@82 875 Im[WS(rs, 7)] = T3O - T3N;
Chris@82 876 }
Chris@82 877 }
Chris@82 878 }
Chris@82 879 }
Chris@82 880 }
Chris@82 881
Chris@82 882 static const tw_instr twinstr[] = {
Chris@82 883 {TW_FULL, 1, 16},
Chris@82 884 {TW_NEXT, 1, 0}
Chris@82 885 };
Chris@82 886
Chris@82 887 static const hc2c_desc desc = { 16, "hc2cbdft_16", twinstr, &GENUS, {168, 46, 38, 0} };
Chris@82 888
Chris@82 889 void X(codelet_hc2cbdft_16) (planner *p) {
Chris@82 890 X(khc2c_register) (p, hc2cbdft_16, &desc, HC2C_VIA_DFT);
Chris@82 891 }
Chris@82 892 #endif