Mercurial > hg > sv-dependency-builds

annotate src/fftw-3.3.8/rdft/scalar/r2cb/hb_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 08:07:32 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_hc2hc.native -fma -compact -variables 4 -pipeline-latency 4 -sign 1 -n 16 -dif -name hb_16 -include rdft/scalar/hb.h */
Chris@82 29
Chris@82 30 /*
Chris@82 31 * This function contains 174 FP additions, 100 FP multiplications,
Chris@82 32 * (or, 104 additions, 30 multiplications, 70 fused multiply/add),
Chris@82 33 * 63 stack variables, 3 constants, and 64 memory accesses
Chris@82 34 */
Chris@82 35 #include "rdft/scalar/hb.h"
Chris@82 36
Chris@82 37 static void hb_16(R *cr, R *ci, 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(KP707106781, +0.707106781186547524400844362104849039284835938);
Chris@82 41 DK(KP414213562, +0.414213562373095048801688724209698078569671875);
Chris@82 42 {
Chris@82 43 INT m;
Chris@82 44 for (m = mb, W = W + ((mb - 1) * 30); m < me; m = m + 1, cr = cr + ms, ci = ci - ms, W = W + 30, MAKE_VOLATILE_STRIDE(32, rs)) {
Chris@82 45 E TA, T1O, T21, T1h, T2P, T2S, T3b, T3p, T3q, T3D, T1k, T1P, Tf, T3y, T2A;
Chris@82 46 E T36, TL, T22, T3s, T3t, T3z, T2F, T2U, T2K, T2V, Tu, T3E, TX, T1n, T1T;
Chris@82 47 E T24, T1W, T25, T18, T1m;
Chris@82 48 {
Chris@82 49 E T3, Tw, TJ, T2x, T1g, T2Q, T6, T1d, Ta, TB, Tz, T2R, TE, T2y, Td;
Chris@82 50 E TG;
Chris@82 51 {
Chris@82 52 E T1, T2, TH, TI;
Chris@82 53 T1 = cr[0];
Chris@82 54 T2 = ci[WS(rs, 7)];
Chris@82 55 T3 = T1 + T2;
Chris@82 56 Tw = T1 - T2;
Chris@82 57 TH = ci[WS(rs, 9)];
Chris@82 58 TI = cr[WS(rs, 14)];
Chris@82 59 TJ = TH + TI;
Chris@82 60 T2x = TH - TI;
Chris@82 61 }
Chris@82 62 {
Chris@82 63 E T1e, T1f, T4, T5;
Chris@82 64 T1e = ci[WS(rs, 15)];
Chris@82 65 T1f = cr[WS(rs, 8)];
Chris@82 66 T1g = T1e + T1f;
Chris@82 67 T2Q = T1e - T1f;
Chris@82 68 T4 = cr[WS(rs, 4)];
Chris@82 69 T5 = ci[WS(rs, 3)];
Chris@82 70 T6 = T4 + T5;
Chris@82 71 T1d = T4 - T5;
Chris@82 72 }
Chris@82 73 {
Chris@82 74 E T8, T9, Tx, Ty;
Chris@82 75 T8 = cr[WS(rs, 2)];
Chris@82 76 T9 = ci[WS(rs, 5)];
Chris@82 77 Ta = T8 + T9;
Chris@82 78 TB = T8 - T9;
Chris@82 79 Tx = ci[WS(rs, 11)];
Chris@82 80 Ty = cr[WS(rs, 12)];
Chris@82 81 Tz = Tx + Ty;
Chris@82 82 T2R = Tx - Ty;
Chris@82 83 }
Chris@82 84 {
Chris@82 85 E TC, TD, Tb, Tc;
Chris@82 86 TC = ci[WS(rs, 13)];
Chris@82 87 TD = cr[WS(rs, 10)];
Chris@82 88 TE = TC + TD;
Chris@82 89 T2y = TC - TD;
Chris@82 90 Tb = ci[WS(rs, 1)];
Chris@82 91 Tc = cr[WS(rs, 6)];
Chris@82 92 Td = Tb + Tc;
Chris@82 93 TG = Tb - Tc;
Chris@82 94 }
Chris@82 95 TA = Tw - Tz;
Chris@82 96 T1O = Tw + Tz;
Chris@82 97 T21 = T1g - T1d;
Chris@82 98 T1h = T1d + T1g;
Chris@82 99 T2P = Ta - Td;
Chris@82 100 T2S = T2Q - T2R;
Chris@82 101 T3b = T2S - T2P;
Chris@82 102 {
Chris@82 103 E T1i, T1j, T7, Te;
Chris@82 104 T3p = T2Q + T2R;
Chris@82 105 T3q = T2y + T2x;
Chris@82 106 T3D = T3p - T3q;
Chris@82 107 T1i = TB + TE;
Chris@82 108 T1j = TG + TJ;
Chris@82 109 T1k = T1i - T1j;
Chris@82 110 T1P = T1i + T1j;
Chris@82 111 T7 = T3 + T6;
Chris@82 112 Te = Ta + Td;
Chris@82 113 Tf = T7 + Te;
Chris@82 114 T3y = T7 - Te;
Chris@82 115 {
Chris@82 116 E T2w, T2z, TF, TK;
Chris@82 117 T2w = T3 - T6;
Chris@82 118 T2z = T2x - T2y;
Chris@82 119 T2A = T2w + T2z;
Chris@82 120 T36 = T2w - T2z;
Chris@82 121 TF = TB - TE;
Chris@82 122 TK = TG - TJ;
Chris@82 123 TL = TF + TK;
Chris@82 124 T22 = TF - TK;
Chris@82 125 }
Chris@82 126 }
Chris@82 127 }
Chris@82 128 {
Chris@82 129 E Ti, T13, T11, T2C, T16, T2D, Tl, TY, Tp, TS, TQ, T2H, TV, T2I, Ts;
Chris@82 130 E TN, T2B, T2E;
Chris@82 131 {
Chris@82 132 E Tg, Th, TZ, T10;
Chris@82 133 Tg = cr[WS(rs, 1)];
Chris@82 134 Th = ci[WS(rs, 6)];
Chris@82 135 Ti = Tg + Th;
Chris@82 136 T13 = Tg - Th;
Chris@82 137 TZ = ci[WS(rs, 14)];
Chris@82 138 T10 = cr[WS(rs, 9)];
Chris@82 139 T11 = TZ + T10;
Chris@82 140 T2C = TZ - T10;
Chris@82 141 }
Chris@82 142 {
Chris@82 143 E T14, T15, Tj, Tk;
Chris@82 144 T14 = ci[WS(rs, 10)];
Chris@82 145 T15 = cr[WS(rs, 13)];
Chris@82 146 T16 = T14 + T15;
Chris@82 147 T2D = T14 - T15;
Chris@82 148 Tj = cr[WS(rs, 5)];
Chris@82 149 Tk = ci[WS(rs, 2)];
Chris@82 150 Tl = Tj + Tk;
Chris@82 151 TY = Tj - Tk;
Chris@82 152 }
Chris@82 153 {
Chris@82 154 E Tn, To, TO, TP;
Chris@82 155 Tn = ci[0];
Chris@82 156 To = cr[WS(rs, 7)];
Chris@82 157 Tp = Tn + To;
Chris@82 158 TS = Tn - To;
Chris@82 159 TO = ci[WS(rs, 8)];
Chris@82 160 TP = cr[WS(rs, 15)];
Chris@82 161 TQ = TO + TP;
Chris@82 162 T2H = TO - TP;
Chris@82 163 }
Chris@82 164 {
Chris@82 165 E TT, TU, Tq, Tr;
Chris@82 166 TT = ci[WS(rs, 12)];
Chris@82 167 TU = cr[WS(rs, 11)];
Chris@82 168 TV = TT + TU;
Chris@82 169 T2I = TT - TU;
Chris@82 170 Tq = cr[WS(rs, 3)];
Chris@82 171 Tr = ci[WS(rs, 4)];
Chris@82 172 Ts = Tq + Tr;
Chris@82 173 TN = Tq - Tr;
Chris@82 174 }
Chris@82 175 T3s = T2C + T2D;
Chris@82 176 T3t = T2H + T2I;
Chris@82 177 T3z = T3t - T3s;
Chris@82 178 T2B = Ti - Tl;
Chris@82 179 T2E = T2C - T2D;
Chris@82 180 T2F = T2B - T2E;
Chris@82 181 T2U = T2B + T2E;
Chris@82 182 {
Chris@82 183 E T2G, T2J, Tm, Tt;
Chris@82 184 T2G = Tp - Ts;
Chris@82 185 T2J = T2H - T2I;
Chris@82 186 T2K = T2G + T2J;
Chris@82 187 T2V = T2J - T2G;
Chris@82 188 Tm = Ti + Tl;
Chris@82 189 Tt = Tp + Ts;
Chris@82 190 Tu = Tm + Tt;
Chris@82 191 T3E = Tm - Tt;
Chris@82 192 }
Chris@82 193 {
Chris@82 194 E TR, TW, T1R, T1S;
Chris@82 195 TR = TN - TQ;
Chris@82 196 TW = TS - TV;
Chris@82 197 TX = FNMS(KP414213562, TW, TR);
Chris@82 198 T1n = FMA(KP414213562, TR, TW);
Chris@82 199 T1R = T11 - TY;
Chris@82 200 T1S = T13 + T16;
Chris@82 201 T1T = FNMS(KP414213562, T1S, T1R);
Chris@82 202 T24 = FMA(KP414213562, T1R, T1S);
Chris@82 203 }
Chris@82 204 {
Chris@82 205 E T1U, T1V, T12, T17;
Chris@82 206 T1U = TN + TQ;
Chris@82 207 T1V = TS + TV;
Chris@82 208 T1W = FNMS(KP414213562, T1V, T1U);
Chris@82 209 T25 = FMA(KP414213562, T1U, T1V);
Chris@82 210 T12 = TY + T11;
Chris@82 211 T17 = T13 - T16;
Chris@82 212 T18 = FMA(KP414213562, T17, T12);
Chris@82 213 T1m = FNMS(KP414213562, T12, T17);
Chris@82 214 }
Chris@82 215 }
Chris@82 216 cr[0] = Tf + Tu;
Chris@82 217 {
Chris@82 218 E T3r, T3u, T3v, T3l, T3n, T3o, T3w, T3m;
Chris@82 219 T3r = T3p + T3q;
Chris@82 220 T3u = T3s + T3t;
Chris@82 221 T3v = T3r - T3u;
Chris@82 222 T3m = Tf - Tu;
Chris@82 223 T3l = W[14];
Chris@82 224 T3n = T3l * T3m;
Chris@82 225 T3o = W[15];
Chris@82 226 T3w = T3o * T3m;
Chris@82 227 ci[0] = T3r + T3u;
Chris@82 228 ci[WS(rs, 8)] = FMA(T3l, T3v, T3w);
Chris@82 229 cr[WS(rs, 8)] = FNMS(T3o, T3v, T3n);
Chris@82 230 }
Chris@82 231 {
Chris@82 232 E T3A, T3F, T3B, T3G, T3x, T3C;
Chris@82 233 T3A = T3y - T3z;
Chris@82 234 T3F = T3D - T3E;
Chris@82 235 T3x = W[22];
Chris@82 236 T3B = T3x * T3A;
Chris@82 237 T3G = T3x * T3F;
Chris@82 238 T3C = W[23];
Chris@82 239 cr[WS(rs, 12)] = FNMS(T3C, T3F, T3B);
Chris@82 240 ci[WS(rs, 12)] = FMA(T3C, T3A, T3G);
Chris@82 241 }
Chris@82 242 {
Chris@82 243 E T3I, T3L, T3J, T3M, T3H, T3K;
Chris@82 244 T3I = T3y + T3z;
Chris@82 245 T3L = T3E + T3D;
Chris@82 246 T3H = W[6];
Chris@82 247 T3J = T3H * T3I;
Chris@82 248 T3M = T3H * T3L;
Chris@82 249 T3K = W[7];
Chris@82 250 cr[WS(rs, 4)] = FNMS(T3K, T3L, T3J);
Chris@82 251 ci[WS(rs, 4)] = FMA(T3K, T3I, T3M);
Chris@82 252 }
Chris@82 253 {
Chris@82 254 E T38, T3g, T3d, T3j, T37, T3c;
Chris@82 255 T37 = T2V - T2U;
Chris@82 256 T38 = FNMS(KP707106781, T37, T36);
Chris@82 257 T3g = FMA(KP707106781, T37, T36);
Chris@82 258 T3c = T2F - T2K;
Chris@82 259 T3d = FNMS(KP707106781, T3c, T3b);
Chris@82 260 T3j = FMA(KP707106781, T3c, T3b);
Chris@82 261 {
Chris@82 262 E T39, T3e, T35, T3a;
Chris@82 263 T35 = W[26];
Chris@82 264 T39 = T35 * T38;
Chris@82 265 T3e = T35 * T3d;
Chris@82 266 T3a = W[27];
Chris@82 267 cr[WS(rs, 14)] = FNMS(T3a, T3d, T39);
Chris@82 268 ci[WS(rs, 14)] = FMA(T3a, T38, T3e);
Chris@82 269 }
Chris@82 270 {
Chris@82 271 E T3h, T3k, T3f, T3i;
Chris@82 272 T3f = W[10];
Chris@82 273 T3h = T3f * T3g;
Chris@82 274 T3k = T3f * T3j;
Chris@82 275 T3i = W[11];
Chris@82 276 cr[WS(rs, 6)] = FNMS(T3i, T3j, T3h);
Chris@82 277 ci[WS(rs, 6)] = FMA(T3i, T3g, T3k);
Chris@82 278 }
Chris@82 279 }
Chris@82 280 {
Chris@82 281 E T2M, T30, T2X, T33, T2L, T2T, T2W;
Chris@82 282 T2L = T2F + T2K;
Chris@82 283 T2M = FNMS(KP707106781, T2L, T2A);
Chris@82 284 T30 = FMA(KP707106781, T2L, T2A);
Chris@82 285 T2T = T2P + T2S;
Chris@82 286 T2W = T2U + T2V;
Chris@82 287 T2X = FNMS(KP707106781, T2W, T2T);
Chris@82 288 T33 = FMA(KP707106781, T2W, T2T);
Chris@82 289 {
Chris@82 290 E T2v, T2N, T2O, T2Y;
Chris@82 291 T2v = W[18];
Chris@82 292 T2N = T2v * T2M;
Chris@82 293 T2O = W[19];
Chris@82 294 T2Y = T2O * T2M;
Chris@82 295 cr[WS(rs, 10)] = FNMS(T2O, T2X, T2N);
Chris@82 296 ci[WS(rs, 10)] = FMA(T2v, T2X, T2Y);
Chris@82 297 }
Chris@82 298 {
Chris@82 299 E T2Z, T31, T32, T34;
Chris@82 300 T2Z = W[2];
Chris@82 301 T31 = T2Z * T30;
Chris@82 302 T32 = W[3];
Chris@82 303 T34 = T32 * T30;
Chris@82 304 cr[WS(rs, 2)] = FNMS(T32, T33, T31);
Chris@82 305 ci[WS(rs, 2)] = FMA(T2Z, T33, T34);
Chris@82 306 }
Chris@82 307 }
Chris@82 308 {
Chris@82 309 E T1Y, T2a, T27, T2d;
Chris@82 310 {
Chris@82 311 E T1Q, T1X, T23, T26;
Chris@82 312 T1Q = FNMS(KP707106781, T1P, T1O);
Chris@82 313 T1X = T1T + T1W;
Chris@82 314 T1Y = FMA(KP923879532, T1X, T1Q);
Chris@82 315 T2a = FNMS(KP923879532, T1X, T1Q);
Chris@82 316 T23 = FMA(KP707106781, T22, T21);
Chris@82 317 T26 = T24 - T25;
Chris@82 318 T27 = FNMS(KP923879532, T26, T23);
Chris@82 319 T2d = FMA(KP923879532, T26, T23);
Chris@82 320 }
Chris@82 321 {
Chris@82 322 E T1N, T1Z, T20, T28;
Chris@82 323 T1N = W[20];
Chris@82 324 T1Z = T1N * T1Y;
Chris@82 325 T20 = W[21];
Chris@82 326 T28 = T20 * T1Y;
Chris@82 327 cr[WS(rs, 11)] = FNMS(T20, T27, T1Z);
Chris@82 328 ci[WS(rs, 11)] = FMA(T1N, T27, T28);
Chris@82 329 }
Chris@82 330 {
Chris@82 331 E T29, T2b, T2c, T2e;
Chris@82 332 T29 = W[4];
Chris@82 333 T2b = T29 * T2a;
Chris@82 334 T2c = W[5];
Chris@82 335 T2e = T2c * T2a;
Chris@82 336 cr[WS(rs, 3)] = FNMS(T2c, T2d, T2b);
Chris@82 337 ci[WS(rs, 3)] = FMA(T29, T2d, T2e);
Chris@82 338 }
Chris@82 339 }
Chris@82 340 {
Chris@82 341 E T1a, T1s, T1p, T1v;
Chris@82 342 {
Chris@82 343 E TM, T19, T1l, T1o;
Chris@82 344 TM = FNMS(KP707106781, TL, TA);
Chris@82 345 T19 = TX - T18;
Chris@82 346 T1a = FNMS(KP923879532, T19, TM);
Chris@82 347 T1s = FMA(KP923879532, T19, TM);
Chris@82 348 T1l = FNMS(KP707106781, T1k, T1h);
Chris@82 349 T1o = T1m - T1n;
Chris@82 350 T1p = FNMS(KP923879532, T1o, T1l);
Chris@82 351 T1v = FMA(KP923879532, T1o, T1l);
Chris@82 352 }
Chris@82 353 {
Chris@82 354 E Tv, T1b, T1c, T1q;
Chris@82 355 Tv = W[24];
Chris@82 356 T1b = Tv * T1a;
Chris@82 357 T1c = W[25];
Chris@82 358 T1q = T1c * T1a;
Chris@82 359 cr[WS(rs, 13)] = FNMS(T1c, T1p, T1b);
Chris@82 360 ci[WS(rs, 13)] = FMA(Tv, T1p, T1q);
Chris@82 361 }
Chris@82 362 {
Chris@82 363 E T1r, T1t, T1u, T1w;
Chris@82 364 T1r = W[8];
Chris@82 365 T1t = T1r * T1s;
Chris@82 366 T1u = W[9];
Chris@82 367 T1w = T1u * T1s;
Chris@82 368 cr[WS(rs, 5)] = FNMS(T1u, T1v, T1t);
Chris@82 369 ci[WS(rs, 5)] = FMA(T1r, T1v, T1w);
Chris@82 370 }
Chris@82 371 }
Chris@82 372 {
Chris@82 373 E T2i, T2q, T2n, T2t;
Chris@82 374 {
Chris@82 375 E T2g, T2h, T2l, T2m;
Chris@82 376 T2g = FMA(KP707106781, T1P, T1O);
Chris@82 377 T2h = T24 + T25;
Chris@82 378 T2i = FNMS(KP923879532, T2h, T2g);
Chris@82 379 T2q = FMA(KP923879532, T2h, T2g);
Chris@82 380 T2l = FNMS(KP707106781, T22, T21);
Chris@82 381 T2m = T1W - T1T;
Chris@82 382 T2n = FMA(KP923879532, T2m, T2l);
Chris@82 383 T2t = FNMS(KP923879532, T2m, T2l);
Chris@82 384 }
Chris@82 385 {
Chris@82 386 E T2j, T2o, T2f, T2k;
Chris@82 387 T2f = W[12];
Chris@82 388 T2j = T2f * T2i;
Chris@82 389 T2o = T2f * T2n;
Chris@82 390 T2k = W[13];
Chris@82 391 cr[WS(rs, 7)] = FNMS(T2k, T2n, T2j);
Chris@82 392 ci[WS(rs, 7)] = FMA(T2k, T2i, T2o);
Chris@82 393 }
Chris@82 394 {
Chris@82 395 E T2r, T2u, T2p, T2s;
Chris@82 396 T2p = W[28];
Chris@82 397 T2r = T2p * T2q;
Chris@82 398 T2u = T2p * T2t;
Chris@82 399 T2s = W[29];
Chris@82 400 cr[WS(rs, 15)] = FNMS(T2s, T2t, T2r);
Chris@82 401 ci[WS(rs, 15)] = FMA(T2s, T2q, T2u);
Chris@82 402 }
Chris@82 403 }
Chris@82 404 {
Chris@82 405 E T1A, T1I, T1F, T1L;
Chris@82 406 {
Chris@82 407 E T1y, T1z, T1D, T1E;
Chris@82 408 T1y = FMA(KP707106781, TL, TA);
Chris@82 409 T1z = T1m + T1n;
Chris@82 410 T1A = FNMS(KP923879532, T1z, T1y);
Chris@82 411 T1I = FMA(KP923879532, T1z, T1y);
Chris@82 412 T1D = FMA(KP707106781, T1k, T1h);
Chris@82 413 T1E = T18 + TX;
Chris@82 414 T1F = FNMS(KP923879532, T1E, T1D);
Chris@82 415 T1L = FMA(KP923879532, T1E, T1D);
Chris@82 416 }
Chris@82 417 {
Chris@82 418 E T1B, T1G, T1x, T1C;
Chris@82 419 T1x = W[16];
Chris@82 420 T1B = T1x * T1A;
Chris@82 421 T1G = T1x * T1F;
Chris@82 422 T1C = W[17];
Chris@82 423 cr[WS(rs, 9)] = FNMS(T1C, T1F, T1B);
Chris@82 424 ci[WS(rs, 9)] = FMA(T1C, T1A, T1G);
Chris@82 425 }
Chris@82 426 {
Chris@82 427 E T1J, T1M, T1H, T1K;
Chris@82 428 T1H = W[0];
Chris@82 429 T1J = T1H * T1I;
Chris@82 430 T1M = T1H * T1L;
Chris@82 431 T1K = W[1];
Chris@82 432 cr[WS(rs, 1)] = FNMS(T1K, T1L, T1J);
Chris@82 433 ci[WS(rs, 1)] = FMA(T1K, T1I, T1M);
Chris@82 434 }
Chris@82 435 }
Chris@82 436 }
Chris@82 437 }
Chris@82 438 }
Chris@82 439
Chris@82 440 static const tw_instr twinstr[] = {
Chris@82 441 {TW_FULL, 1, 16},
Chris@82 442 {TW_NEXT, 1, 0}
Chris@82 443 };
Chris@82 444
Chris@82 445 static const hc2hc_desc desc = { 16, "hb_16", twinstr, &GENUS, {104, 30, 70, 0} };
Chris@82 446
Chris@82 447 void X(codelet_hb_16) (planner *p) {
Chris@82 448 X(khc2hc_register) (p, hb_16, &desc);
Chris@82 449 }
Chris@82 450 #else
Chris@82 451
Chris@82 452 /* Generated by: ../../../genfft/gen_hc2hc.native -compact -variables 4 -pipeline-latency 4 -sign 1 -n 16 -dif -name hb_16 -include rdft/scalar/hb.h */
Chris@82 453
Chris@82 454 /*
Chris@82 455 * This function contains 174 FP additions, 84 FP multiplications,
Chris@82 456 * (or, 136 additions, 46 multiplications, 38 fused multiply/add),
Chris@82 457 * 50 stack variables, 3 constants, and 64 memory accesses
Chris@82 458 */
Chris@82 459 #include "rdft/scalar/hb.h"
Chris@82 460
Chris@82 461 static void hb_16(R *cr, R *ci, const R *W, stride rs, INT mb, INT me, INT ms)
Chris@82 462 {
Chris@82 463 DK(KP382683432, +0.382683432365089771728459984030398866761344562);
Chris@82 464 DK(KP923879532, +0.923879532511286756128183189396788286822416626);
Chris@82 465 DK(KP707106781, +0.707106781186547524400844362104849039284835938);
Chris@82 466 {
Chris@82 467 INT m;
Chris@82 468 for (m = mb, W = W + ((mb - 1) * 30); m < me; m = m + 1, cr = cr + ms, ci = ci - ms, W = W + 30, MAKE_VOLATILE_STRIDE(32, rs)) {
Chris@82 469 E T7, T2K, T2W, Tw, T17, T1S, T2k, T1w, Te, TD, T1x, T10, T2n, T2L, T1Z;
Chris@82 470 E T2X, Tm, T1z, TN, T19, T2e, T2p, T2P, T2Z, Tt, T1A, TW, T1a, T27, T2q;
Chris@82 471 E T2S, T30;
Chris@82 472 {
Chris@82 473 E T3, T1Q, T16, T1R, T6, T2i, T13, T2j;
Chris@82 474 {
Chris@82 475 E T1, T2, T14, T15;
Chris@82 476 T1 = cr[0];
Chris@82 477 T2 = ci[WS(rs, 7)];
Chris@82 478 T3 = T1 + T2;
Chris@82 479 T1Q = T1 - T2;
Chris@82 480 T14 = ci[WS(rs, 11)];
Chris@82 481 T15 = cr[WS(rs, 12)];
Chris@82 482 T16 = T14 - T15;
Chris@82 483 T1R = T14 + T15;
Chris@82 484 }
Chris@82 485 {
Chris@82 486 E T4, T5, T11, T12;
Chris@82 487 T4 = cr[WS(rs, 4)];
Chris@82 488 T5 = ci[WS(rs, 3)];
Chris@82 489 T6 = T4 + T5;
Chris@82 490 T2i = T4 - T5;
Chris@82 491 T11 = ci[WS(rs, 15)];
Chris@82 492 T12 = cr[WS(rs, 8)];
Chris@82 493 T13 = T11 - T12;
Chris@82 494 T2j = T11 + T12;
Chris@82 495 }
Chris@82 496 T7 = T3 + T6;
Chris@82 497 T2K = T1Q + T1R;
Chris@82 498 T2W = T2j - T2i;
Chris@82 499 Tw = T3 - T6;
Chris@82 500 T17 = T13 - T16;
Chris@82 501 T1S = T1Q - T1R;
Chris@82 502 T2k = T2i + T2j;
Chris@82 503 T1w = T13 + T16;
Chris@82 504 }
Chris@82 505 {
Chris@82 506 E Ta, T1T, TC, T1U, Td, T1W, Tz, T1X;
Chris@82 507 {
Chris@82 508 E T8, T9, TA, TB;
Chris@82 509 T8 = cr[WS(rs, 2)];
Chris@82 510 T9 = ci[WS(rs, 5)];
Chris@82 511 Ta = T8 + T9;
Chris@82 512 T1T = T8 - T9;
Chris@82 513 TA = ci[WS(rs, 13)];
Chris@82 514 TB = cr[WS(rs, 10)];
Chris@82 515 TC = TA - TB;
Chris@82 516 T1U = TA + TB;
Chris@82 517 }
Chris@82 518 {
Chris@82 519 E Tb, Tc, Tx, Ty;
Chris@82 520 Tb = ci[WS(rs, 1)];
Chris@82 521 Tc = cr[WS(rs, 6)];
Chris@82 522 Td = Tb + Tc;
Chris@82 523 T1W = Tb - Tc;
Chris@82 524 Tx = ci[WS(rs, 9)];
Chris@82 525 Ty = cr[WS(rs, 14)];
Chris@82 526 Tz = Tx - Ty;
Chris@82 527 T1X = Tx + Ty;
Chris@82 528 }
Chris@82 529 Te = Ta + Td;
Chris@82 530 TD = Tz - TC;
Chris@82 531 T1x = TC + Tz;
Chris@82 532 T10 = Ta - Td;
Chris@82 533 {
Chris@82 534 E T2l, T2m, T1V, T1Y;
Chris@82 535 T2l = T1T + T1U;
Chris@82 536 T2m = T1W + T1X;
Chris@82 537 T2n = KP707106781 * (T2l - T2m);
Chris@82 538 T2L = KP707106781 * (T2l + T2m);
Chris@82 539 T1V = T1T - T1U;
Chris@82 540 T1Y = T1W - T1X;
Chris@82 541 T1Z = KP707106781 * (T1V + T1Y);
Chris@82 542 T2X = KP707106781 * (T1V - T1Y);
Chris@82 543 }
Chris@82 544 }
Chris@82 545 {
Chris@82 546 E Ti, T2b, TL, T2c, Tl, T28, TI, T29, TF, TM;
Chris@82 547 {
Chris@82 548 E Tg, Th, TJ, TK;
Chris@82 549 Tg = cr[WS(rs, 1)];
Chris@82 550 Th = ci[WS(rs, 6)];
Chris@82 551 Ti = Tg + Th;
Chris@82 552 T2b = Tg - Th;
Chris@82 553 TJ = ci[WS(rs, 10)];
Chris@82 554 TK = cr[WS(rs, 13)];
Chris@82 555 TL = TJ - TK;
Chris@82 556 T2c = TJ + TK;
Chris@82 557 }
Chris@82 558 {
Chris@82 559 E Tj, Tk, TG, TH;
Chris@82 560 Tj = cr[WS(rs, 5)];
Chris@82 561 Tk = ci[WS(rs, 2)];
Chris@82 562 Tl = Tj + Tk;
Chris@82 563 T28 = Tj - Tk;
Chris@82 564 TG = ci[WS(rs, 14)];
Chris@82 565 TH = cr[WS(rs, 9)];
Chris@82 566 TI = TG - TH;
Chris@82 567 T29 = TG + TH;
Chris@82 568 }
Chris@82 569 Tm = Ti + Tl;
Chris@82 570 T1z = TI + TL;
Chris@82 571 TF = Ti - Tl;
Chris@82 572 TM = TI - TL;
Chris@82 573 TN = TF - TM;
Chris@82 574 T19 = TF + TM;
Chris@82 575 {
Chris@82 576 E T2a, T2d, T2N, T2O;
Chris@82 577 T2a = T28 + T29;
Chris@82 578 T2d = T2b - T2c;
Chris@82 579 T2e = FMA(KP923879532, T2a, KP382683432 * T2d);
Chris@82 580 T2p = FNMS(KP382683432, T2a, KP923879532 * T2d);
Chris@82 581 T2N = T2b + T2c;
Chris@82 582 T2O = T29 - T28;
Chris@82 583 T2P = FNMS(KP923879532, T2O, KP382683432 * T2N);
Chris@82 584 T2Z = FMA(KP382683432, T2O, KP923879532 * T2N);
Chris@82 585 }
Chris@82 586 }
Chris@82 587 {
Chris@82 588 E Tp, T24, TU, T25, Ts, T21, TR, T22, TO, TV;
Chris@82 589 {
Chris@82 590 E Tn, To, TS, TT;
Chris@82 591 Tn = ci[0];
Chris@82 592 To = cr[WS(rs, 7)];
Chris@82 593 Tp = Tn + To;
Chris@82 594 T24 = Tn - To;
Chris@82 595 TS = ci[WS(rs, 12)];
Chris@82 596 TT = cr[WS(rs, 11)];
Chris@82 597 TU = TS - TT;
Chris@82 598 T25 = TS + TT;
Chris@82 599 }
Chris@82 600 {
Chris@82 601 E Tq, Tr, TP, TQ;
Chris@82 602 Tq = cr[WS(rs, 3)];
Chris@82 603 Tr = ci[WS(rs, 4)];
Chris@82 604 Ts = Tq + Tr;
Chris@82 605 T21 = Tq - Tr;
Chris@82 606 TP = ci[WS(rs, 8)];
Chris@82 607 TQ = cr[WS(rs, 15)];
Chris@82 608 TR = TP - TQ;
Chris@82 609 T22 = TP + TQ;
Chris@82 610 }
Chris@82 611 Tt = Tp + Ts;
Chris@82 612 T1A = TR + TU;
Chris@82 613 TO = Tp - Ts;
Chris@82 614 TV = TR - TU;
Chris@82 615 TW = TO + TV;
Chris@82 616 T1a = TV - TO;
Chris@82 617 {
Chris@82 618 E T23, T26, T2Q, T2R;
Chris@82 619 T23 = T21 - T22;
Chris@82 620 T26 = T24 - T25;
Chris@82 621 T27 = FNMS(KP382683432, T26, KP923879532 * T23);
Chris@82 622 T2q = FMA(KP382683432, T23, KP923879532 * T26);
Chris@82 623 T2Q = T24 + T25;
Chris@82 624 T2R = T21 + T22;
Chris@82 625 T2S = FNMS(KP923879532, T2R, KP382683432 * T2Q);
Chris@82 626 T30 = FMA(KP382683432, T2R, KP923879532 * T2Q);
Chris@82 627 }
Chris@82 628 }
Chris@82 629 {
Chris@82 630 E Tf, Tu, T1u, T1y, T1B, T1C, T1t, T1v;
Chris@82 631 Tf = T7 + Te;
Chris@82 632 Tu = Tm + Tt;
Chris@82 633 T1u = Tf - Tu;
Chris@82 634 T1y = T1w + T1x;
Chris@82 635 T1B = T1z + T1A;
Chris@82 636 T1C = T1y - T1B;
Chris@82 637 cr[0] = Tf + Tu;
Chris@82 638 ci[0] = T1y + T1B;
Chris@82 639 T1t = W[14];
Chris@82 640 T1v = W[15];
Chris@82 641 cr[WS(rs, 8)] = FNMS(T1v, T1C, T1t * T1u);
Chris@82 642 ci[WS(rs, 8)] = FMA(T1v, T1u, T1t * T1C);
Chris@82 643 }
Chris@82 644 {
Chris@82 645 E T2U, T34, T32, T36;
Chris@82 646 {
Chris@82 647 E T2M, T2T, T2Y, T31;
Chris@82 648 T2M = T2K - T2L;
Chris@82 649 T2T = T2P + T2S;
Chris@82 650 T2U = T2M - T2T;
Chris@82 651 T34 = T2M + T2T;
Chris@82 652 T2Y = T2W + T2X;
Chris@82 653 T31 = T2Z - T30;
Chris@82 654 T32 = T2Y - T31;
Chris@82 655 T36 = T2Y + T31;
Chris@82 656 }
Chris@82 657 {
Chris@82 658 E T2J, T2V, T33, T35;
Chris@82 659 T2J = W[20];
Chris@82 660 T2V = W[21];
Chris@82 661 cr[WS(rs, 11)] = FNMS(T2V, T32, T2J * T2U);
Chris@82 662 ci[WS(rs, 11)] = FMA(T2V, T2U, T2J * T32);
Chris@82 663 T33 = W[4];
Chris@82 664 T35 = W[5];
Chris@82 665 cr[WS(rs, 3)] = FNMS(T35, T36, T33 * T34);
Chris@82 666 ci[WS(rs, 3)] = FMA(T35, T34, T33 * T36);
Chris@82 667 }
Chris@82 668 }
Chris@82 669 {
Chris@82 670 E T3a, T3g, T3e, T3i;
Chris@82 671 {
Chris@82 672 E T38, T39, T3c, T3d;
Chris@82 673 T38 = T2K + T2L;
Chris@82 674 T39 = T2Z + T30;
Chris@82 675 T3a = T38 - T39;
Chris@82 676 T3g = T38 + T39;
Chris@82 677 T3c = T2W - T2X;
Chris@82 678 T3d = T2P - T2S;
Chris@82 679 T3e = T3c + T3d;
Chris@82 680 T3i = T3c - T3d;
Chris@82 681 }
Chris@82 682 {
Chris@82 683 E T37, T3b, T3f, T3h;
Chris@82 684 T37 = W[12];
Chris@82 685 T3b = W[13];
Chris@82 686 cr[WS(rs, 7)] = FNMS(T3b, T3e, T37 * T3a);
Chris@82 687 ci[WS(rs, 7)] = FMA(T37, T3e, T3b * T3a);
Chris@82 688 T3f = W[28];
Chris@82 689 T3h = W[29];
Chris@82 690 cr[WS(rs, 15)] = FNMS(T3h, T3i, T3f * T3g);
Chris@82 691 ci[WS(rs, 15)] = FMA(T3f, T3i, T3h * T3g);
Chris@82 692 }
Chris@82 693 }
Chris@82 694 {
Chris@82 695 E TY, T1e, T1c, T1g;
Chris@82 696 {
Chris@82 697 E TE, TX, T18, T1b;
Chris@82 698 TE = Tw + TD;
Chris@82 699 TX = KP707106781 * (TN + TW);
Chris@82 700 TY = TE - TX;
Chris@82 701 T1e = TE + TX;
Chris@82 702 T18 = T10 + T17;
Chris@82 703 T1b = KP707106781 * (T19 + T1a);
Chris@82 704 T1c = T18 - T1b;
Chris@82 705 T1g = T18 + T1b;
Chris@82 706 }
Chris@82 707 {
Chris@82 708 E Tv, TZ, T1d, T1f;
Chris@82 709 Tv = W[18];
Chris@82 710 TZ = W[19];
Chris@82 711 cr[WS(rs, 10)] = FNMS(TZ, T1c, Tv * TY);
Chris@82 712 ci[WS(rs, 10)] = FMA(TZ, TY, Tv * T1c);
Chris@82 713 T1d = W[2];
Chris@82 714 T1f = W[3];
Chris@82 715 cr[WS(rs, 2)] = FNMS(T1f, T1g, T1d * T1e);
Chris@82 716 ci[WS(rs, 2)] = FMA(T1f, T1e, T1d * T1g);
Chris@82 717 }
Chris@82 718 }
Chris@82 719 {
Chris@82 720 E T1k, T1q, T1o, T1s;
Chris@82 721 {
Chris@82 722 E T1i, T1j, T1m, T1n;
Chris@82 723 T1i = Tw - TD;
Chris@82 724 T1j = KP707106781 * (T1a - T19);
Chris@82 725 T1k = T1i - T1j;
Chris@82 726 T1q = T1i + T1j;
Chris@82 727 T1m = T17 - T10;
Chris@82 728 T1n = KP707106781 * (TN - TW);
Chris@82 729 T1o = T1m - T1n;
Chris@82 730 T1s = T1m + T1n;
Chris@82 731 }
Chris@82 732 {
Chris@82 733 E T1h, T1l, T1p, T1r;
Chris@82 734 T1h = W[26];
Chris@82 735 T1l = W[27];
Chris@82 736 cr[WS(rs, 14)] = FNMS(T1l, T1o, T1h * T1k);
Chris@82 737 ci[WS(rs, 14)] = FMA(T1h, T1o, T1l * T1k);
Chris@82 738 T1p = W[10];
Chris@82 739 T1r = W[11];
Chris@82 740 cr[WS(rs, 6)] = FNMS(T1r, T1s, T1p * T1q);
Chris@82 741 ci[WS(rs, 6)] = FMA(T1p, T1s, T1r * T1q);
Chris@82 742 }
Chris@82 743 }
Chris@82 744 {
Chris@82 745 E T2g, T2u, T2s, T2w;
Chris@82 746 {
Chris@82 747 E T20, T2f, T2o, T2r;
Chris@82 748 T20 = T1S - T1Z;
Chris@82 749 T2f = T27 - T2e;
Chris@82 750 T2g = T20 - T2f;
Chris@82 751 T2u = T20 + T2f;
Chris@82 752 T2o = T2k - T2n;
Chris@82 753 T2r = T2p - T2q;
Chris@82 754 T2s = T2o - T2r;
Chris@82 755 T2w = T2o + T2r;
Chris@82 756 }
Chris@82 757 {
Chris@82 758 E T1P, T2h, T2t, T2v;
Chris@82 759 T1P = W[24];
Chris@82 760 T2h = W[25];
Chris@82 761 cr[WS(rs, 13)] = FNMS(T2h, T2s, T1P * T2g);
Chris@82 762 ci[WS(rs, 13)] = FMA(T2h, T2g, T1P * T2s);
Chris@82 763 T2t = W[8];
Chris@82 764 T2v = W[9];
Chris@82 765 cr[WS(rs, 5)] = FNMS(T2v, T2w, T2t * T2u);
Chris@82 766 ci[WS(rs, 5)] = FMA(T2v, T2u, T2t * T2w);
Chris@82 767 }
Chris@82 768 }
Chris@82 769 {
Chris@82 770 E T2A, T2G, T2E, T2I;
Chris@82 771 {
Chris@82 772 E T2y, T2z, T2C, T2D;
Chris@82 773 T2y = T1S + T1Z;
Chris@82 774 T2z = T2p + T2q;
Chris@82 775 T2A = T2y - T2z;
Chris@82 776 T2G = T2y + T2z;
Chris@82 777 T2C = T2k + T2n;
Chris@82 778 T2D = T2e + T27;
Chris@82 779 T2E = T2C - T2D;
Chris@82 780 T2I = T2C + T2D;
Chris@82 781 }
Chris@82 782 {
Chris@82 783 E T2x, T2B, T2F, T2H;
Chris@82 784 T2x = W[16];
Chris@82 785 T2B = W[17];
Chris@82 786 cr[WS(rs, 9)] = FNMS(T2B, T2E, T2x * T2A);
Chris@82 787 ci[WS(rs, 9)] = FMA(T2x, T2E, T2B * T2A);
Chris@82 788 T2F = W[0];
Chris@82 789 T2H = W[1];
Chris@82 790 cr[WS(rs, 1)] = FNMS(T2H, T2I, T2F * T2G);
Chris@82 791 ci[WS(rs, 1)] = FMA(T2F, T2I, T2H * T2G);
Chris@82 792 }
Chris@82 793 }
Chris@82 794 {
Chris@82 795 E T1G, T1M, T1K, T1O;
Chris@82 796 {
Chris@82 797 E T1E, T1F, T1I, T1J;
Chris@82 798 T1E = T7 - Te;
Chris@82 799 T1F = T1A - T1z;
Chris@82 800 T1G = T1E - T1F;
Chris@82 801 T1M = T1E + T1F;
Chris@82 802 T1I = T1w - T1x;
Chris@82 803 T1J = Tm - Tt;
Chris@82 804 T1K = T1I - T1J;
Chris@82 805 T1O = T1J + T1I;
Chris@82 806 }
Chris@82 807 {
Chris@82 808 E T1D, T1H, T1L, T1N;
Chris@82 809 T1D = W[22];
Chris@82 810 T1H = W[23];
Chris@82 811 cr[WS(rs, 12)] = FNMS(T1H, T1K, T1D * T1G);
Chris@82 812 ci[WS(rs, 12)] = FMA(T1D, T1K, T1H * T1G);
Chris@82 813 T1L = W[6];
Chris@82 814 T1N = W[7];
Chris@82 815 cr[WS(rs, 4)] = FNMS(T1N, T1O, T1L * T1M);
Chris@82 816 ci[WS(rs, 4)] = FMA(T1L, T1O, T1N * T1M);
Chris@82 817 }
Chris@82 818 }
Chris@82 819 }
Chris@82 820 }
Chris@82 821 }
Chris@82 822
Chris@82 823 static const tw_instr twinstr[] = {
Chris@82 824 {TW_FULL, 1, 16},
Chris@82 825 {TW_NEXT, 1, 0}
Chris@82 826 };
Chris@82 827
Chris@82 828 static const hc2hc_desc desc = { 16, "hb_16", twinstr, &GENUS, {136, 46, 38, 0} };
Chris@82 829
Chris@82 830 void X(codelet_hb_16) (planner *p) {
Chris@82 831 X(khc2hc_register) (p, hb_16, &desc);
Chris@82 832 }
Chris@82 833 #endif