Mercurial > hg > sv-dependency-builds

annotate src/fftw-3.3.8/rdft/scalar/r2cb/hb_12.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 12 -dif -name hb_12 -include rdft/scalar/hb.h */
Chris@82 29
Chris@82 30 /*
Chris@82 31 * This function contains 118 FP additions, 68 FP multiplications,
Chris@82 32 * (or, 72 additions, 22 multiplications, 46 fused multiply/add),
Chris@82 33 * 47 stack variables, 2 constants, and 48 memory accesses
Chris@82 34 */
Chris@82 35 #include "rdft/scalar/hb.h"
Chris@82 36
Chris@82 37 static void hb_12(R *cr, R *ci, const R *W, stride rs, INT mb, INT me, INT ms)
Chris@82 38 {
Chris@82 39 DK(KP866025403, +0.866025403784438646763723170752936183471402627);
Chris@82 40 DK(KP500000000, +0.500000000000000000000000000000000000000000000);
Chris@82 41 {
Chris@82 42 INT m;
Chris@82 43 for (m = mb, W = W + ((mb - 1) * 22); m < me; m = m + 1, cr = cr + ms, ci = ci - ms, W = W + 22, MAKE_VOLATILE_STRIDE(24, rs)) {
Chris@82 44 E T18, T20, T1b, T21, T1s, T2a, T1p, T29, TI, TN, TO, Tb, To, T1f, T23;
Chris@82 45 E T1i, T24, T1z, T2d, T1w, T2c, Tt, Ty, Tz, Tm, TD;
Chris@82 46 {
Chris@82 47 E T1, TE, TM, T6, T4, T1o, TH, T17, TL, T1a, T9, T1r;
Chris@82 48 T1 = cr[0];
Chris@82 49 TE = ci[WS(rs, 11)];
Chris@82 50 TM = cr[WS(rs, 6)];
Chris@82 51 T6 = ci[WS(rs, 5)];
Chris@82 52 {
Chris@82 53 E T2, T3, TF, TG;
Chris@82 54 T2 = cr[WS(rs, 4)];
Chris@82 55 T3 = ci[WS(rs, 3)];
Chris@82 56 T4 = T2 + T3;
Chris@82 57 T1o = T2 - T3;
Chris@82 58 TF = ci[WS(rs, 7)];
Chris@82 59 TG = cr[WS(rs, 8)];
Chris@82 60 TH = TF - TG;
Chris@82 61 T17 = TF + TG;
Chris@82 62 }
Chris@82 63 {
Chris@82 64 E TJ, TK, T7, T8;
Chris@82 65 TJ = ci[WS(rs, 9)];
Chris@82 66 TK = cr[WS(rs, 10)];
Chris@82 67 TL = TJ - TK;
Chris@82 68 T1a = TJ + TK;
Chris@82 69 T7 = ci[WS(rs, 1)];
Chris@82 70 T8 = cr[WS(rs, 2)];
Chris@82 71 T9 = T7 + T8;
Chris@82 72 T1r = T7 - T8;
Chris@82 73 }
Chris@82 74 {
Chris@82 75 E T16, T19, T1q, T1n, T5, Ta;
Chris@82 76 T16 = FNMS(KP500000000, T4, T1);
Chris@82 77 T18 = FNMS(KP866025403, T17, T16);
Chris@82 78 T20 = FMA(KP866025403, T17, T16);
Chris@82 79 T19 = FNMS(KP500000000, T9, T6);
Chris@82 80 T1b = FMA(KP866025403, T1a, T19);
Chris@82 81 T21 = FNMS(KP866025403, T1a, T19);
Chris@82 82 T1q = FMA(KP500000000, TL, TM);
Chris@82 83 T1s = FNMS(KP866025403, T1r, T1q);
Chris@82 84 T2a = FMA(KP866025403, T1r, T1q);
Chris@82 85 T1n = FNMS(KP500000000, TH, TE);
Chris@82 86 T1p = FMA(KP866025403, T1o, T1n);
Chris@82 87 T29 = FNMS(KP866025403, T1o, T1n);
Chris@82 88 TI = TE + TH;
Chris@82 89 TN = TL - TM;
Chris@82 90 TO = TI - TN;
Chris@82 91 T5 = T1 + T4;
Chris@82 92 Ta = T6 + T9;
Chris@82 93 Tb = T5 + Ta;
Chris@82 94 To = T5 - Ta;
Chris@82 95 }
Chris@82 96 }
Chris@82 97 {
Chris@82 98 E Tc, Tp, Tx, Th, Tf, T1v, Ts, T1e, Tw, T1h, Tk, T1y;
Chris@82 99 Tc = cr[WS(rs, 3)];
Chris@82 100 Tp = ci[WS(rs, 8)];
Chris@82 101 Tx = cr[WS(rs, 9)];
Chris@82 102 Th = ci[WS(rs, 2)];
Chris@82 103 {
Chris@82 104 E Td, Te, Tq, Tr;
Chris@82 105 Td = ci[WS(rs, 4)];
Chris@82 106 Te = ci[0];
Chris@82 107 Tf = Td + Te;
Chris@82 108 T1v = Td - Te;
Chris@82 109 Tq = cr[WS(rs, 7)];
Chris@82 110 Tr = cr[WS(rs, 11)];
Chris@82 111 Ts = Tq + Tr;
Chris@82 112 T1e = Tq - Tr;
Chris@82 113 }
Chris@82 114 {
Chris@82 115 E Tu, Tv, Ti, Tj;
Chris@82 116 Tu = ci[WS(rs, 10)];
Chris@82 117 Tv = ci[WS(rs, 6)];
Chris@82 118 Tw = Tu + Tv;
Chris@82 119 T1h = Tv - Tu;
Chris@82 120 Ti = cr[WS(rs, 1)];
Chris@82 121 Tj = cr[WS(rs, 5)];
Chris@82 122 Tk = Ti + Tj;
Chris@82 123 T1y = Ti - Tj;
Chris@82 124 }
Chris@82 125 {
Chris@82 126 E T1d, T1g, T1x, T1u, Tg, Tl;
Chris@82 127 T1d = FNMS(KP500000000, Tf, Tc);
Chris@82 128 T1f = FMA(KP866025403, T1e, T1d);
Chris@82 129 T23 = FNMS(KP866025403, T1e, T1d);
Chris@82 130 T1g = FNMS(KP500000000, Tk, Th);
Chris@82 131 T1i = FMA(KP866025403, T1h, T1g);
Chris@82 132 T24 = FNMS(KP866025403, T1h, T1g);
Chris@82 133 T1x = FMA(KP500000000, Tw, Tx);
Chris@82 134 T1z = FNMS(KP866025403, T1y, T1x);
Chris@82 135 T2d = FMA(KP866025403, T1y, T1x);
Chris@82 136 T1u = FMA(KP500000000, Ts, Tp);
Chris@82 137 T1w = FMA(KP866025403, T1v, T1u);
Chris@82 138 T2c = FNMS(KP866025403, T1v, T1u);
Chris@82 139 Tt = Tp - Ts;
Chris@82 140 Ty = Tw - Tx;
Chris@82 141 Tz = Tt - Ty;
Chris@82 142 Tg = Tc + Tf;
Chris@82 143 Tl = Th + Tk;
Chris@82 144 Tm = Tg + Tl;
Chris@82 145 TD = Tg - Tl;
Chris@82 146 }
Chris@82 147 }
Chris@82 148 cr[0] = Tb + Tm;
Chris@82 149 {
Chris@82 150 E TA, TP, TB, TQ, Tn, TC;
Chris@82 151 TA = To - Tz;
Chris@82 152 TP = TD + TO;
Chris@82 153 Tn = W[16];
Chris@82 154 TB = Tn * TA;
Chris@82 155 TQ = Tn * TP;
Chris@82 156 TC = W[17];
Chris@82 157 cr[WS(rs, 9)] = FNMS(TC, TP, TB);
Chris@82 158 ci[WS(rs, 9)] = FMA(TC, TA, TQ);
Chris@82 159 }
Chris@82 160 {
Chris@82 161 E TS, TV, TT, TW, TR, TU;
Chris@82 162 TS = To + Tz;
Chris@82 163 TV = TO - TD;
Chris@82 164 TR = W[4];
Chris@82 165 TT = TR * TS;
Chris@82 166 TW = TR * TV;
Chris@82 167 TU = W[5];
Chris@82 168 cr[WS(rs, 3)] = FNMS(TU, TV, TT);
Chris@82 169 ci[WS(rs, 3)] = FMA(TU, TS, TW);
Chris@82 170 }
Chris@82 171 {
Chris@82 172 E T11, T12, T13, TX, TZ, T10, T14, TY;
Chris@82 173 T11 = TI + TN;
Chris@82 174 T12 = Tt + Ty;
Chris@82 175 T13 = T11 - T12;
Chris@82 176 TY = Tb - Tm;
Chris@82 177 TX = W[10];
Chris@82 178 TZ = TX * TY;
Chris@82 179 T10 = W[11];
Chris@82 180 T14 = T10 * TY;
Chris@82 181 ci[0] = T11 + T12;
Chris@82 182 ci[WS(rs, 6)] = FMA(TX, T13, T14);
Chris@82 183 cr[WS(rs, 6)] = FNMS(T10, T13, TZ);
Chris@82 184 }
Chris@82 185 {
Chris@82 186 E T1k, T1E, T1B, T1H;
Chris@82 187 {
Chris@82 188 E T1c, T1j, T1t, T1A;
Chris@82 189 T1c = T18 + T1b;
Chris@82 190 T1j = T1f + T1i;
Chris@82 191 T1k = T1c - T1j;
Chris@82 192 T1E = T1c + T1j;
Chris@82 193 T1t = T1p - T1s;
Chris@82 194 T1A = T1w - T1z;
Chris@82 195 T1B = T1t - T1A;
Chris@82 196 T1H = T1t + T1A;
Chris@82 197 }
Chris@82 198 {
Chris@82 199 E T15, T1l, T1m, T1C;
Chris@82 200 T15 = W[18];
Chris@82 201 T1l = T15 * T1k;
Chris@82 202 T1m = W[19];
Chris@82 203 T1C = T1m * T1k;
Chris@82 204 cr[WS(rs, 10)] = FNMS(T1m, T1B, T1l);
Chris@82 205 ci[WS(rs, 10)] = FMA(T15, T1B, T1C);
Chris@82 206 }
Chris@82 207 {
Chris@82 208 E T1D, T1F, T1G, T1I;
Chris@82 209 T1D = W[6];
Chris@82 210 T1F = T1D * T1E;
Chris@82 211 T1G = W[7];
Chris@82 212 T1I = T1G * T1E;
Chris@82 213 cr[WS(rs, 4)] = FNMS(T1G, T1H, T1F);
Chris@82 214 ci[WS(rs, 4)] = FMA(T1D, T1H, T1I);
Chris@82 215 }
Chris@82 216 }
Chris@82 217 {
Chris@82 218 E T26, T2i, T2f, T2l;
Chris@82 219 {
Chris@82 220 E T22, T25, T2b, T2e;
Chris@82 221 T22 = T20 + T21;
Chris@82 222 T25 = T23 + T24;
Chris@82 223 T26 = T22 - T25;
Chris@82 224 T2i = T22 + T25;
Chris@82 225 T2b = T29 - T2a;
Chris@82 226 T2e = T2c - T2d;
Chris@82 227 T2f = T2b - T2e;
Chris@82 228 T2l = T2b + T2e;
Chris@82 229 }
Chris@82 230 {
Chris@82 231 E T1Z, T27, T28, T2g;
Chris@82 232 T1Z = W[2];
Chris@82 233 T27 = T1Z * T26;
Chris@82 234 T28 = W[3];
Chris@82 235 T2g = T28 * T26;
Chris@82 236 cr[WS(rs, 2)] = FNMS(T28, T2f, T27);
Chris@82 237 ci[WS(rs, 2)] = FMA(T1Z, T2f, T2g);
Chris@82 238 }
Chris@82 239 {
Chris@82 240 E T2h, T2j, T2k, T2m;
Chris@82 241 T2h = W[14];
Chris@82 242 T2j = T2h * T2i;
Chris@82 243 T2k = W[15];
Chris@82 244 T2m = T2k * T2i;
Chris@82 245 cr[WS(rs, 8)] = FNMS(T2k, T2l, T2j);
Chris@82 246 ci[WS(rs, 8)] = FMA(T2h, T2l, T2m);
Chris@82 247 }
Chris@82 248 }
Chris@82 249 {
Chris@82 250 E T2q, T2y, T2v, T2B;
Chris@82 251 {
Chris@82 252 E T2o, T2p, T2t, T2u;
Chris@82 253 T2o = T20 - T21;
Chris@82 254 T2p = T2c + T2d;
Chris@82 255 T2q = T2o - T2p;
Chris@82 256 T2y = T2o + T2p;
Chris@82 257 T2t = T29 + T2a;
Chris@82 258 T2u = T23 - T24;
Chris@82 259 T2v = T2t + T2u;
Chris@82 260 T2B = T2t - T2u;
Chris@82 261 }
Chris@82 262 {
Chris@82 263 E T2r, T2w, T2n, T2s;
Chris@82 264 T2n = W[8];
Chris@82 265 T2r = T2n * T2q;
Chris@82 266 T2w = T2n * T2v;
Chris@82 267 T2s = W[9];
Chris@82 268 cr[WS(rs, 5)] = FNMS(T2s, T2v, T2r);
Chris@82 269 ci[WS(rs, 5)] = FMA(T2s, T2q, T2w);
Chris@82 270 }
Chris@82 271 {
Chris@82 272 E T2z, T2C, T2x, T2A;
Chris@82 273 T2x = W[20];
Chris@82 274 T2z = T2x * T2y;
Chris@82 275 T2C = T2x * T2B;
Chris@82 276 T2A = W[21];
Chris@82 277 cr[WS(rs, 11)] = FNMS(T2A, T2B, T2z);
Chris@82 278 ci[WS(rs, 11)] = FMA(T2A, T2y, T2C);
Chris@82 279 }
Chris@82 280 }
Chris@82 281 {
Chris@82 282 E T1M, T1U, T1R, T1X;
Chris@82 283 {
Chris@82 284 E T1K, T1L, T1P, T1Q;
Chris@82 285 T1K = T18 - T1b;
Chris@82 286 T1L = T1w + T1z;
Chris@82 287 T1M = T1K - T1L;
Chris@82 288 T1U = T1K + T1L;
Chris@82 289 T1P = T1p + T1s;
Chris@82 290 T1Q = T1f - T1i;
Chris@82 291 T1R = T1P + T1Q;
Chris@82 292 T1X = T1P - T1Q;
Chris@82 293 }
Chris@82 294 {
Chris@82 295 E T1N, T1S, T1J, T1O;
Chris@82 296 T1J = W[0];
Chris@82 297 T1N = T1J * T1M;
Chris@82 298 T1S = T1J * T1R;
Chris@82 299 T1O = W[1];
Chris@82 300 cr[WS(rs, 1)] = FNMS(T1O, T1R, T1N);
Chris@82 301 ci[WS(rs, 1)] = FMA(T1O, T1M, T1S);
Chris@82 302 }
Chris@82 303 {
Chris@82 304 E T1V, T1Y, T1T, T1W;
Chris@82 305 T1T = W[12];
Chris@82 306 T1V = T1T * T1U;
Chris@82 307 T1Y = T1T * T1X;
Chris@82 308 T1W = W[13];
Chris@82 309 cr[WS(rs, 7)] = FNMS(T1W, T1X, T1V);
Chris@82 310 ci[WS(rs, 7)] = FMA(T1W, T1U, T1Y);
Chris@82 311 }
Chris@82 312 }
Chris@82 313 }
Chris@82 314 }
Chris@82 315 }
Chris@82 316
Chris@82 317 static const tw_instr twinstr[] = {
Chris@82 318 {TW_FULL, 1, 12},
Chris@82 319 {TW_NEXT, 1, 0}
Chris@82 320 };
Chris@82 321
Chris@82 322 static const hc2hc_desc desc = { 12, "hb_12", twinstr, &GENUS, {72, 22, 46, 0} };
Chris@82 323
Chris@82 324 void X(codelet_hb_12) (planner *p) {
Chris@82 325 X(khc2hc_register) (p, hb_12, &desc);
Chris@82 326 }
Chris@82 327 #else
Chris@82 328
Chris@82 329 /* Generated by: ../../../genfft/gen_hc2hc.native -compact -variables 4 -pipeline-latency 4 -sign 1 -n 12 -dif -name hb_12 -include rdft/scalar/hb.h */
Chris@82 330
Chris@82 331 /*
Chris@82 332 * This function contains 118 FP additions, 60 FP multiplications,
Chris@82 333 * (or, 88 additions, 30 multiplications, 30 fused multiply/add),
Chris@82 334 * 39 stack variables, 2 constants, and 48 memory accesses
Chris@82 335 */
Chris@82 336 #include "rdft/scalar/hb.h"
Chris@82 337
Chris@82 338 static void hb_12(R *cr, R *ci, const R *W, stride rs, INT mb, INT me, INT ms)
Chris@82 339 {
Chris@82 340 DK(KP500000000, +0.500000000000000000000000000000000000000000000);
Chris@82 341 DK(KP866025403, +0.866025403784438646763723170752936183471402627);
Chris@82 342 {
Chris@82 343 INT m;
Chris@82 344 for (m = mb, W = W + ((mb - 1) * 22); m < me; m = m + 1, cr = cr + ms, ci = ci - ms, W = W + 22, MAKE_VOLATILE_STRIDE(24, rs)) {
Chris@82 345 E T5, TH, T12, T1M, T1i, T1U, Tg, Tt, T19, T1X, T1p, T1P, Ta, TM, T15;
Chris@82 346 E T1N, T1l, T1V, Tl, Ty, T1c, T1Y, T1s, T1Q;
Chris@82 347 {
Chris@82 348 E T1, TD, T4, T1g, TG, T11, T10, T1h;
Chris@82 349 T1 = cr[0];
Chris@82 350 TD = ci[WS(rs, 11)];
Chris@82 351 {
Chris@82 352 E T2, T3, TE, TF;
Chris@82 353 T2 = cr[WS(rs, 4)];
Chris@82 354 T3 = ci[WS(rs, 3)];
Chris@82 355 T4 = T2 + T3;
Chris@82 356 T1g = KP866025403 * (T2 - T3);
Chris@82 357 TE = ci[WS(rs, 7)];
Chris@82 358 TF = cr[WS(rs, 8)];
Chris@82 359 TG = TE - TF;
Chris@82 360 T11 = KP866025403 * (TE + TF);
Chris@82 361 }
Chris@82 362 T5 = T1 + T4;
Chris@82 363 TH = TD + TG;
Chris@82 364 T10 = FNMS(KP500000000, T4, T1);
Chris@82 365 T12 = T10 - T11;
Chris@82 366 T1M = T10 + T11;
Chris@82 367 T1h = FNMS(KP500000000, TG, TD);
Chris@82 368 T1i = T1g + T1h;
Chris@82 369 T1U = T1h - T1g;
Chris@82 370 }
Chris@82 371 {
Chris@82 372 E Tc, Tp, Tf, T17, Ts, T1o, T18, T1n;
Chris@82 373 Tc = cr[WS(rs, 3)];
Chris@82 374 Tp = ci[WS(rs, 8)];
Chris@82 375 {
Chris@82 376 E Td, Te, Tq, Tr;
Chris@82 377 Td = ci[WS(rs, 4)];
Chris@82 378 Te = ci[0];
Chris@82 379 Tf = Td + Te;
Chris@82 380 T17 = KP866025403 * (Td - Te);
Chris@82 381 Tq = cr[WS(rs, 7)];
Chris@82 382 Tr = cr[WS(rs, 11)];
Chris@82 383 Ts = Tq + Tr;
Chris@82 384 T1o = KP866025403 * (Tq - Tr);
Chris@82 385 }
Chris@82 386 Tg = Tc + Tf;
Chris@82 387 Tt = Tp - Ts;
Chris@82 388 T18 = FMA(KP500000000, Ts, Tp);
Chris@82 389 T19 = T17 + T18;
Chris@82 390 T1X = T18 - T17;
Chris@82 391 T1n = FNMS(KP500000000, Tf, Tc);
Chris@82 392 T1p = T1n + T1o;
Chris@82 393 T1P = T1n - T1o;
Chris@82 394 }
Chris@82 395 {
Chris@82 396 E T6, TL, T9, T1j, TK, T14, T13, T1k;
Chris@82 397 T6 = ci[WS(rs, 5)];
Chris@82 398 TL = cr[WS(rs, 6)];
Chris@82 399 {
Chris@82 400 E T7, T8, TI, TJ;
Chris@82 401 T7 = ci[WS(rs, 1)];
Chris@82 402 T8 = cr[WS(rs, 2)];
Chris@82 403 T9 = T7 + T8;
Chris@82 404 T1j = KP866025403 * (T7 - T8);
Chris@82 405 TI = ci[WS(rs, 9)];
Chris@82 406 TJ = cr[WS(rs, 10)];
Chris@82 407 TK = TI - TJ;
Chris@82 408 T14 = KP866025403 * (TI + TJ);
Chris@82 409 }
Chris@82 410 Ta = T6 + T9;
Chris@82 411 TM = TK - TL;
Chris@82 412 T13 = FNMS(KP500000000, T9, T6);
Chris@82 413 T15 = T13 + T14;
Chris@82 414 T1N = T13 - T14;
Chris@82 415 T1k = FMA(KP500000000, TK, TL);
Chris@82 416 T1l = T1j - T1k;
Chris@82 417 T1V = T1j + T1k;
Chris@82 418 }
Chris@82 419 {
Chris@82 420 E Th, Tx, Tk, T1a, Tw, T1r, T1b, T1q;
Chris@82 421 Th = ci[WS(rs, 2)];
Chris@82 422 Tx = cr[WS(rs, 9)];
Chris@82 423 {
Chris@82 424 E Ti, Tj, Tu, Tv;
Chris@82 425 Ti = cr[WS(rs, 1)];
Chris@82 426 Tj = cr[WS(rs, 5)];
Chris@82 427 Tk = Ti + Tj;
Chris@82 428 T1a = KP866025403 * (Ti - Tj);
Chris@82 429 Tu = ci[WS(rs, 10)];
Chris@82 430 Tv = ci[WS(rs, 6)];
Chris@82 431 Tw = Tu + Tv;
Chris@82 432 T1r = KP866025403 * (Tv - Tu);
Chris@82 433 }
Chris@82 434 Tl = Th + Tk;
Chris@82 435 Ty = Tw - Tx;
Chris@82 436 T1b = FMA(KP500000000, Tw, Tx);
Chris@82 437 T1c = T1a - T1b;
Chris@82 438 T1Y = T1a + T1b;
Chris@82 439 T1q = FNMS(KP500000000, Tk, Th);
Chris@82 440 T1s = T1q + T1r;
Chris@82 441 T1Q = T1q - T1r;
Chris@82 442 }
Chris@82 443 {
Chris@82 444 E Tb, Tm, TU, TW, TX, TY, TT, TV;
Chris@82 445 Tb = T5 + Ta;
Chris@82 446 Tm = Tg + Tl;
Chris@82 447 TU = Tb - Tm;
Chris@82 448 TW = TH + TM;
Chris@82 449 TX = Tt + Ty;
Chris@82 450 TY = TW - TX;
Chris@82 451 cr[0] = Tb + Tm;
Chris@82 452 ci[0] = TW + TX;
Chris@82 453 TT = W[10];
Chris@82 454 TV = W[11];
Chris@82 455 cr[WS(rs, 6)] = FNMS(TV, TY, TT * TU);
Chris@82 456 ci[WS(rs, 6)] = FMA(TV, TU, TT * TY);
Chris@82 457 }
Chris@82 458 {
Chris@82 459 E TA, TQ, TO, TS;
Chris@82 460 {
Chris@82 461 E To, Tz, TC, TN;
Chris@82 462 To = T5 - Ta;
Chris@82 463 Tz = Tt - Ty;
Chris@82 464 TA = To - Tz;
Chris@82 465 TQ = To + Tz;
Chris@82 466 TC = Tg - Tl;
Chris@82 467 TN = TH - TM;
Chris@82 468 TO = TC + TN;
Chris@82 469 TS = TN - TC;
Chris@82 470 }
Chris@82 471 {
Chris@82 472 E Tn, TB, TP, TR;
Chris@82 473 Tn = W[16];
Chris@82 474 TB = W[17];
Chris@82 475 cr[WS(rs, 9)] = FNMS(TB, TO, Tn * TA);
Chris@82 476 ci[WS(rs, 9)] = FMA(Tn, TO, TB * TA);
Chris@82 477 TP = W[4];
Chris@82 478 TR = W[5];
Chris@82 479 cr[WS(rs, 3)] = FNMS(TR, TS, TP * TQ);
Chris@82 480 ci[WS(rs, 3)] = FMA(TP, TS, TR * TQ);
Chris@82 481 }
Chris@82 482 }
Chris@82 483 {
Chris@82 484 E T28, T2e, T2c, T2g;
Chris@82 485 {
Chris@82 486 E T26, T27, T2a, T2b;
Chris@82 487 T26 = T1M - T1N;
Chris@82 488 T27 = T1X + T1Y;
Chris@82 489 T28 = T26 - T27;
Chris@82 490 T2e = T26 + T27;
Chris@82 491 T2a = T1U + T1V;
Chris@82 492 T2b = T1P - T1Q;
Chris@82 493 T2c = T2a + T2b;
Chris@82 494 T2g = T2a - T2b;
Chris@82 495 }
Chris@82 496 {
Chris@82 497 E T25, T29, T2d, T2f;
Chris@82 498 T25 = W[8];
Chris@82 499 T29 = W[9];
Chris@82 500 cr[WS(rs, 5)] = FNMS(T29, T2c, T25 * T28);
Chris@82 501 ci[WS(rs, 5)] = FMA(T25, T2c, T29 * T28);
Chris@82 502 T2d = W[20];
Chris@82 503 T2f = W[21];
Chris@82 504 cr[WS(rs, 11)] = FNMS(T2f, T2g, T2d * T2e);
Chris@82 505 ci[WS(rs, 11)] = FMA(T2d, T2g, T2f * T2e);
Chris@82 506 }
Chris@82 507 }
Chris@82 508 {
Chris@82 509 E T1S, T22, T20, T24;
Chris@82 510 {
Chris@82 511 E T1O, T1R, T1W, T1Z;
Chris@82 512 T1O = T1M + T1N;
Chris@82 513 T1R = T1P + T1Q;
Chris@82 514 T1S = T1O - T1R;
Chris@82 515 T22 = T1O + T1R;
Chris@82 516 T1W = T1U - T1V;
Chris@82 517 T1Z = T1X - T1Y;
Chris@82 518 T20 = T1W - T1Z;
Chris@82 519 T24 = T1W + T1Z;
Chris@82 520 }
Chris@82 521 {
Chris@82 522 E T1L, T1T, T21, T23;
Chris@82 523 T1L = W[2];
Chris@82 524 T1T = W[3];
Chris@82 525 cr[WS(rs, 2)] = FNMS(T1T, T20, T1L * T1S);
Chris@82 526 ci[WS(rs, 2)] = FMA(T1T, T1S, T1L * T20);
Chris@82 527 T21 = W[14];
Chris@82 528 T23 = W[15];
Chris@82 529 cr[WS(rs, 8)] = FNMS(T23, T24, T21 * T22);
Chris@82 530 ci[WS(rs, 8)] = FMA(T23, T22, T21 * T24);
Chris@82 531 }
Chris@82 532 }
Chris@82 533 {
Chris@82 534 E T1C, T1I, T1G, T1K;
Chris@82 535 {
Chris@82 536 E T1A, T1B, T1E, T1F;
Chris@82 537 T1A = T12 + T15;
Chris@82 538 T1B = T1p + T1s;
Chris@82 539 T1C = T1A - T1B;
Chris@82 540 T1I = T1A + T1B;
Chris@82 541 T1E = T1i + T1l;
Chris@82 542 T1F = T19 + T1c;
Chris@82 543 T1G = T1E - T1F;
Chris@82 544 T1K = T1E + T1F;
Chris@82 545 }
Chris@82 546 {
Chris@82 547 E T1z, T1D, T1H, T1J;
Chris@82 548 T1z = W[18];
Chris@82 549 T1D = W[19];
Chris@82 550 cr[WS(rs, 10)] = FNMS(T1D, T1G, T1z * T1C);
Chris@82 551 ci[WS(rs, 10)] = FMA(T1D, T1C, T1z * T1G);
Chris@82 552 T1H = W[6];
Chris@82 553 T1J = W[7];
Chris@82 554 cr[WS(rs, 4)] = FNMS(T1J, T1K, T1H * T1I);
Chris@82 555 ci[WS(rs, 4)] = FMA(T1J, T1I, T1H * T1K);
Chris@82 556 }
Chris@82 557 }
Chris@82 558 {
Chris@82 559 E T1e, T1w, T1u, T1y;
Chris@82 560 {
Chris@82 561 E T16, T1d, T1m, T1t;
Chris@82 562 T16 = T12 - T15;
Chris@82 563 T1d = T19 - T1c;
Chris@82 564 T1e = T16 - T1d;
Chris@82 565 T1w = T16 + T1d;
Chris@82 566 T1m = T1i - T1l;
Chris@82 567 T1t = T1p - T1s;
Chris@82 568 T1u = T1m + T1t;
Chris@82 569 T1y = T1m - T1t;
Chris@82 570 }
Chris@82 571 {
Chris@82 572 E TZ, T1f, T1v, T1x;
Chris@82 573 TZ = W[0];
Chris@82 574 T1f = W[1];
Chris@82 575 cr[WS(rs, 1)] = FNMS(T1f, T1u, TZ * T1e);
Chris@82 576 ci[WS(rs, 1)] = FMA(TZ, T1u, T1f * T1e);
Chris@82 577 T1v = W[12];
Chris@82 578 T1x = W[13];
Chris@82 579 cr[WS(rs, 7)] = FNMS(T1x, T1y, T1v * T1w);
Chris@82 580 ci[WS(rs, 7)] = FMA(T1v, T1y, T1x * T1w);
Chris@82 581 }
Chris@82 582 }
Chris@82 583 }
Chris@82 584 }
Chris@82 585 }
Chris@82 586
Chris@82 587 static const tw_instr twinstr[] = {
Chris@82 588 {TW_FULL, 1, 12},
Chris@82 589 {TW_NEXT, 1, 0}
Chris@82 590 };
Chris@82 591
Chris@82 592 static const hc2hc_desc desc = { 12, "hb_12", twinstr, &GENUS, {88, 30, 30, 0} };
Chris@82 593
Chris@82 594 void X(codelet_hb_12) (planner *p) {
Chris@82 595 X(khc2hc_register) (p, hb_12, &desc);
Chris@82 596 }
Chris@82 597 #endif