Mercurial > hg > sv-dependency-builds

annotate src/fftw-3.3.5/dft/scalar/codelets/q1_4.c @ 84:08ae793730bd

Find changesets by keywords (author, files, the commit message), revision number or hash, or revset expression.
Add null config files
author Chris Cannam
date Mon, 02 Mar 2020 14:03:47 +0000
parents 2cd0e3b3e1fd
children
rev   line source
Chris@42 1 /*
Chris@42 2 * Copyright (c) 2003, 2007-14 Matteo Frigo
Chris@42 3 * Copyright (c) 2003, 2007-14 Massachusetts Institute of Technology
Chris@42 4 *
Chris@42 5 * This program is free software; you can redistribute it and/or modify
Chris@42 6 * it under the terms of the GNU General Public License as published by
Chris@42 7 * the Free Software Foundation; either version 2 of the License, or
Chris@42 8 * (at your option) any later version.
Chris@42 9 *
Chris@42 10 * This program is distributed in the hope that it will be useful,
Chris@42 11 * but WITHOUT ANY WARRANTY; without even the implied warranty of
Chris@42 12 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
Chris@42 13 * GNU General Public License for more details.
Chris@42 14 *
Chris@42 15 * You should have received a copy of the GNU General Public License
Chris@42 16 * along with this program; if not, write to the Free Software
Chris@42 17 * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
Chris@42 18 *
Chris@42 19 */
Chris@42 20
Chris@42 21 /* This file was automatically generated --- DO NOT EDIT */
Chris@42 22 /* Generated on Sat Jul 30 16:37:39 EDT 2016 */
Chris@42 23
Chris@42 24 #include "codelet-dft.h"
Chris@42 25
Chris@42 26 #ifdef HAVE_FMA
Chris@42 27
Chris@42 28 /* Generated by: ../../../genfft/gen_twidsq.native -fma -reorder-insns -schedule-for-pipeline -compact -variables 4 -pipeline-latency 4 -reload-twiddle -dif -n 4 -name q1_4 -include q.h */
Chris@42 29
Chris@42 30 /*
Chris@42 31 * This function contains 88 FP additions, 48 FP multiplications,
Chris@42 32 * (or, 64 additions, 24 multiplications, 24 fused multiply/add),
Chris@42 33 * 76 stack variables, 0 constants, and 64 memory accesses
Chris@42 34 */
Chris@42 35 #include "q.h"
Chris@42 36
Chris@42 37 static void q1_4(R *rio, R *iio, const R *W, stride rs, stride vs, INT mb, INT me, INT ms)
Chris@42 38 {
Chris@42 39 {
Chris@42 40 INT m;
Chris@42 41 for (m = mb, W = W + (mb * 6); m < me; m = m + 1, rio = rio + ms, iio = iio + ms, W = W + 6, MAKE_VOLATILE_STRIDE(8, rs), MAKE_VOLATILE_STRIDE(0, vs)) {
Chris@42 42 E T1X, T1S, T1L, T1Y, T1R;
Chris@42 43 {
Chris@42 44 E T3, Tf, Tv, Ti, Tw, Tx, T6, Tm, Tc, Ts, T1T, T1H, T29, T1W, T2a;
Chris@42 45 E T2b, T1K, T20, T1Q, T26, TN, TB, T13, TQ, T14, T15, TE, TU, TK, T10;
Chris@42 46 E T1l, T19, T1a, T1h, T1B, T1o, T1C, T1b, T1D, T1e, T1c;
Chris@42 47 {
Chris@42 48 E T1I, T1P, T1J, T1M;
Chris@42 49 {
Chris@42 50 E Tb, T4, T5, T8;
Chris@42 51 {
Chris@42 52 E T1, T2, T9, Ta, Tg, Th;
Chris@42 53 T1 = rio[0];
Chris@42 54 T2 = rio[WS(rs, 2)];
Chris@42 55 T9 = iio[0];
Chris@42 56 Ta = iio[WS(rs, 2)];
Chris@42 57 Tg = iio[WS(rs, 1)];
Chris@42 58 T3 = T1 + T2;
Chris@42 59 Tf = T1 - T2;
Chris@42 60 Th = iio[WS(rs, 3)];
Chris@42 61 Tv = T9 + Ta;
Chris@42 62 Tb = T9 - Ta;
Chris@42 63 T4 = rio[WS(rs, 1)];
Chris@42 64 Ti = Tg - Th;
Chris@42 65 Tw = Tg + Th;
Chris@42 66 T5 = rio[WS(rs, 3)];
Chris@42 67 }
Chris@42 68 Tx = Tv - Tw;
Chris@42 69 T8 = T4 - T5;
Chris@42 70 T6 = T4 + T5;
Chris@42 71 {
Chris@42 72 E T1N, T1O, T1F, T1G, T1U, T1V;
Chris@42 73 T1F = rio[WS(vs, 3)];
Chris@42 74 T1G = rio[WS(vs, 3) + WS(rs, 2)];
Chris@42 75 Tm = Tb - T8;
Chris@42 76 Tc = T8 + Tb;
Chris@42 77 Ts = T3 - T6;
Chris@42 78 T1T = T1F - T1G;
Chris@42 79 T1H = T1F + T1G;
Chris@42 80 T1N = iio[WS(vs, 3)];
Chris@42 81 T1O = iio[WS(vs, 3) + WS(rs, 2)];
Chris@42 82 T1U = iio[WS(vs, 3) + WS(rs, 1)];
Chris@42 83 T1V = iio[WS(vs, 3) + WS(rs, 3)];
Chris@42 84 T1I = rio[WS(vs, 3) + WS(rs, 1)];
Chris@42 85 T1P = T1N - T1O;
Chris@42 86 T29 = T1N + T1O;
Chris@42 87 T1W = T1U - T1V;
Chris@42 88 T2a = T1U + T1V;
Chris@42 89 T1J = rio[WS(vs, 3) + WS(rs, 3)];
Chris@42 90 }
Chris@42 91 }
Chris@42 92 T2b = T29 - T2a;
Chris@42 93 T1M = T1I - T1J;
Chris@42 94 T1K = T1I + T1J;
Chris@42 95 {
Chris@42 96 E TC, TJ, TD, TG;
Chris@42 97 {
Chris@42 98 E TH, TI, Tz, TA, TO, TP;
Chris@42 99 Tz = rio[WS(vs, 1)];
Chris@42 100 TA = rio[WS(vs, 1) + WS(rs, 2)];
Chris@42 101 T20 = T1P - T1M;
Chris@42 102 T1Q = T1M + T1P;
Chris@42 103 T26 = T1H - T1K;
Chris@42 104 TN = Tz - TA;
Chris@42 105 TB = Tz + TA;
Chris@42 106 TH = iio[WS(vs, 1)];
Chris@42 107 TI = iio[WS(vs, 1) + WS(rs, 2)];
Chris@42 108 TO = iio[WS(vs, 1) + WS(rs, 1)];
Chris@42 109 TP = iio[WS(vs, 1) + WS(rs, 3)];
Chris@42 110 TC = rio[WS(vs, 1) + WS(rs, 1)];
Chris@42 111 TJ = TH - TI;
Chris@42 112 T13 = TH + TI;
Chris@42 113 TQ = TO - TP;
Chris@42 114 T14 = TO + TP;
Chris@42 115 TD = rio[WS(vs, 1) + WS(rs, 3)];
Chris@42 116 }
Chris@42 117 T15 = T13 - T14;
Chris@42 118 TG = TC - TD;
Chris@42 119 TE = TC + TD;
Chris@42 120 {
Chris@42 121 E T1f, T1g, T17, T18, T1m, T1n;
Chris@42 122 T17 = rio[WS(vs, 2)];
Chris@42 123 T18 = rio[WS(vs, 2) + WS(rs, 2)];
Chris@42 124 TU = TJ - TG;
Chris@42 125 TK = TG + TJ;
Chris@42 126 T10 = TB - TE;
Chris@42 127 T1l = T17 - T18;
Chris@42 128 T19 = T17 + T18;
Chris@42 129 T1f = iio[WS(vs, 2)];
Chris@42 130 T1g = iio[WS(vs, 2) + WS(rs, 2)];
Chris@42 131 T1m = iio[WS(vs, 2) + WS(rs, 1)];
Chris@42 132 T1n = iio[WS(vs, 2) + WS(rs, 3)];
Chris@42 133 T1a = rio[WS(vs, 2) + WS(rs, 1)];
Chris@42 134 T1h = T1f - T1g;
Chris@42 135 T1B = T1f + T1g;
Chris@42 136 T1o = T1m - T1n;
Chris@42 137 T1C = T1m + T1n;
Chris@42 138 T1b = rio[WS(vs, 2) + WS(rs, 3)];
Chris@42 139 }
Chris@42 140 }
Chris@42 141 }
Chris@42 142 T1D = T1B - T1C;
Chris@42 143 T1e = T1a - T1b;
Chris@42 144 T1c = T1a + T1b;
Chris@42 145 {
Chris@42 146 E T1s, T1i, T1y, T28, T27, Tr, Tu;
Chris@42 147 rio[0] = T3 + T6;
Chris@42 148 iio[0] = Tv + Tw;
Chris@42 149 T1s = T1h - T1e;
Chris@42 150 T1i = T1e + T1h;
Chris@42 151 T1y = T19 - T1c;
Chris@42 152 rio[WS(rs, 1)] = TB + TE;
Chris@42 153 iio[WS(rs, 1)] = T13 + T14;
Chris@42 154 rio[WS(rs, 2)] = T19 + T1c;
Chris@42 155 iio[WS(rs, 2)] = T1B + T1C;
Chris@42 156 iio[WS(rs, 3)] = T29 + T2a;
Chris@42 157 rio[WS(rs, 3)] = T1H + T1K;
Chris@42 158 Tr = W[2];
Chris@42 159 Tu = W[3];
Chris@42 160 {
Chris@42 161 E T25, Ty, Tt, T2c;
Chris@42 162 T25 = W[2];
Chris@42 163 T28 = W[3];
Chris@42 164 Ty = Tr * Tx;
Chris@42 165 Tt = Tr * Ts;
Chris@42 166 T2c = T25 * T2b;
Chris@42 167 T27 = T25 * T26;
Chris@42 168 iio[WS(vs, 2)] = FNMS(Tu, Ts, Ty);
Chris@42 169 rio[WS(vs, 2)] = FMA(Tu, Tx, Tt);
Chris@42 170 iio[WS(vs, 2) + WS(rs, 3)] = FNMS(T28, T26, T2c);
Chris@42 171 }
Chris@42 172 rio[WS(vs, 2) + WS(rs, 3)] = FMA(T28, T2b, T27);
Chris@42 173 {
Chris@42 174 E Tp, T1v, T23, T22, T1Z, TR, TM, TF;
Chris@42 175 {
Chris@42 176 E T1A, T1z, TZ, T12;
Chris@42 177 TZ = W[2];
Chris@42 178 T12 = W[3];
Chris@42 179 {
Chris@42 180 E T1x, T16, T11, T1E;
Chris@42 181 T1x = W[2];
Chris@42 182 T1A = W[3];
Chris@42 183 T16 = TZ * T15;
Chris@42 184 T11 = TZ * T10;
Chris@42 185 T1E = T1x * T1D;
Chris@42 186 T1z = T1x * T1y;
Chris@42 187 iio[WS(vs, 2) + WS(rs, 1)] = FNMS(T12, T10, T16);
Chris@42 188 rio[WS(vs, 2) + WS(rs, 1)] = FMA(T12, T15, T11);
Chris@42 189 iio[WS(vs, 2) + WS(rs, 2)] = FNMS(T1A, T1y, T1E);
Chris@42 190 }
Chris@42 191 rio[WS(vs, 2) + WS(rs, 2)] = FMA(T1A, T1D, T1z);
Chris@42 192 {
Chris@42 193 E Tj, Te, T7, T1p, T1k, T1j;
Chris@42 194 Tp = Tf + Ti;
Chris@42 195 Tj = Tf - Ti;
Chris@42 196 Te = W[5];
Chris@42 197 T7 = W[4];
Chris@42 198 {
Chris@42 199 E T1d, T1q, Tk, Td;
Chris@42 200 T1p = T1l - T1o;
Chris@42 201 T1v = T1l + T1o;
Chris@42 202 T1k = W[5];
Chris@42 203 Tk = Te * Tc;
Chris@42 204 Td = T7 * Tc;
Chris@42 205 T1d = W[4];
Chris@42 206 T1q = T1k * T1i;
Chris@42 207 rio[WS(vs, 3)] = FMA(T7, Tj, Tk);
Chris@42 208 iio[WS(vs, 3)] = FNMS(Te, Tj, Td);
Chris@42 209 T1j = T1d * T1i;
Chris@42 210 rio[WS(vs, 3) + WS(rs, 2)] = FMA(T1d, T1p, T1q);
Chris@42 211 }
Chris@42 212 T23 = T1T + T1W;
Chris@42 213 T1X = T1T - T1W;
Chris@42 214 T22 = W[1];
Chris@42 215 iio[WS(vs, 3) + WS(rs, 2)] = FNMS(T1k, T1p, T1j);
Chris@42 216 T1Z = W[0];
Chris@42 217 }
Chris@42 218 }
Chris@42 219 {
Chris@42 220 E TX, TW, TT, TY, TV, T24, T21;
Chris@42 221 TX = TN + TQ;
Chris@42 222 TR = TN - TQ;
Chris@42 223 T24 = T22 * T20;
Chris@42 224 TW = W[1];
Chris@42 225 T21 = T1Z * T20;
Chris@42 226 TT = W[0];
Chris@42 227 rio[WS(vs, 1) + WS(rs, 3)] = FMA(T1Z, T23, T24);
Chris@42 228 TY = TW * TU;
Chris@42 229 iio[WS(vs, 1) + WS(rs, 3)] = FNMS(T22, T23, T21);
Chris@42 230 TV = TT * TU;
Chris@42 231 rio[WS(vs, 1) + WS(rs, 1)] = FMA(TT, TX, TY);
Chris@42 232 TM = W[5];
Chris@42 233 iio[WS(vs, 1) + WS(rs, 1)] = FNMS(TW, TX, TV);
Chris@42 234 TF = W[4];
Chris@42 235 }
Chris@42 236 {
Chris@42 237 E To, Tl, Tq, Tn, TS, TL;
Chris@42 238 TS = TM * TK;
Chris@42 239 To = W[1];
Chris@42 240 TL = TF * TK;
Chris@42 241 Tl = W[0];
Chris@42 242 rio[WS(vs, 3) + WS(rs, 1)] = FMA(TF, TR, TS);
Chris@42 243 Tq = To * Tm;
Chris@42 244 iio[WS(vs, 3) + WS(rs, 1)] = FNMS(TM, TR, TL);
Chris@42 245 Tn = Tl * Tm;
Chris@42 246 {
Chris@42 247 E T1u, T1r, T1w, T1t;
Chris@42 248 rio[WS(vs, 1)] = FMA(Tl, Tp, Tq);
Chris@42 249 T1u = W[1];
Chris@42 250 iio[WS(vs, 1)] = FNMS(To, Tp, Tn);
Chris@42 251 T1r = W[0];
Chris@42 252 T1w = T1u * T1s;
Chris@42 253 T1S = W[5];
Chris@42 254 T1t = T1r * T1s;
Chris@42 255 T1L = W[4];
Chris@42 256 rio[WS(vs, 1) + WS(rs, 2)] = FMA(T1r, T1v, T1w);
Chris@42 257 T1Y = T1S * T1Q;
Chris@42 258 iio[WS(vs, 1) + WS(rs, 2)] = FNMS(T1u, T1v, T1t);
Chris@42 259 T1R = T1L * T1Q;
Chris@42 260 }
Chris@42 261 }
Chris@42 262 }
Chris@42 263 }
Chris@42 264 }
Chris@42 265 rio[WS(vs, 3) + WS(rs, 3)] = FMA(T1L, T1X, T1Y);
Chris@42 266 iio[WS(vs, 3) + WS(rs, 3)] = FNMS(T1S, T1X, T1R);
Chris@42 267 }
Chris@42 268 }
Chris@42 269 }
Chris@42 270
Chris@42 271 static const tw_instr twinstr[] = {
Chris@42 272 {TW_FULL, 0, 4},
Chris@42 273 {TW_NEXT, 1, 0}
Chris@42 274 };
Chris@42 275
Chris@42 276 static const ct_desc desc = { 4, "q1_4", twinstr, &GENUS, {64, 24, 24, 0}, 0, 0, 0 };
Chris@42 277
Chris@42 278 void X(codelet_q1_4) (planner *p) {
Chris@42 279 X(kdft_difsq_register) (p, q1_4, &desc);
Chris@42 280 }
Chris@42 281 #else /* HAVE_FMA */
Chris@42 282
Chris@42 283 /* Generated by: ../../../genfft/gen_twidsq.native -compact -variables 4 -pipeline-latency 4 -reload-twiddle -dif -n 4 -name q1_4 -include q.h */
Chris@42 284
Chris@42 285 /*
Chris@42 286 * This function contains 88 FP additions, 48 FP multiplications,
Chris@42 287 * (or, 64 additions, 24 multiplications, 24 fused multiply/add),
Chris@42 288 * 37 stack variables, 0 constants, and 64 memory accesses
Chris@42 289 */
Chris@42 290 #include "q.h"
Chris@42 291
Chris@42 292 static void q1_4(R *rio, R *iio, const R *W, stride rs, stride vs, INT mb, INT me, INT ms)
Chris@42 293 {
Chris@42 294 {
Chris@42 295 INT m;
Chris@42 296 for (m = mb, W = W + (mb * 6); m < me; m = m + 1, rio = rio + ms, iio = iio + ms, W = W + 6, MAKE_VOLATILE_STRIDE(8, rs), MAKE_VOLATILE_STRIDE(0, vs)) {
Chris@42 297 E T3, Te, Tb, Tq, T6, T8, Th, Tr, Tv, TG, TD, TS, Ty, TA, TJ;
Chris@42 298 E TT, TX, T18, T15, T1k, T10, T12, T1b, T1l, T1p, T1A, T1x, T1M, T1s, T1u;
Chris@42 299 E T1D, T1N;
Chris@42 300 {
Chris@42 301 E T1, T2, T9, Ta;
Chris@42 302 T1 = rio[0];
Chris@42 303 T2 = rio[WS(rs, 2)];
Chris@42 304 T3 = T1 + T2;
Chris@42 305 Te = T1 - T2;
Chris@42 306 T9 = iio[0];
Chris@42 307 Ta = iio[WS(rs, 2)];
Chris@42 308 Tb = T9 - Ta;
Chris@42 309 Tq = T9 + Ta;
Chris@42 310 }
Chris@42 311 {
Chris@42 312 E T4, T5, Tf, Tg;
Chris@42 313 T4 = rio[WS(rs, 1)];
Chris@42 314 T5 = rio[WS(rs, 3)];
Chris@42 315 T6 = T4 + T5;
Chris@42 316 T8 = T4 - T5;
Chris@42 317 Tf = iio[WS(rs, 1)];
Chris@42 318 Tg = iio[WS(rs, 3)];
Chris@42 319 Th = Tf - Tg;
Chris@42 320 Tr = Tf + Tg;
Chris@42 321 }
Chris@42 322 {
Chris@42 323 E Tt, Tu, TB, TC;
Chris@42 324 Tt = rio[WS(vs, 1)];
Chris@42 325 Tu = rio[WS(vs, 1) + WS(rs, 2)];
Chris@42 326 Tv = Tt + Tu;
Chris@42 327 TG = Tt - Tu;
Chris@42 328 TB = iio[WS(vs, 1)];
Chris@42 329 TC = iio[WS(vs, 1) + WS(rs, 2)];
Chris@42 330 TD = TB - TC;
Chris@42 331 TS = TB + TC;
Chris@42 332 }
Chris@42 333 {
Chris@42 334 E Tw, Tx, TH, TI;
Chris@42 335 Tw = rio[WS(vs, 1) + WS(rs, 1)];
Chris@42 336 Tx = rio[WS(vs, 1) + WS(rs, 3)];
Chris@42 337 Ty = Tw + Tx;
Chris@42 338 TA = Tw - Tx;
Chris@42 339 TH = iio[WS(vs, 1) + WS(rs, 1)];
Chris@42 340 TI = iio[WS(vs, 1) + WS(rs, 3)];
Chris@42 341 TJ = TH - TI;
Chris@42 342 TT = TH + TI;
Chris@42 343 }
Chris@42 344 {
Chris@42 345 E TV, TW, T13, T14;
Chris@42 346 TV = rio[WS(vs, 2)];
Chris@42 347 TW = rio[WS(vs, 2) + WS(rs, 2)];
Chris@42 348 TX = TV + TW;
Chris@42 349 T18 = TV - TW;
Chris@42 350 T13 = iio[WS(vs, 2)];
Chris@42 351 T14 = iio[WS(vs, 2) + WS(rs, 2)];
Chris@42 352 T15 = T13 - T14;
Chris@42 353 T1k = T13 + T14;
Chris@42 354 }
Chris@42 355 {
Chris@42 356 E TY, TZ, T19, T1a;
Chris@42 357 TY = rio[WS(vs, 2) + WS(rs, 1)];
Chris@42 358 TZ = rio[WS(vs, 2) + WS(rs, 3)];
Chris@42 359 T10 = TY + TZ;
Chris@42 360 T12 = TY - TZ;
Chris@42 361 T19 = iio[WS(vs, 2) + WS(rs, 1)];
Chris@42 362 T1a = iio[WS(vs, 2) + WS(rs, 3)];
Chris@42 363 T1b = T19 - T1a;
Chris@42 364 T1l = T19 + T1a;
Chris@42 365 }
Chris@42 366 {
Chris@42 367 E T1n, T1o, T1v, T1w;
Chris@42 368 T1n = rio[WS(vs, 3)];
Chris@42 369 T1o = rio[WS(vs, 3) + WS(rs, 2)];
Chris@42 370 T1p = T1n + T1o;
Chris@42 371 T1A = T1n - T1o;
Chris@42 372 T1v = iio[WS(vs, 3)];
Chris@42 373 T1w = iio[WS(vs, 3) + WS(rs, 2)];
Chris@42 374 T1x = T1v - T1w;
Chris@42 375 T1M = T1v + T1w;
Chris@42 376 }
Chris@42 377 {
Chris@42 378 E T1q, T1r, T1B, T1C;
Chris@42 379 T1q = rio[WS(vs, 3) + WS(rs, 1)];
Chris@42 380 T1r = rio[WS(vs, 3) + WS(rs, 3)];
Chris@42 381 T1s = T1q + T1r;
Chris@42 382 T1u = T1q - T1r;
Chris@42 383 T1B = iio[WS(vs, 3) + WS(rs, 1)];
Chris@42 384 T1C = iio[WS(vs, 3) + WS(rs, 3)];
Chris@42 385 T1D = T1B - T1C;
Chris@42 386 T1N = T1B + T1C;
Chris@42 387 }
Chris@42 388 rio[0] = T3 + T6;
Chris@42 389 iio[0] = Tq + Tr;
Chris@42 390 rio[WS(rs, 1)] = Tv + Ty;
Chris@42 391 iio[WS(rs, 1)] = TS + TT;
Chris@42 392 rio[WS(rs, 2)] = TX + T10;
Chris@42 393 iio[WS(rs, 2)] = T1k + T1l;
Chris@42 394 iio[WS(rs, 3)] = T1M + T1N;
Chris@42 395 rio[WS(rs, 3)] = T1p + T1s;
Chris@42 396 {
Chris@42 397 E Tc, Ti, T7, Td;
Chris@42 398 Tc = T8 + Tb;
Chris@42 399 Ti = Te - Th;
Chris@42 400 T7 = W[4];
Chris@42 401 Td = W[5];
Chris@42 402 iio[WS(vs, 3)] = FNMS(Td, Ti, T7 * Tc);
Chris@42 403 rio[WS(vs, 3)] = FMA(Td, Tc, T7 * Ti);
Chris@42 404 }
Chris@42 405 {
Chris@42 406 E T1K, T1O, T1J, T1L;
Chris@42 407 T1K = T1p - T1s;
Chris@42 408 T1O = T1M - T1N;
Chris@42 409 T1J = W[2];
Chris@42 410 T1L = W[3];
Chris@42 411 rio[WS(vs, 2) + WS(rs, 3)] = FMA(T1J, T1K, T1L * T1O);
Chris@42 412 iio[WS(vs, 2) + WS(rs, 3)] = FNMS(T1L, T1K, T1J * T1O);
Chris@42 413 }
Chris@42 414 {
Chris@42 415 E Tk, Tm, Tj, Tl;
Chris@42 416 Tk = Tb - T8;
Chris@42 417 Tm = Te + Th;
Chris@42 418 Tj = W[0];
Chris@42 419 Tl = W[1];
Chris@42 420 iio[WS(vs, 1)] = FNMS(Tl, Tm, Tj * Tk);
Chris@42 421 rio[WS(vs, 1)] = FMA(Tl, Tk, Tj * Tm);
Chris@42 422 }
Chris@42 423 {
Chris@42 424 E To, Ts, Tn, Tp;
Chris@42 425 To = T3 - T6;
Chris@42 426 Ts = Tq - Tr;
Chris@42 427 Tn = W[2];
Chris@42 428 Tp = W[3];
Chris@42 429 rio[WS(vs, 2)] = FMA(Tn, To, Tp * Ts);
Chris@42 430 iio[WS(vs, 2)] = FNMS(Tp, To, Tn * Ts);
Chris@42 431 }
Chris@42 432 {
Chris@42 433 E T16, T1c, T11, T17;
Chris@42 434 T16 = T12 + T15;
Chris@42 435 T1c = T18 - T1b;
Chris@42 436 T11 = W[4];
Chris@42 437 T17 = W[5];
Chris@42 438 iio[WS(vs, 3) + WS(rs, 2)] = FNMS(T17, T1c, T11 * T16);
Chris@42 439 rio[WS(vs, 3) + WS(rs, 2)] = FMA(T17, T16, T11 * T1c);
Chris@42 440 }
Chris@42 441 {
Chris@42 442 E T1G, T1I, T1F, T1H;
Chris@42 443 T1G = T1x - T1u;
Chris@42 444 T1I = T1A + T1D;
Chris@42 445 T1F = W[0];
Chris@42 446 T1H = W[1];
Chris@42 447 iio[WS(vs, 1) + WS(rs, 3)] = FNMS(T1H, T1I, T1F * T1G);
Chris@42 448 rio[WS(vs, 1) + WS(rs, 3)] = FMA(T1H, T1G, T1F * T1I);
Chris@42 449 }
Chris@42 450 {
Chris@42 451 E TQ, TU, TP, TR;
Chris@42 452 TQ = Tv - Ty;
Chris@42 453 TU = TS - TT;
Chris@42 454 TP = W[2];
Chris@42 455 TR = W[3];
Chris@42 456 rio[WS(vs, 2) + WS(rs, 1)] = FMA(TP, TQ, TR * TU);
Chris@42 457 iio[WS(vs, 2) + WS(rs, 1)] = FNMS(TR, TQ, TP * TU);
Chris@42 458 }
Chris@42 459 {
Chris@42 460 E T1e, T1g, T1d, T1f;
Chris@42 461 T1e = T15 - T12;
Chris@42 462 T1g = T18 + T1b;
Chris@42 463 T1d = W[0];
Chris@42 464 T1f = W[1];
Chris@42 465 iio[WS(vs, 1) + WS(rs, 2)] = FNMS(T1f, T1g, T1d * T1e);
Chris@42 466 rio[WS(vs, 1) + WS(rs, 2)] = FMA(T1f, T1e, T1d * T1g);
Chris@42 467 }
Chris@42 468 {
Chris@42 469 E T1i, T1m, T1h, T1j;
Chris@42 470 T1i = TX - T10;
Chris@42 471 T1m = T1k - T1l;
Chris@42 472 T1h = W[2];
Chris@42 473 T1j = W[3];
Chris@42 474 rio[WS(vs, 2) + WS(rs, 2)] = FMA(T1h, T1i, T1j * T1m);
Chris@42 475 iio[WS(vs, 2) + WS(rs, 2)] = FNMS(T1j, T1i, T1h * T1m);
Chris@42 476 }
Chris@42 477 {
Chris@42 478 E T1y, T1E, T1t, T1z;
Chris@42 479 T1y = T1u + T1x;
Chris@42 480 T1E = T1A - T1D;
Chris@42 481 T1t = W[4];
Chris@42 482 T1z = W[5];
Chris@42 483 iio[WS(vs, 3) + WS(rs, 3)] = FNMS(T1z, T1E, T1t * T1y);
Chris@42 484 rio[WS(vs, 3) + WS(rs, 3)] = FMA(T1z, T1y, T1t * T1E);
Chris@42 485 }
Chris@42 486 {
Chris@42 487 E TM, TO, TL, TN;
Chris@42 488 TM = TD - TA;
Chris@42 489 TO = TG + TJ;
Chris@42 490 TL = W[0];
Chris@42 491 TN = W[1];
Chris@42 492 iio[WS(vs, 1) + WS(rs, 1)] = FNMS(TN, TO, TL * TM);
Chris@42 493 rio[WS(vs, 1) + WS(rs, 1)] = FMA(TN, TM, TL * TO);
Chris@42 494 }
Chris@42 495 {
Chris@42 496 E TE, TK, Tz, TF;
Chris@42 497 TE = TA + TD;
Chris@42 498 TK = TG - TJ;
Chris@42 499 Tz = W[4];
Chris@42 500 TF = W[5];
Chris@42 501 iio[WS(vs, 3) + WS(rs, 1)] = FNMS(TF, TK, Tz * TE);
Chris@42 502 rio[WS(vs, 3) + WS(rs, 1)] = FMA(TF, TE, Tz * TK);
Chris@42 503 }
Chris@42 504 }
Chris@42 505 }
Chris@42 506 }
Chris@42 507
Chris@42 508 static const tw_instr twinstr[] = {
Chris@42 509 {TW_FULL, 0, 4},
Chris@42 510 {TW_NEXT, 1, 0}
Chris@42 511 };
Chris@42 512
Chris@42 513 static const ct_desc desc = { 4, "q1_4", twinstr, &GENUS, {64, 24, 24, 0}, 0, 0, 0 };
Chris@42 514
Chris@42 515 void X(codelet_q1_4) (planner *p) {
Chris@42 516 X(kdft_difsq_register) (p, q1_4, &desc);
Chris@42 517 }
Chris@42 518 #endif /* HAVE_FMA */