Mercurial > hg > sv-dependency-builds

annotate src/fftw-3.3.8/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 d0c2a83c1364
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 08:04:29 EDT 2018 */
Chris@82 23
Chris@82 24 #include "dft/codelet-dft.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_twidsq.native -fma -compact -variables 4 -pipeline-latency 4 -reload-twiddle -dif -n 4 -name q1_4 -include dft/scalar/q.h */
Chris@82 29
Chris@82 30 /*
Chris@82 31 * This function contains 88 FP additions, 48 FP multiplications,
Chris@82 32 * (or, 64 additions, 24 multiplications, 24 fused multiply/add),
Chris@82 33 * 51 stack variables, 0 constants, and 64 memory accesses
Chris@82 34 */
Chris@82 35 #include "dft/scalar/q.h"
Chris@82 36
Chris@82 37 static void q1_4(R *rio, R *iio, const R *W, stride rs, stride vs, INT mb, INT me, INT ms)
Chris@82 38 {
Chris@82 39 {
Chris@82 40 INT m;
Chris@82 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@82 42 E T3, Tv, Tw, T6, Tc, Tf, Tx, Ts, Tm, Ti, T1H, T29, T2a, T1K, T1Q;
Chris@82 43 E T1T, T2b, T26, T20, T1W, TB, T13, T14, TE, TK, TN, T15, T10, TU, TQ;
Chris@82 44 E T19, T1B, T1C, T1c, T1i, T1l, T1D, T1y, T1s, T1o;
Chris@82 45 {
Chris@82 46 E T1, T2, Tb, Tg, Th, T8;
Chris@82 47 {
Chris@82 48 E T9, Ta, T4, T5;
Chris@82 49 T1 = rio[0];
Chris@82 50 T2 = rio[WS(rs, 2)];
Chris@82 51 T3 = T1 + T2;
Chris@82 52 T9 = iio[0];
Chris@82 53 Ta = iio[WS(rs, 2)];
Chris@82 54 Tb = T9 - Ta;
Chris@82 55 Tv = T9 + Ta;
Chris@82 56 Tg = iio[WS(rs, 1)];
Chris@82 57 Th = iio[WS(rs, 3)];
Chris@82 58 Tw = Tg + Th;
Chris@82 59 T4 = rio[WS(rs, 1)];
Chris@82 60 T5 = rio[WS(rs, 3)];
Chris@82 61 T6 = T4 + T5;
Chris@82 62 T8 = T4 - T5;
Chris@82 63 }
Chris@82 64 Tc = T8 + Tb;
Chris@82 65 Tf = T1 - T2;
Chris@82 66 Tx = Tv - Tw;
Chris@82 67 Ts = T3 - T6;
Chris@82 68 Tm = Tb - T8;
Chris@82 69 Ti = Tg - Th;
Chris@82 70 }
Chris@82 71 {
Chris@82 72 E T1F, T1G, T1P, T1U, T1V, T1M;
Chris@82 73 {
Chris@82 74 E T1N, T1O, T1I, T1J;
Chris@82 75 T1F = rio[WS(vs, 3)];
Chris@82 76 T1G = rio[WS(vs, 3) + WS(rs, 2)];
Chris@82 77 T1H = T1F + T1G;
Chris@82 78 T1N = iio[WS(vs, 3)];
Chris@82 79 T1O = iio[WS(vs, 3) + WS(rs, 2)];
Chris@82 80 T1P = T1N - T1O;
Chris@82 81 T29 = T1N + T1O;
Chris@82 82 T1U = iio[WS(vs, 3) + WS(rs, 1)];
Chris@82 83 T1V = iio[WS(vs, 3) + WS(rs, 3)];
Chris@82 84 T2a = T1U + T1V;
Chris@82 85 T1I = rio[WS(vs, 3) + WS(rs, 1)];
Chris@82 86 T1J = rio[WS(vs, 3) + WS(rs, 3)];
Chris@82 87 T1K = T1I + T1J;
Chris@82 88 T1M = T1I - T1J;
Chris@82 89 }
Chris@82 90 T1Q = T1M + T1P;
Chris@82 91 T1T = T1F - T1G;
Chris@82 92 T2b = T29 - T2a;
Chris@82 93 T26 = T1H - T1K;
Chris@82 94 T20 = T1P - T1M;
Chris@82 95 T1W = T1U - T1V;
Chris@82 96 }
Chris@82 97 {
Chris@82 98 E Tz, TA, TJ, TO, TP, TG;
Chris@82 99 {
Chris@82 100 E TH, TI, TC, TD;
Chris@82 101 Tz = rio[WS(vs, 1)];
Chris@82 102 TA = rio[WS(vs, 1) + WS(rs, 2)];
Chris@82 103 TB = Tz + TA;
Chris@82 104 TH = iio[WS(vs, 1)];
Chris@82 105 TI = iio[WS(vs, 1) + WS(rs, 2)];
Chris@82 106 TJ = TH - TI;
Chris@82 107 T13 = TH + TI;
Chris@82 108 TO = iio[WS(vs, 1) + WS(rs, 1)];
Chris@82 109 TP = iio[WS(vs, 1) + WS(rs, 3)];
Chris@82 110 T14 = TO + TP;
Chris@82 111 TC = rio[WS(vs, 1) + WS(rs, 1)];
Chris@82 112 TD = rio[WS(vs, 1) + WS(rs, 3)];
Chris@82 113 TE = TC + TD;
Chris@82 114 TG = TC - TD;
Chris@82 115 }
Chris@82 116 TK = TG + TJ;
Chris@82 117 TN = Tz - TA;
Chris@82 118 T15 = T13 - T14;
Chris@82 119 T10 = TB - TE;
Chris@82 120 TU = TJ - TG;
Chris@82 121 TQ = TO - TP;
Chris@82 122 }
Chris@82 123 {
Chris@82 124 E T17, T18, T1h, T1m, T1n, T1e;
Chris@82 125 {
Chris@82 126 E T1f, T1g, T1a, T1b;
Chris@82 127 T17 = rio[WS(vs, 2)];
Chris@82 128 T18 = rio[WS(vs, 2) + WS(rs, 2)];
Chris@82 129 T19 = T17 + T18;
Chris@82 130 T1f = iio[WS(vs, 2)];
Chris@82 131 T1g = iio[WS(vs, 2) + WS(rs, 2)];
Chris@82 132 T1h = T1f - T1g;
Chris@82 133 T1B = T1f + T1g;
Chris@82 134 T1m = iio[WS(vs, 2) + WS(rs, 1)];
Chris@82 135 T1n = iio[WS(vs, 2) + WS(rs, 3)];
Chris@82 136 T1C = T1m + T1n;
Chris@82 137 T1a = rio[WS(vs, 2) + WS(rs, 1)];
Chris@82 138 T1b = rio[WS(vs, 2) + WS(rs, 3)];
Chris@82 139 T1c = T1a + T1b;
Chris@82 140 T1e = T1a - T1b;
Chris@82 141 }
Chris@82 142 T1i = T1e + T1h;
Chris@82 143 T1l = T17 - T18;
Chris@82 144 T1D = T1B - T1C;
Chris@82 145 T1y = T19 - T1c;
Chris@82 146 T1s = T1h - T1e;
Chris@82 147 T1o = T1m - T1n;
Chris@82 148 }
Chris@82 149 rio[0] = T3 + T6;
Chris@82 150 iio[0] = Tv + Tw;
Chris@82 151 rio[WS(rs, 1)] = TB + TE;
Chris@82 152 iio[WS(rs, 1)] = T13 + T14;
Chris@82 153 rio[WS(rs, 2)] = T19 + T1c;
Chris@82 154 iio[WS(rs, 2)] = T1B + T1C;
Chris@82 155 iio[WS(rs, 3)] = T29 + T2a;
Chris@82 156 rio[WS(rs, 3)] = T1H + T1K;
Chris@82 157 {
Chris@82 158 E Tt, Ty, Tr, Tu;
Chris@82 159 Tr = W[2];
Chris@82 160 Tt = Tr * Ts;
Chris@82 161 Ty = Tr * Tx;
Chris@82 162 Tu = W[3];
Chris@82 163 rio[WS(vs, 2)] = FMA(Tu, Tx, Tt);
Chris@82 164 iio[WS(vs, 2)] = FNMS(Tu, Ts, Ty);
Chris@82 165 }
Chris@82 166 {
Chris@82 167 E T27, T2c, T25, T28;
Chris@82 168 T25 = W[2];
Chris@82 169 T27 = T25 * T26;
Chris@82 170 T2c = T25 * T2b;
Chris@82 171 T28 = W[3];
Chris@82 172 rio[WS(vs, 2) + WS(rs, 3)] = FMA(T28, T2b, T27);
Chris@82 173 iio[WS(vs, 2) + WS(rs, 3)] = FNMS(T28, T26, T2c);
Chris@82 174 }
Chris@82 175 {
Chris@82 176 E T11, T16, TZ, T12;
Chris@82 177 TZ = W[2];
Chris@82 178 T11 = TZ * T10;
Chris@82 179 T16 = TZ * T15;
Chris@82 180 T12 = W[3];
Chris@82 181 rio[WS(vs, 2) + WS(rs, 1)] = FMA(T12, T15, T11);
Chris@82 182 iio[WS(vs, 2) + WS(rs, 1)] = FNMS(T12, T10, T16);
Chris@82 183 }
Chris@82 184 {
Chris@82 185 E T1z, T1E, T1x, T1A;
Chris@82 186 T1x = W[2];
Chris@82 187 T1z = T1x * T1y;
Chris@82 188 T1E = T1x * T1D;
Chris@82 189 T1A = W[3];
Chris@82 190 rio[WS(vs, 2) + WS(rs, 2)] = FMA(T1A, T1D, T1z);
Chris@82 191 iio[WS(vs, 2) + WS(rs, 2)] = FNMS(T1A, T1y, T1E);
Chris@82 192 }
Chris@82 193 {
Chris@82 194 E Tj, Te, Tk, T7, Td;
Chris@82 195 Tj = Tf - Ti;
Chris@82 196 Te = W[5];
Chris@82 197 Tk = Te * Tc;
Chris@82 198 T7 = W[4];
Chris@82 199 Td = T7 * Tc;
Chris@82 200 iio[WS(vs, 3)] = FNMS(Te, Tj, Td);
Chris@82 201 rio[WS(vs, 3)] = FMA(T7, Tj, Tk);
Chris@82 202 }
Chris@82 203 {
Chris@82 204 E T1p, T1k, T1q, T1d, T1j;
Chris@82 205 T1p = T1l - T1o;
Chris@82 206 T1k = W[5];
Chris@82 207 T1q = T1k * T1i;
Chris@82 208 T1d = W[4];
Chris@82 209 T1j = T1d * T1i;
Chris@82 210 iio[WS(vs, 3) + WS(rs, 2)] = FNMS(T1k, T1p, T1j);
Chris@82 211 rio[WS(vs, 3) + WS(rs, 2)] = FMA(T1d, T1p, T1q);
Chris@82 212 }
Chris@82 213 {
Chris@82 214 E T23, T22, T24, T1Z, T21;
Chris@82 215 T23 = T1T + T1W;
Chris@82 216 T22 = W[1];
Chris@82 217 T24 = T22 * T20;
Chris@82 218 T1Z = W[0];
Chris@82 219 T21 = T1Z * T20;
Chris@82 220 iio[WS(vs, 1) + WS(rs, 3)] = FNMS(T22, T23, T21);
Chris@82 221 rio[WS(vs, 1) + WS(rs, 3)] = FMA(T1Z, T23, T24);
Chris@82 222 }
Chris@82 223 {
Chris@82 224 E TX, TW, TY, TT, TV;
Chris@82 225 TX = TN + TQ;
Chris@82 226 TW = W[1];
Chris@82 227 TY = TW * TU;
Chris@82 228 TT = W[0];
Chris@82 229 TV = TT * TU;
Chris@82 230 iio[WS(vs, 1) + WS(rs, 1)] = FNMS(TW, TX, TV);
Chris@82 231 rio[WS(vs, 1) + WS(rs, 1)] = FMA(TT, TX, TY);
Chris@82 232 }
Chris@82 233 {
Chris@82 234 E TR, TM, TS, TF, TL;
Chris@82 235 TR = TN - TQ;
Chris@82 236 TM = W[5];
Chris@82 237 TS = TM * TK;
Chris@82 238 TF = W[4];
Chris@82 239 TL = TF * TK;
Chris@82 240 iio[WS(vs, 3) + WS(rs, 1)] = FNMS(TM, TR, TL);
Chris@82 241 rio[WS(vs, 3) + WS(rs, 1)] = FMA(TF, TR, TS);
Chris@82 242 }
Chris@82 243 {
Chris@82 244 E Tp, To, Tq, Tl, Tn;
Chris@82 245 Tp = Tf + Ti;
Chris@82 246 To = W[1];
Chris@82 247 Tq = To * Tm;
Chris@82 248 Tl = W[0];
Chris@82 249 Tn = Tl * Tm;
Chris@82 250 iio[WS(vs, 1)] = FNMS(To, Tp, Tn);
Chris@82 251 rio[WS(vs, 1)] = FMA(Tl, Tp, Tq);
Chris@82 252 }
Chris@82 253 {
Chris@82 254 E T1v, T1u, T1w, T1r, T1t;
Chris@82 255 T1v = T1l + T1o;
Chris@82 256 T1u = W[1];
Chris@82 257 T1w = T1u * T1s;
Chris@82 258 T1r = W[0];
Chris@82 259 T1t = T1r * T1s;
Chris@82 260 iio[WS(vs, 1) + WS(rs, 2)] = FNMS(T1u, T1v, T1t);
Chris@82 261 rio[WS(vs, 1) + WS(rs, 2)] = FMA(T1r, T1v, T1w);
Chris@82 262 }
Chris@82 263 {
Chris@82 264 E T1X, T1S, T1Y, T1L, T1R;
Chris@82 265 T1X = T1T - T1W;
Chris@82 266 T1S = W[5];
Chris@82 267 T1Y = T1S * T1Q;
Chris@82 268 T1L = W[4];
Chris@82 269 T1R = T1L * T1Q;
Chris@82 270 iio[WS(vs, 3) + WS(rs, 3)] = FNMS(T1S, T1X, T1R);
Chris@82 271 rio[WS(vs, 3) + WS(rs, 3)] = FMA(T1L, T1X, T1Y);
Chris@82 272 }
Chris@82 273 }
Chris@82 274 }
Chris@82 275 }
Chris@82 276
Chris@82 277 static const tw_instr twinstr[] = {
Chris@82 278 {TW_FULL, 0, 4},
Chris@82 279 {TW_NEXT, 1, 0}
Chris@82 280 };
Chris@82 281
Chris@82 282 static const ct_desc desc = { 4, "q1_4", twinstr, &GENUS, {64, 24, 24, 0}, 0, 0, 0 };
Chris@82 283
Chris@82 284 void X(codelet_q1_4) (planner *p) {
Chris@82 285 X(kdft_difsq_register) (p, q1_4, &desc);
Chris@82 286 }
Chris@82 287 #else
Chris@82 288
Chris@82 289 /* Generated by: ../../../genfft/gen_twidsq.native -compact -variables 4 -pipeline-latency 4 -reload-twiddle -dif -n 4 -name q1_4 -include dft/scalar/q.h */
Chris@82 290
Chris@82 291 /*
Chris@82 292 * This function contains 88 FP additions, 48 FP multiplications,
Chris@82 293 * (or, 64 additions, 24 multiplications, 24 fused multiply/add),
Chris@82 294 * 37 stack variables, 0 constants, and 64 memory accesses
Chris@82 295 */
Chris@82 296 #include "dft/scalar/q.h"
Chris@82 297
Chris@82 298 static void q1_4(R *rio, R *iio, const R *W, stride rs, stride vs, INT mb, INT me, INT ms)
Chris@82 299 {
Chris@82 300 {
Chris@82 301 INT m;
Chris@82 302 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@82 303 E T3, Te, Tb, Tq, T6, T8, Th, Tr, Tv, TG, TD, TS, Ty, TA, TJ;
Chris@82 304 E TT, TX, T18, T15, T1k, T10, T12, T1b, T1l, T1p, T1A, T1x, T1M, T1s, T1u;
Chris@82 305 E T1D, T1N;
Chris@82 306 {
Chris@82 307 E T1, T2, T9, Ta;
Chris@82 308 T1 = rio[0];
Chris@82 309 T2 = rio[WS(rs, 2)];
Chris@82 310 T3 = T1 + T2;
Chris@82 311 Te = T1 - T2;
Chris@82 312 T9 = iio[0];
Chris@82 313 Ta = iio[WS(rs, 2)];
Chris@82 314 Tb = T9 - Ta;
Chris@82 315 Tq = T9 + Ta;
Chris@82 316 }
Chris@82 317 {
Chris@82 318 E T4, T5, Tf, Tg;
Chris@82 319 T4 = rio[WS(rs, 1)];
Chris@82 320 T5 = rio[WS(rs, 3)];
Chris@82 321 T6 = T4 + T5;
Chris@82 322 T8 = T4 - T5;
Chris@82 323 Tf = iio[WS(rs, 1)];
Chris@82 324 Tg = iio[WS(rs, 3)];
Chris@82 325 Th = Tf - Tg;
Chris@82 326 Tr = Tf + Tg;
Chris@82 327 }
Chris@82 328 {
Chris@82 329 E Tt, Tu, TB, TC;
Chris@82 330 Tt = rio[WS(vs, 1)];
Chris@82 331 Tu = rio[WS(vs, 1) + WS(rs, 2)];
Chris@82 332 Tv = Tt + Tu;
Chris@82 333 TG = Tt - Tu;
Chris@82 334 TB = iio[WS(vs, 1)];
Chris@82 335 TC = iio[WS(vs, 1) + WS(rs, 2)];
Chris@82 336 TD = TB - TC;
Chris@82 337 TS = TB + TC;
Chris@82 338 }
Chris@82 339 {
Chris@82 340 E Tw, Tx, TH, TI;
Chris@82 341 Tw = rio[WS(vs, 1) + WS(rs, 1)];
Chris@82 342 Tx = rio[WS(vs, 1) + WS(rs, 3)];
Chris@82 343 Ty = Tw + Tx;
Chris@82 344 TA = Tw - Tx;
Chris@82 345 TH = iio[WS(vs, 1) + WS(rs, 1)];
Chris@82 346 TI = iio[WS(vs, 1) + WS(rs, 3)];
Chris@82 347 TJ = TH - TI;
Chris@82 348 TT = TH + TI;
Chris@82 349 }
Chris@82 350 {
Chris@82 351 E TV, TW, T13, T14;
Chris@82 352 TV = rio[WS(vs, 2)];
Chris@82 353 TW = rio[WS(vs, 2) + WS(rs, 2)];
Chris@82 354 TX = TV + TW;
Chris@82 355 T18 = TV - TW;
Chris@82 356 T13 = iio[WS(vs, 2)];
Chris@82 357 T14 = iio[WS(vs, 2) + WS(rs, 2)];
Chris@82 358 T15 = T13 - T14;
Chris@82 359 T1k = T13 + T14;
Chris@82 360 }
Chris@82 361 {
Chris@82 362 E TY, TZ, T19, T1a;
Chris@82 363 TY = rio[WS(vs, 2) + WS(rs, 1)];
Chris@82 364 TZ = rio[WS(vs, 2) + WS(rs, 3)];
Chris@82 365 T10 = TY + TZ;
Chris@82 366 T12 = TY - TZ;
Chris@82 367 T19 = iio[WS(vs, 2) + WS(rs, 1)];
Chris@82 368 T1a = iio[WS(vs, 2) + WS(rs, 3)];
Chris@82 369 T1b = T19 - T1a;
Chris@82 370 T1l = T19 + T1a;
Chris@82 371 }
Chris@82 372 {
Chris@82 373 E T1n, T1o, T1v, T1w;
Chris@82 374 T1n = rio[WS(vs, 3)];
Chris@82 375 T1o = rio[WS(vs, 3) + WS(rs, 2)];
Chris@82 376 T1p = T1n + T1o;
Chris@82 377 T1A = T1n - T1o;
Chris@82 378 T1v = iio[WS(vs, 3)];
Chris@82 379 T1w = iio[WS(vs, 3) + WS(rs, 2)];
Chris@82 380 T1x = T1v - T1w;
Chris@82 381 T1M = T1v + T1w;
Chris@82 382 }
Chris@82 383 {
Chris@82 384 E T1q, T1r, T1B, T1C;
Chris@82 385 T1q = rio[WS(vs, 3) + WS(rs, 1)];
Chris@82 386 T1r = rio[WS(vs, 3) + WS(rs, 3)];
Chris@82 387 T1s = T1q + T1r;
Chris@82 388 T1u = T1q - T1r;
Chris@82 389 T1B = iio[WS(vs, 3) + WS(rs, 1)];
Chris@82 390 T1C = iio[WS(vs, 3) + WS(rs, 3)];
Chris@82 391 T1D = T1B - T1C;
Chris@82 392 T1N = T1B + T1C;
Chris@82 393 }
Chris@82 394 rio[0] = T3 + T6;
Chris@82 395 iio[0] = Tq + Tr;
Chris@82 396 rio[WS(rs, 1)] = Tv + Ty;
Chris@82 397 iio[WS(rs, 1)] = TS + TT;
Chris@82 398 rio[WS(rs, 2)] = TX + T10;
Chris@82 399 iio[WS(rs, 2)] = T1k + T1l;
Chris@82 400 iio[WS(rs, 3)] = T1M + T1N;
Chris@82 401 rio[WS(rs, 3)] = T1p + T1s;
Chris@82 402 {
Chris@82 403 E Tc, Ti, T7, Td;
Chris@82 404 Tc = T8 + Tb;
Chris@82 405 Ti = Te - Th;
Chris@82 406 T7 = W[4];
Chris@82 407 Td = W[5];
Chris@82 408 iio[WS(vs, 3)] = FNMS(Td, Ti, T7 * Tc);
Chris@82 409 rio[WS(vs, 3)] = FMA(Td, Tc, T7 * Ti);
Chris@82 410 }
Chris@82 411 {
Chris@82 412 E T1K, T1O, T1J, T1L;
Chris@82 413 T1K = T1p - T1s;
Chris@82 414 T1O = T1M - T1N;
Chris@82 415 T1J = W[2];
Chris@82 416 T1L = W[3];
Chris@82 417 rio[WS(vs, 2) + WS(rs, 3)] = FMA(T1J, T1K, T1L * T1O);
Chris@82 418 iio[WS(vs, 2) + WS(rs, 3)] = FNMS(T1L, T1K, T1J * T1O);
Chris@82 419 }
Chris@82 420 {
Chris@82 421 E Tk, Tm, Tj, Tl;
Chris@82 422 Tk = Tb - T8;
Chris@82 423 Tm = Te + Th;
Chris@82 424 Tj = W[0];
Chris@82 425 Tl = W[1];
Chris@82 426 iio[WS(vs, 1)] = FNMS(Tl, Tm, Tj * Tk);
Chris@82 427 rio[WS(vs, 1)] = FMA(Tl, Tk, Tj * Tm);
Chris@82 428 }
Chris@82 429 {
Chris@82 430 E To, Ts, Tn, Tp;
Chris@82 431 To = T3 - T6;
Chris@82 432 Ts = Tq - Tr;
Chris@82 433 Tn = W[2];
Chris@82 434 Tp = W[3];
Chris@82 435 rio[WS(vs, 2)] = FMA(Tn, To, Tp * Ts);
Chris@82 436 iio[WS(vs, 2)] = FNMS(Tp, To, Tn * Ts);
Chris@82 437 }
Chris@82 438 {
Chris@82 439 E T16, T1c, T11, T17;
Chris@82 440 T16 = T12 + T15;
Chris@82 441 T1c = T18 - T1b;
Chris@82 442 T11 = W[4];
Chris@82 443 T17 = W[5];
Chris@82 444 iio[WS(vs, 3) + WS(rs, 2)] = FNMS(T17, T1c, T11 * T16);
Chris@82 445 rio[WS(vs, 3) + WS(rs, 2)] = FMA(T17, T16, T11 * T1c);
Chris@82 446 }
Chris@82 447 {
Chris@82 448 E T1G, T1I, T1F, T1H;
Chris@82 449 T1G = T1x - T1u;
Chris@82 450 T1I = T1A + T1D;
Chris@82 451 T1F = W[0];
Chris@82 452 T1H = W[1];
Chris@82 453 iio[WS(vs, 1) + WS(rs, 3)] = FNMS(T1H, T1I, T1F * T1G);
Chris@82 454 rio[WS(vs, 1) + WS(rs, 3)] = FMA(T1H, T1G, T1F * T1I);
Chris@82 455 }
Chris@82 456 {
Chris@82 457 E TQ, TU, TP, TR;
Chris@82 458 TQ = Tv - Ty;
Chris@82 459 TU = TS - TT;
Chris@82 460 TP = W[2];
Chris@82 461 TR = W[3];
Chris@82 462 rio[WS(vs, 2) + WS(rs, 1)] = FMA(TP, TQ, TR * TU);
Chris@82 463 iio[WS(vs, 2) + WS(rs, 1)] = FNMS(TR, TQ, TP * TU);
Chris@82 464 }
Chris@82 465 {
Chris@82 466 E T1e, T1g, T1d, T1f;
Chris@82 467 T1e = T15 - T12;
Chris@82 468 T1g = T18 + T1b;
Chris@82 469 T1d = W[0];
Chris@82 470 T1f = W[1];
Chris@82 471 iio[WS(vs, 1) + WS(rs, 2)] = FNMS(T1f, T1g, T1d * T1e);
Chris@82 472 rio[WS(vs, 1) + WS(rs, 2)] = FMA(T1f, T1e, T1d * T1g);
Chris@82 473 }
Chris@82 474 {
Chris@82 475 E T1i, T1m, T1h, T1j;
Chris@82 476 T1i = TX - T10;
Chris@82 477 T1m = T1k - T1l;
Chris@82 478 T1h = W[2];
Chris@82 479 T1j = W[3];
Chris@82 480 rio[WS(vs, 2) + WS(rs, 2)] = FMA(T1h, T1i, T1j * T1m);
Chris@82 481 iio[WS(vs, 2) + WS(rs, 2)] = FNMS(T1j, T1i, T1h * T1m);
Chris@82 482 }
Chris@82 483 {
Chris@82 484 E T1y, T1E, T1t, T1z;
Chris@82 485 T1y = T1u + T1x;
Chris@82 486 T1E = T1A - T1D;
Chris@82 487 T1t = W[4];
Chris@82 488 T1z = W[5];
Chris@82 489 iio[WS(vs, 3) + WS(rs, 3)] = FNMS(T1z, T1E, T1t * T1y);
Chris@82 490 rio[WS(vs, 3) + WS(rs, 3)] = FMA(T1z, T1y, T1t * T1E);
Chris@82 491 }
Chris@82 492 {
Chris@82 493 E TM, TO, TL, TN;
Chris@82 494 TM = TD - TA;
Chris@82 495 TO = TG + TJ;
Chris@82 496 TL = W[0];
Chris@82 497 TN = W[1];
Chris@82 498 iio[WS(vs, 1) + WS(rs, 1)] = FNMS(TN, TO, TL * TM);
Chris@82 499 rio[WS(vs, 1) + WS(rs, 1)] = FMA(TN, TM, TL * TO);
Chris@82 500 }
Chris@82 501 {
Chris@82 502 E TE, TK, Tz, TF;
Chris@82 503 TE = TA + TD;
Chris@82 504 TK = TG - TJ;
Chris@82 505 Tz = W[4];
Chris@82 506 TF = W[5];
Chris@82 507 iio[WS(vs, 3) + WS(rs, 1)] = FNMS(TF, TK, Tz * TE);
Chris@82 508 rio[WS(vs, 3) + WS(rs, 1)] = FMA(TF, TE, Tz * TK);
Chris@82 509 }
Chris@82 510 }
Chris@82 511 }
Chris@82 512 }
Chris@82 513
Chris@82 514 static const tw_instr twinstr[] = {
Chris@82 515 {TW_FULL, 0, 4},
Chris@82 516 {TW_NEXT, 1, 0}
Chris@82 517 };
Chris@82 518
Chris@82 519 static const ct_desc desc = { 4, "q1_4", twinstr, &GENUS, {64, 24, 24, 0}, 0, 0, 0 };
Chris@82 520
Chris@82 521 void X(codelet_q1_4) (planner *p) {
Chris@82 522 X(kdft_difsq_register) (p, q1_4, &desc);
Chris@82 523 }
Chris@82 524 #endif