Mercurial > hg > sv-dependency-builds

annotate src/fftw-3.3.8/dft/simd/common/t1fv_32.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 08:05:29 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_twiddle_c.native -fma -simd -compact -variables 4 -pipeline-latency 8 -n 32 -name t1fv_32 -include dft/simd/t1f.h */
Chris@82 29
Chris@82 30 /*
Chris@82 31 * This function contains 217 FP additions, 160 FP multiplications,
Chris@82 32 * (or, 119 additions, 62 multiplications, 98 fused multiply/add),
Chris@82 33 * 59 stack variables, 7 constants, and 64 memory accesses
Chris@82 34 */
Chris@82 35 #include "dft/simd/t1f.h"
Chris@82 36
Chris@82 37 static void t1fv_32(R *ri, R *ii, const R *W, stride rs, INT mb, INT me, INT ms)
Chris@82 38 {
Chris@82 39 DVK(KP831469612, +0.831469612302545237078788377617905756738560812);
Chris@82 40 DVK(KP668178637, +0.668178637919298919997757686523080761552472251);
Chris@82 41 DVK(KP980785280, +0.980785280403230449126182236134239036973933731);
Chris@82 42 DVK(KP198912367, +0.198912367379658006911597622644676228597850501);
Chris@82 43 DVK(KP923879532, +0.923879532511286756128183189396788286822416626);
Chris@82 44 DVK(KP707106781, +0.707106781186547524400844362104849039284835938);
Chris@82 45 DVK(KP414213562, +0.414213562373095048801688724209698078569671875);
Chris@82 46 {
Chris@82 47 INT m;
Chris@82 48 R *x;
Chris@82 49 x = ri;
Chris@82 50 for (m = mb, W = W + (mb * ((TWVL / VL) * 62)); m < me; m = m + VL, x = x + (VL * ms), W = W + (TWVL * 62), MAKE_VOLATILE_STRIDE(32, rs)) {
Chris@82 51 V T4, T1z, T2o, T32, Tf, T1A, T2r, T3f, TC, T1D, T2L, T34, Tr, T1C, T2O;
Chris@82 52 V T33, T1k, T20, T2F, T3b, T1r, T21, T2C, T3a, TV, T1X, T2y, T38, T12, T1Y;
Chris@82 53 V T2v, T37;
Chris@82 54 {
Chris@82 55 V T1, T1y, T3, T1w, T1x, T2, T1v, T2m, T2n;
Chris@82 56 T1 = LD(&(x[0]), ms, &(x[0]));
Chris@82 57 T1x = LD(&(x[WS(rs, 24)]), ms, &(x[0]));
Chris@82 58 T1y = BYTWJ(&(W[TWVL * 46]), T1x);
Chris@82 59 T2 = LD(&(x[WS(rs, 16)]), ms, &(x[0]));
Chris@82 60 T3 = BYTWJ(&(W[TWVL * 30]), T2);
Chris@82 61 T1v = LD(&(x[WS(rs, 8)]), ms, &(x[0]));
Chris@82 62 T1w = BYTWJ(&(W[TWVL * 14]), T1v);
Chris@82 63 T4 = VSUB(T1, T3);
Chris@82 64 T1z = VSUB(T1w, T1y);
Chris@82 65 T2m = VADD(T1, T3);
Chris@82 66 T2n = VADD(T1w, T1y);
Chris@82 67 T2o = VADD(T2m, T2n);
Chris@82 68 T32 = VSUB(T2m, T2n);
Chris@82 69 }
Chris@82 70 {
Chris@82 71 V T6, Td, T8, Tb;
Chris@82 72 {
Chris@82 73 V T5, Tc, T7, Ta;
Chris@82 74 T5 = LD(&(x[WS(rs, 4)]), ms, &(x[0]));
Chris@82 75 T6 = BYTWJ(&(W[TWVL * 6]), T5);
Chris@82 76 Tc = LD(&(x[WS(rs, 12)]), ms, &(x[0]));
Chris@82 77 Td = BYTWJ(&(W[TWVL * 22]), Tc);
Chris@82 78 T7 = LD(&(x[WS(rs, 20)]), ms, &(x[0]));
Chris@82 79 T8 = BYTWJ(&(W[TWVL * 38]), T7);
Chris@82 80 Ta = LD(&(x[WS(rs, 28)]), ms, &(x[0]));
Chris@82 81 Tb = BYTWJ(&(W[TWVL * 54]), Ta);
Chris@82 82 }
Chris@82 83 {
Chris@82 84 V T9, Te, T2p, T2q;
Chris@82 85 T9 = VSUB(T6, T8);
Chris@82 86 Te = VSUB(Tb, Td);
Chris@82 87 Tf = VADD(T9, Te);
Chris@82 88 T1A = VSUB(Te, T9);
Chris@82 89 T2p = VADD(T6, T8);
Chris@82 90 T2q = VADD(Tb, Td);
Chris@82 91 T2r = VADD(T2p, T2q);
Chris@82 92 T3f = VSUB(T2q, T2p);
Chris@82 93 }
Chris@82 94 }
Chris@82 95 {
Chris@82 96 V Tt, TA, Tv, Ty;
Chris@82 97 {
Chris@82 98 V Ts, Tz, Tu, Tx;
Chris@82 99 Ts = LD(&(x[WS(rs, 30)]), ms, &(x[0]));
Chris@82 100 Tt = BYTWJ(&(W[TWVL * 58]), Ts);
Chris@82 101 Tz = LD(&(x[WS(rs, 6)]), ms, &(x[0]));
Chris@82 102 TA = BYTWJ(&(W[TWVL * 10]), Tz);
Chris@82 103 Tu = LD(&(x[WS(rs, 14)]), ms, &(x[0]));
Chris@82 104 Tv = BYTWJ(&(W[TWVL * 26]), Tu);
Chris@82 105 Tx = LD(&(x[WS(rs, 22)]), ms, &(x[0]));
Chris@82 106 Ty = BYTWJ(&(W[TWVL * 42]), Tx);
Chris@82 107 }
Chris@82 108 {
Chris@82 109 V Tw, TB, T2J, T2K;
Chris@82 110 Tw = VSUB(Tt, Tv);
Chris@82 111 TB = VSUB(Ty, TA);
Chris@82 112 TC = VFNMS(LDK(KP414213562), TB, Tw);
Chris@82 113 T1D = VFMA(LDK(KP414213562), Tw, TB);
Chris@82 114 T2J = VADD(Tt, Tv);
Chris@82 115 T2K = VADD(TA, Ty);
Chris@82 116 T2L = VADD(T2J, T2K);
Chris@82 117 T34 = VSUB(T2J, T2K);
Chris@82 118 }
Chris@82 119 }
Chris@82 120 {
Chris@82 121 V Ti, Tp, Tk, Tn;
Chris@82 122 {
Chris@82 123 V Th, To, Tj, Tm;
Chris@82 124 Th = LD(&(x[WS(rs, 2)]), ms, &(x[0]));
Chris@82 125 Ti = BYTWJ(&(W[TWVL * 2]), Th);
Chris@82 126 To = LD(&(x[WS(rs, 26)]), ms, &(x[0]));
Chris@82 127 Tp = BYTWJ(&(W[TWVL * 50]), To);
Chris@82 128 Tj = LD(&(x[WS(rs, 18)]), ms, &(x[0]));
Chris@82 129 Tk = BYTWJ(&(W[TWVL * 34]), Tj);
Chris@82 130 Tm = LD(&(x[WS(rs, 10)]), ms, &(x[0]));
Chris@82 131 Tn = BYTWJ(&(W[TWVL * 18]), Tm);
Chris@82 132 }
Chris@82 133 {
Chris@82 134 V Tl, Tq, T2M, T2N;
Chris@82 135 Tl = VSUB(Ti, Tk);
Chris@82 136 Tq = VSUB(Tn, Tp);
Chris@82 137 Tr = VFNMS(LDK(KP414213562), Tq, Tl);
Chris@82 138 T1C = VFMA(LDK(KP414213562), Tl, Tq);
Chris@82 139 T2M = VADD(Ti, Tk);
Chris@82 140 T2N = VADD(Tn, Tp);
Chris@82 141 T2O = VADD(T2M, T2N);
Chris@82 142 T33 = VSUB(T2M, T2N);
Chris@82 143 }
Chris@82 144 }
Chris@82 145 {
Chris@82 146 V T15, T17, T1o, T1m, T1f, T1h, T1i, T1a, T1c, T1d;
Chris@82 147 {
Chris@82 148 V T14, T16, T1n, T1l;
Chris@82 149 T14 = LD(&(x[WS(rs, 31)]), ms, &(x[WS(rs, 1)]));
Chris@82 150 T15 = BYTWJ(&(W[TWVL * 60]), T14);
Chris@82 151 T16 = LD(&(x[WS(rs, 15)]), ms, &(x[WS(rs, 1)]));
Chris@82 152 T17 = BYTWJ(&(W[TWVL * 28]), T16);
Chris@82 153 T1n = LD(&(x[WS(rs, 7)]), ms, &(x[WS(rs, 1)]));
Chris@82 154 T1o = BYTWJ(&(W[TWVL * 12]), T1n);
Chris@82 155 T1l = LD(&(x[WS(rs, 23)]), ms, &(x[WS(rs, 1)]));
Chris@82 156 T1m = BYTWJ(&(W[TWVL * 44]), T1l);
Chris@82 157 {
Chris@82 158 V T1e, T1g, T19, T1b;
Chris@82 159 T1e = LD(&(x[WS(rs, 27)]), ms, &(x[WS(rs, 1)]));
Chris@82 160 T1f = BYTWJ(&(W[TWVL * 52]), T1e);
Chris@82 161 T1g = LD(&(x[WS(rs, 11)]), ms, &(x[WS(rs, 1)]));
Chris@82 162 T1h = BYTWJ(&(W[TWVL * 20]), T1g);
Chris@82 163 T1i = VSUB(T1f, T1h);
Chris@82 164 T19 = LD(&(x[WS(rs, 3)]), ms, &(x[WS(rs, 1)]));
Chris@82 165 T1a = BYTWJ(&(W[TWVL * 4]), T19);
Chris@82 166 T1b = LD(&(x[WS(rs, 19)]), ms, &(x[WS(rs, 1)]));
Chris@82 167 T1c = BYTWJ(&(W[TWVL * 36]), T1b);
Chris@82 168 T1d = VSUB(T1a, T1c);
Chris@82 169 }
Chris@82 170 }
Chris@82 171 {
Chris@82 172 V T18, T1j, T2D, T2E;
Chris@82 173 T18 = VSUB(T15, T17);
Chris@82 174 T1j = VADD(T1d, T1i);
Chris@82 175 T1k = VFMA(LDK(KP707106781), T1j, T18);
Chris@82 176 T20 = VFNMS(LDK(KP707106781), T1j, T18);
Chris@82 177 T2D = VADD(T1a, T1c);
Chris@82 178 T2E = VADD(T1f, T1h);
Chris@82 179 T2F = VADD(T2D, T2E);
Chris@82 180 T3b = VSUB(T2E, T2D);
Chris@82 181 }
Chris@82 182 {
Chris@82 183 V T1p, T1q, T2A, T2B;
Chris@82 184 T1p = VSUB(T1m, T1o);
Chris@82 185 T1q = VSUB(T1i, T1d);
Chris@82 186 T1r = VFMA(LDK(KP707106781), T1q, T1p);
Chris@82 187 T21 = VFNMS(LDK(KP707106781), T1q, T1p);
Chris@82 188 T2A = VADD(T15, T17);
Chris@82 189 T2B = VADD(T1o, T1m);
Chris@82 190 T2C = VADD(T2A, T2B);
Chris@82 191 T3a = VSUB(T2A, T2B);
Chris@82 192 }
Chris@82 193 }
Chris@82 194 {
Chris@82 195 V TG, TI, TZ, TX, TQ, TS, TT, TL, TN, TO;
Chris@82 196 {
Chris@82 197 V TF, TH, TY, TW;
Chris@82 198 TF = LD(&(x[WS(rs, 1)]), ms, &(x[WS(rs, 1)]));
Chris@82 199 TG = BYTWJ(&(W[0]), TF);
Chris@82 200 TH = LD(&(x[WS(rs, 17)]), ms, &(x[WS(rs, 1)]));
Chris@82 201 TI = BYTWJ(&(W[TWVL * 32]), TH);
Chris@82 202 TY = LD(&(x[WS(rs, 25)]), ms, &(x[WS(rs, 1)]));
Chris@82 203 TZ = BYTWJ(&(W[TWVL * 48]), TY);
Chris@82 204 TW = LD(&(x[WS(rs, 9)]), ms, &(x[WS(rs, 1)]));
Chris@82 205 TX = BYTWJ(&(W[TWVL * 16]), TW);
Chris@82 206 {
Chris@82 207 V TP, TR, TK, TM;
Chris@82 208 TP = LD(&(x[WS(rs, 29)]), ms, &(x[WS(rs, 1)]));
Chris@82 209 TQ = BYTWJ(&(W[TWVL * 56]), TP);
Chris@82 210 TR = LD(&(x[WS(rs, 13)]), ms, &(x[WS(rs, 1)]));
Chris@82 211 TS = BYTWJ(&(W[TWVL * 24]), TR);
Chris@82 212 TT = VSUB(TQ, TS);
Chris@82 213 TK = LD(&(x[WS(rs, 5)]), ms, &(x[WS(rs, 1)]));
Chris@82 214 TL = BYTWJ(&(W[TWVL * 8]), TK);
Chris@82 215 TM = LD(&(x[WS(rs, 21)]), ms, &(x[WS(rs, 1)]));
Chris@82 216 TN = BYTWJ(&(W[TWVL * 40]), TM);
Chris@82 217 TO = VSUB(TL, TN);
Chris@82 218 }
Chris@82 219 }
Chris@82 220 {
Chris@82 221 V TJ, TU, T2w, T2x;
Chris@82 222 TJ = VSUB(TG, TI);
Chris@82 223 TU = VADD(TO, TT);
Chris@82 224 TV = VFMA(LDK(KP707106781), TU, TJ);
Chris@82 225 T1X = VFNMS(LDK(KP707106781), TU, TJ);
Chris@82 226 T2w = VADD(TL, TN);
Chris@82 227 T2x = VADD(TQ, TS);
Chris@82 228 T2y = VADD(T2w, T2x);
Chris@82 229 T38 = VSUB(T2w, T2x);
Chris@82 230 }
Chris@82 231 {
Chris@82 232 V T10, T11, T2t, T2u;
Chris@82 233 T10 = VSUB(TX, TZ);
Chris@82 234 T11 = VSUB(TO, TT);
Chris@82 235 T12 = VFMA(LDK(KP707106781), T11, T10);
Chris@82 236 T1Y = VFNMS(LDK(KP707106781), T11, T10);
Chris@82 237 T2t = VADD(TG, TI);
Chris@82 238 T2u = VADD(TX, TZ);
Chris@82 239 T2v = VADD(T2t, T2u);
Chris@82 240 T37 = VSUB(T2t, T2u);
Chris@82 241 }
Chris@82 242 }
Chris@82 243 {
Chris@82 244 V T2W, T30, T2Z, T31;
Chris@82 245 {
Chris@82 246 V T2U, T2V, T2X, T2Y;
Chris@82 247 T2U = VADD(T2o, T2r);
Chris@82 248 T2V = VADD(T2O, T2L);
Chris@82 249 T2W = VADD(T2U, T2V);
Chris@82 250 T30 = VSUB(T2U, T2V);
Chris@82 251 T2X = VADD(T2v, T2y);
Chris@82 252 T2Y = VADD(T2C, T2F);
Chris@82 253 T2Z = VADD(T2X, T2Y);
Chris@82 254 T31 = VSUB(T2Y, T2X);
Chris@82 255 }
Chris@82 256 ST(&(x[WS(rs, 16)]), VSUB(T2W, T2Z), ms, &(x[0]));
Chris@82 257 ST(&(x[WS(rs, 8)]), VFMAI(T31, T30), ms, &(x[0]));
Chris@82 258 ST(&(x[0]), VADD(T2W, T2Z), ms, &(x[0]));
Chris@82 259 ST(&(x[WS(rs, 24)]), VFNMSI(T31, T30), ms, &(x[0]));
Chris@82 260 }
Chris@82 261 {
Chris@82 262 V T2s, T2P, T2H, T2Q, T2z, T2G;
Chris@82 263 T2s = VSUB(T2o, T2r);
Chris@82 264 T2P = VSUB(T2L, T2O);
Chris@82 265 T2z = VSUB(T2v, T2y);
Chris@82 266 T2G = VSUB(T2C, T2F);
Chris@82 267 T2H = VADD(T2z, T2G);
Chris@82 268 T2Q = VSUB(T2G, T2z);
Chris@82 269 {
Chris@82 270 V T2I, T2R, T2S, T2T;
Chris@82 271 T2I = VFNMS(LDK(KP707106781), T2H, T2s);
Chris@82 272 T2R = VFNMS(LDK(KP707106781), T2Q, T2P);
Chris@82 273 ST(&(x[WS(rs, 12)]), VFNMSI(T2R, T2I), ms, &(x[0]));
Chris@82 274 ST(&(x[WS(rs, 20)]), VFMAI(T2R, T2I), ms, &(x[0]));
Chris@82 275 T2S = VFMA(LDK(KP707106781), T2H, T2s);
Chris@82 276 T2T = VFMA(LDK(KP707106781), T2Q, T2P);
Chris@82 277 ST(&(x[WS(rs, 28)]), VFNMSI(T2T, T2S), ms, &(x[0]));
Chris@82 278 ST(&(x[WS(rs, 4)]), VFMAI(T2T, T2S), ms, &(x[0]));
Chris@82 279 }
Chris@82 280 }
Chris@82 281 {
Chris@82 282 V T36, T3o, T3h, T3r, T3d, T3s, T3k, T3p, T35, T3g;
Chris@82 283 T35 = VADD(T33, T34);
Chris@82 284 T36 = VFMA(LDK(KP707106781), T35, T32);
Chris@82 285 T3o = VFNMS(LDK(KP707106781), T35, T32);
Chris@82 286 T3g = VSUB(T34, T33);
Chris@82 287 T3h = VFMA(LDK(KP707106781), T3g, T3f);
Chris@82 288 T3r = VFNMS(LDK(KP707106781), T3g, T3f);
Chris@82 289 {
Chris@82 290 V T39, T3c, T3i, T3j;
Chris@82 291 T39 = VFNMS(LDK(KP414213562), T38, T37);
Chris@82 292 T3c = VFNMS(LDK(KP414213562), T3b, T3a);
Chris@82 293 T3d = VADD(T39, T3c);
Chris@82 294 T3s = VSUB(T3c, T39);
Chris@82 295 T3i = VFMA(LDK(KP414213562), T3a, T3b);
Chris@82 296 T3j = VFMA(LDK(KP414213562), T37, T38);
Chris@82 297 T3k = VSUB(T3i, T3j);
Chris@82 298 T3p = VADD(T3j, T3i);
Chris@82 299 }
Chris@82 300 {
Chris@82 301 V T3e, T3l, T3u, T3v;
Chris@82 302 T3e = VFNMS(LDK(KP923879532), T3d, T36);
Chris@82 303 T3l = VFNMS(LDK(KP923879532), T3k, T3h);
Chris@82 304 ST(&(x[WS(rs, 14)]), VFNMSI(T3l, T3e), ms, &(x[0]));
Chris@82 305 ST(&(x[WS(rs, 18)]), VFMAI(T3l, T3e), ms, &(x[0]));
Chris@82 306 T3u = VFMA(LDK(KP923879532), T3p, T3o);
Chris@82 307 T3v = VFNMS(LDK(KP923879532), T3s, T3r);
Chris@82 308 ST(&(x[WS(rs, 6)]), VFNMSI(T3v, T3u), ms, &(x[0]));
Chris@82 309 ST(&(x[WS(rs, 26)]), VFMAI(T3v, T3u), ms, &(x[0]));
Chris@82 310 }
Chris@82 311 {
Chris@82 312 V T3m, T3n, T3q, T3t;
Chris@82 313 T3m = VFMA(LDK(KP923879532), T3d, T36);
Chris@82 314 T3n = VFMA(LDK(KP923879532), T3k, T3h);
Chris@82 315 ST(&(x[WS(rs, 30)]), VFNMSI(T3n, T3m), ms, &(x[0]));
Chris@82 316 ST(&(x[WS(rs, 2)]), VFMAI(T3n, T3m), ms, &(x[0]));
Chris@82 317 T3q = VFNMS(LDK(KP923879532), T3p, T3o);
Chris@82 318 T3t = VFMA(LDK(KP923879532), T3s, T3r);
Chris@82 319 ST(&(x[WS(rs, 10)]), VFMAI(T3t, T3q), ms, &(x[0]));
Chris@82 320 ST(&(x[WS(rs, 22)]), VFNMSI(T3t, T3q), ms, &(x[0]));
Chris@82 321 }
Chris@82 322 }
Chris@82 323 {
Chris@82 324 V TE, T1M, T1I, T1N, T1t, T1Q, T1F, T1P;
Chris@82 325 {
Chris@82 326 V Tg, TD, T1G, T1H;
Chris@82 327 Tg = VFMA(LDK(KP707106781), Tf, T4);
Chris@82 328 TD = VADD(Tr, TC);
Chris@82 329 TE = VFMA(LDK(KP923879532), TD, Tg);
Chris@82 330 T1M = VFNMS(LDK(KP923879532), TD, Tg);
Chris@82 331 T1G = VFMA(LDK(KP198912367), TV, T12);
Chris@82 332 T1H = VFMA(LDK(KP198912367), T1k, T1r);
Chris@82 333 T1I = VSUB(T1G, T1H);
Chris@82 334 T1N = VADD(T1G, T1H);
Chris@82 335 }
Chris@82 336 {
Chris@82 337 V T13, T1s, T1B, T1E;
Chris@82 338 T13 = VFNMS(LDK(KP198912367), T12, TV);
Chris@82 339 T1s = VFNMS(LDK(KP198912367), T1r, T1k);
Chris@82 340 T1t = VADD(T13, T1s);
Chris@82 341 T1Q = VSUB(T1s, T13);
Chris@82 342 T1B = VFNMS(LDK(KP707106781), T1A, T1z);
Chris@82 343 T1E = VSUB(T1C, T1D);
Chris@82 344 T1F = VFMA(LDK(KP923879532), T1E, T1B);
Chris@82 345 T1P = VFNMS(LDK(KP923879532), T1E, T1B);
Chris@82 346 }
Chris@82 347 {
Chris@82 348 V T1u, T1J, T1S, T1T;
Chris@82 349 T1u = VFNMS(LDK(KP980785280), T1t, TE);
Chris@82 350 T1J = VFNMS(LDK(KP980785280), T1I, T1F);
Chris@82 351 ST(&(x[WS(rs, 17)]), VFNMSI(T1J, T1u), ms, &(x[WS(rs, 1)]));
Chris@82 352 ST(&(x[WS(rs, 15)]), VFMAI(T1J, T1u), ms, &(x[WS(rs, 1)]));
Chris@82 353 T1S = VFMA(LDK(KP980785280), T1N, T1M);
Chris@82 354 T1T = VFMA(LDK(KP980785280), T1Q, T1P);
Chris@82 355 ST(&(x[WS(rs, 7)]), VFMAI(T1T, T1S), ms, &(x[WS(rs, 1)]));
Chris@82 356 ST(&(x[WS(rs, 25)]), VFNMSI(T1T, T1S), ms, &(x[WS(rs, 1)]));
Chris@82 357 }
Chris@82 358 {
Chris@82 359 V T1K, T1L, T1O, T1R;
Chris@82 360 T1K = VFMA(LDK(KP980785280), T1t, TE);
Chris@82 361 T1L = VFMA(LDK(KP980785280), T1I, T1F);
Chris@82 362 ST(&(x[WS(rs, 1)]), VFNMSI(T1L, T1K), ms, &(x[WS(rs, 1)]));
Chris@82 363 ST(&(x[WS(rs, 31)]), VFMAI(T1L, T1K), ms, &(x[WS(rs, 1)]));
Chris@82 364 T1O = VFNMS(LDK(KP980785280), T1N, T1M);
Chris@82 365 T1R = VFNMS(LDK(KP980785280), T1Q, T1P);
Chris@82 366 ST(&(x[WS(rs, 9)]), VFNMSI(T1R, T1O), ms, &(x[WS(rs, 1)]));
Chris@82 367 ST(&(x[WS(rs, 23)]), VFMAI(T1R, T1O), ms, &(x[WS(rs, 1)]));
Chris@82 368 }
Chris@82 369 }
Chris@82 370 {
Chris@82 371 V T1W, T2e, T2a, T2f, T23, T2i, T27, T2h;
Chris@82 372 {
Chris@82 373 V T1U, T1V, T28, T29;
Chris@82 374 T1U = VFNMS(LDK(KP707106781), Tf, T4);
Chris@82 375 T1V = VADD(T1C, T1D);
Chris@82 376 T1W = VFMA(LDK(KP923879532), T1V, T1U);
Chris@82 377 T2e = VFNMS(LDK(KP923879532), T1V, T1U);
Chris@82 378 T28 = VFNMS(LDK(KP668178637), T1X, T1Y);
Chris@82 379 T29 = VFNMS(LDK(KP668178637), T20, T21);
Chris@82 380 T2a = VSUB(T28, T29);
Chris@82 381 T2f = VADD(T28, T29);
Chris@82 382 }
Chris@82 383 {
Chris@82 384 V T1Z, T22, T25, T26;
Chris@82 385 T1Z = VFMA(LDK(KP668178637), T1Y, T1X);
Chris@82 386 T22 = VFMA(LDK(KP668178637), T21, T20);
Chris@82 387 T23 = VADD(T1Z, T22);
Chris@82 388 T2i = VSUB(T22, T1Z);
Chris@82 389 T25 = VFMA(LDK(KP707106781), T1A, T1z);
Chris@82 390 T26 = VSUB(TC, Tr);
Chris@82 391 T27 = VFMA(LDK(KP923879532), T26, T25);
Chris@82 392 T2h = VFNMS(LDK(KP923879532), T26, T25);
Chris@82 393 }
Chris@82 394 {
Chris@82 395 V T24, T2b, T2k, T2l;
Chris@82 396 T24 = VFNMS(LDK(KP831469612), T23, T1W);
Chris@82 397 T2b = VFNMS(LDK(KP831469612), T2a, T27);
Chris@82 398 ST(&(x[WS(rs, 13)]), VFNMSI(T2b, T24), ms, &(x[WS(rs, 1)]));
Chris@82 399 ST(&(x[WS(rs, 19)]), VFMAI(T2b, T24), ms, &(x[WS(rs, 1)]));
Chris@82 400 T2k = VFNMS(LDK(KP831469612), T2f, T2e);
Chris@82 401 T2l = VFNMS(LDK(KP831469612), T2i, T2h);
Chris@82 402 ST(&(x[WS(rs, 5)]), VFNMSI(T2l, T2k), ms, &(x[WS(rs, 1)]));
Chris@82 403 ST(&(x[WS(rs, 27)]), VFMAI(T2l, T2k), ms, &(x[WS(rs, 1)]));
Chris@82 404 }
Chris@82 405 {
Chris@82 406 V T2c, T2d, T2g, T2j;
Chris@82 407 T2c = VFMA(LDK(KP831469612), T23, T1W);
Chris@82 408 T2d = VFMA(LDK(KP831469612), T2a, T27);
Chris@82 409 ST(&(x[WS(rs, 29)]), VFNMSI(T2d, T2c), ms, &(x[WS(rs, 1)]));
Chris@82 410 ST(&(x[WS(rs, 3)]), VFMAI(T2d, T2c), ms, &(x[WS(rs, 1)]));
Chris@82 411 T2g = VFMA(LDK(KP831469612), T2f, T2e);
Chris@82 412 T2j = VFMA(LDK(KP831469612), T2i, T2h);
Chris@82 413 ST(&(x[WS(rs, 11)]), VFMAI(T2j, T2g), ms, &(x[WS(rs, 1)]));
Chris@82 414 ST(&(x[WS(rs, 21)]), VFNMSI(T2j, T2g), ms, &(x[WS(rs, 1)]));
Chris@82 415 }
Chris@82 416 }
Chris@82 417 }
Chris@82 418 }
Chris@82 419 VLEAVE();
Chris@82 420 }
Chris@82 421
Chris@82 422 static const tw_instr twinstr[] = {
Chris@82 423 VTW(0, 1),
Chris@82 424 VTW(0, 2),
Chris@82 425 VTW(0, 3),
Chris@82 426 VTW(0, 4),
Chris@82 427 VTW(0, 5),
Chris@82 428 VTW(0, 6),
Chris@82 429 VTW(0, 7),
Chris@82 430 VTW(0, 8),
Chris@82 431 VTW(0, 9),
Chris@82 432 VTW(0, 10),
Chris@82 433 VTW(0, 11),
Chris@82 434 VTW(0, 12),
Chris@82 435 VTW(0, 13),
Chris@82 436 VTW(0, 14),
Chris@82 437 VTW(0, 15),
Chris@82 438 VTW(0, 16),
Chris@82 439 VTW(0, 17),
Chris@82 440 VTW(0, 18),
Chris@82 441 VTW(0, 19),
Chris@82 442 VTW(0, 20),
Chris@82 443 VTW(0, 21),
Chris@82 444 VTW(0, 22),
Chris@82 445 VTW(0, 23),
Chris@82 446 VTW(0, 24),
Chris@82 447 VTW(0, 25),
Chris@82 448 VTW(0, 26),
Chris@82 449 VTW(0, 27),
Chris@82 450 VTW(0, 28),
Chris@82 451 VTW(0, 29),
Chris@82 452 VTW(0, 30),
Chris@82 453 VTW(0, 31),
Chris@82 454 {TW_NEXT, VL, 0}
Chris@82 455 };
Chris@82 456
Chris@82 457 static const ct_desc desc = { 32, XSIMD_STRING("t1fv_32"), twinstr, &GENUS, {119, 62, 98, 0}, 0, 0, 0 };
Chris@82 458
Chris@82 459 void XSIMD(codelet_t1fv_32) (planner *p) {
Chris@82 460 X(kdft_dit_register) (p, t1fv_32, &desc);
Chris@82 461 }
Chris@82 462 #else
Chris@82 463
Chris@82 464 /* Generated by: ../../../genfft/gen_twiddle_c.native -simd -compact -variables 4 -pipeline-latency 8 -n 32 -name t1fv_32 -include dft/simd/t1f.h */
Chris@82 465
Chris@82 466 /*
Chris@82 467 * This function contains 217 FP additions, 104 FP multiplications,
Chris@82 468 * (or, 201 additions, 88 multiplications, 16 fused multiply/add),
Chris@82 469 * 59 stack variables, 7 constants, and 64 memory accesses
Chris@82 470 */
Chris@82 471 #include "dft/simd/t1f.h"
Chris@82 472
Chris@82 473 static void t1fv_32(R *ri, R *ii, const R *W, stride rs, INT mb, INT me, INT ms)
Chris@82 474 {
Chris@82 475 DVK(KP555570233, +0.555570233019602224742830813948532874374937191);
Chris@82 476 DVK(KP831469612, +0.831469612302545237078788377617905756738560812);
Chris@82 477 DVK(KP195090322, +0.195090322016128267848284868477022240927691618);
Chris@82 478 DVK(KP980785280, +0.980785280403230449126182236134239036973933731);
Chris@82 479 DVK(KP382683432, +0.382683432365089771728459984030398866761344562);
Chris@82 480 DVK(KP923879532, +0.923879532511286756128183189396788286822416626);
Chris@82 481 DVK(KP707106781, +0.707106781186547524400844362104849039284835938);
Chris@82 482 {
Chris@82 483 INT m;
Chris@82 484 R *x;
Chris@82 485 x = ri;
Chris@82 486 for (m = mb, W = W + (mb * ((TWVL / VL) * 62)); m < me; m = m + VL, x = x + (VL * ms), W = W + (TWVL * 62), MAKE_VOLATILE_STRIDE(32, rs)) {
Chris@82 487 V T4, T1A, T2o, T32, Tf, T1v, T2r, T3f, TC, T1C, T2L, T34, Tr, T1D, T2O;
Chris@82 488 V T33, T1k, T20, T2F, T3b, T1r, T21, T2C, T3a, TV, T1X, T2y, T38, T12, T1Y;
Chris@82 489 V T2v, T37;
Chris@82 490 {
Chris@82 491 V T1, T1z, T3, T1x, T1y, T2, T1w, T2m, T2n;
Chris@82 492 T1 = LD(&(x[0]), ms, &(x[0]));
Chris@82 493 T1y = LD(&(x[WS(rs, 24)]), ms, &(x[0]));
Chris@82 494 T1z = BYTWJ(&(W[TWVL * 46]), T1y);
Chris@82 495 T2 = LD(&(x[WS(rs, 16)]), ms, &(x[0]));
Chris@82 496 T3 = BYTWJ(&(W[TWVL * 30]), T2);
Chris@82 497 T1w = LD(&(x[WS(rs, 8)]), ms, &(x[0]));
Chris@82 498 T1x = BYTWJ(&(W[TWVL * 14]), T1w);
Chris@82 499 T4 = VSUB(T1, T3);
Chris@82 500 T1A = VSUB(T1x, T1z);
Chris@82 501 T2m = VADD(T1, T3);
Chris@82 502 T2n = VADD(T1x, T1z);
Chris@82 503 T2o = VADD(T2m, T2n);
Chris@82 504 T32 = VSUB(T2m, T2n);
Chris@82 505 }
Chris@82 506 {
Chris@82 507 V T6, Td, T8, Tb;
Chris@82 508 {
Chris@82 509 V T5, Tc, T7, Ta;
Chris@82 510 T5 = LD(&(x[WS(rs, 4)]), ms, &(x[0]));
Chris@82 511 T6 = BYTWJ(&(W[TWVL * 6]), T5);
Chris@82 512 Tc = LD(&(x[WS(rs, 12)]), ms, &(x[0]));
Chris@82 513 Td = BYTWJ(&(W[TWVL * 22]), Tc);
Chris@82 514 T7 = LD(&(x[WS(rs, 20)]), ms, &(x[0]));
Chris@82 515 T8 = BYTWJ(&(W[TWVL * 38]), T7);
Chris@82 516 Ta = LD(&(x[WS(rs, 28)]), ms, &(x[0]));
Chris@82 517 Tb = BYTWJ(&(W[TWVL * 54]), Ta);
Chris@82 518 }
Chris@82 519 {
Chris@82 520 V T9, Te, T2p, T2q;
Chris@82 521 T9 = VSUB(T6, T8);
Chris@82 522 Te = VSUB(Tb, Td);
Chris@82 523 Tf = VMUL(LDK(KP707106781), VADD(T9, Te));
Chris@82 524 T1v = VMUL(LDK(KP707106781), VSUB(Te, T9));
Chris@82 525 T2p = VADD(T6, T8);
Chris@82 526 T2q = VADD(Tb, Td);
Chris@82 527 T2r = VADD(T2p, T2q);
Chris@82 528 T3f = VSUB(T2q, T2p);
Chris@82 529 }
Chris@82 530 }
Chris@82 531 {
Chris@82 532 V Tt, TA, Tv, Ty;
Chris@82 533 {
Chris@82 534 V Ts, Tz, Tu, Tx;
Chris@82 535 Ts = LD(&(x[WS(rs, 30)]), ms, &(x[0]));
Chris@82 536 Tt = BYTWJ(&(W[TWVL * 58]), Ts);
Chris@82 537 Tz = LD(&(x[WS(rs, 22)]), ms, &(x[0]));
Chris@82 538 TA = BYTWJ(&(W[TWVL * 42]), Tz);
Chris@82 539 Tu = LD(&(x[WS(rs, 14)]), ms, &(x[0]));
Chris@82 540 Tv = BYTWJ(&(W[TWVL * 26]), Tu);
Chris@82 541 Tx = LD(&(x[WS(rs, 6)]), ms, &(x[0]));
Chris@82 542 Ty = BYTWJ(&(W[TWVL * 10]), Tx);
Chris@82 543 }
Chris@82 544 {
Chris@82 545 V Tw, TB, T2J, T2K;
Chris@82 546 Tw = VSUB(Tt, Tv);
Chris@82 547 TB = VSUB(Ty, TA);
Chris@82 548 TC = VFMA(LDK(KP923879532), Tw, VMUL(LDK(KP382683432), TB));
Chris@82 549 T1C = VFNMS(LDK(KP923879532), TB, VMUL(LDK(KP382683432), Tw));
Chris@82 550 T2J = VADD(Tt, Tv);
Chris@82 551 T2K = VADD(Ty, TA);
Chris@82 552 T2L = VADD(T2J, T2K);
Chris@82 553 T34 = VSUB(T2J, T2K);
Chris@82 554 }
Chris@82 555 }
Chris@82 556 {
Chris@82 557 V Ti, Tp, Tk, Tn;
Chris@82 558 {
Chris@82 559 V Th, To, Tj, Tm;
Chris@82 560 Th = LD(&(x[WS(rs, 2)]), ms, &(x[0]));
Chris@82 561 Ti = BYTWJ(&(W[TWVL * 2]), Th);
Chris@82 562 To = LD(&(x[WS(rs, 26)]), ms, &(x[0]));
Chris@82 563 Tp = BYTWJ(&(W[TWVL * 50]), To);
Chris@82 564 Tj = LD(&(x[WS(rs, 18)]), ms, &(x[0]));
Chris@82 565 Tk = BYTWJ(&(W[TWVL * 34]), Tj);
Chris@82 566 Tm = LD(&(x[WS(rs, 10)]), ms, &(x[0]));
Chris@82 567 Tn = BYTWJ(&(W[TWVL * 18]), Tm);
Chris@82 568 }
Chris@82 569 {
Chris@82 570 V Tl, Tq, T2M, T2N;
Chris@82 571 Tl = VSUB(Ti, Tk);
Chris@82 572 Tq = VSUB(Tn, Tp);
Chris@82 573 Tr = VFNMS(LDK(KP382683432), Tq, VMUL(LDK(KP923879532), Tl));
Chris@82 574 T1D = VFMA(LDK(KP382683432), Tl, VMUL(LDK(KP923879532), Tq));
Chris@82 575 T2M = VADD(Ti, Tk);
Chris@82 576 T2N = VADD(Tn, Tp);
Chris@82 577 T2O = VADD(T2M, T2N);
Chris@82 578 T33 = VSUB(T2M, T2N);
Chris@82 579 }
Chris@82 580 }
Chris@82 581 {
Chris@82 582 V T15, T17, T1p, T1n, T1f, T1h, T1i, T1a, T1c, T1d;
Chris@82 583 {
Chris@82 584 V T14, T16, T1o, T1m;
Chris@82 585 T14 = LD(&(x[WS(rs, 31)]), ms, &(x[WS(rs, 1)]));
Chris@82 586 T15 = BYTWJ(&(W[TWVL * 60]), T14);
Chris@82 587 T16 = LD(&(x[WS(rs, 15)]), ms, &(x[WS(rs, 1)]));
Chris@82 588 T17 = BYTWJ(&(W[TWVL * 28]), T16);
Chris@82 589 T1o = LD(&(x[WS(rs, 23)]), ms, &(x[WS(rs, 1)]));
Chris@82 590 T1p = BYTWJ(&(W[TWVL * 44]), T1o);
Chris@82 591 T1m = LD(&(x[WS(rs, 7)]), ms, &(x[WS(rs, 1)]));
Chris@82 592 T1n = BYTWJ(&(W[TWVL * 12]), T1m);
Chris@82 593 {
Chris@82 594 V T1e, T1g, T19, T1b;
Chris@82 595 T1e = LD(&(x[WS(rs, 27)]), ms, &(x[WS(rs, 1)]));
Chris@82 596 T1f = BYTWJ(&(W[TWVL * 52]), T1e);
Chris@82 597 T1g = LD(&(x[WS(rs, 11)]), ms, &(x[WS(rs, 1)]));
Chris@82 598 T1h = BYTWJ(&(W[TWVL * 20]), T1g);
Chris@82 599 T1i = VSUB(T1f, T1h);
Chris@82 600 T19 = LD(&(x[WS(rs, 3)]), ms, &(x[WS(rs, 1)]));
Chris@82 601 T1a = BYTWJ(&(W[TWVL * 4]), T19);
Chris@82 602 T1b = LD(&(x[WS(rs, 19)]), ms, &(x[WS(rs, 1)]));
Chris@82 603 T1c = BYTWJ(&(W[TWVL * 36]), T1b);
Chris@82 604 T1d = VSUB(T1a, T1c);
Chris@82 605 }
Chris@82 606 }
Chris@82 607 {
Chris@82 608 V T18, T1j, T2D, T2E;
Chris@82 609 T18 = VSUB(T15, T17);
Chris@82 610 T1j = VMUL(LDK(KP707106781), VADD(T1d, T1i));
Chris@82 611 T1k = VADD(T18, T1j);
Chris@82 612 T20 = VSUB(T18, T1j);
Chris@82 613 T2D = VADD(T1a, T1c);
Chris@82 614 T2E = VADD(T1f, T1h);
Chris@82 615 T2F = VADD(T2D, T2E);
Chris@82 616 T3b = VSUB(T2E, T2D);
Chris@82 617 }
Chris@82 618 {
Chris@82 619 V T1l, T1q, T2A, T2B;
Chris@82 620 T1l = VMUL(LDK(KP707106781), VSUB(T1i, T1d));
Chris@82 621 T1q = VSUB(T1n, T1p);
Chris@82 622 T1r = VSUB(T1l, T1q);
Chris@82 623 T21 = VADD(T1q, T1l);
Chris@82 624 T2A = VADD(T15, T17);
Chris@82 625 T2B = VADD(T1n, T1p);
Chris@82 626 T2C = VADD(T2A, T2B);
Chris@82 627 T3a = VSUB(T2A, T2B);
Chris@82 628 }
Chris@82 629 }
Chris@82 630 {
Chris@82 631 V TG, TI, T10, TY, TQ, TS, TT, TL, TN, TO;
Chris@82 632 {
Chris@82 633 V TF, TH, TZ, TX;
Chris@82 634 TF = LD(&(x[WS(rs, 1)]), ms, &(x[WS(rs, 1)]));
Chris@82 635 TG = BYTWJ(&(W[0]), TF);
Chris@82 636 TH = LD(&(x[WS(rs, 17)]), ms, &(x[WS(rs, 1)]));
Chris@82 637 TI = BYTWJ(&(W[TWVL * 32]), TH);
Chris@82 638 TZ = LD(&(x[WS(rs, 25)]), ms, &(x[WS(rs, 1)]));
Chris@82 639 T10 = BYTWJ(&(W[TWVL * 48]), TZ);
Chris@82 640 TX = LD(&(x[WS(rs, 9)]), ms, &(x[WS(rs, 1)]));
Chris@82 641 TY = BYTWJ(&(W[TWVL * 16]), TX);
Chris@82 642 {
Chris@82 643 V TP, TR, TK, TM;
Chris@82 644 TP = LD(&(x[WS(rs, 29)]), ms, &(x[WS(rs, 1)]));
Chris@82 645 TQ = BYTWJ(&(W[TWVL * 56]), TP);
Chris@82 646 TR = LD(&(x[WS(rs, 13)]), ms, &(x[WS(rs, 1)]));
Chris@82 647 TS = BYTWJ(&(W[TWVL * 24]), TR);
Chris@82 648 TT = VSUB(TQ, TS);
Chris@82 649 TK = LD(&(x[WS(rs, 5)]), ms, &(x[WS(rs, 1)]));
Chris@82 650 TL = BYTWJ(&(W[TWVL * 8]), TK);
Chris@82 651 TM = LD(&(x[WS(rs, 21)]), ms, &(x[WS(rs, 1)]));
Chris@82 652 TN = BYTWJ(&(W[TWVL * 40]), TM);
Chris@82 653 TO = VSUB(TL, TN);
Chris@82 654 }
Chris@82 655 }
Chris@82 656 {
Chris@82 657 V TJ, TU, T2w, T2x;
Chris@82 658 TJ = VSUB(TG, TI);
Chris@82 659 TU = VMUL(LDK(KP707106781), VADD(TO, TT));
Chris@82 660 TV = VADD(TJ, TU);
Chris@82 661 T1X = VSUB(TJ, TU);
Chris@82 662 T2w = VADD(TL, TN);
Chris@82 663 T2x = VADD(TQ, TS);
Chris@82 664 T2y = VADD(T2w, T2x);
Chris@82 665 T38 = VSUB(T2x, T2w);
Chris@82 666 }
Chris@82 667 {
Chris@82 668 V TW, T11, T2t, T2u;
Chris@82 669 TW = VMUL(LDK(KP707106781), VSUB(TT, TO));
Chris@82 670 T11 = VSUB(TY, T10);
Chris@82 671 T12 = VSUB(TW, T11);
Chris@82 672 T1Y = VADD(T11, TW);
Chris@82 673 T2t = VADD(TG, TI);
Chris@82 674 T2u = VADD(TY, T10);
Chris@82 675 T2v = VADD(T2t, T2u);
Chris@82 676 T37 = VSUB(T2t, T2u);
Chris@82 677 }
Chris@82 678 }
Chris@82 679 {
Chris@82 680 V T2W, T30, T2Z, T31;
Chris@82 681 {
Chris@82 682 V T2U, T2V, T2X, T2Y;
Chris@82 683 T2U = VADD(T2o, T2r);
Chris@82 684 T2V = VADD(T2O, T2L);
Chris@82 685 T2W = VADD(T2U, T2V);
Chris@82 686 T30 = VSUB(T2U, T2V);
Chris@82 687 T2X = VADD(T2v, T2y);
Chris@82 688 T2Y = VADD(T2C, T2F);
Chris@82 689 T2Z = VADD(T2X, T2Y);
Chris@82 690 T31 = VBYI(VSUB(T2Y, T2X));
Chris@82 691 }
Chris@82 692 ST(&(x[WS(rs, 16)]), VSUB(T2W, T2Z), ms, &(x[0]));
Chris@82 693 ST(&(x[WS(rs, 8)]), VADD(T30, T31), ms, &(x[0]));
Chris@82 694 ST(&(x[0]), VADD(T2W, T2Z), ms, &(x[0]));
Chris@82 695 ST(&(x[WS(rs, 24)]), VSUB(T30, T31), ms, &(x[0]));
Chris@82 696 }
Chris@82 697 {
Chris@82 698 V T2s, T2P, T2H, T2Q, T2z, T2G;
Chris@82 699 T2s = VSUB(T2o, T2r);
Chris@82 700 T2P = VSUB(T2L, T2O);
Chris@82 701 T2z = VSUB(T2v, T2y);
Chris@82 702 T2G = VSUB(T2C, T2F);
Chris@82 703 T2H = VMUL(LDK(KP707106781), VADD(T2z, T2G));
Chris@82 704 T2Q = VMUL(LDK(KP707106781), VSUB(T2G, T2z));
Chris@82 705 {
Chris@82 706 V T2I, T2R, T2S, T2T;
Chris@82 707 T2I = VADD(T2s, T2H);
Chris@82 708 T2R = VBYI(VADD(T2P, T2Q));
Chris@82 709 ST(&(x[WS(rs, 28)]), VSUB(T2I, T2R), ms, &(x[0]));
Chris@82 710 ST(&(x[WS(rs, 4)]), VADD(T2I, T2R), ms, &(x[0]));
Chris@82 711 T2S = VSUB(T2s, T2H);
Chris@82 712 T2T = VBYI(VSUB(T2Q, T2P));
Chris@82 713 ST(&(x[WS(rs, 20)]), VSUB(T2S, T2T), ms, &(x[0]));
Chris@82 714 ST(&(x[WS(rs, 12)]), VADD(T2S, T2T), ms, &(x[0]));
Chris@82 715 }
Chris@82 716 }
Chris@82 717 {
Chris@82 718 V T36, T3r, T3h, T3p, T3d, T3o, T3k, T3s, T35, T3g;
Chris@82 719 T35 = VMUL(LDK(KP707106781), VADD(T33, T34));
Chris@82 720 T36 = VADD(T32, T35);
Chris@82 721 T3r = VSUB(T32, T35);
Chris@82 722 T3g = VMUL(LDK(KP707106781), VSUB(T34, T33));
Chris@82 723 T3h = VADD(T3f, T3g);
Chris@82 724 T3p = VSUB(T3g, T3f);
Chris@82 725 {
Chris@82 726 V T39, T3c, T3i, T3j;
Chris@82 727 T39 = VFMA(LDK(KP923879532), T37, VMUL(LDK(KP382683432), T38));
Chris@82 728 T3c = VFNMS(LDK(KP382683432), T3b, VMUL(LDK(KP923879532), T3a));
Chris@82 729 T3d = VADD(T39, T3c);
Chris@82 730 T3o = VSUB(T3c, T39);
Chris@82 731 T3i = VFNMS(LDK(KP382683432), T37, VMUL(LDK(KP923879532), T38));
Chris@82 732 T3j = VFMA(LDK(KP382683432), T3a, VMUL(LDK(KP923879532), T3b));
Chris@82 733 T3k = VADD(T3i, T3j);
Chris@82 734 T3s = VSUB(T3j, T3i);
Chris@82 735 }
Chris@82 736 {
Chris@82 737 V T3e, T3l, T3u, T3v;
Chris@82 738 T3e = VADD(T36, T3d);
Chris@82 739 T3l = VBYI(VADD(T3h, T3k));
Chris@82 740 ST(&(x[WS(rs, 30)]), VSUB(T3e, T3l), ms, &(x[0]));
Chris@82 741 ST(&(x[WS(rs, 2)]), VADD(T3e, T3l), ms, &(x[0]));
Chris@82 742 T3u = VBYI(VADD(T3p, T3o));
Chris@82 743 T3v = VADD(T3r, T3s);
Chris@82 744 ST(&(x[WS(rs, 6)]), VADD(T3u, T3v), ms, &(x[0]));
Chris@82 745 ST(&(x[WS(rs, 26)]), VSUB(T3v, T3u), ms, &(x[0]));
Chris@82 746 }
Chris@82 747 {
Chris@82 748 V T3m, T3n, T3q, T3t;
Chris@82 749 T3m = VSUB(T36, T3d);
Chris@82 750 T3n = VBYI(VSUB(T3k, T3h));
Chris@82 751 ST(&(x[WS(rs, 18)]), VSUB(T3m, T3n), ms, &(x[0]));
Chris@82 752 ST(&(x[WS(rs, 14)]), VADD(T3m, T3n), ms, &(x[0]));
Chris@82 753 T3q = VBYI(VSUB(T3o, T3p));
Chris@82 754 T3t = VSUB(T3r, T3s);
Chris@82 755 ST(&(x[WS(rs, 10)]), VADD(T3q, T3t), ms, &(x[0]));
Chris@82 756 ST(&(x[WS(rs, 22)]), VSUB(T3t, T3q), ms, &(x[0]));
Chris@82 757 }
Chris@82 758 }
Chris@82 759 {
Chris@82 760 V TE, T1P, T1I, T1Q, T1t, T1M, T1F, T1N;
Chris@82 761 {
Chris@82 762 V Tg, TD, T1G, T1H;
Chris@82 763 Tg = VADD(T4, Tf);
Chris@82 764 TD = VADD(Tr, TC);
Chris@82 765 TE = VADD(Tg, TD);
Chris@82 766 T1P = VSUB(Tg, TD);
Chris@82 767 T1G = VFNMS(LDK(KP195090322), TV, VMUL(LDK(KP980785280), T12));
Chris@82 768 T1H = VFMA(LDK(KP195090322), T1k, VMUL(LDK(KP980785280), T1r));
Chris@82 769 T1I = VADD(T1G, T1H);
Chris@82 770 T1Q = VSUB(T1H, T1G);
Chris@82 771 }
Chris@82 772 {
Chris@82 773 V T13, T1s, T1B, T1E;
Chris@82 774 T13 = VFMA(LDK(KP980785280), TV, VMUL(LDK(KP195090322), T12));
Chris@82 775 T1s = VFNMS(LDK(KP195090322), T1r, VMUL(LDK(KP980785280), T1k));
Chris@82 776 T1t = VADD(T13, T1s);
Chris@82 777 T1M = VSUB(T1s, T13);
Chris@82 778 T1B = VSUB(T1v, T1A);
Chris@82 779 T1E = VSUB(T1C, T1D);
Chris@82 780 T1F = VADD(T1B, T1E);
Chris@82 781 T1N = VSUB(T1E, T1B);
Chris@82 782 }
Chris@82 783 {
Chris@82 784 V T1u, T1J, T1S, T1T;
Chris@82 785 T1u = VADD(TE, T1t);
Chris@82 786 T1J = VBYI(VADD(T1F, T1I));
Chris@82 787 ST(&(x[WS(rs, 31)]), VSUB(T1u, T1J), ms, &(x[WS(rs, 1)]));
Chris@82 788 ST(&(x[WS(rs, 1)]), VADD(T1u, T1J), ms, &(x[WS(rs, 1)]));
Chris@82 789 T1S = VBYI(VADD(T1N, T1M));
Chris@82 790 T1T = VADD(T1P, T1Q);
Chris@82 791 ST(&(x[WS(rs, 7)]), VADD(T1S, T1T), ms, &(x[WS(rs, 1)]));
Chris@82 792 ST(&(x[WS(rs, 25)]), VSUB(T1T, T1S), ms, &(x[WS(rs, 1)]));
Chris@82 793 }
Chris@82 794 {
Chris@82 795 V T1K, T1L, T1O, T1R;
Chris@82 796 T1K = VSUB(TE, T1t);
Chris@82 797 T1L = VBYI(VSUB(T1I, T1F));
Chris@82 798 ST(&(x[WS(rs, 17)]), VSUB(T1K, T1L), ms, &(x[WS(rs, 1)]));
Chris@82 799 ST(&(x[WS(rs, 15)]), VADD(T1K, T1L), ms, &(x[WS(rs, 1)]));
Chris@82 800 T1O = VBYI(VSUB(T1M, T1N));
Chris@82 801 T1R = VSUB(T1P, T1Q);
Chris@82 802 ST(&(x[WS(rs, 9)]), VADD(T1O, T1R), ms, &(x[WS(rs, 1)]));
Chris@82 803 ST(&(x[WS(rs, 23)]), VSUB(T1R, T1O), ms, &(x[WS(rs, 1)]));
Chris@82 804 }
Chris@82 805 }
Chris@82 806 {
Chris@82 807 V T1W, T2h, T2a, T2i, T23, T2e, T27, T2f;
Chris@82 808 {
Chris@82 809 V T1U, T1V, T28, T29;
Chris@82 810 T1U = VSUB(T4, Tf);
Chris@82 811 T1V = VADD(T1D, T1C);
Chris@82 812 T1W = VADD(T1U, T1V);
Chris@82 813 T2h = VSUB(T1U, T1V);
Chris@82 814 T28 = VFNMS(LDK(KP555570233), T1X, VMUL(LDK(KP831469612), T1Y));
Chris@82 815 T29 = VFMA(LDK(KP555570233), T20, VMUL(LDK(KP831469612), T21));
Chris@82 816 T2a = VADD(T28, T29);
Chris@82 817 T2i = VSUB(T29, T28);
Chris@82 818 }
Chris@82 819 {
Chris@82 820 V T1Z, T22, T25, T26;
Chris@82 821 T1Z = VFMA(LDK(KP831469612), T1X, VMUL(LDK(KP555570233), T1Y));
Chris@82 822 T22 = VFNMS(LDK(KP555570233), T21, VMUL(LDK(KP831469612), T20));
Chris@82 823 T23 = VADD(T1Z, T22);
Chris@82 824 T2e = VSUB(T22, T1Z);
Chris@82 825 T25 = VADD(T1A, T1v);
Chris@82 826 T26 = VSUB(TC, Tr);
Chris@82 827 T27 = VADD(T25, T26);
Chris@82 828 T2f = VSUB(T26, T25);
Chris@82 829 }
Chris@82 830 {
Chris@82 831 V T24, T2b, T2k, T2l;
Chris@82 832 T24 = VADD(T1W, T23);
Chris@82 833 T2b = VBYI(VADD(T27, T2a));
Chris@82 834 ST(&(x[WS(rs, 29)]), VSUB(T24, T2b), ms, &(x[WS(rs, 1)]));
Chris@82 835 ST(&(x[WS(rs, 3)]), VADD(T24, T2b), ms, &(x[WS(rs, 1)]));
Chris@82 836 T2k = VBYI(VADD(T2f, T2e));
Chris@82 837 T2l = VADD(T2h, T2i);
Chris@82 838 ST(&(x[WS(rs, 5)]), VADD(T2k, T2l), ms, &(x[WS(rs, 1)]));
Chris@82 839 ST(&(x[WS(rs, 27)]), VSUB(T2l, T2k), ms, &(x[WS(rs, 1)]));
Chris@82 840 }
Chris@82 841 {
Chris@82 842 V T2c, T2d, T2g, T2j;
Chris@82 843 T2c = VSUB(T1W, T23);
Chris@82 844 T2d = VBYI(VSUB(T2a, T27));
Chris@82 845 ST(&(x[WS(rs, 19)]), VSUB(T2c, T2d), ms, &(x[WS(rs, 1)]));
Chris@82 846 ST(&(x[WS(rs, 13)]), VADD(T2c, T2d), ms, &(x[WS(rs, 1)]));
Chris@82 847 T2g = VBYI(VSUB(T2e, T2f));
Chris@82 848 T2j = VSUB(T2h, T2i);
Chris@82 849 ST(&(x[WS(rs, 11)]), VADD(T2g, T2j), ms, &(x[WS(rs, 1)]));
Chris@82 850 ST(&(x[WS(rs, 21)]), VSUB(T2j, T2g), ms, &(x[WS(rs, 1)]));
Chris@82 851 }
Chris@82 852 }
Chris@82 853 }
Chris@82 854 }
Chris@82 855 VLEAVE();
Chris@82 856 }
Chris@82 857
Chris@82 858 static const tw_instr twinstr[] = {
Chris@82 859 VTW(0, 1),
Chris@82 860 VTW(0, 2),
Chris@82 861 VTW(0, 3),
Chris@82 862 VTW(0, 4),
Chris@82 863 VTW(0, 5),
Chris@82 864 VTW(0, 6),
Chris@82 865 VTW(0, 7),
Chris@82 866 VTW(0, 8),
Chris@82 867 VTW(0, 9),
Chris@82 868 VTW(0, 10),
Chris@82 869 VTW(0, 11),
Chris@82 870 VTW(0, 12),
Chris@82 871 VTW(0, 13),
Chris@82 872 VTW(0, 14),
Chris@82 873 VTW(0, 15),
Chris@82 874 VTW(0, 16),
Chris@82 875 VTW(0, 17),
Chris@82 876 VTW(0, 18),
Chris@82 877 VTW(0, 19),
Chris@82 878 VTW(0, 20),
Chris@82 879 VTW(0, 21),
Chris@82 880 VTW(0, 22),
Chris@82 881 VTW(0, 23),
Chris@82 882 VTW(0, 24),
Chris@82 883 VTW(0, 25),
Chris@82 884 VTW(0, 26),
Chris@82 885 VTW(0, 27),
Chris@82 886 VTW(0, 28),
Chris@82 887 VTW(0, 29),
Chris@82 888 VTW(0, 30),
Chris@82 889 VTW(0, 31),
Chris@82 890 {TW_NEXT, VL, 0}
Chris@82 891 };
Chris@82 892
Chris@82 893 static const ct_desc desc = { 32, XSIMD_STRING("t1fv_32"), twinstr, &GENUS, {201, 88, 16, 0}, 0, 0, 0 };
Chris@82 894
Chris@82 895 void XSIMD(codelet_t1fv_32) (planner *p) {
Chris@82 896 X(kdft_dit_register) (p, t1fv_32, &desc);
Chris@82 897 }
Chris@82 898 #endif