Mercurial > hg > sv-dependency-builds

annotate src/fftw-3.3.8/dft/simd/common/t1sv_16.c @ 82:d0c2a83c1364

Find changesets by keywords (author, files, the commit message), revision number or hash, or revset expression.
Add FFTW 3.3.8 source, and a Linux build
author Chris Cannam
date Tue, 19 Nov 2019 14:52:55 +0000
parents
children
rev   line source
Chris@82 1 /*
Chris@82 2 * Copyright (c) 2003, 2007-14 Matteo Frigo
Chris@82 3 * Copyright (c) 2003, 2007-14 Massachusetts Institute of Technology
Chris@82 4 *
Chris@82 5 * This program is free software; you can redistribute it and/or modify
Chris@82 6 * it under the terms of the GNU General Public License as published by
Chris@82 7 * the Free Software Foundation; either version 2 of the License, or
Chris@82 8 * (at your option) any later version.
Chris@82 9 *
Chris@82 10 * This program is distributed in the hope that it will be useful,
Chris@82 11 * but WITHOUT ANY WARRANTY; without even the implied warranty of
Chris@82 12 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
Chris@82 13 * GNU General Public License for more details.
Chris@82 14 *
Chris@82 15 * You should have received a copy of the GNU General Public License
Chris@82 16 * along with this program; if not, write to the Free Software
Chris@82 17 * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
Chris@82 18 *
Chris@82 19 */
Chris@82 20
Chris@82 21 /* This file was automatically generated --- DO NOT EDIT */
Chris@82 22 /* Generated on Thu May 24 08:06:10 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.native -fma -simd -compact -variables 4 -pipeline-latency 8 -n 16 -name t1sv_16 -include dft/simd/ts.h */
Chris@82 29
Chris@82 30 /*
Chris@82 31 * This function contains 174 FP additions, 100 FP multiplications,
Chris@82 32 * (or, 104 additions, 30 multiplications, 70 fused multiply/add),
Chris@82 33 * 60 stack variables, 3 constants, and 64 memory accesses
Chris@82 34 */
Chris@82 35 #include "dft/simd/ts.h"
Chris@82 36
Chris@82 37 static void t1sv_16(R *ri, R *ii, const R *W, stride rs, INT mb, INT me, INT ms)
Chris@82 38 {
Chris@82 39 DVK(KP923879532, +0.923879532511286756128183189396788286822416626);
Chris@82 40 DVK(KP414213562, +0.414213562373095048801688724209698078569671875);
Chris@82 41 DVK(KP707106781, +0.707106781186547524400844362104849039284835938);
Chris@82 42 {
Chris@82 43 INT m;
Chris@82 44 for (m = mb, W = W + (mb * 30); m < me; m = m + (2 * VL), ri = ri + ((2 * VL) * ms), ii = ii + ((2 * VL) * ms), W = W + ((2 * VL) * 30), MAKE_VOLATILE_STRIDE(32, rs)) {
Chris@82 45 V T8, T3z, T1I, T3o, T1s, T35, T2o, T2r, T1F, T36, T2p, T2w, Tl, T3A, T1N;
Chris@82 46 V T3k, Tz, T2V, T1T, T1U, T11, T30, T29, T2c, T1e, T31, T2a, T2h, TM, T2W;
Chris@82 47 V T1W, T21;
Chris@82 48 {
Chris@82 49 V T1, T3n, T3, T6, T4, T3l, T2, T7, T3m, T5;
Chris@82 50 T1 = LD(&(ri[0]), ms, &(ri[0]));
Chris@82 51 T3n = LD(&(ii[0]), ms, &(ii[0]));
Chris@82 52 T3 = LD(&(ri[WS(rs, 8)]), ms, &(ri[0]));
Chris@82 53 T6 = LD(&(ii[WS(rs, 8)]), ms, &(ii[0]));
Chris@82 54 T2 = LDW(&(W[TWVL * 14]));
Chris@82 55 T4 = VMUL(T2, T3);
Chris@82 56 T3l = VMUL(T2, T6);
Chris@82 57 T5 = LDW(&(W[TWVL * 15]));
Chris@82 58 T7 = VFMA(T5, T6, T4);
Chris@82 59 T3m = VFNMS(T5, T3, T3l);
Chris@82 60 T8 = VADD(T1, T7);
Chris@82 61 T3z = VSUB(T3n, T3m);
Chris@82 62 T1I = VSUB(T1, T7);
Chris@82 63 T3o = VADD(T3m, T3n);
Chris@82 64 }
Chris@82 65 {
Chris@82 66 V T1h, T1k, T1i, T2k, T1n, T1q, T1o, T2m, T1g, T1m;
Chris@82 67 T1h = LD(&(ri[WS(rs, 15)]), ms, &(ri[WS(rs, 1)]));
Chris@82 68 T1k = LD(&(ii[WS(rs, 15)]), ms, &(ii[WS(rs, 1)]));
Chris@82 69 T1g = LDW(&(W[TWVL * 28]));
Chris@82 70 T1i = VMUL(T1g, T1h);
Chris@82 71 T2k = VMUL(T1g, T1k);
Chris@82 72 T1n = LD(&(ri[WS(rs, 7)]), ms, &(ri[WS(rs, 1)]));
Chris@82 73 T1q = LD(&(ii[WS(rs, 7)]), ms, &(ii[WS(rs, 1)]));
Chris@82 74 T1m = LDW(&(W[TWVL * 12]));
Chris@82 75 T1o = VMUL(T1m, T1n);
Chris@82 76 T2m = VMUL(T1m, T1q);
Chris@82 77 {
Chris@82 78 V T1l, T2l, T1r, T2n, T1j, T1p;
Chris@82 79 T1j = LDW(&(W[TWVL * 29]));
Chris@82 80 T1l = VFMA(T1j, T1k, T1i);
Chris@82 81 T2l = VFNMS(T1j, T1h, T2k);
Chris@82 82 T1p = LDW(&(W[TWVL * 13]));
Chris@82 83 T1r = VFMA(T1p, T1q, T1o);
Chris@82 84 T2n = VFNMS(T1p, T1n, T2m);
Chris@82 85 T1s = VADD(T1l, T1r);
Chris@82 86 T35 = VADD(T2l, T2n);
Chris@82 87 T2o = VSUB(T2l, T2n);
Chris@82 88 T2r = VSUB(T1l, T1r);
Chris@82 89 }
Chris@82 90 }
Chris@82 91 {
Chris@82 92 V T1u, T1x, T1v, T2s, T1A, T1D, T1B, T2u, T1t, T1z;
Chris@82 93 T1u = LD(&(ri[WS(rs, 3)]), ms, &(ri[WS(rs, 1)]));
Chris@82 94 T1x = LD(&(ii[WS(rs, 3)]), ms, &(ii[WS(rs, 1)]));
Chris@82 95 T1t = LDW(&(W[TWVL * 4]));
Chris@82 96 T1v = VMUL(T1t, T1u);
Chris@82 97 T2s = VMUL(T1t, T1x);
Chris@82 98 T1A = LD(&(ri[WS(rs, 11)]), ms, &(ri[WS(rs, 1)]));
Chris@82 99 T1D = LD(&(ii[WS(rs, 11)]), ms, &(ii[WS(rs, 1)]));
Chris@82 100 T1z = LDW(&(W[TWVL * 20]));
Chris@82 101 T1B = VMUL(T1z, T1A);
Chris@82 102 T2u = VMUL(T1z, T1D);
Chris@82 103 {
Chris@82 104 V T1y, T2t, T1E, T2v, T1w, T1C;
Chris@82 105 T1w = LDW(&(W[TWVL * 5]));
Chris@82 106 T1y = VFMA(T1w, T1x, T1v);
Chris@82 107 T2t = VFNMS(T1w, T1u, T2s);
Chris@82 108 T1C = LDW(&(W[TWVL * 21]));
Chris@82 109 T1E = VFMA(T1C, T1D, T1B);
Chris@82 110 T2v = VFNMS(T1C, T1A, T2u);
Chris@82 111 T1F = VADD(T1y, T1E);
Chris@82 112 T36 = VADD(T2t, T2v);
Chris@82 113 T2p = VSUB(T1y, T1E);
Chris@82 114 T2w = VSUB(T2t, T2v);
Chris@82 115 }
Chris@82 116 }
Chris@82 117 {
Chris@82 118 V Ta, Td, Tb, T1J, Tg, Tj, Th, T1L, T9, Tf;
Chris@82 119 Ta = LD(&(ri[WS(rs, 4)]), ms, &(ri[0]));
Chris@82 120 Td = LD(&(ii[WS(rs, 4)]), ms, &(ii[0]));
Chris@82 121 T9 = LDW(&(W[TWVL * 6]));
Chris@82 122 Tb = VMUL(T9, Ta);
Chris@82 123 T1J = VMUL(T9, Td);
Chris@82 124 Tg = LD(&(ri[WS(rs, 12)]), ms, &(ri[0]));
Chris@82 125 Tj = LD(&(ii[WS(rs, 12)]), ms, &(ii[0]));
Chris@82 126 Tf = LDW(&(W[TWVL * 22]));
Chris@82 127 Th = VMUL(Tf, Tg);
Chris@82 128 T1L = VMUL(Tf, Tj);
Chris@82 129 {
Chris@82 130 V Te, T1K, Tk, T1M, Tc, Ti;
Chris@82 131 Tc = LDW(&(W[TWVL * 7]));
Chris@82 132 Te = VFMA(Tc, Td, Tb);
Chris@82 133 T1K = VFNMS(Tc, Ta, T1J);
Chris@82 134 Ti = LDW(&(W[TWVL * 23]));
Chris@82 135 Tk = VFMA(Ti, Tj, Th);
Chris@82 136 T1M = VFNMS(Ti, Tg, T1L);
Chris@82 137 Tl = VADD(Te, Tk);
Chris@82 138 T3A = VSUB(Te, Tk);
Chris@82 139 T1N = VSUB(T1K, T1M);
Chris@82 140 T3k = VADD(T1K, T1M);
Chris@82 141 }
Chris@82 142 }
Chris@82 143 {
Chris@82 144 V To, Tr, Tp, T1P, Tu, Tx, Tv, T1R, Tn, Tt;
Chris@82 145 To = LD(&(ri[WS(rs, 2)]), ms, &(ri[0]));
Chris@82 146 Tr = LD(&(ii[WS(rs, 2)]), ms, &(ii[0]));
Chris@82 147 Tn = LDW(&(W[TWVL * 2]));
Chris@82 148 Tp = VMUL(Tn, To);
Chris@82 149 T1P = VMUL(Tn, Tr);
Chris@82 150 Tu = LD(&(ri[WS(rs, 10)]), ms, &(ri[0]));
Chris@82 151 Tx = LD(&(ii[WS(rs, 10)]), ms, &(ii[0]));
Chris@82 152 Tt = LDW(&(W[TWVL * 18]));
Chris@82 153 Tv = VMUL(Tt, Tu);
Chris@82 154 T1R = VMUL(Tt, Tx);
Chris@82 155 {
Chris@82 156 V Ts, T1Q, Ty, T1S, Tq, Tw;
Chris@82 157 Tq = LDW(&(W[TWVL * 3]));
Chris@82 158 Ts = VFMA(Tq, Tr, Tp);
Chris@82 159 T1Q = VFNMS(Tq, To, T1P);
Chris@82 160 Tw = LDW(&(W[TWVL * 19]));
Chris@82 161 Ty = VFMA(Tw, Tx, Tv);
Chris@82 162 T1S = VFNMS(Tw, Tu, T1R);
Chris@82 163 Tz = VADD(Ts, Ty);
Chris@82 164 T2V = VADD(T1Q, T1S);
Chris@82 165 T1T = VSUB(T1Q, T1S);
Chris@82 166 T1U = VSUB(Ts, Ty);
Chris@82 167 }
Chris@82 168 }
Chris@82 169 {
Chris@82 170 V TQ, TT, TR, T25, TW, TZ, TX, T27, TP, TV;
Chris@82 171 TQ = LD(&(ri[WS(rs, 1)]), ms, &(ri[WS(rs, 1)]));
Chris@82 172 TT = LD(&(ii[WS(rs, 1)]), ms, &(ii[WS(rs, 1)]));
Chris@82 173 TP = LDW(&(W[0]));
Chris@82 174 TR = VMUL(TP, TQ);
Chris@82 175 T25 = VMUL(TP, TT);
Chris@82 176 TW = LD(&(ri[WS(rs, 9)]), ms, &(ri[WS(rs, 1)]));
Chris@82 177 TZ = LD(&(ii[WS(rs, 9)]), ms, &(ii[WS(rs, 1)]));
Chris@82 178 TV = LDW(&(W[TWVL * 16]));
Chris@82 179 TX = VMUL(TV, TW);
Chris@82 180 T27 = VMUL(TV, TZ);
Chris@82 181 {
Chris@82 182 V TU, T26, T10, T28, TS, TY;
Chris@82 183 TS = LDW(&(W[TWVL * 1]));
Chris@82 184 TU = VFMA(TS, TT, TR);
Chris@82 185 T26 = VFNMS(TS, TQ, T25);
Chris@82 186 TY = LDW(&(W[TWVL * 17]));
Chris@82 187 T10 = VFMA(TY, TZ, TX);
Chris@82 188 T28 = VFNMS(TY, TW, T27);
Chris@82 189 T11 = VADD(TU, T10);
Chris@82 190 T30 = VADD(T26, T28);
Chris@82 191 T29 = VSUB(T26, T28);
Chris@82 192 T2c = VSUB(TU, T10);
Chris@82 193 }
Chris@82 194 }
Chris@82 195 {
Chris@82 196 V T13, T16, T14, T2d, T19, T1c, T1a, T2f, T12, T18;
Chris@82 197 T13 = LD(&(ri[WS(rs, 5)]), ms, &(ri[WS(rs, 1)]));
Chris@82 198 T16 = LD(&(ii[WS(rs, 5)]), ms, &(ii[WS(rs, 1)]));
Chris@82 199 T12 = LDW(&(W[TWVL * 8]));
Chris@82 200 T14 = VMUL(T12, T13);
Chris@82 201 T2d = VMUL(T12, T16);
Chris@82 202 T19 = LD(&(ri[WS(rs, 13)]), ms, &(ri[WS(rs, 1)]));
Chris@82 203 T1c = LD(&(ii[WS(rs, 13)]), ms, &(ii[WS(rs, 1)]));
Chris@82 204 T18 = LDW(&(W[TWVL * 24]));
Chris@82 205 T1a = VMUL(T18, T19);
Chris@82 206 T2f = VMUL(T18, T1c);
Chris@82 207 {
Chris@82 208 V T17, T2e, T1d, T2g, T15, T1b;
Chris@82 209 T15 = LDW(&(W[TWVL * 9]));
Chris@82 210 T17 = VFMA(T15, T16, T14);
Chris@82 211 T2e = VFNMS(T15, T13, T2d);
Chris@82 212 T1b = LDW(&(W[TWVL * 25]));
Chris@82 213 T1d = VFMA(T1b, T1c, T1a);
Chris@82 214 T2g = VFNMS(T1b, T19, T2f);
Chris@82 215 T1e = VADD(T17, T1d);
Chris@82 216 T31 = VADD(T2e, T2g);
Chris@82 217 T2a = VSUB(T17, T1d);
Chris@82 218 T2h = VSUB(T2e, T2g);
Chris@82 219 }
Chris@82 220 }
Chris@82 221 {
Chris@82 222 V TB, TE, TC, T1X, TH, TK, TI, T1Z, TA, TG;
Chris@82 223 TB = LD(&(ri[WS(rs, 14)]), ms, &(ri[0]));
Chris@82 224 TE = LD(&(ii[WS(rs, 14)]), ms, &(ii[0]));
Chris@82 225 TA = LDW(&(W[TWVL * 26]));
Chris@82 226 TC = VMUL(TA, TB);
Chris@82 227 T1X = VMUL(TA, TE);
Chris@82 228 TH = LD(&(ri[WS(rs, 6)]), ms, &(ri[0]));
Chris@82 229 TK = LD(&(ii[WS(rs, 6)]), ms, &(ii[0]));
Chris@82 230 TG = LDW(&(W[TWVL * 10]));
Chris@82 231 TI = VMUL(TG, TH);
Chris@82 232 T1Z = VMUL(TG, TK);
Chris@82 233 {
Chris@82 234 V TF, T1Y, TL, T20, TD, TJ;
Chris@82 235 TD = LDW(&(W[TWVL * 27]));
Chris@82 236 TF = VFMA(TD, TE, TC);
Chris@82 237 T1Y = VFNMS(TD, TB, T1X);
Chris@82 238 TJ = LDW(&(W[TWVL * 11]));
Chris@82 239 TL = VFMA(TJ, TK, TI);
Chris@82 240 T20 = VFNMS(TJ, TH, T1Z);
Chris@82 241 TM = VADD(TF, TL);
Chris@82 242 T2W = VADD(T1Y, T20);
Chris@82 243 T1W = VSUB(TF, TL);
Chris@82 244 T21 = VSUB(T1Y, T20);
Chris@82 245 }
Chris@82 246 }
Chris@82 247 {
Chris@82 248 V TO, T3e, T3q, T3s, T1H, T3r, T3h, T3i;
Chris@82 249 {
Chris@82 250 V Tm, TN, T3j, T3p;
Chris@82 251 Tm = VADD(T8, Tl);
Chris@82 252 TN = VADD(Tz, TM);
Chris@82 253 TO = VADD(Tm, TN);
Chris@82 254 T3e = VSUB(Tm, TN);
Chris@82 255 T3j = VADD(T2V, T2W);
Chris@82 256 T3p = VADD(T3k, T3o);
Chris@82 257 T3q = VADD(T3j, T3p);
Chris@82 258 T3s = VSUB(T3p, T3j);
Chris@82 259 }
Chris@82 260 {
Chris@82 261 V T1f, T1G, T3f, T3g;
Chris@82 262 T1f = VADD(T11, T1e);
Chris@82 263 T1G = VADD(T1s, T1F);
Chris@82 264 T1H = VADD(T1f, T1G);
Chris@82 265 T3r = VSUB(T1G, T1f);
Chris@82 266 T3f = VADD(T30, T31);
Chris@82 267 T3g = VADD(T35, T36);
Chris@82 268 T3h = VSUB(T3f, T3g);
Chris@82 269 T3i = VADD(T3f, T3g);
Chris@82 270 }
Chris@82 271 ST(&(ri[WS(rs, 8)]), VSUB(TO, T1H), ms, &(ri[0]));
Chris@82 272 ST(&(ii[WS(rs, 8)]), VSUB(T3q, T3i), ms, &(ii[0]));
Chris@82 273 ST(&(ri[0]), VADD(TO, T1H), ms, &(ri[0]));
Chris@82 274 ST(&(ii[0]), VADD(T3i, T3q), ms, &(ii[0]));
Chris@82 275 ST(&(ri[WS(rs, 12)]), VSUB(T3e, T3h), ms, &(ri[0]));
Chris@82 276 ST(&(ii[WS(rs, 12)]), VSUB(T3s, T3r), ms, &(ii[0]));
Chris@82 277 ST(&(ri[WS(rs, 4)]), VADD(T3e, T3h), ms, &(ri[0]));
Chris@82 278 ST(&(ii[WS(rs, 4)]), VADD(T3r, T3s), ms, &(ii[0]));
Chris@82 279 }
Chris@82 280 {
Chris@82 281 V T2Y, T3a, T3v, T3x, T33, T3b, T38, T3c;
Chris@82 282 {
Chris@82 283 V T2U, T2X, T3t, T3u;
Chris@82 284 T2U = VSUB(T8, Tl);
Chris@82 285 T2X = VSUB(T2V, T2W);
Chris@82 286 T2Y = VADD(T2U, T2X);
Chris@82 287 T3a = VSUB(T2U, T2X);
Chris@82 288 T3t = VSUB(TM, Tz);
Chris@82 289 T3u = VSUB(T3o, T3k);
Chris@82 290 T3v = VADD(T3t, T3u);
Chris@82 291 T3x = VSUB(T3u, T3t);
Chris@82 292 }
Chris@82 293 {
Chris@82 294 V T2Z, T32, T34, T37;
Chris@82 295 T2Z = VSUB(T11, T1e);
Chris@82 296 T32 = VSUB(T30, T31);
Chris@82 297 T33 = VADD(T2Z, T32);
Chris@82 298 T3b = VSUB(T32, T2Z);
Chris@82 299 T34 = VSUB(T1s, T1F);
Chris@82 300 T37 = VSUB(T35, T36);
Chris@82 301 T38 = VSUB(T34, T37);
Chris@82 302 T3c = VADD(T34, T37);
Chris@82 303 }
Chris@82 304 {
Chris@82 305 V T39, T3w, T3d, T3y;
Chris@82 306 T39 = VADD(T33, T38);
Chris@82 307 ST(&(ri[WS(rs, 10)]), VFNMS(LDK(KP707106781), T39, T2Y), ms, &(ri[0]));
Chris@82 308 ST(&(ri[WS(rs, 2)]), VFMA(LDK(KP707106781), T39, T2Y), ms, &(ri[0]));
Chris@82 309 T3w = VADD(T3b, T3c);
Chris@82 310 ST(&(ii[WS(rs, 2)]), VFMA(LDK(KP707106781), T3w, T3v), ms, &(ii[0]));
Chris@82 311 ST(&(ii[WS(rs, 10)]), VFNMS(LDK(KP707106781), T3w, T3v), ms, &(ii[0]));
Chris@82 312 T3d = VSUB(T3b, T3c);
Chris@82 313 ST(&(ri[WS(rs, 14)]), VFNMS(LDK(KP707106781), T3d, T3a), ms, &(ri[0]));
Chris@82 314 ST(&(ri[WS(rs, 6)]), VFMA(LDK(KP707106781), T3d, T3a), ms, &(ri[0]));
Chris@82 315 T3y = VSUB(T38, T33);
Chris@82 316 ST(&(ii[WS(rs, 6)]), VFMA(LDK(KP707106781), T3y, T3x), ms, &(ii[0]));
Chris@82 317 ST(&(ii[WS(rs, 14)]), VFNMS(LDK(KP707106781), T3y, T3x), ms, &(ii[0]));
Chris@82 318 }
Chris@82 319 }
Chris@82 320 {
Chris@82 321 V T1O, T3B, T3H, T2E, T23, T3C, T2O, T2S, T2H, T3I, T2j, T2B, T2L, T2R, T2y;
Chris@82 322 V T2C;
Chris@82 323 {
Chris@82 324 V T1V, T22, T2b, T2i;
Chris@82 325 T1O = VSUB(T1I, T1N);
Chris@82 326 T3B = VSUB(T3z, T3A);
Chris@82 327 T3H = VADD(T3A, T3z);
Chris@82 328 T2E = VADD(T1I, T1N);
Chris@82 329 T1V = VSUB(T1T, T1U);
Chris@82 330 T22 = VADD(T1W, T21);
Chris@82 331 T23 = VSUB(T1V, T22);
Chris@82 332 T3C = VADD(T1V, T22);
Chris@82 333 {
Chris@82 334 V T2M, T2N, T2F, T2G;
Chris@82 335 T2M = VADD(T2r, T2w);
Chris@82 336 T2N = VSUB(T2o, T2p);
Chris@82 337 T2O = VFNMS(LDK(KP414213562), T2N, T2M);
Chris@82 338 T2S = VFMA(LDK(KP414213562), T2M, T2N);
Chris@82 339 T2F = VADD(T1U, T1T);
Chris@82 340 T2G = VSUB(T1W, T21);
Chris@82 341 T2H = VADD(T2F, T2G);
Chris@82 342 T3I = VSUB(T2G, T2F);
Chris@82 343 }
Chris@82 344 T2b = VADD(T29, T2a);
Chris@82 345 T2i = VSUB(T2c, T2h);
Chris@82 346 T2j = VFMA(LDK(KP414213562), T2i, T2b);
Chris@82 347 T2B = VFNMS(LDK(KP414213562), T2b, T2i);
Chris@82 348 {
Chris@82 349 V T2J, T2K, T2q, T2x;
Chris@82 350 T2J = VADD(T2c, T2h);
Chris@82 351 T2K = VSUB(T29, T2a);
Chris@82 352 T2L = VFMA(LDK(KP414213562), T2K, T2J);
Chris@82 353 T2R = VFNMS(LDK(KP414213562), T2J, T2K);
Chris@82 354 T2q = VADD(T2o, T2p);
Chris@82 355 T2x = VSUB(T2r, T2w);
Chris@82 356 T2y = VFNMS(LDK(KP414213562), T2x, T2q);
Chris@82 357 T2C = VFMA(LDK(KP414213562), T2q, T2x);
Chris@82 358 }
Chris@82 359 }
Chris@82 360 {
Chris@82 361 V T24, T2z, T3J, T3K;
Chris@82 362 T24 = VFMA(LDK(KP707106781), T23, T1O);
Chris@82 363 T2z = VSUB(T2j, T2y);
Chris@82 364 ST(&(ri[WS(rs, 11)]), VFNMS(LDK(KP923879532), T2z, T24), ms, &(ri[WS(rs, 1)]));
Chris@82 365 ST(&(ri[WS(rs, 3)]), VFMA(LDK(KP923879532), T2z, T24), ms, &(ri[WS(rs, 1)]));
Chris@82 366 T3J = VFMA(LDK(KP707106781), T3I, T3H);
Chris@82 367 T3K = VSUB(T2C, T2B);
Chris@82 368 ST(&(ii[WS(rs, 3)]), VFMA(LDK(KP923879532), T3K, T3J), ms, &(ii[WS(rs, 1)]));
Chris@82 369 ST(&(ii[WS(rs, 11)]), VFNMS(LDK(KP923879532), T3K, T3J), ms, &(ii[WS(rs, 1)]));
Chris@82 370 }
Chris@82 371 {
Chris@82 372 V T2A, T2D, T3L, T3M;
Chris@82 373 T2A = VFNMS(LDK(KP707106781), T23, T1O);
Chris@82 374 T2D = VADD(T2B, T2C);
Chris@82 375 ST(&(ri[WS(rs, 7)]), VFNMS(LDK(KP923879532), T2D, T2A), ms, &(ri[WS(rs, 1)]));
Chris@82 376 ST(&(ri[WS(rs, 15)]), VFMA(LDK(KP923879532), T2D, T2A), ms, &(ri[WS(rs, 1)]));
Chris@82 377 T3L = VFNMS(LDK(KP707106781), T3I, T3H);
Chris@82 378 T3M = VADD(T2j, T2y);
Chris@82 379 ST(&(ii[WS(rs, 7)]), VFNMS(LDK(KP923879532), T3M, T3L), ms, &(ii[WS(rs, 1)]));
Chris@82 380 ST(&(ii[WS(rs, 15)]), VFMA(LDK(KP923879532), T3M, T3L), ms, &(ii[WS(rs, 1)]));
Chris@82 381 }
Chris@82 382 {
Chris@82 383 V T2I, T2P, T3D, T3E;
Chris@82 384 T2I = VFMA(LDK(KP707106781), T2H, T2E);
Chris@82 385 T2P = VADD(T2L, T2O);
Chris@82 386 ST(&(ri[WS(rs, 9)]), VFNMS(LDK(KP923879532), T2P, T2I), ms, &(ri[WS(rs, 1)]));
Chris@82 387 ST(&(ri[WS(rs, 1)]), VFMA(LDK(KP923879532), T2P, T2I), ms, &(ri[WS(rs, 1)]));
Chris@82 388 T3D = VFMA(LDK(KP707106781), T3C, T3B);
Chris@82 389 T3E = VADD(T2R, T2S);
Chris@82 390 ST(&(ii[WS(rs, 1)]), VFMA(LDK(KP923879532), T3E, T3D), ms, &(ii[WS(rs, 1)]));
Chris@82 391 ST(&(ii[WS(rs, 9)]), VFNMS(LDK(KP923879532), T3E, T3D), ms, &(ii[WS(rs, 1)]));
Chris@82 392 }
Chris@82 393 {
Chris@82 394 V T2Q, T2T, T3F, T3G;
Chris@82 395 T2Q = VFNMS(LDK(KP707106781), T2H, T2E);
Chris@82 396 T2T = VSUB(T2R, T2S);
Chris@82 397 ST(&(ri[WS(rs, 13)]), VFNMS(LDK(KP923879532), T2T, T2Q), ms, &(ri[WS(rs, 1)]));
Chris@82 398 ST(&(ri[WS(rs, 5)]), VFMA(LDK(KP923879532), T2T, T2Q), ms, &(ri[WS(rs, 1)]));
Chris@82 399 T3F = VFNMS(LDK(KP707106781), T3C, T3B);
Chris@82 400 T3G = VSUB(T2O, T2L);
Chris@82 401 ST(&(ii[WS(rs, 5)]), VFMA(LDK(KP923879532), T3G, T3F), ms, &(ii[WS(rs, 1)]));
Chris@82 402 ST(&(ii[WS(rs, 13)]), VFNMS(LDK(KP923879532), T3G, T3F), ms, &(ii[WS(rs, 1)]));
Chris@82 403 }
Chris@82 404 }
Chris@82 405 }
Chris@82 406 }
Chris@82 407 VLEAVE();
Chris@82 408 }
Chris@82 409
Chris@82 410 static const tw_instr twinstr[] = {
Chris@82 411 VTW(0, 1),
Chris@82 412 VTW(0, 2),
Chris@82 413 VTW(0, 3),
Chris@82 414 VTW(0, 4),
Chris@82 415 VTW(0, 5),
Chris@82 416 VTW(0, 6),
Chris@82 417 VTW(0, 7),
Chris@82 418 VTW(0, 8),
Chris@82 419 VTW(0, 9),
Chris@82 420 VTW(0, 10),
Chris@82 421 VTW(0, 11),
Chris@82 422 VTW(0, 12),
Chris@82 423 VTW(0, 13),
Chris@82 424 VTW(0, 14),
Chris@82 425 VTW(0, 15),
Chris@82 426 {TW_NEXT, (2 * VL), 0}
Chris@82 427 };
Chris@82 428
Chris@82 429 static const ct_desc desc = { 16, XSIMD_STRING("t1sv_16"), twinstr, &GENUS, {104, 30, 70, 0}, 0, 0, 0 };
Chris@82 430
Chris@82 431 void XSIMD(codelet_t1sv_16) (planner *p) {
Chris@82 432 X(kdft_dit_register) (p, t1sv_16, &desc);
Chris@82 433 }
Chris@82 434 #else
Chris@82 435
Chris@82 436 /* Generated by: ../../../genfft/gen_twiddle.native -simd -compact -variables 4 -pipeline-latency 8 -n 16 -name t1sv_16 -include dft/simd/ts.h */
Chris@82 437
Chris@82 438 /*
Chris@82 439 * This function contains 174 FP additions, 84 FP multiplications,
Chris@82 440 * (or, 136 additions, 46 multiplications, 38 fused multiply/add),
Chris@82 441 * 52 stack variables, 3 constants, and 64 memory accesses
Chris@82 442 */
Chris@82 443 #include "dft/simd/ts.h"
Chris@82 444
Chris@82 445 static void t1sv_16(R *ri, R *ii, const R *W, stride rs, INT mb, INT me, INT ms)
Chris@82 446 {
Chris@82 447 DVK(KP382683432, +0.382683432365089771728459984030398866761344562);
Chris@82 448 DVK(KP923879532, +0.923879532511286756128183189396788286822416626);
Chris@82 449 DVK(KP707106781, +0.707106781186547524400844362104849039284835938);
Chris@82 450 {
Chris@82 451 INT m;
Chris@82 452 for (m = mb, W = W + (mb * 30); m < me; m = m + (2 * VL), ri = ri + ((2 * VL) * ms), ii = ii + ((2 * VL) * ms), W = W + ((2 * VL) * 30), MAKE_VOLATILE_STRIDE(32, rs)) {
Chris@82 453 V T7, T37, T1t, T2U, Ti, T38, T1w, T2R, Tu, T2s, T1C, T2c, TF, T2t, T1H;
Chris@82 454 V T2d, T1f, T1q, T2B, T2C, T2D, T2E, T1Z, T2j, T24, T2k, TS, T13, T2w, T2x;
Chris@82 455 V T2y, T2z, T1O, T2g, T1T, T2h;
Chris@82 456 {
Chris@82 457 V T1, T2T, T6, T2S;
Chris@82 458 T1 = LD(&(ri[0]), ms, &(ri[0]));
Chris@82 459 T2T = LD(&(ii[0]), ms, &(ii[0]));
Chris@82 460 {
Chris@82 461 V T3, T5, T2, T4;
Chris@82 462 T3 = LD(&(ri[WS(rs, 8)]), ms, &(ri[0]));
Chris@82 463 T5 = LD(&(ii[WS(rs, 8)]), ms, &(ii[0]));
Chris@82 464 T2 = LDW(&(W[TWVL * 14]));
Chris@82 465 T4 = LDW(&(W[TWVL * 15]));
Chris@82 466 T6 = VFMA(T2, T3, VMUL(T4, T5));
Chris@82 467 T2S = VFNMS(T4, T3, VMUL(T2, T5));
Chris@82 468 }
Chris@82 469 T7 = VADD(T1, T6);
Chris@82 470 T37 = VSUB(T2T, T2S);
Chris@82 471 T1t = VSUB(T1, T6);
Chris@82 472 T2U = VADD(T2S, T2T);
Chris@82 473 }
Chris@82 474 {
Chris@82 475 V Tc, T1u, Th, T1v;
Chris@82 476 {
Chris@82 477 V T9, Tb, T8, Ta;
Chris@82 478 T9 = LD(&(ri[WS(rs, 4)]), ms, &(ri[0]));
Chris@82 479 Tb = LD(&(ii[WS(rs, 4)]), ms, &(ii[0]));
Chris@82 480 T8 = LDW(&(W[TWVL * 6]));
Chris@82 481 Ta = LDW(&(W[TWVL * 7]));
Chris@82 482 Tc = VFMA(T8, T9, VMUL(Ta, Tb));
Chris@82 483 T1u = VFNMS(Ta, T9, VMUL(T8, Tb));
Chris@82 484 }
Chris@82 485 {
Chris@82 486 V Te, Tg, Td, Tf;
Chris@82 487 Te = LD(&(ri[WS(rs, 12)]), ms, &(ri[0]));
Chris@82 488 Tg = LD(&(ii[WS(rs, 12)]), ms, &(ii[0]));
Chris@82 489 Td = LDW(&(W[TWVL * 22]));
Chris@82 490 Tf = LDW(&(W[TWVL * 23]));
Chris@82 491 Th = VFMA(Td, Te, VMUL(Tf, Tg));
Chris@82 492 T1v = VFNMS(Tf, Te, VMUL(Td, Tg));
Chris@82 493 }
Chris@82 494 Ti = VADD(Tc, Th);
Chris@82 495 T38 = VSUB(Tc, Th);
Chris@82 496 T1w = VSUB(T1u, T1v);
Chris@82 497 T2R = VADD(T1u, T1v);
Chris@82 498 }
Chris@82 499 {
Chris@82 500 V To, T1y, Tt, T1z, T1A, T1B;
Chris@82 501 {
Chris@82 502 V Tl, Tn, Tk, Tm;
Chris@82 503 Tl = LD(&(ri[WS(rs, 2)]), ms, &(ri[0]));
Chris@82 504 Tn = LD(&(ii[WS(rs, 2)]), ms, &(ii[0]));
Chris@82 505 Tk = LDW(&(W[TWVL * 2]));
Chris@82 506 Tm = LDW(&(W[TWVL * 3]));
Chris@82 507 To = VFMA(Tk, Tl, VMUL(Tm, Tn));
Chris@82 508 T1y = VFNMS(Tm, Tl, VMUL(Tk, Tn));
Chris@82 509 }
Chris@82 510 {
Chris@82 511 V Tq, Ts, Tp, Tr;
Chris@82 512 Tq = LD(&(ri[WS(rs, 10)]), ms, &(ri[0]));
Chris@82 513 Ts = LD(&(ii[WS(rs, 10)]), ms, &(ii[0]));
Chris@82 514 Tp = LDW(&(W[TWVL * 18]));
Chris@82 515 Tr = LDW(&(W[TWVL * 19]));
Chris@82 516 Tt = VFMA(Tp, Tq, VMUL(Tr, Ts));
Chris@82 517 T1z = VFNMS(Tr, Tq, VMUL(Tp, Ts));
Chris@82 518 }
Chris@82 519 Tu = VADD(To, Tt);
Chris@82 520 T2s = VADD(T1y, T1z);
Chris@82 521 T1A = VSUB(T1y, T1z);
Chris@82 522 T1B = VSUB(To, Tt);
Chris@82 523 T1C = VSUB(T1A, T1B);
Chris@82 524 T2c = VADD(T1B, T1A);
Chris@82 525 }
Chris@82 526 {
Chris@82 527 V Tz, T1E, TE, T1F, T1D, T1G;
Chris@82 528 {
Chris@82 529 V Tw, Ty, Tv, Tx;
Chris@82 530 Tw = LD(&(ri[WS(rs, 14)]), ms, &(ri[0]));
Chris@82 531 Ty = LD(&(ii[WS(rs, 14)]), ms, &(ii[0]));
Chris@82 532 Tv = LDW(&(W[TWVL * 26]));
Chris@82 533 Tx = LDW(&(W[TWVL * 27]));
Chris@82 534 Tz = VFMA(Tv, Tw, VMUL(Tx, Ty));
Chris@82 535 T1E = VFNMS(Tx, Tw, VMUL(Tv, Ty));
Chris@82 536 }
Chris@82 537 {
Chris@82 538 V TB, TD, TA, TC;
Chris@82 539 TB = LD(&(ri[WS(rs, 6)]), ms, &(ri[0]));
Chris@82 540 TD = LD(&(ii[WS(rs, 6)]), ms, &(ii[0]));
Chris@82 541 TA = LDW(&(W[TWVL * 10]));
Chris@82 542 TC = LDW(&(W[TWVL * 11]));
Chris@82 543 TE = VFMA(TA, TB, VMUL(TC, TD));
Chris@82 544 T1F = VFNMS(TC, TB, VMUL(TA, TD));
Chris@82 545 }
Chris@82 546 TF = VADD(Tz, TE);
Chris@82 547 T2t = VADD(T1E, T1F);
Chris@82 548 T1D = VSUB(Tz, TE);
Chris@82 549 T1G = VSUB(T1E, T1F);
Chris@82 550 T1H = VADD(T1D, T1G);
Chris@82 551 T2d = VSUB(T1D, T1G);
Chris@82 552 }
Chris@82 553 {
Chris@82 554 V T19, T20, T1p, T1X, T1e, T21, T1k, T1W;
Chris@82 555 {
Chris@82 556 V T16, T18, T15, T17;
Chris@82 557 T16 = LD(&(ri[WS(rs, 15)]), ms, &(ri[WS(rs, 1)]));
Chris@82 558 T18 = LD(&(ii[WS(rs, 15)]), ms, &(ii[WS(rs, 1)]));
Chris@82 559 T15 = LDW(&(W[TWVL * 28]));
Chris@82 560 T17 = LDW(&(W[TWVL * 29]));
Chris@82 561 T19 = VFMA(T15, T16, VMUL(T17, T18));
Chris@82 562 T20 = VFNMS(T17, T16, VMUL(T15, T18));
Chris@82 563 }
Chris@82 564 {
Chris@82 565 V T1m, T1o, T1l, T1n;
Chris@82 566 T1m = LD(&(ri[WS(rs, 11)]), ms, &(ri[WS(rs, 1)]));
Chris@82 567 T1o = LD(&(ii[WS(rs, 11)]), ms, &(ii[WS(rs, 1)]));
Chris@82 568 T1l = LDW(&(W[TWVL * 20]));
Chris@82 569 T1n = LDW(&(W[TWVL * 21]));
Chris@82 570 T1p = VFMA(T1l, T1m, VMUL(T1n, T1o));
Chris@82 571 T1X = VFNMS(T1n, T1m, VMUL(T1l, T1o));
Chris@82 572 }
Chris@82 573 {
Chris@82 574 V T1b, T1d, T1a, T1c;
Chris@82 575 T1b = LD(&(ri[WS(rs, 7)]), ms, &(ri[WS(rs, 1)]));
Chris@82 576 T1d = LD(&(ii[WS(rs, 7)]), ms, &(ii[WS(rs, 1)]));
Chris@82 577 T1a = LDW(&(W[TWVL * 12]));
Chris@82 578 T1c = LDW(&(W[TWVL * 13]));
Chris@82 579 T1e = VFMA(T1a, T1b, VMUL(T1c, T1d));
Chris@82 580 T21 = VFNMS(T1c, T1b, VMUL(T1a, T1d));
Chris@82 581 }
Chris@82 582 {
Chris@82 583 V T1h, T1j, T1g, T1i;
Chris@82 584 T1h = LD(&(ri[WS(rs, 3)]), ms, &(ri[WS(rs, 1)]));
Chris@82 585 T1j = LD(&(ii[WS(rs, 3)]), ms, &(ii[WS(rs, 1)]));
Chris@82 586 T1g = LDW(&(W[TWVL * 4]));
Chris@82 587 T1i = LDW(&(W[TWVL * 5]));
Chris@82 588 T1k = VFMA(T1g, T1h, VMUL(T1i, T1j));
Chris@82 589 T1W = VFNMS(T1i, T1h, VMUL(T1g, T1j));
Chris@82 590 }
Chris@82 591 T1f = VADD(T19, T1e);
Chris@82 592 T1q = VADD(T1k, T1p);
Chris@82 593 T2B = VSUB(T1f, T1q);
Chris@82 594 T2C = VADD(T20, T21);
Chris@82 595 T2D = VADD(T1W, T1X);
Chris@82 596 T2E = VSUB(T2C, T2D);
Chris@82 597 {
Chris@82 598 V T1V, T1Y, T22, T23;
Chris@82 599 T1V = VSUB(T19, T1e);
Chris@82 600 T1Y = VSUB(T1W, T1X);
Chris@82 601 T1Z = VSUB(T1V, T1Y);
Chris@82 602 T2j = VADD(T1V, T1Y);
Chris@82 603 T22 = VSUB(T20, T21);
Chris@82 604 T23 = VSUB(T1k, T1p);
Chris@82 605 T24 = VADD(T22, T23);
Chris@82 606 T2k = VSUB(T22, T23);
Chris@82 607 }
Chris@82 608 }
Chris@82 609 {
Chris@82 610 V TM, T1K, T12, T1R, TR, T1L, TX, T1Q;
Chris@82 611 {
Chris@82 612 V TJ, TL, TI, TK;
Chris@82 613 TJ = LD(&(ri[WS(rs, 1)]), ms, &(ri[WS(rs, 1)]));
Chris@82 614 TL = LD(&(ii[WS(rs, 1)]), ms, &(ii[WS(rs, 1)]));
Chris@82 615 TI = LDW(&(W[0]));
Chris@82 616 TK = LDW(&(W[TWVL * 1]));
Chris@82 617 TM = VFMA(TI, TJ, VMUL(TK, TL));
Chris@82 618 T1K = VFNMS(TK, TJ, VMUL(TI, TL));
Chris@82 619 }
Chris@82 620 {
Chris@82 621 V TZ, T11, TY, T10;
Chris@82 622 TZ = LD(&(ri[WS(rs, 13)]), ms, &(ri[WS(rs, 1)]));
Chris@82 623 T11 = LD(&(ii[WS(rs, 13)]), ms, &(ii[WS(rs, 1)]));
Chris@82 624 TY = LDW(&(W[TWVL * 24]));
Chris@82 625 T10 = LDW(&(W[TWVL * 25]));
Chris@82 626 T12 = VFMA(TY, TZ, VMUL(T10, T11));
Chris@82 627 T1R = VFNMS(T10, TZ, VMUL(TY, T11));
Chris@82 628 }
Chris@82 629 {
Chris@82 630 V TO, TQ, TN, TP;
Chris@82 631 TO = LD(&(ri[WS(rs, 9)]), ms, &(ri[WS(rs, 1)]));
Chris@82 632 TQ = LD(&(ii[WS(rs, 9)]), ms, &(ii[WS(rs, 1)]));
Chris@82 633 TN = LDW(&(W[TWVL * 16]));
Chris@82 634 TP = LDW(&(W[TWVL * 17]));
Chris@82 635 TR = VFMA(TN, TO, VMUL(TP, TQ));
Chris@82 636 T1L = VFNMS(TP, TO, VMUL(TN, TQ));
Chris@82 637 }
Chris@82 638 {
Chris@82 639 V TU, TW, TT, TV;
Chris@82 640 TU = LD(&(ri[WS(rs, 5)]), ms, &(ri[WS(rs, 1)]));
Chris@82 641 TW = LD(&(ii[WS(rs, 5)]), ms, &(ii[WS(rs, 1)]));
Chris@82 642 TT = LDW(&(W[TWVL * 8]));
Chris@82 643 TV = LDW(&(W[TWVL * 9]));
Chris@82 644 TX = VFMA(TT, TU, VMUL(TV, TW));
Chris@82 645 T1Q = VFNMS(TV, TU, VMUL(TT, TW));
Chris@82 646 }
Chris@82 647 TS = VADD(TM, TR);
Chris@82 648 T13 = VADD(TX, T12);
Chris@82 649 T2w = VSUB(TS, T13);
Chris@82 650 T2x = VADD(T1K, T1L);
Chris@82 651 T2y = VADD(T1Q, T1R);
Chris@82 652 T2z = VSUB(T2x, T2y);
Chris@82 653 {
Chris@82 654 V T1M, T1N, T1P, T1S;
Chris@82 655 T1M = VSUB(T1K, T1L);
Chris@82 656 T1N = VSUB(TX, T12);
Chris@82 657 T1O = VADD(T1M, T1N);
Chris@82 658 T2g = VSUB(T1M, T1N);
Chris@82 659 T1P = VSUB(TM, TR);
Chris@82 660 T1S = VSUB(T1Q, T1R);
Chris@82 661 T1T = VSUB(T1P, T1S);
Chris@82 662 T2h = VADD(T1P, T1S);
Chris@82 663 }
Chris@82 664 }
Chris@82 665 {
Chris@82 666 V T1J, T27, T3g, T3i, T26, T3h, T2a, T3d;
Chris@82 667 {
Chris@82 668 V T1x, T1I, T3e, T3f;
Chris@82 669 T1x = VSUB(T1t, T1w);
Chris@82 670 T1I = VMUL(LDK(KP707106781), VSUB(T1C, T1H));
Chris@82 671 T1J = VADD(T1x, T1I);
Chris@82 672 T27 = VSUB(T1x, T1I);
Chris@82 673 T3e = VMUL(LDK(KP707106781), VSUB(T2d, T2c));
Chris@82 674 T3f = VADD(T38, T37);
Chris@82 675 T3g = VADD(T3e, T3f);
Chris@82 676 T3i = VSUB(T3f, T3e);
Chris@82 677 }
Chris@82 678 {
Chris@82 679 V T1U, T25, T28, T29;
Chris@82 680 T1U = VFMA(LDK(KP923879532), T1O, VMUL(LDK(KP382683432), T1T));
Chris@82 681 T25 = VFNMS(LDK(KP923879532), T24, VMUL(LDK(KP382683432), T1Z));
Chris@82 682 T26 = VADD(T1U, T25);
Chris@82 683 T3h = VSUB(T25, T1U);
Chris@82 684 T28 = VFNMS(LDK(KP923879532), T1T, VMUL(LDK(KP382683432), T1O));
Chris@82 685 T29 = VFMA(LDK(KP382683432), T24, VMUL(LDK(KP923879532), T1Z));
Chris@82 686 T2a = VSUB(T28, T29);
Chris@82 687 T3d = VADD(T28, T29);
Chris@82 688 }
Chris@82 689 ST(&(ri[WS(rs, 11)]), VSUB(T1J, T26), ms, &(ri[WS(rs, 1)]));
Chris@82 690 ST(&(ii[WS(rs, 11)]), VSUB(T3g, T3d), ms, &(ii[WS(rs, 1)]));
Chris@82 691 ST(&(ri[WS(rs, 3)]), VADD(T1J, T26), ms, &(ri[WS(rs, 1)]));
Chris@82 692 ST(&(ii[WS(rs, 3)]), VADD(T3d, T3g), ms, &(ii[WS(rs, 1)]));
Chris@82 693 ST(&(ri[WS(rs, 15)]), VSUB(T27, T2a), ms, &(ri[WS(rs, 1)]));
Chris@82 694 ST(&(ii[WS(rs, 15)]), VSUB(T3i, T3h), ms, &(ii[WS(rs, 1)]));
Chris@82 695 ST(&(ri[WS(rs, 7)]), VADD(T27, T2a), ms, &(ri[WS(rs, 1)]));
Chris@82 696 ST(&(ii[WS(rs, 7)]), VADD(T3h, T3i), ms, &(ii[WS(rs, 1)]));
Chris@82 697 }
Chris@82 698 {
Chris@82 699 V T2v, T2H, T32, T34, T2G, T33, T2K, T2Z;
Chris@82 700 {
Chris@82 701 V T2r, T2u, T30, T31;
Chris@82 702 T2r = VSUB(T7, Ti);
Chris@82 703 T2u = VSUB(T2s, T2t);
Chris@82 704 T2v = VADD(T2r, T2u);
Chris@82 705 T2H = VSUB(T2r, T2u);
Chris@82 706 T30 = VSUB(TF, Tu);
Chris@82 707 T31 = VSUB(T2U, T2R);
Chris@82 708 T32 = VADD(T30, T31);
Chris@82 709 T34 = VSUB(T31, T30);
Chris@82 710 }
Chris@82 711 {
Chris@82 712 V T2A, T2F, T2I, T2J;
Chris@82 713 T2A = VADD(T2w, T2z);
Chris@82 714 T2F = VSUB(T2B, T2E);
Chris@82 715 T2G = VMUL(LDK(KP707106781), VADD(T2A, T2F));
Chris@82 716 T33 = VMUL(LDK(KP707106781), VSUB(T2F, T2A));
Chris@82 717 T2I = VSUB(T2z, T2w);
Chris@82 718 T2J = VADD(T2B, T2E);
Chris@82 719 T2K = VMUL(LDK(KP707106781), VSUB(T2I, T2J));
Chris@82 720 T2Z = VMUL(LDK(KP707106781), VADD(T2I, T2J));
Chris@82 721 }
Chris@82 722 ST(&(ri[WS(rs, 10)]), VSUB(T2v, T2G), ms, &(ri[0]));
Chris@82 723 ST(&(ii[WS(rs, 10)]), VSUB(T32, T2Z), ms, &(ii[0]));
Chris@82 724 ST(&(ri[WS(rs, 2)]), VADD(T2v, T2G), ms, &(ri[0]));
Chris@82 725 ST(&(ii[WS(rs, 2)]), VADD(T2Z, T32), ms, &(ii[0]));
Chris@82 726 ST(&(ri[WS(rs, 14)]), VSUB(T2H, T2K), ms, &(ri[0]));
Chris@82 727 ST(&(ii[WS(rs, 14)]), VSUB(T34, T33), ms, &(ii[0]));
Chris@82 728 ST(&(ri[WS(rs, 6)]), VADD(T2H, T2K), ms, &(ri[0]));
Chris@82 729 ST(&(ii[WS(rs, 6)]), VADD(T33, T34), ms, &(ii[0]));
Chris@82 730 }
Chris@82 731 {
Chris@82 732 V T2f, T2n, T3a, T3c, T2m, T3b, T2q, T35;
Chris@82 733 {
Chris@82 734 V T2b, T2e, T36, T39;
Chris@82 735 T2b = VADD(T1t, T1w);
Chris@82 736 T2e = VMUL(LDK(KP707106781), VADD(T2c, T2d));
Chris@82 737 T2f = VADD(T2b, T2e);
Chris@82 738 T2n = VSUB(T2b, T2e);
Chris@82 739 T36 = VMUL(LDK(KP707106781), VADD(T1C, T1H));
Chris@82 740 T39 = VSUB(T37, T38);
Chris@82 741 T3a = VADD(T36, T39);
Chris@82 742 T3c = VSUB(T39, T36);
Chris@82 743 }
Chris@82 744 {
Chris@82 745 V T2i, T2l, T2o, T2p;
Chris@82 746 T2i = VFMA(LDK(KP382683432), T2g, VMUL(LDK(KP923879532), T2h));
Chris@82 747 T2l = VFNMS(LDK(KP382683432), T2k, VMUL(LDK(KP923879532), T2j));
Chris@82 748 T2m = VADD(T2i, T2l);
Chris@82 749 T3b = VSUB(T2l, T2i);
Chris@82 750 T2o = VFNMS(LDK(KP382683432), T2h, VMUL(LDK(KP923879532), T2g));
Chris@82 751 T2p = VFMA(LDK(KP923879532), T2k, VMUL(LDK(KP382683432), T2j));
Chris@82 752 T2q = VSUB(T2o, T2p);
Chris@82 753 T35 = VADD(T2o, T2p);
Chris@82 754 }
Chris@82 755 ST(&(ri[WS(rs, 9)]), VSUB(T2f, T2m), ms, &(ri[WS(rs, 1)]));
Chris@82 756 ST(&(ii[WS(rs, 9)]), VSUB(T3a, T35), ms, &(ii[WS(rs, 1)]));
Chris@82 757 ST(&(ri[WS(rs, 1)]), VADD(T2f, T2m), ms, &(ri[WS(rs, 1)]));
Chris@82 758 ST(&(ii[WS(rs, 1)]), VADD(T35, T3a), ms, &(ii[WS(rs, 1)]));
Chris@82 759 ST(&(ri[WS(rs, 13)]), VSUB(T2n, T2q), ms, &(ri[WS(rs, 1)]));
Chris@82 760 ST(&(ii[WS(rs, 13)]), VSUB(T3c, T3b), ms, &(ii[WS(rs, 1)]));
Chris@82 761 ST(&(ri[WS(rs, 5)]), VADD(T2n, T2q), ms, &(ri[WS(rs, 1)]));
Chris@82 762 ST(&(ii[WS(rs, 5)]), VADD(T3b, T3c), ms, &(ii[WS(rs, 1)]));
Chris@82 763 }
Chris@82 764 {
Chris@82 765 V TH, T2L, T2W, T2Y, T1s, T2X, T2O, T2P;
Chris@82 766 {
Chris@82 767 V Tj, TG, T2Q, T2V;
Chris@82 768 Tj = VADD(T7, Ti);
Chris@82 769 TG = VADD(Tu, TF);
Chris@82 770 TH = VADD(Tj, TG);
Chris@82 771 T2L = VSUB(Tj, TG);
Chris@82 772 T2Q = VADD(T2s, T2t);
Chris@82 773 T2V = VADD(T2R, T2U);
Chris@82 774 T2W = VADD(T2Q, T2V);
Chris@82 775 T2Y = VSUB(T2V, T2Q);
Chris@82 776 }
Chris@82 777 {
Chris@82 778 V T14, T1r, T2M, T2N;
Chris@82 779 T14 = VADD(TS, T13);
Chris@82 780 T1r = VADD(T1f, T1q);
Chris@82 781 T1s = VADD(T14, T1r);
Chris@82 782 T2X = VSUB(T1r, T14);
Chris@82 783 T2M = VADD(T2x, T2y);
Chris@82 784 T2N = VADD(T2C, T2D);
Chris@82 785 T2O = VSUB(T2M, T2N);
Chris@82 786 T2P = VADD(T2M, T2N);
Chris@82 787 }
Chris@82 788 ST(&(ri[WS(rs, 8)]), VSUB(TH, T1s), ms, &(ri[0]));
Chris@82 789 ST(&(ii[WS(rs, 8)]), VSUB(T2W, T2P), ms, &(ii[0]));
Chris@82 790 ST(&(ri[0]), VADD(TH, T1s), ms, &(ri[0]));
Chris@82 791 ST(&(ii[0]), VADD(T2P, T2W), ms, &(ii[0]));
Chris@82 792 ST(&(ri[WS(rs, 12)]), VSUB(T2L, T2O), ms, &(ri[0]));
Chris@82 793 ST(&(ii[WS(rs, 12)]), VSUB(T2Y, T2X), ms, &(ii[0]));
Chris@82 794 ST(&(ri[WS(rs, 4)]), VADD(T2L, T2O), ms, &(ri[0]));
Chris@82 795 ST(&(ii[WS(rs, 4)]), VADD(T2X, T2Y), ms, &(ii[0]));
Chris@82 796 }
Chris@82 797 }
Chris@82 798 }
Chris@82 799 VLEAVE();
Chris@82 800 }
Chris@82 801
Chris@82 802 static const tw_instr twinstr[] = {
Chris@82 803 VTW(0, 1),
Chris@82 804 VTW(0, 2),
Chris@82 805 VTW(0, 3),
Chris@82 806 VTW(0, 4),
Chris@82 807 VTW(0, 5),
Chris@82 808 VTW(0, 6),
Chris@82 809 VTW(0, 7),
Chris@82 810 VTW(0, 8),
Chris@82 811 VTW(0, 9),
Chris@82 812 VTW(0, 10),
Chris@82 813 VTW(0, 11),
Chris@82 814 VTW(0, 12),
Chris@82 815 VTW(0, 13),
Chris@82 816 VTW(0, 14),
Chris@82 817 VTW(0, 15),
Chris@82 818 {TW_NEXT, (2 * VL), 0}
Chris@82 819 };
Chris@82 820
Chris@82 821 static const ct_desc desc = { 16, XSIMD_STRING("t1sv_16"), twinstr, &GENUS, {136, 46, 38, 0}, 0, 0, 0 };
Chris@82 822
Chris@82 823 void XSIMD(codelet_t1sv_16) (planner *p) {
Chris@82 824 X(kdft_dit_register) (p, t1sv_16, &desc);
Chris@82 825 }
Chris@82 826 #endif