Mercurial > hg > sv-dependency-builds

annotate src/fftw-3.3.5/dft/simd/common/t2sv_16.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 16:45:19 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_twiddle.native -fma -reorder-insns -schedule-for-pipeline -simd -compact -variables 4 -pipeline-latency 8 -twiddle-log3 -precompute-twiddles -n 16 -name t2sv_16 -include ts.h */
Chris@42 29
Chris@42 30 /*
Chris@42 31 * This function contains 196 FP additions, 134 FP multiplications,
Chris@42 32 * (or, 104 additions, 42 multiplications, 92 fused multiply/add),
Chris@42 33 * 120 stack variables, 3 constants, and 64 memory accesses
Chris@42 34 */
Chris@42 35 #include "ts.h"
Chris@42 36
Chris@42 37 static void t2sv_16(R *ri, R *ii, const R *W, stride rs, INT mb, INT me, INT ms)
Chris@42 38 {
Chris@42 39 DVK(KP923879532, +0.923879532511286756128183189396788286822416626);
Chris@42 40 DVK(KP414213562, +0.414213562373095048801688724209698078569671875);
Chris@42 41 DVK(KP707106781, +0.707106781186547524400844362104849039284835938);
Chris@42 42 {
Chris@42 43 INT m;
Chris@42 44 for (m = mb, W = W + (mb * 8); m < me; m = m + (2 * VL), ri = ri + ((2 * VL) * ms), ii = ii + ((2 * VL) * ms), W = W + ((2 * VL) * 8), MAKE_VOLATILE_STRIDE(32, rs)) {
Chris@42 45 V T34, T30, T2N, T2v, T2M, T2g, T3V, T3X, T32, T2U, T33, T2X, T2O, T2K, T3P;
Chris@42 46 V T3R;
Chris@42 47 {
Chris@42 48 V T2, Tf, TM, TO, T3, T6, T5, Th;
Chris@42 49 T2 = LDW(&(W[0]));
Chris@42 50 Tf = LDW(&(W[TWVL * 2]));
Chris@42 51 TM = LDW(&(W[TWVL * 6]));
Chris@42 52 TO = LDW(&(W[TWVL * 7]));
Chris@42 53 T3 = LDW(&(W[TWVL * 4]));
Chris@42 54 T6 = LDW(&(W[TWVL * 5]));
Chris@42 55 T5 = LDW(&(W[TWVL * 1]));
Chris@42 56 Th = LDW(&(W[TWVL * 3]));
Chris@42 57 {
Chris@42 58 V TW, TZ, Te, T1U, T3A, T3L, T2D, T1G, T3h, T2A, T2B, T1R, T3i, T2I, Tx;
Chris@42 59 V T3M, T1Z, T3w, TL, T26, T25, T37, T1l, T2q, T1d, T2o, T2l, T3c, T1r, T2s;
Chris@42 60 V TX, T10, TV, T2a;
Chris@42 61 {
Chris@42 62 V Tz, TP, TT, Tq, TF, Tu, TI, Tm, TC, T1j, T1p, T1m, T1f, T1O, T1M;
Chris@42 63 V T1K, T2F, Tj, Tn, T1Q, T2G, Tk, T1V, Tr, Tv;
Chris@42 64 {
Chris@42 65 V T1, Ti, Tb, T3z, T8, Tc, T1u, T1D, T1L, T1z, T9, T3x, T1v, T1w, T1A;
Chris@42 66 V T1E;
Chris@42 67 {
Chris@42 68 V T7, T1i, T1e, T1C, T1y;
Chris@42 69 T1 = LD(&(ri[0]), ms, &(ri[0]));
Chris@42 70 {
Chris@42 71 V Tg, TN, TS, Tp;
Chris@42 72 Tg = VMUL(T2, Tf);
Chris@42 73 TN = VMUL(T2, TM);
Chris@42 74 TS = VMUL(T2, TO);
Chris@42 75 Tp = VMUL(Tf, T3);
Chris@42 76 {
Chris@42 77 V T4, Tt, Ta, Tl;
Chris@42 78 T4 = VMUL(T2, T3);
Chris@42 79 Tt = VMUL(Tf, T6);
Chris@42 80 Ta = VMUL(T2, T6);
Chris@42 81 Tl = VMUL(T2, Th);
Chris@42 82 Ti = VFNMS(T5, Th, Tg);
Chris@42 83 Tz = VFMA(T5, Th, Tg);
Chris@42 84 TP = VFMA(T5, TO, TN);
Chris@42 85 TT = VFNMS(T5, TM, TS);
Chris@42 86 TW = VFMA(Th, T6, Tp);
Chris@42 87 Tq = VFNMS(Th, T6, Tp);
Chris@42 88 TF = VFNMS(T5, T6, T4);
Chris@42 89 T7 = VFMA(T5, T6, T4);
Chris@42 90 Tu = VFMA(Th, T3, Tt);
Chris@42 91 TZ = VFNMS(Th, T3, Tt);
Chris@42 92 TI = VFMA(T5, T3, Ta);
Chris@42 93 Tb = VFNMS(T5, T3, Ta);
Chris@42 94 Tm = VFMA(T5, Tf, Tl);
Chris@42 95 TC = VFNMS(T5, Tf, Tl);
Chris@42 96 T1i = VMUL(Ti, T6);
Chris@42 97 T1e = VMUL(Ti, T3);
Chris@42 98 T1C = VMUL(Tz, T6);
Chris@42 99 T1y = VMUL(Tz, T3);
Chris@42 100 T3z = LD(&(ii[0]), ms, &(ii[0]));
Chris@42 101 }
Chris@42 102 }
Chris@42 103 T8 = LD(&(ri[WS(rs, 8)]), ms, &(ri[0]));
Chris@42 104 Tc = LD(&(ii[WS(rs, 8)]), ms, &(ii[0]));
Chris@42 105 T1u = LD(&(ri[WS(rs, 15)]), ms, &(ri[WS(rs, 1)]));
Chris@42 106 T1j = VFNMS(Tm, T3, T1i);
Chris@42 107 T1p = VFMA(Tm, T3, T1i);
Chris@42 108 T1m = VFNMS(Tm, T6, T1e);
Chris@42 109 T1f = VFMA(Tm, T6, T1e);
Chris@42 110 T1D = VFNMS(TC, T3, T1C);
Chris@42 111 T1O = VFMA(TC, T3, T1C);
Chris@42 112 T1L = VFNMS(TC, T6, T1y);
Chris@42 113 T1z = VFMA(TC, T6, T1y);
Chris@42 114 T9 = VMUL(T7, T8);
Chris@42 115 T3x = VMUL(T7, Tc);
Chris@42 116 T1v = VMUL(TM, T1u);
Chris@42 117 T1w = LD(&(ii[WS(rs, 15)]), ms, &(ii[WS(rs, 1)]));
Chris@42 118 T1A = LD(&(ri[WS(rs, 7)]), ms, &(ri[WS(rs, 1)]));
Chris@42 119 T1E = LD(&(ii[WS(rs, 7)]), ms, &(ii[WS(rs, 1)]));
Chris@42 120 }
Chris@42 121 {
Chris@42 122 V T1x, T2x, T1F, T2z, T1N, T1P;
Chris@42 123 {
Chris@42 124 V T1H, T1J, T1I, T2E;
Chris@42 125 {
Chris@42 126 V Td, T3y, T2w, T1B, T2y;
Chris@42 127 T1H = LD(&(ri[WS(rs, 3)]), ms, &(ri[WS(rs, 1)]));
Chris@42 128 T1J = LD(&(ii[WS(rs, 3)]), ms, &(ii[WS(rs, 1)]));
Chris@42 129 Td = VFMA(Tb, Tc, T9);
Chris@42 130 T3y = VFNMS(Tb, T8, T3x);
Chris@42 131 T1M = LD(&(ri[WS(rs, 11)]), ms, &(ri[WS(rs, 1)]));
Chris@42 132 T1x = VFMA(TO, T1w, T1v);
Chris@42 133 T2w = VMUL(TM, T1w);
Chris@42 134 T1B = VMUL(T1z, T1A);
Chris@42 135 T2y = VMUL(T1z, T1E);
Chris@42 136 T1I = VMUL(Tf, T1H);
Chris@42 137 T2E = VMUL(Tf, T1J);
Chris@42 138 Te = VADD(T1, Td);
Chris@42 139 T1U = VSUB(T1, Td);
Chris@42 140 T3A = VADD(T3y, T3z);
Chris@42 141 T3L = VSUB(T3z, T3y);
Chris@42 142 T2x = VFNMS(TO, T1u, T2w);
Chris@42 143 T1F = VFMA(T1D, T1E, T1B);
Chris@42 144 T2z = VFNMS(T1D, T1A, T2y);
Chris@42 145 T1N = VMUL(T1L, T1M);
Chris@42 146 T1P = LD(&(ii[WS(rs, 11)]), ms, &(ii[WS(rs, 1)]));
Chris@42 147 }
Chris@42 148 T1K = VFMA(Th, T1J, T1I);
Chris@42 149 T2F = VFNMS(Th, T1H, T2E);
Chris@42 150 }
Chris@42 151 Tj = LD(&(ri[WS(rs, 4)]), ms, &(ri[0]));
Chris@42 152 Tn = LD(&(ii[WS(rs, 4)]), ms, &(ii[0]));
Chris@42 153 T2D = VSUB(T1x, T1F);
Chris@42 154 T1G = VADD(T1x, T1F);
Chris@42 155 T3h = VADD(T2x, T2z);
Chris@42 156 T2A = VSUB(T2x, T2z);
Chris@42 157 T1Q = VFMA(T1O, T1P, T1N);
Chris@42 158 T2G = VMUL(T1L, T1P);
Chris@42 159 Tk = VMUL(Ti, Tj);
Chris@42 160 T1V = VMUL(Ti, Tn);
Chris@42 161 Tr = LD(&(ri[WS(rs, 12)]), ms, &(ri[0]));
Chris@42 162 Tv = LD(&(ii[WS(rs, 12)]), ms, &(ii[0]));
Chris@42 163 }
Chris@42 164 }
Chris@42 165 {
Chris@42 166 V TE, T22, T15, T17, TK, T16, T2h, T24, T19, T1b;
Chris@42 167 {
Chris@42 168 V To, T1W, TG, TJ, Tw, T1Y, TH, T23;
Chris@42 169 {
Chris@42 170 V TA, TD, TB, T21, T2H, Ts, T1X;
Chris@42 171 TA = LD(&(ri[WS(rs, 2)]), ms, &(ri[0]));
Chris@42 172 TD = LD(&(ii[WS(rs, 2)]), ms, &(ii[0]));
Chris@42 173 T2B = VSUB(T1K, T1Q);
Chris@42 174 T1R = VADD(T1K, T1Q);
Chris@42 175 T2H = VFNMS(T1O, T1M, T2G);
Chris@42 176 To = VFMA(Tm, Tn, Tk);
Chris@42 177 T1W = VFNMS(Tm, Tj, T1V);
Chris@42 178 Ts = VMUL(Tq, Tr);
Chris@42 179 T1X = VMUL(Tq, Tv);
Chris@42 180 TB = VMUL(Tz, TA);
Chris@42 181 T21 = VMUL(Tz, TD);
Chris@42 182 TG = LD(&(ri[WS(rs, 10)]), ms, &(ri[0]));
Chris@42 183 T3i = VADD(T2F, T2H);
Chris@42 184 T2I = VSUB(T2F, T2H);
Chris@42 185 TJ = LD(&(ii[WS(rs, 10)]), ms, &(ii[0]));
Chris@42 186 Tw = VFMA(Tu, Tv, Ts);
Chris@42 187 T1Y = VFNMS(Tu, Tr, T1X);
Chris@42 188 TE = VFMA(TC, TD, TB);
Chris@42 189 T22 = VFNMS(TC, TA, T21);
Chris@42 190 TH = VMUL(TF, TG);
Chris@42 191 }
Chris@42 192 T15 = LD(&(ri[WS(rs, 1)]), ms, &(ri[WS(rs, 1)]));
Chris@42 193 T17 = LD(&(ii[WS(rs, 1)]), ms, &(ii[WS(rs, 1)]));
Chris@42 194 T23 = VMUL(TF, TJ);
Chris@42 195 Tx = VADD(To, Tw);
Chris@42 196 T3M = VSUB(To, Tw);
Chris@42 197 T1Z = VSUB(T1W, T1Y);
Chris@42 198 T3w = VADD(T1W, T1Y);
Chris@42 199 TK = VFMA(TI, TJ, TH);
Chris@42 200 T16 = VMUL(T2, T15);
Chris@42 201 T2h = VMUL(T2, T17);
Chris@42 202 T24 = VFNMS(TI, TG, T23);
Chris@42 203 T19 = LD(&(ri[WS(rs, 9)]), ms, &(ri[WS(rs, 1)]));
Chris@42 204 T1b = LD(&(ii[WS(rs, 9)]), ms, &(ii[WS(rs, 1)]));
Chris@42 205 }
Chris@42 206 {
Chris@42 207 V T1g, T1k, T18, T2i, T1a, T2j, T1h, T2p, T1n, T1q;
Chris@42 208 T1g = LD(&(ri[WS(rs, 5)]), ms, &(ri[WS(rs, 1)]));
Chris@42 209 T1k = LD(&(ii[WS(rs, 5)]), ms, &(ii[WS(rs, 1)]));
Chris@42 210 TL = VADD(TE, TK);
Chris@42 211 T26 = VSUB(TE, TK);
Chris@42 212 T18 = VFMA(T5, T17, T16);
Chris@42 213 T2i = VFNMS(T5, T15, T2h);
Chris@42 214 T25 = VSUB(T22, T24);
Chris@42 215 T37 = VADD(T22, T24);
Chris@42 216 T1a = VMUL(T3, T19);
Chris@42 217 T2j = VMUL(T3, T1b);
Chris@42 218 T1h = VMUL(T1f, T1g);
Chris@42 219 T2p = VMUL(T1f, T1k);
Chris@42 220 T1n = LD(&(ri[WS(rs, 13)]), ms, &(ri[WS(rs, 1)]));
Chris@42 221 T1q = LD(&(ii[WS(rs, 13)]), ms, &(ii[WS(rs, 1)]));
Chris@42 222 {
Chris@42 223 V TQ, TU, TR, T29;
Chris@42 224 {
Chris@42 225 V T1c, T2k, T1o, T2r;
Chris@42 226 TQ = LD(&(ri[WS(rs, 14)]), ms, &(ri[0]));
Chris@42 227 TU = LD(&(ii[WS(rs, 14)]), ms, &(ii[0]));
Chris@42 228 T1c = VFMA(T6, T1b, T1a);
Chris@42 229 T2k = VFNMS(T6, T19, T2j);
Chris@42 230 T1l = VFMA(T1j, T1k, T1h);
Chris@42 231 T2q = VFNMS(T1j, T1g, T2p);
Chris@42 232 T1o = VMUL(T1m, T1n);
Chris@42 233 T2r = VMUL(T1m, T1q);
Chris@42 234 TR = VMUL(TP, TQ);
Chris@42 235 T29 = VMUL(TP, TU);
Chris@42 236 T1d = VADD(T18, T1c);
Chris@42 237 T2o = VSUB(T18, T1c);
Chris@42 238 T2l = VSUB(T2i, T2k);
Chris@42 239 T3c = VADD(T2i, T2k);
Chris@42 240 T1r = VFMA(T1p, T1q, T1o);
Chris@42 241 T2s = VFNMS(T1p, T1n, T2r);
Chris@42 242 TX = LD(&(ri[WS(rs, 6)]), ms, &(ri[0]));
Chris@42 243 T10 = LD(&(ii[WS(rs, 6)]), ms, &(ii[0]));
Chris@42 244 }
Chris@42 245 TV = VFMA(TT, TU, TR);
Chris@42 246 T2a = VFNMS(TT, TQ, T29);
Chris@42 247 }
Chris@42 248 }
Chris@42 249 }
Chris@42 250 }
Chris@42 251 {
Chris@42 252 V T36, Ty, T3B, T3G, T1s, T2m, T2t, T3d, TY, T2b, T3g, T1S, T3s, T3j;
Chris@42 253 T36 = VSUB(Te, Tx);
Chris@42 254 Ty = VADD(Te, Tx);
Chris@42 255 T3B = VADD(T3w, T3A);
Chris@42 256 T3G = VSUB(T3A, T3w);
Chris@42 257 T1s = VADD(T1l, T1r);
Chris@42 258 T2m = VSUB(T1l, T1r);
Chris@42 259 T2t = VSUB(T2q, T2s);
Chris@42 260 T3d = VADD(T2q, T2s);
Chris@42 261 TY = VMUL(TW, TX);
Chris@42 262 T2b = VMUL(TW, T10);
Chris@42 263 T3g = VSUB(T1G, T1R);
Chris@42 264 T1S = VADD(T1G, T1R);
Chris@42 265 T3s = VADD(T3h, T3i);
Chris@42 266 T3j = VSUB(T3h, T3i);
Chris@42 267 {
Chris@42 268 V T3D, T1T, T3u, T3t, T28, T12, T38, T2d, T3n, T3f;
Chris@42 269 {
Chris@42 270 V T1t, T3b, T3e, T3r, T11, T2c;
Chris@42 271 T1t = VADD(T1d, T1s);
Chris@42 272 T3b = VSUB(T1d, T1s);
Chris@42 273 T3e = VSUB(T3c, T3d);
Chris@42 274 T3r = VADD(T3c, T3d);
Chris@42 275 T11 = VFMA(TZ, T10, TY);
Chris@42 276 T2c = VFNMS(TZ, TX, T2b);
Chris@42 277 T3D = VSUB(T1S, T1t);
Chris@42 278 T1T = VADD(T1t, T1S);
Chris@42 279 T3u = VADD(T3r, T3s);
Chris@42 280 T3t = VSUB(T3r, T3s);
Chris@42 281 T28 = VSUB(TV, T11);
Chris@42 282 T12 = VADD(TV, T11);
Chris@42 283 T38 = VADD(T2a, T2c);
Chris@42 284 T2d = VSUB(T2a, T2c);
Chris@42 285 T3n = VSUB(T3e, T3b);
Chris@42 286 T3f = VADD(T3b, T3e);
Chris@42 287 }
Chris@42 288 {
Chris@42 289 V T2Q, T20, T3N, T3T, T2J, T2C, T2W, T2V, T3O, T2f, T3U, T2T;
Chris@42 290 {
Chris@42 291 V T2R, T27, T2e, T2S, T13, T3F;
Chris@42 292 T2Q = VADD(T1U, T1Z);
Chris@42 293 T20 = VSUB(T1U, T1Z);
Chris@42 294 T3N = VSUB(T3L, T3M);
Chris@42 295 T3T = VADD(T3M, T3L);
Chris@42 296 T13 = VADD(TL, T12);
Chris@42 297 T3F = VSUB(T12, TL);
Chris@42 298 {
Chris@42 299 V T3v, T39, T3o, T3k;
Chris@42 300 T3v = VADD(T37, T38);
Chris@42 301 T39 = VSUB(T37, T38);
Chris@42 302 T3o = VADD(T3g, T3j);
Chris@42 303 T3k = VSUB(T3g, T3j);
Chris@42 304 {
Chris@42 305 V T3H, T3J, T14, T3q;
Chris@42 306 T3H = VADD(T3F, T3G);
Chris@42 307 T3J = VSUB(T3G, T3F);
Chris@42 308 T14 = VADD(Ty, T13);
Chris@42 309 T3q = VSUB(Ty, T13);
Chris@42 310 {
Chris@42 311 V T3a, T3m, T3C, T3E;
Chris@42 312 T3a = VADD(T36, T39);
Chris@42 313 T3m = VSUB(T36, T39);
Chris@42 314 T3C = VADD(T3v, T3B);
Chris@42 315 T3E = VSUB(T3B, T3v);
Chris@42 316 {
Chris@42 317 V T3I, T3p, T3l, T3K;
Chris@42 318 T3I = VADD(T3n, T3o);
Chris@42 319 T3p = VSUB(T3n, T3o);
Chris@42 320 T3l = VADD(T3f, T3k);
Chris@42 321 T3K = VSUB(T3k, T3f);
Chris@42 322 ST(&(ri[WS(rs, 4)]), VADD(T3q, T3t), ms, &(ri[0]));
Chris@42 323 ST(&(ri[WS(rs, 12)]), VSUB(T3q, T3t), ms, &(ri[0]));
Chris@42 324 ST(&(ri[0]), VADD(T14, T1T), ms, &(ri[0]));
Chris@42 325 ST(&(ri[WS(rs, 8)]), VSUB(T14, T1T), ms, &(ri[0]));
Chris@42 326 ST(&(ii[WS(rs, 4)]), VADD(T3D, T3E), ms, &(ii[0]));
Chris@42 327 ST(&(ii[WS(rs, 12)]), VSUB(T3E, T3D), ms, &(ii[0]));
Chris@42 328 ST(&(ii[0]), VADD(T3u, T3C), ms, &(ii[0]));
Chris@42 329 ST(&(ii[WS(rs, 8)]), VSUB(T3C, T3u), ms, &(ii[0]));
Chris@42 330 ST(&(ri[WS(rs, 6)]), VFMA(LDK(KP707106781), T3p, T3m), ms, &(ri[0]));
Chris@42 331 ST(&(ri[WS(rs, 14)]), VFNMS(LDK(KP707106781), T3p, T3m), ms, &(ri[0]));
Chris@42 332 ST(&(ii[WS(rs, 10)]), VFNMS(LDK(KP707106781), T3I, T3H), ms, &(ii[0]));
Chris@42 333 ST(&(ii[WS(rs, 2)]), VFMA(LDK(KP707106781), T3I, T3H), ms, &(ii[0]));
Chris@42 334 ST(&(ii[WS(rs, 14)]), VFNMS(LDK(KP707106781), T3K, T3J), ms, &(ii[0]));
Chris@42 335 ST(&(ii[WS(rs, 6)]), VFMA(LDK(KP707106781), T3K, T3J), ms, &(ii[0]));
Chris@42 336 ST(&(ri[WS(rs, 2)]), VFMA(LDK(KP707106781), T3l, T3a), ms, &(ri[0]));
Chris@42 337 ST(&(ri[WS(rs, 10)]), VFNMS(LDK(KP707106781), T3l, T3a), ms, &(ri[0]));
Chris@42 338 T2R = VADD(T26, T25);
Chris@42 339 T27 = VSUB(T25, T26);
Chris@42 340 T2e = VADD(T28, T2d);
Chris@42 341 T2S = VSUB(T28, T2d);
Chris@42 342 }
Chris@42 343 }
Chris@42 344 }
Chris@42 345 }
Chris@42 346 {
Chris@42 347 V T2Y, T2Z, T2n, T2u;
Chris@42 348 T2J = VSUB(T2D, T2I);
Chris@42 349 T2Y = VADD(T2D, T2I);
Chris@42 350 T2Z = VSUB(T2A, T2B);
Chris@42 351 T2C = VADD(T2A, T2B);
Chris@42 352 T2W = VSUB(T2l, T2m);
Chris@42 353 T2n = VADD(T2l, T2m);
Chris@42 354 T2u = VSUB(T2o, T2t);
Chris@42 355 T2V = VADD(T2o, T2t);
Chris@42 356 T3O = VADD(T27, T2e);
Chris@42 357 T2f = VSUB(T27, T2e);
Chris@42 358 T34 = VFMA(LDK(KP414213562), T2Y, T2Z);
Chris@42 359 T30 = VFNMS(LDK(KP414213562), T2Z, T2Y);
Chris@42 360 T3U = VSUB(T2S, T2R);
Chris@42 361 T2T = VADD(T2R, T2S);
Chris@42 362 T2N = VFNMS(LDK(KP414213562), T2n, T2u);
Chris@42 363 T2v = VFMA(LDK(KP414213562), T2u, T2n);
Chris@42 364 }
Chris@42 365 }
Chris@42 366 T2M = VFNMS(LDK(KP707106781), T2f, T20);
Chris@42 367 T2g = VFMA(LDK(KP707106781), T2f, T20);
Chris@42 368 T3V = VFMA(LDK(KP707106781), T3U, T3T);
Chris@42 369 T3X = VFNMS(LDK(KP707106781), T3U, T3T);
Chris@42 370 T32 = VFNMS(LDK(KP707106781), T2T, T2Q);
Chris@42 371 T2U = VFMA(LDK(KP707106781), T2T, T2Q);
Chris@42 372 T33 = VFNMS(LDK(KP414213562), T2V, T2W);
Chris@42 373 T2X = VFMA(LDK(KP414213562), T2W, T2V);
Chris@42 374 T2O = VFMA(LDK(KP414213562), T2C, T2J);
Chris@42 375 T2K = VFNMS(LDK(KP414213562), T2J, T2C);
Chris@42 376 T3P = VFMA(LDK(KP707106781), T3O, T3N);
Chris@42 377 T3R = VFNMS(LDK(KP707106781), T3O, T3N);
Chris@42 378 }
Chris@42 379 }
Chris@42 380 }
Chris@42 381 }
Chris@42 382 }
Chris@42 383 {
Chris@42 384 V T3Q, T35, T31, T3S;
Chris@42 385 T3Q = VADD(T33, T34);
Chris@42 386 T35 = VSUB(T33, T34);
Chris@42 387 T31 = VADD(T2X, T30);
Chris@42 388 T3S = VSUB(T30, T2X);
Chris@42 389 {
Chris@42 390 V T3W, T2P, T2L, T3Y;
Chris@42 391 T3W = VSUB(T2O, T2N);
Chris@42 392 T2P = VADD(T2N, T2O);
Chris@42 393 T2L = VSUB(T2v, T2K);
Chris@42 394 T3Y = VADD(T2v, T2K);
Chris@42 395 ST(&(ri[WS(rs, 5)]), VFMA(LDK(KP923879532), T35, T32), ms, &(ri[WS(rs, 1)]));
Chris@42 396 ST(&(ri[WS(rs, 13)]), VFNMS(LDK(KP923879532), T35, T32), ms, &(ri[WS(rs, 1)]));
Chris@42 397 ST(&(ii[WS(rs, 9)]), VFNMS(LDK(KP923879532), T3Q, T3P), ms, &(ii[WS(rs, 1)]));
Chris@42 398 ST(&(ii[WS(rs, 1)]), VFMA(LDK(KP923879532), T3Q, T3P), ms, &(ii[WS(rs, 1)]));
Chris@42 399 ST(&(ii[WS(rs, 13)]), VFNMS(LDK(KP923879532), T3S, T3R), ms, &(ii[WS(rs, 1)]));
Chris@42 400 ST(&(ii[WS(rs, 5)]), VFMA(LDK(KP923879532), T3S, T3R), ms, &(ii[WS(rs, 1)]));
Chris@42 401 ST(&(ri[WS(rs, 1)]), VFMA(LDK(KP923879532), T31, T2U), ms, &(ri[WS(rs, 1)]));
Chris@42 402 ST(&(ri[WS(rs, 9)]), VFNMS(LDK(KP923879532), T31, T2U), ms, &(ri[WS(rs, 1)]));
Chris@42 403 ST(&(ri[WS(rs, 15)]), VFMA(LDK(KP923879532), T2P, T2M), ms, &(ri[WS(rs, 1)]));
Chris@42 404 ST(&(ri[WS(rs, 7)]), VFNMS(LDK(KP923879532), T2P, T2M), ms, &(ri[WS(rs, 1)]));
Chris@42 405 ST(&(ii[WS(rs, 11)]), VFNMS(LDK(KP923879532), T3W, T3V), ms, &(ii[WS(rs, 1)]));
Chris@42 406 ST(&(ii[WS(rs, 3)]), VFMA(LDK(KP923879532), T3W, T3V), ms, &(ii[WS(rs, 1)]));
Chris@42 407 ST(&(ii[WS(rs, 15)]), VFMA(LDK(KP923879532), T3Y, T3X), ms, &(ii[WS(rs, 1)]));
Chris@42 408 ST(&(ii[WS(rs, 7)]), VFNMS(LDK(KP923879532), T3Y, T3X), ms, &(ii[WS(rs, 1)]));
Chris@42 409 ST(&(ri[WS(rs, 3)]), VFMA(LDK(KP923879532), T2L, T2g), ms, &(ri[WS(rs, 1)]));
Chris@42 410 ST(&(ri[WS(rs, 11)]), VFNMS(LDK(KP923879532), T2L, T2g), ms, &(ri[WS(rs, 1)]));
Chris@42 411 }
Chris@42 412 }
Chris@42 413 }
Chris@42 414 }
Chris@42 415 VLEAVE();
Chris@42 416 }
Chris@42 417
Chris@42 418 static const tw_instr twinstr[] = {
Chris@42 419 VTW(0, 1),
Chris@42 420 VTW(0, 3),
Chris@42 421 VTW(0, 9),
Chris@42 422 VTW(0, 15),
Chris@42 423 {TW_NEXT, (2 * VL), 0}
Chris@42 424 };
Chris@42 425
Chris@42 426 static const ct_desc desc = { 16, XSIMD_STRING("t2sv_16"), twinstr, &GENUS, {104, 42, 92, 0}, 0, 0, 0 };
Chris@42 427
Chris@42 428 void XSIMD(codelet_t2sv_16) (planner *p) {
Chris@42 429 X(kdft_dit_register) (p, t2sv_16, &desc);
Chris@42 430 }
Chris@42 431 #else /* HAVE_FMA */
Chris@42 432
Chris@42 433 /* Generated by: ../../../genfft/gen_twiddle.native -simd -compact -variables 4 -pipeline-latency 8 -twiddle-log3 -precompute-twiddles -n 16 -name t2sv_16 -include ts.h */
Chris@42 434
Chris@42 435 /*
Chris@42 436 * This function contains 196 FP additions, 108 FP multiplications,
Chris@42 437 * (or, 156 additions, 68 multiplications, 40 fused multiply/add),
Chris@42 438 * 82 stack variables, 3 constants, and 64 memory accesses
Chris@42 439 */
Chris@42 440 #include "ts.h"
Chris@42 441
Chris@42 442 static void t2sv_16(R *ri, R *ii, const R *W, stride rs, INT mb, INT me, INT ms)
Chris@42 443 {
Chris@42 444 DVK(KP382683432, +0.382683432365089771728459984030398866761344562);
Chris@42 445 DVK(KP923879532, +0.923879532511286756128183189396788286822416626);
Chris@42 446 DVK(KP707106781, +0.707106781186547524400844362104849039284835938);
Chris@42 447 {
Chris@42 448 INT m;
Chris@42 449 for (m = mb, W = W + (mb * 8); m < me; m = m + (2 * VL), ri = ri + ((2 * VL) * ms), ii = ii + ((2 * VL) * ms), W = W + ((2 * VL) * 8), MAKE_VOLATILE_STRIDE(32, rs)) {
Chris@42 450 V T2, T5, Tg, Ti, Tk, To, TE, TC, T6, T3, T8, TW, TJ, Tt, TU;
Chris@42 451 V Tc, Tx, TH, TN, TO, TP, TR, T1f, T1k, T1b, T1i, T1y, T1H, T1u, T1F;
Chris@42 452 {
Chris@42 453 V T7, Tv, Ta, Ts, T4, Tw, Tb, Tr;
Chris@42 454 {
Chris@42 455 V Th, Tn, Tj, Tm;
Chris@42 456 T2 = LDW(&(W[0]));
Chris@42 457 T5 = LDW(&(W[TWVL * 1]));
Chris@42 458 Tg = LDW(&(W[TWVL * 2]));
Chris@42 459 Ti = LDW(&(W[TWVL * 3]));
Chris@42 460 Th = VMUL(T2, Tg);
Chris@42 461 Tn = VMUL(T5, Tg);
Chris@42 462 Tj = VMUL(T5, Ti);
Chris@42 463 Tm = VMUL(T2, Ti);
Chris@42 464 Tk = VSUB(Th, Tj);
Chris@42 465 To = VADD(Tm, Tn);
Chris@42 466 TE = VSUB(Tm, Tn);
Chris@42 467 TC = VADD(Th, Tj);
Chris@42 468 T6 = LDW(&(W[TWVL * 5]));
Chris@42 469 T7 = VMUL(T5, T6);
Chris@42 470 Tv = VMUL(Tg, T6);
Chris@42 471 Ta = VMUL(T2, T6);
Chris@42 472 Ts = VMUL(Ti, T6);
Chris@42 473 T3 = LDW(&(W[TWVL * 4]));
Chris@42 474 T4 = VMUL(T2, T3);
Chris@42 475 Tw = VMUL(Ti, T3);
Chris@42 476 Tb = VMUL(T5, T3);
Chris@42 477 Tr = VMUL(Tg, T3);
Chris@42 478 }
Chris@42 479 T8 = VADD(T4, T7);
Chris@42 480 TW = VSUB(Tv, Tw);
Chris@42 481 TJ = VADD(Ta, Tb);
Chris@42 482 Tt = VSUB(Tr, Ts);
Chris@42 483 TU = VADD(Tr, Ts);
Chris@42 484 Tc = VSUB(Ta, Tb);
Chris@42 485 Tx = VADD(Tv, Tw);
Chris@42 486 TH = VSUB(T4, T7);
Chris@42 487 TN = LDW(&(W[TWVL * 6]));
Chris@42 488 TO = LDW(&(W[TWVL * 7]));
Chris@42 489 TP = VFMA(T2, TN, VMUL(T5, TO));
Chris@42 490 TR = VFNMS(T5, TN, VMUL(T2, TO));
Chris@42 491 {
Chris@42 492 V T1d, T1e, T19, T1a;
Chris@42 493 T1d = VMUL(Tk, T6);
Chris@42 494 T1e = VMUL(To, T3);
Chris@42 495 T1f = VSUB(T1d, T1e);
Chris@42 496 T1k = VADD(T1d, T1e);
Chris@42 497 T19 = VMUL(Tk, T3);
Chris@42 498 T1a = VMUL(To, T6);
Chris@42 499 T1b = VADD(T19, T1a);
Chris@42 500 T1i = VSUB(T19, T1a);
Chris@42 501 }
Chris@42 502 {
Chris@42 503 V T1w, T1x, T1s, T1t;
Chris@42 504 T1w = VMUL(TC, T6);
Chris@42 505 T1x = VMUL(TE, T3);
Chris@42 506 T1y = VSUB(T1w, T1x);
Chris@42 507 T1H = VADD(T1w, T1x);
Chris@42 508 T1s = VMUL(TC, T3);
Chris@42 509 T1t = VMUL(TE, T6);
Chris@42 510 T1u = VADD(T1s, T1t);
Chris@42 511 T1F = VSUB(T1s, T1t);
Chris@42 512 }
Chris@42 513 }
Chris@42 514 {
Chris@42 515 V Tf, T3r, T1N, T3e, TA, T3s, T1Q, T3b, TM, T2M, T1W, T2w, TZ, T2N, T21;
Chris@42 516 V T2x, T1B, T1K, T2V, T2W, T2X, T2Y, T2j, T2D, T2o, T2E, T18, T1n, T2Q, T2R;
Chris@42 517 V T2S, T2T, T28, T2A, T2d, T2B;
Chris@42 518 {
Chris@42 519 V T1, T3d, Te, T3c, T9, Td;
Chris@42 520 T1 = LD(&(ri[0]), ms, &(ri[0]));
Chris@42 521 T3d = LD(&(ii[0]), ms, &(ii[0]));
Chris@42 522 T9 = LD(&(ri[WS(rs, 8)]), ms, &(ri[0]));
Chris@42 523 Td = LD(&(ii[WS(rs, 8)]), ms, &(ii[0]));
Chris@42 524 Te = VFMA(T8, T9, VMUL(Tc, Td));
Chris@42 525 T3c = VFNMS(Tc, T9, VMUL(T8, Td));
Chris@42 526 Tf = VADD(T1, Te);
Chris@42 527 T3r = VSUB(T3d, T3c);
Chris@42 528 T1N = VSUB(T1, Te);
Chris@42 529 T3e = VADD(T3c, T3d);
Chris@42 530 }
Chris@42 531 {
Chris@42 532 V Tq, T1O, Tz, T1P;
Chris@42 533 {
Chris@42 534 V Tl, Tp, Tu, Ty;
Chris@42 535 Tl = LD(&(ri[WS(rs, 4)]), ms, &(ri[0]));
Chris@42 536 Tp = LD(&(ii[WS(rs, 4)]), ms, &(ii[0]));
Chris@42 537 Tq = VFMA(Tk, Tl, VMUL(To, Tp));
Chris@42 538 T1O = VFNMS(To, Tl, VMUL(Tk, Tp));
Chris@42 539 Tu = LD(&(ri[WS(rs, 12)]), ms, &(ri[0]));
Chris@42 540 Ty = LD(&(ii[WS(rs, 12)]), ms, &(ii[0]));
Chris@42 541 Tz = VFMA(Tt, Tu, VMUL(Tx, Ty));
Chris@42 542 T1P = VFNMS(Tx, Tu, VMUL(Tt, Ty));
Chris@42 543 }
Chris@42 544 TA = VADD(Tq, Tz);
Chris@42 545 T3s = VSUB(Tq, Tz);
Chris@42 546 T1Q = VSUB(T1O, T1P);
Chris@42 547 T3b = VADD(T1O, T1P);
Chris@42 548 }
Chris@42 549 {
Chris@42 550 V TG, T1S, TL, T1T, T1U, T1V;
Chris@42 551 {
Chris@42 552 V TD, TF, TI, TK;
Chris@42 553 TD = LD(&(ri[WS(rs, 2)]), ms, &(ri[0]));
Chris@42 554 TF = LD(&(ii[WS(rs, 2)]), ms, &(ii[0]));
Chris@42 555 TG = VFMA(TC, TD, VMUL(TE, TF));
Chris@42 556 T1S = VFNMS(TE, TD, VMUL(TC, TF));
Chris@42 557 TI = LD(&(ri[WS(rs, 10)]), ms, &(ri[0]));
Chris@42 558 TK = LD(&(ii[WS(rs, 10)]), ms, &(ii[0]));
Chris@42 559 TL = VFMA(TH, TI, VMUL(TJ, TK));
Chris@42 560 T1T = VFNMS(TJ, TI, VMUL(TH, TK));
Chris@42 561 }
Chris@42 562 TM = VADD(TG, TL);
Chris@42 563 T2M = VADD(T1S, T1T);
Chris@42 564 T1U = VSUB(T1S, T1T);
Chris@42 565 T1V = VSUB(TG, TL);
Chris@42 566 T1W = VSUB(T1U, T1V);
Chris@42 567 T2w = VADD(T1V, T1U);
Chris@42 568 }
Chris@42 569 {
Chris@42 570 V TT, T1Y, TY, T1Z, T1X, T20;
Chris@42 571 {
Chris@42 572 V TQ, TS, TV, TX;
Chris@42 573 TQ = LD(&(ri[WS(rs, 14)]), ms, &(ri[0]));
Chris@42 574 TS = LD(&(ii[WS(rs, 14)]), ms, &(ii[0]));
Chris@42 575 TT = VFMA(TP, TQ, VMUL(TR, TS));
Chris@42 576 T1Y = VFNMS(TR, TQ, VMUL(TP, TS));
Chris@42 577 TV = LD(&(ri[WS(rs, 6)]), ms, &(ri[0]));
Chris@42 578 TX = LD(&(ii[WS(rs, 6)]), ms, &(ii[0]));
Chris@42 579 TY = VFMA(TU, TV, VMUL(TW, TX));
Chris@42 580 T1Z = VFNMS(TW, TV, VMUL(TU, TX));
Chris@42 581 }
Chris@42 582 TZ = VADD(TT, TY);
Chris@42 583 T2N = VADD(T1Y, T1Z);
Chris@42 584 T1X = VSUB(TT, TY);
Chris@42 585 T20 = VSUB(T1Y, T1Z);
Chris@42 586 T21 = VADD(T1X, T20);
Chris@42 587 T2x = VSUB(T1X, T20);
Chris@42 588 }
Chris@42 589 {
Chris@42 590 V T1r, T2k, T1J, T2h, T1A, T2l, T1E, T2g;
Chris@42 591 {
Chris@42 592 V T1p, T1q, T1G, T1I;
Chris@42 593 T1p = LD(&(ri[WS(rs, 15)]), ms, &(ri[WS(rs, 1)]));
Chris@42 594 T1q = LD(&(ii[WS(rs, 15)]), ms, &(ii[WS(rs, 1)]));
Chris@42 595 T1r = VFMA(TN, T1p, VMUL(TO, T1q));
Chris@42 596 T2k = VFNMS(TO, T1p, VMUL(TN, T1q));
Chris@42 597 T1G = LD(&(ri[WS(rs, 11)]), ms, &(ri[WS(rs, 1)]));
Chris@42 598 T1I = LD(&(ii[WS(rs, 11)]), ms, &(ii[WS(rs, 1)]));
Chris@42 599 T1J = VFMA(T1F, T1G, VMUL(T1H, T1I));
Chris@42 600 T2h = VFNMS(T1H, T1G, VMUL(T1F, T1I));
Chris@42 601 }
Chris@42 602 {
Chris@42 603 V T1v, T1z, T1C, T1D;
Chris@42 604 T1v = LD(&(ri[WS(rs, 7)]), ms, &(ri[WS(rs, 1)]));
Chris@42 605 T1z = LD(&(ii[WS(rs, 7)]), ms, &(ii[WS(rs, 1)]));
Chris@42 606 T1A = VFMA(T1u, T1v, VMUL(T1y, T1z));
Chris@42 607 T2l = VFNMS(T1y, T1v, VMUL(T1u, T1z));
Chris@42 608 T1C = LD(&(ri[WS(rs, 3)]), ms, &(ri[WS(rs, 1)]));
Chris@42 609 T1D = LD(&(ii[WS(rs, 3)]), ms, &(ii[WS(rs, 1)]));
Chris@42 610 T1E = VFMA(Tg, T1C, VMUL(Ti, T1D));
Chris@42 611 T2g = VFNMS(Ti, T1C, VMUL(Tg, T1D));
Chris@42 612 }
Chris@42 613 T1B = VADD(T1r, T1A);
Chris@42 614 T1K = VADD(T1E, T1J);
Chris@42 615 T2V = VSUB(T1B, T1K);
Chris@42 616 T2W = VADD(T2k, T2l);
Chris@42 617 T2X = VADD(T2g, T2h);
Chris@42 618 T2Y = VSUB(T2W, T2X);
Chris@42 619 {
Chris@42 620 V T2f, T2i, T2m, T2n;
Chris@42 621 T2f = VSUB(T1r, T1A);
Chris@42 622 T2i = VSUB(T2g, T2h);
Chris@42 623 T2j = VSUB(T2f, T2i);
Chris@42 624 T2D = VADD(T2f, T2i);
Chris@42 625 T2m = VSUB(T2k, T2l);
Chris@42 626 T2n = VSUB(T1E, T1J);
Chris@42 627 T2o = VADD(T2m, T2n);
Chris@42 628 T2E = VSUB(T2m, T2n);
Chris@42 629 }
Chris@42 630 }
Chris@42 631 {
Chris@42 632 V T14, T24, T1m, T2b, T17, T25, T1h, T2a;
Chris@42 633 {
Chris@42 634 V T12, T13, T1j, T1l;
Chris@42 635 T12 = LD(&(ri[WS(rs, 1)]), ms, &(ri[WS(rs, 1)]));
Chris@42 636 T13 = LD(&(ii[WS(rs, 1)]), ms, &(ii[WS(rs, 1)]));
Chris@42 637 T14 = VFMA(T2, T12, VMUL(T5, T13));
Chris@42 638 T24 = VFNMS(T5, T12, VMUL(T2, T13));
Chris@42 639 T1j = LD(&(ri[WS(rs, 13)]), ms, &(ri[WS(rs, 1)]));
Chris@42 640 T1l = LD(&(ii[WS(rs, 13)]), ms, &(ii[WS(rs, 1)]));
Chris@42 641 T1m = VFMA(T1i, T1j, VMUL(T1k, T1l));
Chris@42 642 T2b = VFNMS(T1k, T1j, VMUL(T1i, T1l));
Chris@42 643 }
Chris@42 644 {
Chris@42 645 V T15, T16, T1c, T1g;
Chris@42 646 T15 = LD(&(ri[WS(rs, 9)]), ms, &(ri[WS(rs, 1)]));
Chris@42 647 T16 = LD(&(ii[WS(rs, 9)]), ms, &(ii[WS(rs, 1)]));
Chris@42 648 T17 = VFMA(T3, T15, VMUL(T6, T16));
Chris@42 649 T25 = VFNMS(T6, T15, VMUL(T3, T16));
Chris@42 650 T1c = LD(&(ri[WS(rs, 5)]), ms, &(ri[WS(rs, 1)]));
Chris@42 651 T1g = LD(&(ii[WS(rs, 5)]), ms, &(ii[WS(rs, 1)]));
Chris@42 652 T1h = VFMA(T1b, T1c, VMUL(T1f, T1g));
Chris@42 653 T2a = VFNMS(T1f, T1c, VMUL(T1b, T1g));
Chris@42 654 }
Chris@42 655 T18 = VADD(T14, T17);
Chris@42 656 T1n = VADD(T1h, T1m);
Chris@42 657 T2Q = VSUB(T18, T1n);
Chris@42 658 T2R = VADD(T24, T25);
Chris@42 659 T2S = VADD(T2a, T2b);
Chris@42 660 T2T = VSUB(T2R, T2S);
Chris@42 661 {
Chris@42 662 V T26, T27, T29, T2c;
Chris@42 663 T26 = VSUB(T24, T25);
Chris@42 664 T27 = VSUB(T1h, T1m);
Chris@42 665 T28 = VADD(T26, T27);
Chris@42 666 T2A = VSUB(T26, T27);
Chris@42 667 T29 = VSUB(T14, T17);
Chris@42 668 T2c = VSUB(T2a, T2b);
Chris@42 669 T2d = VSUB(T29, T2c);
Chris@42 670 T2B = VADD(T29, T2c);
Chris@42 671 }
Chris@42 672 }
Chris@42 673 {
Chris@42 674 V T23, T2r, T3A, T3C, T2q, T3B, T2u, T3x;
Chris@42 675 {
Chris@42 676 V T1R, T22, T3y, T3z;
Chris@42 677 T1R = VSUB(T1N, T1Q);
Chris@42 678 T22 = VMUL(LDK(KP707106781), VSUB(T1W, T21));
Chris@42 679 T23 = VADD(T1R, T22);
Chris@42 680 T2r = VSUB(T1R, T22);
Chris@42 681 T3y = VMUL(LDK(KP707106781), VSUB(T2x, T2w));
Chris@42 682 T3z = VADD(T3s, T3r);
Chris@42 683 T3A = VADD(T3y, T3z);
Chris@42 684 T3C = VSUB(T3z, T3y);
Chris@42 685 }
Chris@42 686 {
Chris@42 687 V T2e, T2p, T2s, T2t;
Chris@42 688 T2e = VFMA(LDK(KP923879532), T28, VMUL(LDK(KP382683432), T2d));
Chris@42 689 T2p = VFNMS(LDK(KP923879532), T2o, VMUL(LDK(KP382683432), T2j));
Chris@42 690 T2q = VADD(T2e, T2p);
Chris@42 691 T3B = VSUB(T2p, T2e);
Chris@42 692 T2s = VFNMS(LDK(KP923879532), T2d, VMUL(LDK(KP382683432), T28));
Chris@42 693 T2t = VFMA(LDK(KP382683432), T2o, VMUL(LDK(KP923879532), T2j));
Chris@42 694 T2u = VSUB(T2s, T2t);
Chris@42 695 T3x = VADD(T2s, T2t);
Chris@42 696 }
Chris@42 697 ST(&(ri[WS(rs, 11)]), VSUB(T23, T2q), ms, &(ri[WS(rs, 1)]));
Chris@42 698 ST(&(ii[WS(rs, 11)]), VSUB(T3A, T3x), ms, &(ii[WS(rs, 1)]));
Chris@42 699 ST(&(ri[WS(rs, 3)]), VADD(T23, T2q), ms, &(ri[WS(rs, 1)]));
Chris@42 700 ST(&(ii[WS(rs, 3)]), VADD(T3x, T3A), ms, &(ii[WS(rs, 1)]));
Chris@42 701 ST(&(ri[WS(rs, 15)]), VSUB(T2r, T2u), ms, &(ri[WS(rs, 1)]));
Chris@42 702 ST(&(ii[WS(rs, 15)]), VSUB(T3C, T3B), ms, &(ii[WS(rs, 1)]));
Chris@42 703 ST(&(ri[WS(rs, 7)]), VADD(T2r, T2u), ms, &(ri[WS(rs, 1)]));
Chris@42 704 ST(&(ii[WS(rs, 7)]), VADD(T3B, T3C), ms, &(ii[WS(rs, 1)]));
Chris@42 705 }
Chris@42 706 {
Chris@42 707 V T2P, T31, T3m, T3o, T30, T3n, T34, T3j;
Chris@42 708 {
Chris@42 709 V T2L, T2O, T3k, T3l;
Chris@42 710 T2L = VSUB(Tf, TA);
Chris@42 711 T2O = VSUB(T2M, T2N);
Chris@42 712 T2P = VADD(T2L, T2O);
Chris@42 713 T31 = VSUB(T2L, T2O);
Chris@42 714 T3k = VSUB(TZ, TM);
Chris@42 715 T3l = VSUB(T3e, T3b);
Chris@42 716 T3m = VADD(T3k, T3l);
Chris@42 717 T3o = VSUB(T3l, T3k);
Chris@42 718 }
Chris@42 719 {
Chris@42 720 V T2U, T2Z, T32, T33;
Chris@42 721 T2U = VADD(T2Q, T2T);
Chris@42 722 T2Z = VSUB(T2V, T2Y);
Chris@42 723 T30 = VMUL(LDK(KP707106781), VADD(T2U, T2Z));
Chris@42 724 T3n = VMUL(LDK(KP707106781), VSUB(T2Z, T2U));
Chris@42 725 T32 = VSUB(T2T, T2Q);
Chris@42 726 T33 = VADD(T2V, T2Y);
Chris@42 727 T34 = VMUL(LDK(KP707106781), VSUB(T32, T33));
Chris@42 728 T3j = VMUL(LDK(KP707106781), VADD(T32, T33));
Chris@42 729 }
Chris@42 730 ST(&(ri[WS(rs, 10)]), VSUB(T2P, T30), ms, &(ri[0]));
Chris@42 731 ST(&(ii[WS(rs, 10)]), VSUB(T3m, T3j), ms, &(ii[0]));
Chris@42 732 ST(&(ri[WS(rs, 2)]), VADD(T2P, T30), ms, &(ri[0]));
Chris@42 733 ST(&(ii[WS(rs, 2)]), VADD(T3j, T3m), ms, &(ii[0]));
Chris@42 734 ST(&(ri[WS(rs, 14)]), VSUB(T31, T34), ms, &(ri[0]));
Chris@42 735 ST(&(ii[WS(rs, 14)]), VSUB(T3o, T3n), ms, &(ii[0]));
Chris@42 736 ST(&(ri[WS(rs, 6)]), VADD(T31, T34), ms, &(ri[0]));
Chris@42 737 ST(&(ii[WS(rs, 6)]), VADD(T3n, T3o), ms, &(ii[0]));
Chris@42 738 }
Chris@42 739 {
Chris@42 740 V T2z, T2H, T3u, T3w, T2G, T3v, T2K, T3p;
Chris@42 741 {
Chris@42 742 V T2v, T2y, T3q, T3t;
Chris@42 743 T2v = VADD(T1N, T1Q);
Chris@42 744 T2y = VMUL(LDK(KP707106781), VADD(T2w, T2x));
Chris@42 745 T2z = VADD(T2v, T2y);
Chris@42 746 T2H = VSUB(T2v, T2y);
Chris@42 747 T3q = VMUL(LDK(KP707106781), VADD(T1W, T21));
Chris@42 748 T3t = VSUB(T3r, T3s);
Chris@42 749 T3u = VADD(T3q, T3t);
Chris@42 750 T3w = VSUB(T3t, T3q);
Chris@42 751 }
Chris@42 752 {
Chris@42 753 V T2C, T2F, T2I, T2J;
Chris@42 754 T2C = VFMA(LDK(KP382683432), T2A, VMUL(LDK(KP923879532), T2B));
Chris@42 755 T2F = VFNMS(LDK(KP382683432), T2E, VMUL(LDK(KP923879532), T2D));
Chris@42 756 T2G = VADD(T2C, T2F);
Chris@42 757 T3v = VSUB(T2F, T2C);
Chris@42 758 T2I = VFNMS(LDK(KP382683432), T2B, VMUL(LDK(KP923879532), T2A));
Chris@42 759 T2J = VFMA(LDK(KP923879532), T2E, VMUL(LDK(KP382683432), T2D));
Chris@42 760 T2K = VSUB(T2I, T2J);
Chris@42 761 T3p = VADD(T2I, T2J);
Chris@42 762 }
Chris@42 763 ST(&(ri[WS(rs, 9)]), VSUB(T2z, T2G), ms, &(ri[WS(rs, 1)]));
Chris@42 764 ST(&(ii[WS(rs, 9)]), VSUB(T3u, T3p), ms, &(ii[WS(rs, 1)]));
Chris@42 765 ST(&(ri[WS(rs, 1)]), VADD(T2z, T2G), ms, &(ri[WS(rs, 1)]));
Chris@42 766 ST(&(ii[WS(rs, 1)]), VADD(T3p, T3u), ms, &(ii[WS(rs, 1)]));
Chris@42 767 ST(&(ri[WS(rs, 13)]), VSUB(T2H, T2K), ms, &(ri[WS(rs, 1)]));
Chris@42 768 ST(&(ii[WS(rs, 13)]), VSUB(T3w, T3v), ms, &(ii[WS(rs, 1)]));
Chris@42 769 ST(&(ri[WS(rs, 5)]), VADD(T2H, T2K), ms, &(ri[WS(rs, 1)]));
Chris@42 770 ST(&(ii[WS(rs, 5)]), VADD(T3v, T3w), ms, &(ii[WS(rs, 1)]));
Chris@42 771 }
Chris@42 772 {
Chris@42 773 V T11, T35, T3g, T3i, T1M, T3h, T38, T39;
Chris@42 774 {
Chris@42 775 V TB, T10, T3a, T3f;
Chris@42 776 TB = VADD(Tf, TA);
Chris@42 777 T10 = VADD(TM, TZ);
Chris@42 778 T11 = VADD(TB, T10);
Chris@42 779 T35 = VSUB(TB, T10);
Chris@42 780 T3a = VADD(T2M, T2N);
Chris@42 781 T3f = VADD(T3b, T3e);
Chris@42 782 T3g = VADD(T3a, T3f);
Chris@42 783 T3i = VSUB(T3f, T3a);
Chris@42 784 }
Chris@42 785 {
Chris@42 786 V T1o, T1L, T36, T37;
Chris@42 787 T1o = VADD(T18, T1n);
Chris@42 788 T1L = VADD(T1B, T1K);
Chris@42 789 T1M = VADD(T1o, T1L);
Chris@42 790 T3h = VSUB(T1L, T1o);
Chris@42 791 T36 = VADD(T2R, T2S);
Chris@42 792 T37 = VADD(T2W, T2X);
Chris@42 793 T38 = VSUB(T36, T37);
Chris@42 794 T39 = VADD(T36, T37);
Chris@42 795 }
Chris@42 796 ST(&(ri[WS(rs, 8)]), VSUB(T11, T1M), ms, &(ri[0]));
Chris@42 797 ST(&(ii[WS(rs, 8)]), VSUB(T3g, T39), ms, &(ii[0]));
Chris@42 798 ST(&(ri[0]), VADD(T11, T1M), ms, &(ri[0]));
Chris@42 799 ST(&(ii[0]), VADD(T39, T3g), ms, &(ii[0]));
Chris@42 800 ST(&(ri[WS(rs, 12)]), VSUB(T35, T38), ms, &(ri[0]));
Chris@42 801 ST(&(ii[WS(rs, 12)]), VSUB(T3i, T3h), ms, &(ii[0]));
Chris@42 802 ST(&(ri[WS(rs, 4)]), VADD(T35, T38), ms, &(ri[0]));
Chris@42 803 ST(&(ii[WS(rs, 4)]), VADD(T3h, T3i), ms, &(ii[0]));
Chris@42 804 }
Chris@42 805 }
Chris@42 806 }
Chris@42 807 }
Chris@42 808 VLEAVE();
Chris@42 809 }
Chris@42 810
Chris@42 811 static const tw_instr twinstr[] = {
Chris@42 812 VTW(0, 1),
Chris@42 813 VTW(0, 3),
Chris@42 814 VTW(0, 9),
Chris@42 815 VTW(0, 15),
Chris@42 816 {TW_NEXT, (2 * VL), 0}
Chris@42 817 };
Chris@42 818
Chris@42 819 static const ct_desc desc = { 16, XSIMD_STRING("t2sv_16"), twinstr, &GENUS, {156, 68, 40, 0}, 0, 0, 0 };
Chris@42 820
Chris@42 821 void XSIMD(codelet_t2sv_16) (planner *p) {
Chris@42 822 X(kdft_dit_register) (p, t2sv_16, &desc);
Chris@42 823 }
Chris@42 824 #endif /* HAVE_FMA */