Mercurial > hg > sv-dependency-builds

annotate src/fftw-3.3.5/dft/simd/common/t1sv_16.c @ 42:2cd0e3b3e1fd

Find changesets by keywords (author, files, the commit message), revision number or hash, or revset expression.
Current fftw source
author Chris Cannam
date Tue, 18 Oct 2016 13:40:26 +0100
parents
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 16:45:06 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 -n 16 -name t1sv_16 -include ts.h */
Chris@42 29
Chris@42 30 /*
Chris@42 31 * This function contains 174 FP additions, 100 FP multiplications,
Chris@42 32 * (or, 104 additions, 30 multiplications, 70 fused multiply/add),
Chris@42 33 * 113 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 t1sv_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 * 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@42 45 V T2S, T2O, T2B, T2j, T2A, T24, T3J, T3L, T2Q, T2I, T2R, T2L, T2C, T2y, T3D;
Chris@42 46 V T3F;
Chris@42 47 {
Chris@42 48 V T3o, T3z, T1I, T8, T35, T2o, T1s, T2r, T36, T2w, T1F, T2p, T1N, T3k, Tl;
Chris@42 49 V T3A, T2V, T1T, Tz, T1U, T30, T29, T11, T2c, TH, TK, TJ, T31, T2h, T1e;
Chris@42 50 V T2a, T1Z, TI, T1Y, TF;
Chris@42 51 {
Chris@42 52 V Ta, Td, Tg, Tj, T2t, T1y, Tf, T1J, Tb, Tc, T2v, T1E, Ti;
Chris@42 53 {
Chris@42 54 V T1, T3n, T3, T6, T5, T1h, T1k, T1n, T1q, T1m, T3l, T4, T1j, T1p, T2k;
Chris@42 55 V T1i, T2, T1g;
Chris@42 56 T1 = LD(&(ri[0]), ms, &(ri[0]));
Chris@42 57 T3n = LD(&(ii[0]), ms, &(ii[0]));
Chris@42 58 T3 = LD(&(ri[WS(rs, 8)]), ms, &(ri[0]));
Chris@42 59 T6 = LD(&(ii[WS(rs, 8)]), ms, &(ii[0]));
Chris@42 60 T2 = LDW(&(W[TWVL * 14]));
Chris@42 61 T5 = LDW(&(W[TWVL * 15]));
Chris@42 62 T1h = LD(&(ri[WS(rs, 15)]), ms, &(ri[WS(rs, 1)]));
Chris@42 63 T1k = LD(&(ii[WS(rs, 15)]), ms, &(ii[WS(rs, 1)]));
Chris@42 64 T1g = LDW(&(W[TWVL * 28]));
Chris@42 65 T1n = LD(&(ri[WS(rs, 7)]), ms, &(ri[WS(rs, 1)]));
Chris@42 66 T1q = LD(&(ii[WS(rs, 7)]), ms, &(ii[WS(rs, 1)]));
Chris@42 67 T1m = LDW(&(W[TWVL * 12]));
Chris@42 68 T3l = VMUL(T2, T6);
Chris@42 69 T4 = VMUL(T2, T3);
Chris@42 70 T1j = LDW(&(W[TWVL * 29]));
Chris@42 71 T1p = LDW(&(W[TWVL * 13]));
Chris@42 72 T2k = VMUL(T1g, T1k);
Chris@42 73 T1i = VMUL(T1g, T1h);
Chris@42 74 {
Chris@42 75 V T1u, T1x, T1A, T2s, T1v, T1D, T1z, T1w, T1C, T2u, T1B, T9;
Chris@42 76 {
Chris@42 77 V T2l, T1l, T1t, T2n, T1r;
Chris@42 78 {
Chris@42 79 V T2m, T1o, T3m, T7;
Chris@42 80 T1u = LD(&(ri[WS(rs, 3)]), ms, &(ri[WS(rs, 1)]));
Chris@42 81 T2m = VMUL(T1m, T1q);
Chris@42 82 T1o = VMUL(T1m, T1n);
Chris@42 83 T3m = VFNMS(T5, T3, T3l);
Chris@42 84 T7 = VFMA(T5, T6, T4);
Chris@42 85 T1x = LD(&(ii[WS(rs, 3)]), ms, &(ii[WS(rs, 1)]));
Chris@42 86 T2l = VFNMS(T1j, T1h, T2k);
Chris@42 87 T1l = VFMA(T1j, T1k, T1i);
Chris@42 88 T1t = LDW(&(W[TWVL * 4]));
Chris@42 89 T2n = VFNMS(T1p, T1n, T2m);
Chris@42 90 T1r = VFMA(T1p, T1q, T1o);
Chris@42 91 T3o = VADD(T3m, T3n);
Chris@42 92 T3z = VSUB(T3n, T3m);
Chris@42 93 T1I = VSUB(T1, T7);
Chris@42 94 T8 = VADD(T1, T7);
Chris@42 95 }
Chris@42 96 T1A = LD(&(ri[WS(rs, 11)]), ms, &(ri[WS(rs, 1)]));
Chris@42 97 T2s = VMUL(T1t, T1x);
Chris@42 98 T1v = VMUL(T1t, T1u);
Chris@42 99 T35 = VADD(T2l, T2n);
Chris@42 100 T2o = VSUB(T2l, T2n);
Chris@42 101 T1s = VADD(T1l, T1r);
Chris@42 102 T2r = VSUB(T1l, T1r);
Chris@42 103 T1D = LD(&(ii[WS(rs, 11)]), ms, &(ii[WS(rs, 1)]));
Chris@42 104 T1z = LDW(&(W[TWVL * 20]));
Chris@42 105 }
Chris@42 106 T1w = LDW(&(W[TWVL * 5]));
Chris@42 107 T1C = LDW(&(W[TWVL * 21]));
Chris@42 108 Ta = LD(&(ri[WS(rs, 4)]), ms, &(ri[0]));
Chris@42 109 Td = LD(&(ii[WS(rs, 4)]), ms, &(ii[0]));
Chris@42 110 T9 = LDW(&(W[TWVL * 6]));
Chris@42 111 Tg = LD(&(ri[WS(rs, 12)]), ms, &(ri[0]));
Chris@42 112 Tj = LD(&(ii[WS(rs, 12)]), ms, &(ii[0]));
Chris@42 113 T2u = VMUL(T1z, T1D);
Chris@42 114 T1B = VMUL(T1z, T1A);
Chris@42 115 T2t = VFNMS(T1w, T1u, T2s);
Chris@42 116 T1y = VFMA(T1w, T1x, T1v);
Chris@42 117 Tf = LDW(&(W[TWVL * 22]));
Chris@42 118 T1J = VMUL(T9, Td);
Chris@42 119 Tb = VMUL(T9, Ta);
Chris@42 120 Tc = LDW(&(W[TWVL * 7]));
Chris@42 121 T2v = VFNMS(T1C, T1A, T2u);
Chris@42 122 T1E = VFMA(T1C, T1D, T1B);
Chris@42 123 Ti = LDW(&(W[TWVL * 23]));
Chris@42 124 }
Chris@42 125 }
Chris@42 126 {
Chris@42 127 V TW, TZ, TY, T27, TX, T26, TU;
Chris@42 128 {
Chris@42 129 V To, Tr, Tu, Tx, Tq, Tw, T1P, Tp, T1R, Tv;
Chris@42 130 {
Chris@42 131 V T1K, Te, T1M, Tk, Tn, Tt, T1L, Th;
Chris@42 132 To = LD(&(ri[WS(rs, 2)]), ms, &(ri[0]));
Chris@42 133 T1L = VMUL(Tf, Tj);
Chris@42 134 Th = VMUL(Tf, Tg);
Chris@42 135 Tr = LD(&(ii[WS(rs, 2)]), ms, &(ii[0]));
Chris@42 136 T1K = VFNMS(Tc, Ta, T1J);
Chris@42 137 Te = VFMA(Tc, Td, Tb);
Chris@42 138 T36 = VADD(T2t, T2v);
Chris@42 139 T2w = VSUB(T2t, T2v);
Chris@42 140 T1F = VADD(T1y, T1E);
Chris@42 141 T2p = VSUB(T1y, T1E);
Chris@42 142 T1M = VFNMS(Ti, Tg, T1L);
Chris@42 143 Tk = VFMA(Ti, Tj, Th);
Chris@42 144 Tn = LDW(&(W[TWVL * 2]));
Chris@42 145 Tu = LD(&(ri[WS(rs, 10)]), ms, &(ri[0]));
Chris@42 146 Tx = LD(&(ii[WS(rs, 10)]), ms, &(ii[0]));
Chris@42 147 Tt = LDW(&(W[TWVL * 18]));
Chris@42 148 Tq = LDW(&(W[TWVL * 3]));
Chris@42 149 Tw = LDW(&(W[TWVL * 19]));
Chris@42 150 T1N = VSUB(T1K, T1M);
Chris@42 151 T3k = VADD(T1K, T1M);
Chris@42 152 Tl = VADD(Te, Tk);
Chris@42 153 T3A = VSUB(Te, Tk);
Chris@42 154 T1P = VMUL(Tn, Tr);
Chris@42 155 Tp = VMUL(Tn, To);
Chris@42 156 T1R = VMUL(Tt, Tx);
Chris@42 157 Tv = VMUL(Tt, Tu);
Chris@42 158 }
Chris@42 159 {
Chris@42 160 V TQ, TT, T1Q, Ts, T1S, Ty, TV, T25, TR, TP, TS;
Chris@42 161 TQ = LD(&(ri[WS(rs, 1)]), ms, &(ri[WS(rs, 1)]));
Chris@42 162 TT = LD(&(ii[WS(rs, 1)]), ms, &(ii[WS(rs, 1)]));
Chris@42 163 TP = LDW(&(W[0]));
Chris@42 164 TW = LD(&(ri[WS(rs, 9)]), ms, &(ri[WS(rs, 1)]));
Chris@42 165 T1Q = VFNMS(Tq, To, T1P);
Chris@42 166 Ts = VFMA(Tq, Tr, Tp);
Chris@42 167 T1S = VFNMS(Tw, Tu, T1R);
Chris@42 168 Ty = VFMA(Tw, Tx, Tv);
Chris@42 169 TZ = LD(&(ii[WS(rs, 9)]), ms, &(ii[WS(rs, 1)]));
Chris@42 170 TV = LDW(&(W[TWVL * 16]));
Chris@42 171 T25 = VMUL(TP, TT);
Chris@42 172 TR = VMUL(TP, TQ);
Chris@42 173 TS = LDW(&(W[TWVL * 1]));
Chris@42 174 TY = LDW(&(W[TWVL * 17]));
Chris@42 175 T2V = VADD(T1Q, T1S);
Chris@42 176 T1T = VSUB(T1Q, T1S);
Chris@42 177 Tz = VADD(Ts, Ty);
Chris@42 178 T1U = VSUB(Ts, Ty);
Chris@42 179 T27 = VMUL(TV, TZ);
Chris@42 180 TX = VMUL(TV, TW);
Chris@42 181 T26 = VFNMS(TS, TQ, T25);
Chris@42 182 TU = VFMA(TS, TT, TR);
Chris@42 183 }
Chris@42 184 }
Chris@42 185 {
Chris@42 186 V T19, T1c, T1b, T2f, T1a, T2e, T17;
Chris@42 187 {
Chris@42 188 V T13, T16, T12, T28, T10, T18, T15, T2d, T14;
Chris@42 189 T13 = LD(&(ri[WS(rs, 5)]), ms, &(ri[WS(rs, 1)]));
Chris@42 190 T16 = LD(&(ii[WS(rs, 5)]), ms, &(ii[WS(rs, 1)]));
Chris@42 191 T12 = LDW(&(W[TWVL * 8]));
Chris@42 192 T19 = LD(&(ri[WS(rs, 13)]), ms, &(ri[WS(rs, 1)]));
Chris@42 193 T28 = VFNMS(TY, TW, T27);
Chris@42 194 T10 = VFMA(TY, TZ, TX);
Chris@42 195 T1c = LD(&(ii[WS(rs, 13)]), ms, &(ii[WS(rs, 1)]));
Chris@42 196 T18 = LDW(&(W[TWVL * 24]));
Chris@42 197 T15 = LDW(&(W[TWVL * 9]));
Chris@42 198 T1b = LDW(&(W[TWVL * 25]));
Chris@42 199 T2d = VMUL(T12, T16);
Chris@42 200 T14 = VMUL(T12, T13);
Chris@42 201 T30 = VADD(T26, T28);
Chris@42 202 T29 = VSUB(T26, T28);
Chris@42 203 T11 = VADD(TU, T10);
Chris@42 204 T2c = VSUB(TU, T10);
Chris@42 205 T2f = VMUL(T18, T1c);
Chris@42 206 T1a = VMUL(T18, T19);
Chris@42 207 T2e = VFNMS(T15, T13, T2d);
Chris@42 208 T17 = VFMA(T15, T16, T14);
Chris@42 209 }
Chris@42 210 {
Chris@42 211 V TB, TE, TA, T2g, T1d, TG, TD, T1X, TC;
Chris@42 212 TB = LD(&(ri[WS(rs, 14)]), ms, &(ri[0]));
Chris@42 213 TE = LD(&(ii[WS(rs, 14)]), ms, &(ii[0]));
Chris@42 214 TA = LDW(&(W[TWVL * 26]));
Chris@42 215 TH = LD(&(ri[WS(rs, 6)]), ms, &(ri[0]));
Chris@42 216 T2g = VFNMS(T1b, T19, T2f);
Chris@42 217 T1d = VFMA(T1b, T1c, T1a);
Chris@42 218 TK = LD(&(ii[WS(rs, 6)]), ms, &(ii[0]));
Chris@42 219 TG = LDW(&(W[TWVL * 10]));
Chris@42 220 TD = LDW(&(W[TWVL * 27]));
Chris@42 221 TJ = LDW(&(W[TWVL * 11]));
Chris@42 222 T1X = VMUL(TA, TE);
Chris@42 223 TC = VMUL(TA, TB);
Chris@42 224 T31 = VADD(T2e, T2g);
Chris@42 225 T2h = VSUB(T2e, T2g);
Chris@42 226 T1e = VADD(T17, T1d);
Chris@42 227 T2a = VSUB(T17, T1d);
Chris@42 228 T1Z = VMUL(TG, TK);
Chris@42 229 TI = VMUL(TG, TH);
Chris@42 230 T1Y = VFNMS(TD, TB, T1X);
Chris@42 231 TF = VFMA(TD, TE, TC);
Chris@42 232 }
Chris@42 233 }
Chris@42 234 }
Chris@42 235 }
Chris@42 236 {
Chris@42 237 V T2U, Tm, T3p, T3u, T34, T1G, T1f, T2Z, T20, TL, T32, T3f, T3g, T37;
Chris@42 238 T2U = VSUB(T8, Tl);
Chris@42 239 Tm = VADD(T8, Tl);
Chris@42 240 T3p = VADD(T3k, T3o);
Chris@42 241 T3u = VSUB(T3o, T3k);
Chris@42 242 T34 = VSUB(T1s, T1F);
Chris@42 243 T1G = VADD(T1s, T1F);
Chris@42 244 T1f = VADD(T11, T1e);
Chris@42 245 T2Z = VSUB(T11, T1e);
Chris@42 246 T20 = VFNMS(TJ, TH, T1Z);
Chris@42 247 TL = VFMA(TJ, TK, TI);
Chris@42 248 T32 = VSUB(T30, T31);
Chris@42 249 T3f = VADD(T30, T31);
Chris@42 250 T3g = VADD(T35, T36);
Chris@42 251 T37 = VSUB(T35, T36);
Chris@42 252 {
Chris@42 253 V T3r, T1H, T21, T1W, T3i, T3h, T3j, T2X, TN, T3t, T2W, TM;
Chris@42 254 T3r = VSUB(T1G, T1f);
Chris@42 255 T1H = VADD(T1f, T1G);
Chris@42 256 T21 = VSUB(T1Y, T20);
Chris@42 257 T2W = VADD(T1Y, T20);
Chris@42 258 T1W = VSUB(TF, TL);
Chris@42 259 TM = VADD(TF, TL);
Chris@42 260 T3i = VADD(T3f, T3g);
Chris@42 261 T3h = VSUB(T3f, T3g);
Chris@42 262 T3j = VADD(T2V, T2W);
Chris@42 263 T2X = VSUB(T2V, T2W);
Chris@42 264 TN = VADD(Tz, TM);
Chris@42 265 T3t = VSUB(TM, Tz);
Chris@42 266 {
Chris@42 267 V T2E, T1O, T3B, T3H, T2x, T2q, T2K, T2J, T3C, T23, T3I, T2H;
Chris@42 268 {
Chris@42 269 V T2F, T1V, T22, T2G;
Chris@42 270 T2E = VADD(T1I, T1N);
Chris@42 271 T1O = VSUB(T1I, T1N);
Chris@42 272 {
Chris@42 273 V T3b, T33, T3c, T38;
Chris@42 274 T3b = VSUB(T32, T2Z);
Chris@42 275 T33 = VADD(T2Z, T32);
Chris@42 276 T3c = VADD(T34, T37);
Chris@42 277 T38 = VSUB(T34, T37);
Chris@42 278 {
Chris@42 279 V T3a, T2Y, T3s, T3q;
Chris@42 280 T3a = VSUB(T2U, T2X);
Chris@42 281 T2Y = VADD(T2U, T2X);
Chris@42 282 T3s = VSUB(T3p, T3j);
Chris@42 283 T3q = VADD(T3j, T3p);
Chris@42 284 {
Chris@42 285 V T3x, T3v, T3e, TO;
Chris@42 286 T3x = VSUB(T3u, T3t);
Chris@42 287 T3v = VADD(T3t, T3u);
Chris@42 288 T3e = VSUB(Tm, TN);
Chris@42 289 TO = VADD(Tm, TN);
Chris@42 290 {
Chris@42 291 V T3d, T3w, T3y, T39;
Chris@42 292 T3d = VSUB(T3b, T3c);
Chris@42 293 T3w = VADD(T3b, T3c);
Chris@42 294 T3y = VSUB(T38, T33);
Chris@42 295 T39 = VADD(T33, T38);
Chris@42 296 ST(&(ii[WS(rs, 4)]), VADD(T3r, T3s), ms, &(ii[0]));
Chris@42 297 ST(&(ii[WS(rs, 12)]), VSUB(T3s, T3r), ms, &(ii[0]));
Chris@42 298 ST(&(ii[0]), VADD(T3i, T3q), ms, &(ii[0]));
Chris@42 299 ST(&(ii[WS(rs, 8)]), VSUB(T3q, T3i), ms, &(ii[0]));
Chris@42 300 ST(&(ri[WS(rs, 4)]), VADD(T3e, T3h), ms, &(ri[0]));
Chris@42 301 ST(&(ri[WS(rs, 12)]), VSUB(T3e, T3h), ms, &(ri[0]));
Chris@42 302 ST(&(ri[0]), VADD(TO, T1H), ms, &(ri[0]));
Chris@42 303 ST(&(ri[WS(rs, 8)]), VSUB(TO, T1H), ms, &(ri[0]));
Chris@42 304 ST(&(ri[WS(rs, 6)]), VFMA(LDK(KP707106781), T3d, T3a), ms, &(ri[0]));
Chris@42 305 ST(&(ri[WS(rs, 14)]), VFNMS(LDK(KP707106781), T3d, T3a), ms, &(ri[0]));
Chris@42 306 ST(&(ii[WS(rs, 10)]), VFNMS(LDK(KP707106781), T3w, T3v), ms, &(ii[0]));
Chris@42 307 ST(&(ii[WS(rs, 2)]), VFMA(LDK(KP707106781), T3w, T3v), ms, &(ii[0]));
Chris@42 308 ST(&(ii[WS(rs, 14)]), VFNMS(LDK(KP707106781), T3y, T3x), ms, &(ii[0]));
Chris@42 309 ST(&(ii[WS(rs, 6)]), VFMA(LDK(KP707106781), T3y, T3x), ms, &(ii[0]));
Chris@42 310 ST(&(ri[WS(rs, 2)]), VFMA(LDK(KP707106781), T39, T2Y), ms, &(ri[0]));
Chris@42 311 ST(&(ri[WS(rs, 10)]), VFNMS(LDK(KP707106781), T39, T2Y), ms, &(ri[0]));
Chris@42 312 T3B = VSUB(T3z, T3A);
Chris@42 313 T3H = VADD(T3A, T3z);
Chris@42 314 }
Chris@42 315 }
Chris@42 316 }
Chris@42 317 }
Chris@42 318 T2F = VADD(T1U, T1T);
Chris@42 319 T1V = VSUB(T1T, T1U);
Chris@42 320 T22 = VADD(T1W, T21);
Chris@42 321 T2G = VSUB(T1W, T21);
Chris@42 322 {
Chris@42 323 V T2M, T2N, T2b, T2i;
Chris@42 324 T2x = VSUB(T2r, T2w);
Chris@42 325 T2M = VADD(T2r, T2w);
Chris@42 326 T2N = VSUB(T2o, T2p);
Chris@42 327 T2q = VADD(T2o, T2p);
Chris@42 328 T2K = VSUB(T29, T2a);
Chris@42 329 T2b = VADD(T29, T2a);
Chris@42 330 T2i = VSUB(T2c, T2h);
Chris@42 331 T2J = VADD(T2c, T2h);
Chris@42 332 T3C = VADD(T1V, T22);
Chris@42 333 T23 = VSUB(T1V, T22);
Chris@42 334 T2S = VFMA(LDK(KP414213562), T2M, T2N);
Chris@42 335 T2O = VFNMS(LDK(KP414213562), T2N, T2M);
Chris@42 336 T3I = VSUB(T2G, T2F);
Chris@42 337 T2H = VADD(T2F, T2G);
Chris@42 338 T2B = VFNMS(LDK(KP414213562), T2b, T2i);
Chris@42 339 T2j = VFMA(LDK(KP414213562), T2i, T2b);
Chris@42 340 }
Chris@42 341 }
Chris@42 342 T2A = VFNMS(LDK(KP707106781), T23, T1O);
Chris@42 343 T24 = VFMA(LDK(KP707106781), T23, T1O);
Chris@42 344 T3J = VFMA(LDK(KP707106781), T3I, T3H);
Chris@42 345 T3L = VFNMS(LDK(KP707106781), T3I, T3H);
Chris@42 346 T2Q = VFNMS(LDK(KP707106781), T2H, T2E);
Chris@42 347 T2I = VFMA(LDK(KP707106781), T2H, T2E);
Chris@42 348 T2R = VFNMS(LDK(KP414213562), T2J, T2K);
Chris@42 349 T2L = VFMA(LDK(KP414213562), T2K, T2J);
Chris@42 350 T2C = VFMA(LDK(KP414213562), T2q, T2x);
Chris@42 351 T2y = VFNMS(LDK(KP414213562), T2x, T2q);
Chris@42 352 T3D = VFMA(LDK(KP707106781), T3C, T3B);
Chris@42 353 T3F = VFNMS(LDK(KP707106781), T3C, T3B);
Chris@42 354 }
Chris@42 355 }
Chris@42 356 }
Chris@42 357 }
Chris@42 358 {
Chris@42 359 V T3E, T2T, T2P, T3G;
Chris@42 360 T3E = VADD(T2R, T2S);
Chris@42 361 T2T = VSUB(T2R, T2S);
Chris@42 362 T2P = VADD(T2L, T2O);
Chris@42 363 T3G = VSUB(T2O, T2L);
Chris@42 364 {
Chris@42 365 V T3K, T2D, T2z, T3M;
Chris@42 366 T3K = VSUB(T2C, T2B);
Chris@42 367 T2D = VADD(T2B, T2C);
Chris@42 368 T2z = VSUB(T2j, T2y);
Chris@42 369 T3M = VADD(T2j, T2y);
Chris@42 370 ST(&(ri[WS(rs, 5)]), VFMA(LDK(KP923879532), T2T, T2Q), ms, &(ri[WS(rs, 1)]));
Chris@42 371 ST(&(ri[WS(rs, 13)]), VFNMS(LDK(KP923879532), T2T, T2Q), ms, &(ri[WS(rs, 1)]));
Chris@42 372 ST(&(ii[WS(rs, 9)]), VFNMS(LDK(KP923879532), T3E, T3D), ms, &(ii[WS(rs, 1)]));
Chris@42 373 ST(&(ii[WS(rs, 1)]), VFMA(LDK(KP923879532), T3E, T3D), ms, &(ii[WS(rs, 1)]));
Chris@42 374 ST(&(ii[WS(rs, 13)]), VFNMS(LDK(KP923879532), T3G, T3F), ms, &(ii[WS(rs, 1)]));
Chris@42 375 ST(&(ii[WS(rs, 5)]), VFMA(LDK(KP923879532), T3G, T3F), ms, &(ii[WS(rs, 1)]));
Chris@42 376 ST(&(ri[WS(rs, 1)]), VFMA(LDK(KP923879532), T2P, T2I), ms, &(ri[WS(rs, 1)]));
Chris@42 377 ST(&(ri[WS(rs, 9)]), VFNMS(LDK(KP923879532), T2P, T2I), ms, &(ri[WS(rs, 1)]));
Chris@42 378 ST(&(ri[WS(rs, 15)]), VFMA(LDK(KP923879532), T2D, T2A), ms, &(ri[WS(rs, 1)]));
Chris@42 379 ST(&(ri[WS(rs, 7)]), VFNMS(LDK(KP923879532), T2D, T2A), ms, &(ri[WS(rs, 1)]));
Chris@42 380 ST(&(ii[WS(rs, 11)]), VFNMS(LDK(KP923879532), T3K, T3J), ms, &(ii[WS(rs, 1)]));
Chris@42 381 ST(&(ii[WS(rs, 3)]), VFMA(LDK(KP923879532), T3K, T3J), ms, &(ii[WS(rs, 1)]));
Chris@42 382 ST(&(ii[WS(rs, 15)]), VFMA(LDK(KP923879532), T3M, T3L), ms, &(ii[WS(rs, 1)]));
Chris@42 383 ST(&(ii[WS(rs, 7)]), VFNMS(LDK(KP923879532), T3M, T3L), ms, &(ii[WS(rs, 1)]));
Chris@42 384 ST(&(ri[WS(rs, 3)]), VFMA(LDK(KP923879532), T2z, T24), ms, &(ri[WS(rs, 1)]));
Chris@42 385 ST(&(ri[WS(rs, 11)]), VFNMS(LDK(KP923879532), T2z, T24), ms, &(ri[WS(rs, 1)]));
Chris@42 386 }
Chris@42 387 }
Chris@42 388 }
Chris@42 389 }
Chris@42 390 VLEAVE();
Chris@42 391 }
Chris@42 392
Chris@42 393 static const tw_instr twinstr[] = {
Chris@42 394 VTW(0, 1),
Chris@42 395 VTW(0, 2),
Chris@42 396 VTW(0, 3),
Chris@42 397 VTW(0, 4),
Chris@42 398 VTW(0, 5),
Chris@42 399 VTW(0, 6),
Chris@42 400 VTW(0, 7),
Chris@42 401 VTW(0, 8),
Chris@42 402 VTW(0, 9),
Chris@42 403 VTW(0, 10),
Chris@42 404 VTW(0, 11),
Chris@42 405 VTW(0, 12),
Chris@42 406 VTW(0, 13),
Chris@42 407 VTW(0, 14),
Chris@42 408 VTW(0, 15),
Chris@42 409 {TW_NEXT, (2 * VL), 0}
Chris@42 410 };
Chris@42 411
Chris@42 412 static const ct_desc desc = { 16, XSIMD_STRING("t1sv_16"), twinstr, &GENUS, {104, 30, 70, 0}, 0, 0, 0 };
Chris@42 413
Chris@42 414 void XSIMD(codelet_t1sv_16) (planner *p) {
Chris@42 415 X(kdft_dit_register) (p, t1sv_16, &desc);
Chris@42 416 }
Chris@42 417 #else /* HAVE_FMA */
Chris@42 418
Chris@42 419 /* Generated by: ../../../genfft/gen_twiddle.native -simd -compact -variables 4 -pipeline-latency 8 -n 16 -name t1sv_16 -include ts.h */
Chris@42 420
Chris@42 421 /*
Chris@42 422 * This function contains 174 FP additions, 84 FP multiplications,
Chris@42 423 * (or, 136 additions, 46 multiplications, 38 fused multiply/add),
Chris@42 424 * 52 stack variables, 3 constants, and 64 memory accesses
Chris@42 425 */
Chris@42 426 #include "ts.h"
Chris@42 427
Chris@42 428 static void t1sv_16(R *ri, R *ii, const R *W, stride rs, INT mb, INT me, INT ms)
Chris@42 429 {
Chris@42 430 DVK(KP382683432, +0.382683432365089771728459984030398866761344562);
Chris@42 431 DVK(KP923879532, +0.923879532511286756128183189396788286822416626);
Chris@42 432 DVK(KP707106781, +0.707106781186547524400844362104849039284835938);
Chris@42 433 {
Chris@42 434 INT m;
Chris@42 435 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@42 436 V T7, T37, T1t, T2U, Ti, T38, T1w, T2R, Tu, T2s, T1C, T2c, TF, T2t, T1H;
Chris@42 437 V T2d, T1f, T1q, T2B, T2C, T2D, T2E, T1Z, T2j, T24, T2k, TS, T13, T2w, T2x;
Chris@42 438 V T2y, T2z, T1O, T2g, T1T, T2h;
Chris@42 439 {
Chris@42 440 V T1, T2T, T6, T2S;
Chris@42 441 T1 = LD(&(ri[0]), ms, &(ri[0]));
Chris@42 442 T2T = LD(&(ii[0]), ms, &(ii[0]));
Chris@42 443 {
Chris@42 444 V T3, T5, T2, T4;
Chris@42 445 T3 = LD(&(ri[WS(rs, 8)]), ms, &(ri[0]));
Chris@42 446 T5 = LD(&(ii[WS(rs, 8)]), ms, &(ii[0]));
Chris@42 447 T2 = LDW(&(W[TWVL * 14]));
Chris@42 448 T4 = LDW(&(W[TWVL * 15]));
Chris@42 449 T6 = VFMA(T2, T3, VMUL(T4, T5));
Chris@42 450 T2S = VFNMS(T4, T3, VMUL(T2, T5));
Chris@42 451 }
Chris@42 452 T7 = VADD(T1, T6);
Chris@42 453 T37 = VSUB(T2T, T2S);
Chris@42 454 T1t = VSUB(T1, T6);
Chris@42 455 T2U = VADD(T2S, T2T);
Chris@42 456 }
Chris@42 457 {
Chris@42 458 V Tc, T1u, Th, T1v;
Chris@42 459 {
Chris@42 460 V T9, Tb, T8, Ta;
Chris@42 461 T9 = LD(&(ri[WS(rs, 4)]), ms, &(ri[0]));
Chris@42 462 Tb = LD(&(ii[WS(rs, 4)]), ms, &(ii[0]));
Chris@42 463 T8 = LDW(&(W[TWVL * 6]));
Chris@42 464 Ta = LDW(&(W[TWVL * 7]));
Chris@42 465 Tc = VFMA(T8, T9, VMUL(Ta, Tb));
Chris@42 466 T1u = VFNMS(Ta, T9, VMUL(T8, Tb));
Chris@42 467 }
Chris@42 468 {
Chris@42 469 V Te, Tg, Td, Tf;
Chris@42 470 Te = LD(&(ri[WS(rs, 12)]), ms, &(ri[0]));
Chris@42 471 Tg = LD(&(ii[WS(rs, 12)]), ms, &(ii[0]));
Chris@42 472 Td = LDW(&(W[TWVL * 22]));
Chris@42 473 Tf = LDW(&(W[TWVL * 23]));
Chris@42 474 Th = VFMA(Td, Te, VMUL(Tf, Tg));
Chris@42 475 T1v = VFNMS(Tf, Te, VMUL(Td, Tg));
Chris@42 476 }
Chris@42 477 Ti = VADD(Tc, Th);
Chris@42 478 T38 = VSUB(Tc, Th);
Chris@42 479 T1w = VSUB(T1u, T1v);
Chris@42 480 T2R = VADD(T1u, T1v);
Chris@42 481 }
Chris@42 482 {
Chris@42 483 V To, T1y, Tt, T1z, T1A, T1B;
Chris@42 484 {
Chris@42 485 V Tl, Tn, Tk, Tm;
Chris@42 486 Tl = LD(&(ri[WS(rs, 2)]), ms, &(ri[0]));
Chris@42 487 Tn = LD(&(ii[WS(rs, 2)]), ms, &(ii[0]));
Chris@42 488 Tk = LDW(&(W[TWVL * 2]));
Chris@42 489 Tm = LDW(&(W[TWVL * 3]));
Chris@42 490 To = VFMA(Tk, Tl, VMUL(Tm, Tn));
Chris@42 491 T1y = VFNMS(Tm, Tl, VMUL(Tk, Tn));
Chris@42 492 }
Chris@42 493 {
Chris@42 494 V Tq, Ts, Tp, Tr;
Chris@42 495 Tq = LD(&(ri[WS(rs, 10)]), ms, &(ri[0]));
Chris@42 496 Ts = LD(&(ii[WS(rs, 10)]), ms, &(ii[0]));
Chris@42 497 Tp = LDW(&(W[TWVL * 18]));
Chris@42 498 Tr = LDW(&(W[TWVL * 19]));
Chris@42 499 Tt = VFMA(Tp, Tq, VMUL(Tr, Ts));
Chris@42 500 T1z = VFNMS(Tr, Tq, VMUL(Tp, Ts));
Chris@42 501 }
Chris@42 502 Tu = VADD(To, Tt);
Chris@42 503 T2s = VADD(T1y, T1z);
Chris@42 504 T1A = VSUB(T1y, T1z);
Chris@42 505 T1B = VSUB(To, Tt);
Chris@42 506 T1C = VSUB(T1A, T1B);
Chris@42 507 T2c = VADD(T1B, T1A);
Chris@42 508 }
Chris@42 509 {
Chris@42 510 V Tz, T1E, TE, T1F, T1D, T1G;
Chris@42 511 {
Chris@42 512 V Tw, Ty, Tv, Tx;
Chris@42 513 Tw = LD(&(ri[WS(rs, 14)]), ms, &(ri[0]));
Chris@42 514 Ty = LD(&(ii[WS(rs, 14)]), ms, &(ii[0]));
Chris@42 515 Tv = LDW(&(W[TWVL * 26]));
Chris@42 516 Tx = LDW(&(W[TWVL * 27]));
Chris@42 517 Tz = VFMA(Tv, Tw, VMUL(Tx, Ty));
Chris@42 518 T1E = VFNMS(Tx, Tw, VMUL(Tv, Ty));
Chris@42 519 }
Chris@42 520 {
Chris@42 521 V TB, TD, TA, TC;
Chris@42 522 TB = LD(&(ri[WS(rs, 6)]), ms, &(ri[0]));
Chris@42 523 TD = LD(&(ii[WS(rs, 6)]), ms, &(ii[0]));
Chris@42 524 TA = LDW(&(W[TWVL * 10]));
Chris@42 525 TC = LDW(&(W[TWVL * 11]));
Chris@42 526 TE = VFMA(TA, TB, VMUL(TC, TD));
Chris@42 527 T1F = VFNMS(TC, TB, VMUL(TA, TD));
Chris@42 528 }
Chris@42 529 TF = VADD(Tz, TE);
Chris@42 530 T2t = VADD(T1E, T1F);
Chris@42 531 T1D = VSUB(Tz, TE);
Chris@42 532 T1G = VSUB(T1E, T1F);
Chris@42 533 T1H = VADD(T1D, T1G);
Chris@42 534 T2d = VSUB(T1D, T1G);
Chris@42 535 }
Chris@42 536 {
Chris@42 537 V T19, T20, T1p, T1X, T1e, T21, T1k, T1W;
Chris@42 538 {
Chris@42 539 V T16, T18, T15, T17;
Chris@42 540 T16 = LD(&(ri[WS(rs, 15)]), ms, &(ri[WS(rs, 1)]));
Chris@42 541 T18 = LD(&(ii[WS(rs, 15)]), ms, &(ii[WS(rs, 1)]));
Chris@42 542 T15 = LDW(&(W[TWVL * 28]));
Chris@42 543 T17 = LDW(&(W[TWVL * 29]));
Chris@42 544 T19 = VFMA(T15, T16, VMUL(T17, T18));
Chris@42 545 T20 = VFNMS(T17, T16, VMUL(T15, T18));
Chris@42 546 }
Chris@42 547 {
Chris@42 548 V T1m, T1o, T1l, T1n;
Chris@42 549 T1m = LD(&(ri[WS(rs, 11)]), ms, &(ri[WS(rs, 1)]));
Chris@42 550 T1o = LD(&(ii[WS(rs, 11)]), ms, &(ii[WS(rs, 1)]));
Chris@42 551 T1l = LDW(&(W[TWVL * 20]));
Chris@42 552 T1n = LDW(&(W[TWVL * 21]));
Chris@42 553 T1p = VFMA(T1l, T1m, VMUL(T1n, T1o));
Chris@42 554 T1X = VFNMS(T1n, T1m, VMUL(T1l, T1o));
Chris@42 555 }
Chris@42 556 {
Chris@42 557 V T1b, T1d, T1a, T1c;
Chris@42 558 T1b = LD(&(ri[WS(rs, 7)]), ms, &(ri[WS(rs, 1)]));
Chris@42 559 T1d = LD(&(ii[WS(rs, 7)]), ms, &(ii[WS(rs, 1)]));
Chris@42 560 T1a = LDW(&(W[TWVL * 12]));
Chris@42 561 T1c = LDW(&(W[TWVL * 13]));
Chris@42 562 T1e = VFMA(T1a, T1b, VMUL(T1c, T1d));
Chris@42 563 T21 = VFNMS(T1c, T1b, VMUL(T1a, T1d));
Chris@42 564 }
Chris@42 565 {
Chris@42 566 V T1h, T1j, T1g, T1i;
Chris@42 567 T1h = LD(&(ri[WS(rs, 3)]), ms, &(ri[WS(rs, 1)]));
Chris@42 568 T1j = LD(&(ii[WS(rs, 3)]), ms, &(ii[WS(rs, 1)]));
Chris@42 569 T1g = LDW(&(W[TWVL * 4]));
Chris@42 570 T1i = LDW(&(W[TWVL * 5]));
Chris@42 571 T1k = VFMA(T1g, T1h, VMUL(T1i, T1j));
Chris@42 572 T1W = VFNMS(T1i, T1h, VMUL(T1g, T1j));
Chris@42 573 }
Chris@42 574 T1f = VADD(T19, T1e);
Chris@42 575 T1q = VADD(T1k, T1p);
Chris@42 576 T2B = VSUB(T1f, T1q);
Chris@42 577 T2C = VADD(T20, T21);
Chris@42 578 T2D = VADD(T1W, T1X);
Chris@42 579 T2E = VSUB(T2C, T2D);
Chris@42 580 {
Chris@42 581 V T1V, T1Y, T22, T23;
Chris@42 582 T1V = VSUB(T19, T1e);
Chris@42 583 T1Y = VSUB(T1W, T1X);
Chris@42 584 T1Z = VSUB(T1V, T1Y);
Chris@42 585 T2j = VADD(T1V, T1Y);
Chris@42 586 T22 = VSUB(T20, T21);
Chris@42 587 T23 = VSUB(T1k, T1p);
Chris@42 588 T24 = VADD(T22, T23);
Chris@42 589 T2k = VSUB(T22, T23);
Chris@42 590 }
Chris@42 591 }
Chris@42 592 {
Chris@42 593 V TM, T1K, T12, T1R, TR, T1L, TX, T1Q;
Chris@42 594 {
Chris@42 595 V TJ, TL, TI, TK;
Chris@42 596 TJ = LD(&(ri[WS(rs, 1)]), ms, &(ri[WS(rs, 1)]));
Chris@42 597 TL = LD(&(ii[WS(rs, 1)]), ms, &(ii[WS(rs, 1)]));
Chris@42 598 TI = LDW(&(W[0]));
Chris@42 599 TK = LDW(&(W[TWVL * 1]));
Chris@42 600 TM = VFMA(TI, TJ, VMUL(TK, TL));
Chris@42 601 T1K = VFNMS(TK, TJ, VMUL(TI, TL));
Chris@42 602 }
Chris@42 603 {
Chris@42 604 V TZ, T11, TY, T10;
Chris@42 605 TZ = LD(&(ri[WS(rs, 13)]), ms, &(ri[WS(rs, 1)]));
Chris@42 606 T11 = LD(&(ii[WS(rs, 13)]), ms, &(ii[WS(rs, 1)]));
Chris@42 607 TY = LDW(&(W[TWVL * 24]));
Chris@42 608 T10 = LDW(&(W[TWVL * 25]));
Chris@42 609 T12 = VFMA(TY, TZ, VMUL(T10, T11));
Chris@42 610 T1R = VFNMS(T10, TZ, VMUL(TY, T11));
Chris@42 611 }
Chris@42 612 {
Chris@42 613 V TO, TQ, TN, TP;
Chris@42 614 TO = LD(&(ri[WS(rs, 9)]), ms, &(ri[WS(rs, 1)]));
Chris@42 615 TQ = LD(&(ii[WS(rs, 9)]), ms, &(ii[WS(rs, 1)]));
Chris@42 616 TN = LDW(&(W[TWVL * 16]));
Chris@42 617 TP = LDW(&(W[TWVL * 17]));
Chris@42 618 TR = VFMA(TN, TO, VMUL(TP, TQ));
Chris@42 619 T1L = VFNMS(TP, TO, VMUL(TN, TQ));
Chris@42 620 }
Chris@42 621 {
Chris@42 622 V TU, TW, TT, TV;
Chris@42 623 TU = LD(&(ri[WS(rs, 5)]), ms, &(ri[WS(rs, 1)]));
Chris@42 624 TW = LD(&(ii[WS(rs, 5)]), ms, &(ii[WS(rs, 1)]));
Chris@42 625 TT = LDW(&(W[TWVL * 8]));
Chris@42 626 TV = LDW(&(W[TWVL * 9]));
Chris@42 627 TX = VFMA(TT, TU, VMUL(TV, TW));
Chris@42 628 T1Q = VFNMS(TV, TU, VMUL(TT, TW));
Chris@42 629 }
Chris@42 630 TS = VADD(TM, TR);
Chris@42 631 T13 = VADD(TX, T12);
Chris@42 632 T2w = VSUB(TS, T13);
Chris@42 633 T2x = VADD(T1K, T1L);
Chris@42 634 T2y = VADD(T1Q, T1R);
Chris@42 635 T2z = VSUB(T2x, T2y);
Chris@42 636 {
Chris@42 637 V T1M, T1N, T1P, T1S;
Chris@42 638 T1M = VSUB(T1K, T1L);
Chris@42 639 T1N = VSUB(TX, T12);
Chris@42 640 T1O = VADD(T1M, T1N);
Chris@42 641 T2g = VSUB(T1M, T1N);
Chris@42 642 T1P = VSUB(TM, TR);
Chris@42 643 T1S = VSUB(T1Q, T1R);
Chris@42 644 T1T = VSUB(T1P, T1S);
Chris@42 645 T2h = VADD(T1P, T1S);
Chris@42 646 }
Chris@42 647 }
Chris@42 648 {
Chris@42 649 V T1J, T27, T3g, T3i, T26, T3h, T2a, T3d;
Chris@42 650 {
Chris@42 651 V T1x, T1I, T3e, T3f;
Chris@42 652 T1x = VSUB(T1t, T1w);
Chris@42 653 T1I = VMUL(LDK(KP707106781), VSUB(T1C, T1H));
Chris@42 654 T1J = VADD(T1x, T1I);
Chris@42 655 T27 = VSUB(T1x, T1I);
Chris@42 656 T3e = VMUL(LDK(KP707106781), VSUB(T2d, T2c));
Chris@42 657 T3f = VADD(T38, T37);
Chris@42 658 T3g = VADD(T3e, T3f);
Chris@42 659 T3i = VSUB(T3f, T3e);
Chris@42 660 }
Chris@42 661 {
Chris@42 662 V T1U, T25, T28, T29;
Chris@42 663 T1U = VFMA(LDK(KP923879532), T1O, VMUL(LDK(KP382683432), T1T));
Chris@42 664 T25 = VFNMS(LDK(KP923879532), T24, VMUL(LDK(KP382683432), T1Z));
Chris@42 665 T26 = VADD(T1U, T25);
Chris@42 666 T3h = VSUB(T25, T1U);
Chris@42 667 T28 = VFNMS(LDK(KP923879532), T1T, VMUL(LDK(KP382683432), T1O));
Chris@42 668 T29 = VFMA(LDK(KP382683432), T24, VMUL(LDK(KP923879532), T1Z));
Chris@42 669 T2a = VSUB(T28, T29);
Chris@42 670 T3d = VADD(T28, T29);
Chris@42 671 }
Chris@42 672 ST(&(ri[WS(rs, 11)]), VSUB(T1J, T26), ms, &(ri[WS(rs, 1)]));
Chris@42 673 ST(&(ii[WS(rs, 11)]), VSUB(T3g, T3d), ms, &(ii[WS(rs, 1)]));
Chris@42 674 ST(&(ri[WS(rs, 3)]), VADD(T1J, T26), ms, &(ri[WS(rs, 1)]));
Chris@42 675 ST(&(ii[WS(rs, 3)]), VADD(T3d, T3g), ms, &(ii[WS(rs, 1)]));
Chris@42 676 ST(&(ri[WS(rs, 15)]), VSUB(T27, T2a), ms, &(ri[WS(rs, 1)]));
Chris@42 677 ST(&(ii[WS(rs, 15)]), VSUB(T3i, T3h), ms, &(ii[WS(rs, 1)]));
Chris@42 678 ST(&(ri[WS(rs, 7)]), VADD(T27, T2a), ms, &(ri[WS(rs, 1)]));
Chris@42 679 ST(&(ii[WS(rs, 7)]), VADD(T3h, T3i), ms, &(ii[WS(rs, 1)]));
Chris@42 680 }
Chris@42 681 {
Chris@42 682 V T2v, T2H, T32, T34, T2G, T33, T2K, T2Z;
Chris@42 683 {
Chris@42 684 V T2r, T2u, T30, T31;
Chris@42 685 T2r = VSUB(T7, Ti);
Chris@42 686 T2u = VSUB(T2s, T2t);
Chris@42 687 T2v = VADD(T2r, T2u);
Chris@42 688 T2H = VSUB(T2r, T2u);
Chris@42 689 T30 = VSUB(TF, Tu);
Chris@42 690 T31 = VSUB(T2U, T2R);
Chris@42 691 T32 = VADD(T30, T31);
Chris@42 692 T34 = VSUB(T31, T30);
Chris@42 693 }
Chris@42 694 {
Chris@42 695 V T2A, T2F, T2I, T2J;
Chris@42 696 T2A = VADD(T2w, T2z);
Chris@42 697 T2F = VSUB(T2B, T2E);
Chris@42 698 T2G = VMUL(LDK(KP707106781), VADD(T2A, T2F));
Chris@42 699 T33 = VMUL(LDK(KP707106781), VSUB(T2F, T2A));
Chris@42 700 T2I = VSUB(T2z, T2w);
Chris@42 701 T2J = VADD(T2B, T2E);
Chris@42 702 T2K = VMUL(LDK(KP707106781), VSUB(T2I, T2J));
Chris@42 703 T2Z = VMUL(LDK(KP707106781), VADD(T2I, T2J));
Chris@42 704 }
Chris@42 705 ST(&(ri[WS(rs, 10)]), VSUB(T2v, T2G), ms, &(ri[0]));
Chris@42 706 ST(&(ii[WS(rs, 10)]), VSUB(T32, T2Z), ms, &(ii[0]));
Chris@42 707 ST(&(ri[WS(rs, 2)]), VADD(T2v, T2G), ms, &(ri[0]));
Chris@42 708 ST(&(ii[WS(rs, 2)]), VADD(T2Z, T32), ms, &(ii[0]));
Chris@42 709 ST(&(ri[WS(rs, 14)]), VSUB(T2H, T2K), ms, &(ri[0]));
Chris@42 710 ST(&(ii[WS(rs, 14)]), VSUB(T34, T33), ms, &(ii[0]));
Chris@42 711 ST(&(ri[WS(rs, 6)]), VADD(T2H, T2K), ms, &(ri[0]));
Chris@42 712 ST(&(ii[WS(rs, 6)]), VADD(T33, T34), ms, &(ii[0]));
Chris@42 713 }
Chris@42 714 {
Chris@42 715 V T2f, T2n, T3a, T3c, T2m, T3b, T2q, T35;
Chris@42 716 {
Chris@42 717 V T2b, T2e, T36, T39;
Chris@42 718 T2b = VADD(T1t, T1w);
Chris@42 719 T2e = VMUL(LDK(KP707106781), VADD(T2c, T2d));
Chris@42 720 T2f = VADD(T2b, T2e);
Chris@42 721 T2n = VSUB(T2b, T2e);
Chris@42 722 T36 = VMUL(LDK(KP707106781), VADD(T1C, T1H));
Chris@42 723 T39 = VSUB(T37, T38);
Chris@42 724 T3a = VADD(T36, T39);
Chris@42 725 T3c = VSUB(T39, T36);
Chris@42 726 }
Chris@42 727 {
Chris@42 728 V T2i, T2l, T2o, T2p;
Chris@42 729 T2i = VFMA(LDK(KP382683432), T2g, VMUL(LDK(KP923879532), T2h));
Chris@42 730 T2l = VFNMS(LDK(KP382683432), T2k, VMUL(LDK(KP923879532), T2j));
Chris@42 731 T2m = VADD(T2i, T2l);
Chris@42 732 T3b = VSUB(T2l, T2i);
Chris@42 733 T2o = VFNMS(LDK(KP382683432), T2h, VMUL(LDK(KP923879532), T2g));
Chris@42 734 T2p = VFMA(LDK(KP923879532), T2k, VMUL(LDK(KP382683432), T2j));
Chris@42 735 T2q = VSUB(T2o, T2p);
Chris@42 736 T35 = VADD(T2o, T2p);
Chris@42 737 }
Chris@42 738 ST(&(ri[WS(rs, 9)]), VSUB(T2f, T2m), ms, &(ri[WS(rs, 1)]));
Chris@42 739 ST(&(ii[WS(rs, 9)]), VSUB(T3a, T35), ms, &(ii[WS(rs, 1)]));
Chris@42 740 ST(&(ri[WS(rs, 1)]), VADD(T2f, T2m), ms, &(ri[WS(rs, 1)]));
Chris@42 741 ST(&(ii[WS(rs, 1)]), VADD(T35, T3a), ms, &(ii[WS(rs, 1)]));
Chris@42 742 ST(&(ri[WS(rs, 13)]), VSUB(T2n, T2q), ms, &(ri[WS(rs, 1)]));
Chris@42 743 ST(&(ii[WS(rs, 13)]), VSUB(T3c, T3b), ms, &(ii[WS(rs, 1)]));
Chris@42 744 ST(&(ri[WS(rs, 5)]), VADD(T2n, T2q), ms, &(ri[WS(rs, 1)]));
Chris@42 745 ST(&(ii[WS(rs, 5)]), VADD(T3b, T3c), ms, &(ii[WS(rs, 1)]));
Chris@42 746 }
Chris@42 747 {
Chris@42 748 V TH, T2L, T2W, T2Y, T1s, T2X, T2O, T2P;
Chris@42 749 {
Chris@42 750 V Tj, TG, T2Q, T2V;
Chris@42 751 Tj = VADD(T7, Ti);
Chris@42 752 TG = VADD(Tu, TF);
Chris@42 753 TH = VADD(Tj, TG);
Chris@42 754 T2L = VSUB(Tj, TG);
Chris@42 755 T2Q = VADD(T2s, T2t);
Chris@42 756 T2V = VADD(T2R, T2U);
Chris@42 757 T2W = VADD(T2Q, T2V);
Chris@42 758 T2Y = VSUB(T2V, T2Q);
Chris@42 759 }
Chris@42 760 {
Chris@42 761 V T14, T1r, T2M, T2N;
Chris@42 762 T14 = VADD(TS, T13);
Chris@42 763 T1r = VADD(T1f, T1q);
Chris@42 764 T1s = VADD(T14, T1r);
Chris@42 765 T2X = VSUB(T1r, T14);
Chris@42 766 T2M = VADD(T2x, T2y);
Chris@42 767 T2N = VADD(T2C, T2D);
Chris@42 768 T2O = VSUB(T2M, T2N);
Chris@42 769 T2P = VADD(T2M, T2N);
Chris@42 770 }
Chris@42 771 ST(&(ri[WS(rs, 8)]), VSUB(TH, T1s), ms, &(ri[0]));
Chris@42 772 ST(&(ii[WS(rs, 8)]), VSUB(T2W, T2P), ms, &(ii[0]));
Chris@42 773 ST(&(ri[0]), VADD(TH, T1s), ms, &(ri[0]));
Chris@42 774 ST(&(ii[0]), VADD(T2P, T2W), ms, &(ii[0]));
Chris@42 775 ST(&(ri[WS(rs, 12)]), VSUB(T2L, T2O), ms, &(ri[0]));
Chris@42 776 ST(&(ii[WS(rs, 12)]), VSUB(T2Y, T2X), ms, &(ii[0]));
Chris@42 777 ST(&(ri[WS(rs, 4)]), VADD(T2L, T2O), ms, &(ri[0]));
Chris@42 778 ST(&(ii[WS(rs, 4)]), VADD(T2X, T2Y), ms, &(ii[0]));
Chris@42 779 }
Chris@42 780 }
Chris@42 781 }
Chris@42 782 VLEAVE();
Chris@42 783 }
Chris@42 784
Chris@42 785 static const tw_instr twinstr[] = {
Chris@42 786 VTW(0, 1),
Chris@42 787 VTW(0, 2),
Chris@42 788 VTW(0, 3),
Chris@42 789 VTW(0, 4),
Chris@42 790 VTW(0, 5),
Chris@42 791 VTW(0, 6),
Chris@42 792 VTW(0, 7),
Chris@42 793 VTW(0, 8),
Chris@42 794 VTW(0, 9),
Chris@42 795 VTW(0, 10),
Chris@42 796 VTW(0, 11),
Chris@42 797 VTW(0, 12),
Chris@42 798 VTW(0, 13),
Chris@42 799 VTW(0, 14),
Chris@42 800 VTW(0, 15),
Chris@42 801 {TW_NEXT, (2 * VL), 0}
Chris@42 802 };
Chris@42 803
Chris@42 804 static const ct_desc desc = { 16, XSIMD_STRING("t1sv_16"), twinstr, &GENUS, {136, 46, 38, 0}, 0, 0, 0 };
Chris@42 805
Chris@42 806 void XSIMD(codelet_t1sv_16) (planner *p) {
Chris@42 807 X(kdft_dit_register) (p, t1sv_16, &desc);
Chris@42 808 }
Chris@42 809 #endif /* HAVE_FMA */