Mercurial > hg > sv-dependency-builds

annotate src/fftw-3.3.8/dft/scalar/codelets/t1_7.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 08:04:13 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 -compact -variables 4 -pipeline-latency 4 -n 7 -name t1_7 -include dft/scalar/t.h */
Chris@82 29
Chris@82 30 /*
Chris@82 31 * This function contains 72 FP additions, 66 FP multiplications,
Chris@82 32 * (or, 18 additions, 12 multiplications, 54 fused multiply/add),
Chris@82 33 * 37 stack variables, 6 constants, and 28 memory accesses
Chris@82 34 */
Chris@82 35 #include "dft/scalar/t.h"
Chris@82 36
Chris@82 37 static void t1_7(R *ri, R *ii, const R *W, stride rs, INT mb, INT me, INT ms)
Chris@82 38 {
Chris@82 39 DK(KP974927912, +0.974927912181823607018131682993931217232785801);
Chris@82 40 DK(KP900968867, +0.900968867902419126236102319507445051165919162);
Chris@82 41 DK(KP801937735, +0.801937735804838252472204639014890102331838324);
Chris@82 42 DK(KP554958132, +0.554958132087371191422194871006410481067288862);
Chris@82 43 DK(KP692021471, +0.692021471630095869627814897002069140197260599);
Chris@82 44 DK(KP356895867, +0.356895867892209443894399510021300583399127187);
Chris@82 45 {
Chris@82 46 INT m;
Chris@82 47 for (m = mb, W = W + (mb * 12); m < me; m = m + 1, ri = ri + ms, ii = ii + ms, W = W + 12, MAKE_VOLATILE_STRIDE(14, rs)) {
Chris@82 48 E T1, T1c, Te, T1h, TR, T19, Tr, T1g, TM, T1a, TE, T1i, TW, T1b;
Chris@82 49 T1 = ri[0];
Chris@82 50 T1c = ii[0];
Chris@82 51 {
Chris@82 52 E T3, T6, T4, TN, T9, Tc, Ta, TP, T2, T8;
Chris@82 53 T3 = ri[WS(rs, 1)];
Chris@82 54 T6 = ii[WS(rs, 1)];
Chris@82 55 T2 = W[0];
Chris@82 56 T4 = T2 * T3;
Chris@82 57 TN = T2 * T6;
Chris@82 58 T9 = ri[WS(rs, 6)];
Chris@82 59 Tc = ii[WS(rs, 6)];
Chris@82 60 T8 = W[10];
Chris@82 61 Ta = T8 * T9;
Chris@82 62 TP = T8 * Tc;
Chris@82 63 {
Chris@82 64 E T7, TO, Td, TQ, T5, Tb;
Chris@82 65 T5 = W[1];
Chris@82 66 T7 = FMA(T5, T6, T4);
Chris@82 67 TO = FNMS(T5, T3, TN);
Chris@82 68 Tb = W[11];
Chris@82 69 Td = FMA(Tb, Tc, Ta);
Chris@82 70 TQ = FNMS(Tb, T9, TP);
Chris@82 71 Te = T7 + Td;
Chris@82 72 T1h = Td - T7;
Chris@82 73 TR = TO - TQ;
Chris@82 74 T19 = TO + TQ;
Chris@82 75 }
Chris@82 76 }
Chris@82 77 {
Chris@82 78 E Tg, Tj, Th, TI, Tm, Tp, Tn, TK, Tf, Tl;
Chris@82 79 Tg = ri[WS(rs, 2)];
Chris@82 80 Tj = ii[WS(rs, 2)];
Chris@82 81 Tf = W[2];
Chris@82 82 Th = Tf * Tg;
Chris@82 83 TI = Tf * Tj;
Chris@82 84 Tm = ri[WS(rs, 5)];
Chris@82 85 Tp = ii[WS(rs, 5)];
Chris@82 86 Tl = W[8];
Chris@82 87 Tn = Tl * Tm;
Chris@82 88 TK = Tl * Tp;
Chris@82 89 {
Chris@82 90 E Tk, TJ, Tq, TL, Ti, To;
Chris@82 91 Ti = W[3];
Chris@82 92 Tk = FMA(Ti, Tj, Th);
Chris@82 93 TJ = FNMS(Ti, Tg, TI);
Chris@82 94 To = W[9];
Chris@82 95 Tq = FMA(To, Tp, Tn);
Chris@82 96 TL = FNMS(To, Tm, TK);
Chris@82 97 Tr = Tk + Tq;
Chris@82 98 T1g = Tq - Tk;
Chris@82 99 TM = TJ - TL;
Chris@82 100 T1a = TJ + TL;
Chris@82 101 }
Chris@82 102 }
Chris@82 103 {
Chris@82 104 E Tt, Tw, Tu, TS, Tz, TC, TA, TU, Ts, Ty;
Chris@82 105 Tt = ri[WS(rs, 3)];
Chris@82 106 Tw = ii[WS(rs, 3)];
Chris@82 107 Ts = W[4];
Chris@82 108 Tu = Ts * Tt;
Chris@82 109 TS = Ts * Tw;
Chris@82 110 Tz = ri[WS(rs, 4)];
Chris@82 111 TC = ii[WS(rs, 4)];
Chris@82 112 Ty = W[6];
Chris@82 113 TA = Ty * Tz;
Chris@82 114 TU = Ty * TC;
Chris@82 115 {
Chris@82 116 E Tx, TT, TD, TV, Tv, TB;
Chris@82 117 Tv = W[5];
Chris@82 118 Tx = FMA(Tv, Tw, Tu);
Chris@82 119 TT = FNMS(Tv, Tt, TS);
Chris@82 120 TB = W[7];
Chris@82 121 TD = FMA(TB, TC, TA);
Chris@82 122 TV = FNMS(TB, Tz, TU);
Chris@82 123 TE = Tx + TD;
Chris@82 124 T1i = TD - Tx;
Chris@82 125 TW = TT - TV;
Chris@82 126 T1b = TT + TV;
Chris@82 127 }
Chris@82 128 }
Chris@82 129 ri[0] = T1 + Te + Tr + TE;
Chris@82 130 ii[0] = T19 + T1a + T1b + T1c;
Chris@82 131 {
Chris@82 132 E TG, TY, TF, TX, TH;
Chris@82 133 TF = FNMS(KP356895867, Tr, Te);
Chris@82 134 TG = FNMS(KP692021471, TF, TE);
Chris@82 135 TX = FMA(KP554958132, TW, TR);
Chris@82 136 TY = FMA(KP801937735, TX, TM);
Chris@82 137 TH = FNMS(KP900968867, TG, T1);
Chris@82 138 ri[WS(rs, 6)] = FNMS(KP974927912, TY, TH);
Chris@82 139 ri[WS(rs, 1)] = FMA(KP974927912, TY, TH);
Chris@82 140 }
Chris@82 141 {
Chris@82 142 E T1e, T1k, T1d, T1j, T1f;
Chris@82 143 T1d = FNMS(KP356895867, T1a, T19);
Chris@82 144 T1e = FNMS(KP692021471, T1d, T1b);
Chris@82 145 T1j = FMA(KP554958132, T1i, T1h);
Chris@82 146 T1k = FMA(KP801937735, T1j, T1g);
Chris@82 147 T1f = FNMS(KP900968867, T1e, T1c);
Chris@82 148 ii[WS(rs, 1)] = FMA(KP974927912, T1k, T1f);
Chris@82 149 ii[WS(rs, 6)] = FNMS(KP974927912, T1k, T1f);
Chris@82 150 }
Chris@82 151 {
Chris@82 152 E T10, T13, TZ, T12, T11;
Chris@82 153 TZ = FNMS(KP356895867, Te, TE);
Chris@82 154 T10 = FNMS(KP692021471, TZ, Tr);
Chris@82 155 T12 = FMA(KP554958132, TM, TW);
Chris@82 156 T13 = FNMS(KP801937735, T12, TR);
Chris@82 157 T11 = FNMS(KP900968867, T10, T1);
Chris@82 158 ri[WS(rs, 5)] = FNMS(KP974927912, T13, T11);
Chris@82 159 ri[WS(rs, 2)] = FMA(KP974927912, T13, T11);
Chris@82 160 }
Chris@82 161 {
Chris@82 162 E T1m, T1p, T1l, T1o, T1n;
Chris@82 163 T1l = FNMS(KP356895867, T19, T1b);
Chris@82 164 T1m = FNMS(KP692021471, T1l, T1a);
Chris@82 165 T1o = FMA(KP554958132, T1g, T1i);
Chris@82 166 T1p = FNMS(KP801937735, T1o, T1h);
Chris@82 167 T1n = FNMS(KP900968867, T1m, T1c);
Chris@82 168 ii[WS(rs, 2)] = FMA(KP974927912, T1p, T1n);
Chris@82 169 ii[WS(rs, 5)] = FNMS(KP974927912, T1p, T1n);
Chris@82 170 }
Chris@82 171 {
Chris@82 172 E T15, T18, T14, T17, T16;
Chris@82 173 T14 = FNMS(KP356895867, TE, Tr);
Chris@82 174 T15 = FNMS(KP692021471, T14, Te);
Chris@82 175 T17 = FNMS(KP554958132, TR, TM);
Chris@82 176 T18 = FNMS(KP801937735, T17, TW);
Chris@82 177 T16 = FNMS(KP900968867, T15, T1);
Chris@82 178 ri[WS(rs, 4)] = FNMS(KP974927912, T18, T16);
Chris@82 179 ri[WS(rs, 3)] = FMA(KP974927912, T18, T16);
Chris@82 180 }
Chris@82 181 {
Chris@82 182 E T1r, T1u, T1q, T1t, T1s;
Chris@82 183 T1q = FNMS(KP356895867, T1b, T1a);
Chris@82 184 T1r = FNMS(KP692021471, T1q, T19);
Chris@82 185 T1t = FNMS(KP554958132, T1h, T1g);
Chris@82 186 T1u = FNMS(KP801937735, T1t, T1i);
Chris@82 187 T1s = FNMS(KP900968867, T1r, T1c);
Chris@82 188 ii[WS(rs, 3)] = FMA(KP974927912, T1u, T1s);
Chris@82 189 ii[WS(rs, 4)] = FNMS(KP974927912, T1u, T1s);
Chris@82 190 }
Chris@82 191 }
Chris@82 192 }
Chris@82 193 }
Chris@82 194
Chris@82 195 static const tw_instr twinstr[] = {
Chris@82 196 {TW_FULL, 0, 7},
Chris@82 197 {TW_NEXT, 1, 0}
Chris@82 198 };
Chris@82 199
Chris@82 200 static const ct_desc desc = { 7, "t1_7", twinstr, &GENUS, {18, 12, 54, 0}, 0, 0, 0 };
Chris@82 201
Chris@82 202 void X(codelet_t1_7) (planner *p) {
Chris@82 203 X(kdft_dit_register) (p, t1_7, &desc);
Chris@82 204 }
Chris@82 205 #else
Chris@82 206
Chris@82 207 /* Generated by: ../../../genfft/gen_twiddle.native -compact -variables 4 -pipeline-latency 4 -n 7 -name t1_7 -include dft/scalar/t.h */
Chris@82 208
Chris@82 209 /*
Chris@82 210 * This function contains 72 FP additions, 60 FP multiplications,
Chris@82 211 * (or, 36 additions, 24 multiplications, 36 fused multiply/add),
Chris@82 212 * 29 stack variables, 6 constants, and 28 memory accesses
Chris@82 213 */
Chris@82 214 #include "dft/scalar/t.h"
Chris@82 215
Chris@82 216 static void t1_7(R *ri, R *ii, const R *W, stride rs, INT mb, INT me, INT ms)
Chris@82 217 {
Chris@82 218 DK(KP222520933, +0.222520933956314404288902564496794759466355569);
Chris@82 219 DK(KP900968867, +0.900968867902419126236102319507445051165919162);
Chris@82 220 DK(KP623489801, +0.623489801858733530525004884004239810632274731);
Chris@82 221 DK(KP433883739, +0.433883739117558120475768332848358754609990728);
Chris@82 222 DK(KP781831482, +0.781831482468029808708444526674057750232334519);
Chris@82 223 DK(KP974927912, +0.974927912181823607018131682993931217232785801);
Chris@82 224 {
Chris@82 225 INT m;
Chris@82 226 for (m = mb, W = W + (mb * 12); m < me; m = m + 1, ri = ri + ms, ii = ii + ms, W = W + 12, MAKE_VOLATILE_STRIDE(14, rs)) {
Chris@82 227 E T1, TR, Tc, TS, TC, TO, Tn, TT, TI, TP, Ty, TU, TF, TQ;
Chris@82 228 T1 = ri[0];
Chris@82 229 TR = ii[0];
Chris@82 230 {
Chris@82 231 E T6, TA, Tb, TB;
Chris@82 232 {
Chris@82 233 E T3, T5, T2, T4;
Chris@82 234 T3 = ri[WS(rs, 1)];
Chris@82 235 T5 = ii[WS(rs, 1)];
Chris@82 236 T2 = W[0];
Chris@82 237 T4 = W[1];
Chris@82 238 T6 = FMA(T2, T3, T4 * T5);
Chris@82 239 TA = FNMS(T4, T3, T2 * T5);
Chris@82 240 }
Chris@82 241 {
Chris@82 242 E T8, Ta, T7, T9;
Chris@82 243 T8 = ri[WS(rs, 6)];
Chris@82 244 Ta = ii[WS(rs, 6)];
Chris@82 245 T7 = W[10];
Chris@82 246 T9 = W[11];
Chris@82 247 Tb = FMA(T7, T8, T9 * Ta);
Chris@82 248 TB = FNMS(T9, T8, T7 * Ta);
Chris@82 249 }
Chris@82 250 Tc = T6 + Tb;
Chris@82 251 TS = Tb - T6;
Chris@82 252 TC = TA - TB;
Chris@82 253 TO = TA + TB;
Chris@82 254 }
Chris@82 255 {
Chris@82 256 E Th, TG, Tm, TH;
Chris@82 257 {
Chris@82 258 E Te, Tg, Td, Tf;
Chris@82 259 Te = ri[WS(rs, 2)];
Chris@82 260 Tg = ii[WS(rs, 2)];
Chris@82 261 Td = W[2];
Chris@82 262 Tf = W[3];
Chris@82 263 Th = FMA(Td, Te, Tf * Tg);
Chris@82 264 TG = FNMS(Tf, Te, Td * Tg);
Chris@82 265 }
Chris@82 266 {
Chris@82 267 E Tj, Tl, Ti, Tk;
Chris@82 268 Tj = ri[WS(rs, 5)];
Chris@82 269 Tl = ii[WS(rs, 5)];
Chris@82 270 Ti = W[8];
Chris@82 271 Tk = W[9];
Chris@82 272 Tm = FMA(Ti, Tj, Tk * Tl);
Chris@82 273 TH = FNMS(Tk, Tj, Ti * Tl);
Chris@82 274 }
Chris@82 275 Tn = Th + Tm;
Chris@82 276 TT = Tm - Th;
Chris@82 277 TI = TG - TH;
Chris@82 278 TP = TG + TH;
Chris@82 279 }
Chris@82 280 {
Chris@82 281 E Ts, TD, Tx, TE;
Chris@82 282 {
Chris@82 283 E Tp, Tr, To, Tq;
Chris@82 284 Tp = ri[WS(rs, 3)];
Chris@82 285 Tr = ii[WS(rs, 3)];
Chris@82 286 To = W[4];
Chris@82 287 Tq = W[5];
Chris@82 288 Ts = FMA(To, Tp, Tq * Tr);
Chris@82 289 TD = FNMS(Tq, Tp, To * Tr);
Chris@82 290 }
Chris@82 291 {
Chris@82 292 E Tu, Tw, Tt, Tv;
Chris@82 293 Tu = ri[WS(rs, 4)];
Chris@82 294 Tw = ii[WS(rs, 4)];
Chris@82 295 Tt = W[6];
Chris@82 296 Tv = W[7];
Chris@82 297 Tx = FMA(Tt, Tu, Tv * Tw);
Chris@82 298 TE = FNMS(Tv, Tu, Tt * Tw);
Chris@82 299 }
Chris@82 300 Ty = Ts + Tx;
Chris@82 301 TU = Tx - Ts;
Chris@82 302 TF = TD - TE;
Chris@82 303 TQ = TD + TE;
Chris@82 304 }
Chris@82 305 ri[0] = T1 + Tc + Tn + Ty;
Chris@82 306 ii[0] = TO + TP + TQ + TR;
Chris@82 307 {
Chris@82 308 E TJ, Tz, TX, TY;
Chris@82 309 TJ = FNMS(KP781831482, TF, KP974927912 * TC) - (KP433883739 * TI);
Chris@82 310 Tz = FMA(KP623489801, Ty, T1) + FNMA(KP900968867, Tn, KP222520933 * Tc);
Chris@82 311 ri[WS(rs, 5)] = Tz - TJ;
Chris@82 312 ri[WS(rs, 2)] = Tz + TJ;
Chris@82 313 TX = FNMS(KP781831482, TU, KP974927912 * TS) - (KP433883739 * TT);
Chris@82 314 TY = FMA(KP623489801, TQ, TR) + FNMA(KP900968867, TP, KP222520933 * TO);
Chris@82 315 ii[WS(rs, 2)] = TX + TY;
Chris@82 316 ii[WS(rs, 5)] = TY - TX;
Chris@82 317 }
Chris@82 318 {
Chris@82 319 E TL, TK, TV, TW;
Chris@82 320 TL = FMA(KP781831482, TC, KP974927912 * TI) + (KP433883739 * TF);
Chris@82 321 TK = FMA(KP623489801, Tc, T1) + FNMA(KP900968867, Ty, KP222520933 * Tn);
Chris@82 322 ri[WS(rs, 6)] = TK - TL;
Chris@82 323 ri[WS(rs, 1)] = TK + TL;
Chris@82 324 TV = FMA(KP781831482, TS, KP974927912 * TT) + (KP433883739 * TU);
Chris@82 325 TW = FMA(KP623489801, TO, TR) + FNMA(KP900968867, TQ, KP222520933 * TP);
Chris@82 326 ii[WS(rs, 1)] = TV + TW;
Chris@82 327 ii[WS(rs, 6)] = TW - TV;
Chris@82 328 }
Chris@82 329 {
Chris@82 330 E TN, TM, TZ, T10;
Chris@82 331 TN = FMA(KP433883739, TC, KP974927912 * TF) - (KP781831482 * TI);
Chris@82 332 TM = FMA(KP623489801, Tn, T1) + FNMA(KP222520933, Ty, KP900968867 * Tc);
Chris@82 333 ri[WS(rs, 4)] = TM - TN;
Chris@82 334 ri[WS(rs, 3)] = TM + TN;
Chris@82 335 TZ = FMA(KP433883739, TS, KP974927912 * TU) - (KP781831482 * TT);
Chris@82 336 T10 = FMA(KP623489801, TP, TR) + FNMA(KP222520933, TQ, KP900968867 * TO);
Chris@82 337 ii[WS(rs, 3)] = TZ + T10;
Chris@82 338 ii[WS(rs, 4)] = T10 - TZ;
Chris@82 339 }
Chris@82 340 }
Chris@82 341 }
Chris@82 342 }
Chris@82 343
Chris@82 344 static const tw_instr twinstr[] = {
Chris@82 345 {TW_FULL, 0, 7},
Chris@82 346 {TW_NEXT, 1, 0}
Chris@82 347 };
Chris@82 348
Chris@82 349 static const ct_desc desc = { 7, "t1_7", twinstr, &GENUS, {36, 24, 36, 0}, 0, 0, 0 };
Chris@82 350
Chris@82 351 void X(codelet_t1_7) (planner *p) {
Chris@82 352 X(kdft_dit_register) (p, t1_7, &desc);
Chris@82 353 }
Chris@82 354 #endif