Mercurial > hg > sv-dependency-builds

annotate src/fftw-3.3.8/rdft/scalar/r2cf/hc2cfdft_8.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 08:07:10 EDT 2018 */
Chris@82 23
Chris@82 24 #include "rdft/codelet-rdft.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_hc2cdft.native -fma -compact -variables 4 -pipeline-latency 4 -n 8 -dit -name hc2cfdft_8 -include rdft/scalar/hc2cf.h */
Chris@82 29
Chris@82 30 /*
Chris@82 31 * This function contains 82 FP additions, 52 FP multiplications,
Chris@82 32 * (or, 60 additions, 30 multiplications, 22 fused multiply/add),
Chris@82 33 * 31 stack variables, 2 constants, and 32 memory accesses
Chris@82 34 */
Chris@82 35 #include "rdft/scalar/hc2cf.h"
Chris@82 36
Chris@82 37 static void hc2cfdft_8(R *Rp, R *Ip, R *Rm, R *Im, const R *W, stride rs, INT mb, INT me, INT ms)
Chris@82 38 {
Chris@82 39 DK(KP707106781, +0.707106781186547524400844362104849039284835938);
Chris@82 40 DK(KP500000000, +0.500000000000000000000000000000000000000000000);
Chris@82 41 {
Chris@82 42 INT m;
Chris@82 43 for (m = mb, W = W + ((mb - 1) * 14); m < me; m = m + 1, Rp = Rp + ms, Ip = Ip + ms, Rm = Rm - ms, Im = Im - ms, W = W + 14, MAKE_VOLATILE_STRIDE(32, rs)) {
Chris@82 44 E Ty, T14, TO, T1o, Tv, T16, TG, T1m, Ta, T19, TV, T1h, Tk, T1b, T11;
Chris@82 45 E T1j;
Chris@82 46 {
Chris@82 47 E Tw, Tx, TN, TI, TJ, TK;
Chris@82 48 Tw = Ip[0];
Chris@82 49 Tx = Im[0];
Chris@82 50 TN = Tw + Tx;
Chris@82 51 TI = Rm[0];
Chris@82 52 TJ = Rp[0];
Chris@82 53 TK = TI - TJ;
Chris@82 54 Ty = Tw - Tx;
Chris@82 55 T14 = TJ + TI;
Chris@82 56 {
Chris@82 57 E TH, TL, TM, T1n;
Chris@82 58 TH = W[0];
Chris@82 59 TL = TH * TK;
Chris@82 60 TM = W[1];
Chris@82 61 T1n = TM * TK;
Chris@82 62 TO = FNMS(TM, TN, TL);
Chris@82 63 T1o = FMA(TH, TN, T1n);
Chris@82 64 }
Chris@82 65 }
Chris@82 66 {
Chris@82 67 E Tp, TF, Tu, TC;
Chris@82 68 {
Chris@82 69 E Tn, To, Ts, Tt;
Chris@82 70 Tn = Ip[WS(rs, 2)];
Chris@82 71 To = Im[WS(rs, 2)];
Chris@82 72 Tp = Tn - To;
Chris@82 73 TF = Tn + To;
Chris@82 74 Ts = Rp[WS(rs, 2)];
Chris@82 75 Tt = Rm[WS(rs, 2)];
Chris@82 76 Tu = Ts + Tt;
Chris@82 77 TC = Tt - Ts;
Chris@82 78 }
Chris@82 79 {
Chris@82 80 E Tq, T15, Tm, Tr;
Chris@82 81 Tm = W[6];
Chris@82 82 Tq = Tm * Tp;
Chris@82 83 T15 = Tm * Tu;
Chris@82 84 Tr = W[7];
Chris@82 85 Tv = FNMS(Tr, Tu, Tq);
Chris@82 86 T16 = FMA(Tr, Tp, T15);
Chris@82 87 }
Chris@82 88 {
Chris@82 89 E TB, TD, TE, T1l;
Chris@82 90 TB = W[8];
Chris@82 91 TD = TB * TC;
Chris@82 92 TE = W[9];
Chris@82 93 T1l = TE * TC;
Chris@82 94 TG = FNMS(TE, TF, TD);
Chris@82 95 T1m = FMA(TB, TF, T1l);
Chris@82 96 }
Chris@82 97 }
Chris@82 98 {
Chris@82 99 E T4, TU, T9, TR;
Chris@82 100 {
Chris@82 101 E T2, T3, T7, T8;
Chris@82 102 T2 = Ip[WS(rs, 1)];
Chris@82 103 T3 = Im[WS(rs, 1)];
Chris@82 104 T4 = T2 - T3;
Chris@82 105 TU = T2 + T3;
Chris@82 106 T7 = Rp[WS(rs, 1)];
Chris@82 107 T8 = Rm[WS(rs, 1)];
Chris@82 108 T9 = T7 + T8;
Chris@82 109 TR = T7 - T8;
Chris@82 110 }
Chris@82 111 {
Chris@82 112 E T5, T18, T1, T6;
Chris@82 113 T1 = W[2];
Chris@82 114 T5 = T1 * T4;
Chris@82 115 T18 = T1 * T9;
Chris@82 116 T6 = W[3];
Chris@82 117 Ta = FNMS(T6, T9, T5);
Chris@82 118 T19 = FMA(T6, T4, T18);
Chris@82 119 }
Chris@82 120 {
Chris@82 121 E TS, T1g, TQ, TT;
Chris@82 122 TQ = W[4];
Chris@82 123 TS = TQ * TR;
Chris@82 124 T1g = TQ * TU;
Chris@82 125 TT = W[5];
Chris@82 126 TV = FMA(TT, TU, TS);
Chris@82 127 T1h = FNMS(TT, TR, T1g);
Chris@82 128 }
Chris@82 129 }
Chris@82 130 {
Chris@82 131 E Te, T10, Tj, TX;
Chris@82 132 {
Chris@82 133 E Tc, Td, Th, Ti;
Chris@82 134 Tc = Ip[WS(rs, 3)];
Chris@82 135 Td = Im[WS(rs, 3)];
Chris@82 136 Te = Tc - Td;
Chris@82 137 T10 = Tc + Td;
Chris@82 138 Th = Rp[WS(rs, 3)];
Chris@82 139 Ti = Rm[WS(rs, 3)];
Chris@82 140 Tj = Th + Ti;
Chris@82 141 TX = Th - Ti;
Chris@82 142 }
Chris@82 143 {
Chris@82 144 E Tf, T1a, Tb, Tg;
Chris@82 145 Tb = W[10];
Chris@82 146 Tf = Tb * Te;
Chris@82 147 T1a = Tb * Tj;
Chris@82 148 Tg = W[11];
Chris@82 149 Tk = FNMS(Tg, Tj, Tf);
Chris@82 150 T1b = FMA(Tg, Te, T1a);
Chris@82 151 }
Chris@82 152 {
Chris@82 153 E TY, T1i, TW, TZ;
Chris@82 154 TW = W[12];
Chris@82 155 TY = TW * TX;
Chris@82 156 T1i = TW * T10;
Chris@82 157 TZ = W[13];
Chris@82 158 T11 = FMA(TZ, T10, TY);
Chris@82 159 T1j = FNMS(TZ, TX, T1i);
Chris@82 160 }
Chris@82 161 }
Chris@82 162 {
Chris@82 163 E TA, T1f, T1q, T1s, T13, T1e, T1d, T1r;
Chris@82 164 {
Chris@82 165 E Tl, Tz, T1k, T1p;
Chris@82 166 Tl = Ta + Tk;
Chris@82 167 Tz = Tv + Ty;
Chris@82 168 TA = Tl + Tz;
Chris@82 169 T1f = Tz - Tl;
Chris@82 170 T1k = T1h + T1j;
Chris@82 171 T1p = T1m + T1o;
Chris@82 172 T1q = T1k - T1p;
Chris@82 173 T1s = T1k + T1p;
Chris@82 174 }
Chris@82 175 {
Chris@82 176 E TP, T12, T17, T1c;
Chris@82 177 TP = TG + TO;
Chris@82 178 T12 = TV + T11;
Chris@82 179 T13 = TP - T12;
Chris@82 180 T1e = T12 + TP;
Chris@82 181 T17 = T14 + T16;
Chris@82 182 T1c = T19 + T1b;
Chris@82 183 T1d = T17 - T1c;
Chris@82 184 T1r = T17 + T1c;
Chris@82 185 }
Chris@82 186 Ip[0] = KP500000000 * (TA + T13);
Chris@82 187 Rp[0] = KP500000000 * (T1r + T1s);
Chris@82 188 Im[WS(rs, 3)] = KP500000000 * (T13 - TA);
Chris@82 189 Rm[WS(rs, 3)] = KP500000000 * (T1r - T1s);
Chris@82 190 Rm[WS(rs, 1)] = KP500000000 * (T1d - T1e);
Chris@82 191 Im[WS(rs, 1)] = KP500000000 * (T1q - T1f);
Chris@82 192 Rp[WS(rs, 2)] = KP500000000 * (T1d + T1e);
Chris@82 193 Ip[WS(rs, 2)] = KP500000000 * (T1f + T1q);
Chris@82 194 }
Chris@82 195 {
Chris@82 196 E T1v, T1H, T1F, T1L, T1y, T1I, T1B, T1J;
Chris@82 197 {
Chris@82 198 E T1t, T1u, T1D, T1E;
Chris@82 199 T1t = Ty - Tv;
Chris@82 200 T1u = T19 - T1b;
Chris@82 201 T1v = T1t - T1u;
Chris@82 202 T1H = T1u + T1t;
Chris@82 203 T1D = T14 - T16;
Chris@82 204 T1E = Ta - Tk;
Chris@82 205 T1F = T1D - T1E;
Chris@82 206 T1L = T1D + T1E;
Chris@82 207 }
Chris@82 208 {
Chris@82 209 E T1w, T1x, T1z, T1A;
Chris@82 210 T1w = T1j - T1h;
Chris@82 211 T1x = TV - T11;
Chris@82 212 T1y = T1w + T1x;
Chris@82 213 T1I = T1w - T1x;
Chris@82 214 T1z = TO - TG;
Chris@82 215 T1A = T1o - T1m;
Chris@82 216 T1B = T1z - T1A;
Chris@82 217 T1J = T1z + T1A;
Chris@82 218 }
Chris@82 219 {
Chris@82 220 E T1C, T1M, T1G, T1K;
Chris@82 221 T1C = T1y + T1B;
Chris@82 222 Ip[WS(rs, 1)] = KP500000000 * (FMA(KP707106781, T1C, T1v));
Chris@82 223 Im[WS(rs, 2)] = -(KP500000000 * (FNMS(KP707106781, T1C, T1v)));
Chris@82 224 T1M = T1I + T1J;
Chris@82 225 Rm[WS(rs, 2)] = KP500000000 * (FNMS(KP707106781, T1M, T1L));
Chris@82 226 Rp[WS(rs, 1)] = KP500000000 * (FMA(KP707106781, T1M, T1L));
Chris@82 227 T1G = T1B - T1y;
Chris@82 228 Rm[0] = KP500000000 * (FNMS(KP707106781, T1G, T1F));
Chris@82 229 Rp[WS(rs, 3)] = KP500000000 * (FMA(KP707106781, T1G, T1F));
Chris@82 230 T1K = T1I - T1J;
Chris@82 231 Ip[WS(rs, 3)] = KP500000000 * (FMA(KP707106781, T1K, T1H));
Chris@82 232 Im[0] = -(KP500000000 * (FNMS(KP707106781, T1K, T1H)));
Chris@82 233 }
Chris@82 234 }
Chris@82 235 }
Chris@82 236 }
Chris@82 237 }
Chris@82 238
Chris@82 239 static const tw_instr twinstr[] = {
Chris@82 240 {TW_FULL, 1, 8},
Chris@82 241 {TW_NEXT, 1, 0}
Chris@82 242 };
Chris@82 243
Chris@82 244 static const hc2c_desc desc = { 8, "hc2cfdft_8", twinstr, &GENUS, {60, 30, 22, 0} };
Chris@82 245
Chris@82 246 void X(codelet_hc2cfdft_8) (planner *p) {
Chris@82 247 X(khc2c_register) (p, hc2cfdft_8, &desc, HC2C_VIA_DFT);
Chris@82 248 }
Chris@82 249 #else
Chris@82 250
Chris@82 251 /* Generated by: ../../../genfft/gen_hc2cdft.native -compact -variables 4 -pipeline-latency 4 -n 8 -dit -name hc2cfdft_8 -include rdft/scalar/hc2cf.h */
Chris@82 252
Chris@82 253 /*
Chris@82 254 * This function contains 82 FP additions, 44 FP multiplications,
Chris@82 255 * (or, 68 additions, 30 multiplications, 14 fused multiply/add),
Chris@82 256 * 39 stack variables, 2 constants, and 32 memory accesses
Chris@82 257 */
Chris@82 258 #include "rdft/scalar/hc2cf.h"
Chris@82 259
Chris@82 260 static void hc2cfdft_8(R *Rp, R *Ip, R *Rm, R *Im, const R *W, stride rs, INT mb, INT me, INT ms)
Chris@82 261 {
Chris@82 262 DK(KP353553390, +0.353553390593273762200422181052424519642417969);
Chris@82 263 DK(KP500000000, +0.500000000000000000000000000000000000000000000);
Chris@82 264 {
Chris@82 265 INT m;
Chris@82 266 for (m = mb, W = W + ((mb - 1) * 14); m < me; m = m + 1, Rp = Rp + ms, Ip = Ip + ms, Rm = Rm - ms, Im = Im - ms, W = W + 14, MAKE_VOLATILE_STRIDE(32, rs)) {
Chris@82 267 E Tv, TX, Ts, TY, TE, T1a, TJ, T19, T1l, T1m, T9, T10, Ti, T11, TP;
Chris@82 268 E T16, TU, T17, T1i, T1j;
Chris@82 269 {
Chris@82 270 E Tt, Tu, TD, Tz, TA, TB, Tn, TI, Tr, TG, Tk, To;
Chris@82 271 Tt = Ip[0];
Chris@82 272 Tu = Im[0];
Chris@82 273 TD = Tt + Tu;
Chris@82 274 Tz = Rm[0];
Chris@82 275 TA = Rp[0];
Chris@82 276 TB = Tz - TA;
Chris@82 277 {
Chris@82 278 E Tl, Tm, Tp, Tq;
Chris@82 279 Tl = Ip[WS(rs, 2)];
Chris@82 280 Tm = Im[WS(rs, 2)];
Chris@82 281 Tn = Tl - Tm;
Chris@82 282 TI = Tl + Tm;
Chris@82 283 Tp = Rp[WS(rs, 2)];
Chris@82 284 Tq = Rm[WS(rs, 2)];
Chris@82 285 Tr = Tp + Tq;
Chris@82 286 TG = Tp - Tq;
Chris@82 287 }
Chris@82 288 Tv = Tt - Tu;
Chris@82 289 TX = TA + Tz;
Chris@82 290 Tk = W[6];
Chris@82 291 To = W[7];
Chris@82 292 Ts = FNMS(To, Tr, Tk * Tn);
Chris@82 293 TY = FMA(Tk, Tr, To * Tn);
Chris@82 294 {
Chris@82 295 E Ty, TC, TF, TH;
Chris@82 296 Ty = W[0];
Chris@82 297 TC = W[1];
Chris@82 298 TE = FNMS(TC, TD, Ty * TB);
Chris@82 299 T1a = FMA(TC, TB, Ty * TD);
Chris@82 300 TF = W[8];
Chris@82 301 TH = W[9];
Chris@82 302 TJ = FMA(TF, TG, TH * TI);
Chris@82 303 T19 = FNMS(TH, TG, TF * TI);
Chris@82 304 }
Chris@82 305 T1l = TJ + TE;
Chris@82 306 T1m = T1a - T19;
Chris@82 307 }
Chris@82 308 {
Chris@82 309 E T4, TO, T8, TM, Td, TT, Th, TR;
Chris@82 310 {
Chris@82 311 E T2, T3, T6, T7;
Chris@82 312 T2 = Ip[WS(rs, 1)];
Chris@82 313 T3 = Im[WS(rs, 1)];
Chris@82 314 T4 = T2 - T3;
Chris@82 315 TO = T2 + T3;
Chris@82 316 T6 = Rp[WS(rs, 1)];
Chris@82 317 T7 = Rm[WS(rs, 1)];
Chris@82 318 T8 = T6 + T7;
Chris@82 319 TM = T6 - T7;
Chris@82 320 }
Chris@82 321 {
Chris@82 322 E Tb, Tc, Tf, Tg;
Chris@82 323 Tb = Ip[WS(rs, 3)];
Chris@82 324 Tc = Im[WS(rs, 3)];
Chris@82 325 Td = Tb - Tc;
Chris@82 326 TT = Tb + Tc;
Chris@82 327 Tf = Rp[WS(rs, 3)];
Chris@82 328 Tg = Rm[WS(rs, 3)];
Chris@82 329 Th = Tf + Tg;
Chris@82 330 TR = Tf - Tg;
Chris@82 331 }
Chris@82 332 {
Chris@82 333 E T1, T5, Ta, Te;
Chris@82 334 T1 = W[2];
Chris@82 335 T5 = W[3];
Chris@82 336 T9 = FNMS(T5, T8, T1 * T4);
Chris@82 337 T10 = FMA(T1, T8, T5 * T4);
Chris@82 338 Ta = W[10];
Chris@82 339 Te = W[11];
Chris@82 340 Ti = FNMS(Te, Th, Ta * Td);
Chris@82 341 T11 = FMA(Ta, Th, Te * Td);
Chris@82 342 {
Chris@82 343 E TL, TN, TQ, TS;
Chris@82 344 TL = W[4];
Chris@82 345 TN = W[5];
Chris@82 346 TP = FMA(TL, TM, TN * TO);
Chris@82 347 T16 = FNMS(TN, TM, TL * TO);
Chris@82 348 TQ = W[12];
Chris@82 349 TS = W[13];
Chris@82 350 TU = FMA(TQ, TR, TS * TT);
Chris@82 351 T17 = FNMS(TS, TR, TQ * TT);
Chris@82 352 }
Chris@82 353 T1i = T17 - T16;
Chris@82 354 T1j = TP - TU;
Chris@82 355 }
Chris@82 356 }
Chris@82 357 {
Chris@82 358 E T1h, T1t, T1w, T1y, T1o, T1s, T1r, T1x;
Chris@82 359 {
Chris@82 360 E T1f, T1g, T1u, T1v;
Chris@82 361 T1f = Tv - Ts;
Chris@82 362 T1g = T10 - T11;
Chris@82 363 T1h = KP500000000 * (T1f - T1g);
Chris@82 364 T1t = KP500000000 * (T1g + T1f);
Chris@82 365 T1u = T1i - T1j;
Chris@82 366 T1v = T1l + T1m;
Chris@82 367 T1w = KP353553390 * (T1u - T1v);
Chris@82 368 T1y = KP353553390 * (T1u + T1v);
Chris@82 369 }
Chris@82 370 {
Chris@82 371 E T1k, T1n, T1p, T1q;
Chris@82 372 T1k = T1i + T1j;
Chris@82 373 T1n = T1l - T1m;
Chris@82 374 T1o = KP353553390 * (T1k + T1n);
Chris@82 375 T1s = KP353553390 * (T1n - T1k);
Chris@82 376 T1p = TX - TY;
Chris@82 377 T1q = T9 - Ti;
Chris@82 378 T1r = KP500000000 * (T1p - T1q);
Chris@82 379 T1x = KP500000000 * (T1p + T1q);
Chris@82 380 }
Chris@82 381 Ip[WS(rs, 1)] = T1h + T1o;
Chris@82 382 Rp[WS(rs, 1)] = T1x + T1y;
Chris@82 383 Im[WS(rs, 2)] = T1o - T1h;
Chris@82 384 Rm[WS(rs, 2)] = T1x - T1y;
Chris@82 385 Rm[0] = T1r - T1s;
Chris@82 386 Im[0] = T1w - T1t;
Chris@82 387 Rp[WS(rs, 3)] = T1r + T1s;
Chris@82 388 Ip[WS(rs, 3)] = T1t + T1w;
Chris@82 389 }
Chris@82 390 {
Chris@82 391 E Tx, T15, T1c, T1e, TW, T14, T13, T1d;
Chris@82 392 {
Chris@82 393 E Tj, Tw, T18, T1b;
Chris@82 394 Tj = T9 + Ti;
Chris@82 395 Tw = Ts + Tv;
Chris@82 396 Tx = Tj + Tw;
Chris@82 397 T15 = Tw - Tj;
Chris@82 398 T18 = T16 + T17;
Chris@82 399 T1b = T19 + T1a;
Chris@82 400 T1c = T18 - T1b;
Chris@82 401 T1e = T18 + T1b;
Chris@82 402 }
Chris@82 403 {
Chris@82 404 E TK, TV, TZ, T12;
Chris@82 405 TK = TE - TJ;
Chris@82 406 TV = TP + TU;
Chris@82 407 TW = TK - TV;
Chris@82 408 T14 = TV + TK;
Chris@82 409 TZ = TX + TY;
Chris@82 410 T12 = T10 + T11;
Chris@82 411 T13 = TZ - T12;
Chris@82 412 T1d = TZ + T12;
Chris@82 413 }
Chris@82 414 Ip[0] = KP500000000 * (Tx + TW);
Chris@82 415 Rp[0] = KP500000000 * (T1d + T1e);
Chris@82 416 Im[WS(rs, 3)] = KP500000000 * (TW - Tx);
Chris@82 417 Rm[WS(rs, 3)] = KP500000000 * (T1d - T1e);
Chris@82 418 Rm[WS(rs, 1)] = KP500000000 * (T13 - T14);
Chris@82 419 Im[WS(rs, 1)] = KP500000000 * (T1c - T15);
Chris@82 420 Rp[WS(rs, 2)] = KP500000000 * (T13 + T14);
Chris@82 421 Ip[WS(rs, 2)] = KP500000000 * (T15 + T1c);
Chris@82 422 }
Chris@82 423 }
Chris@82 424 }
Chris@82 425 }
Chris@82 426
Chris@82 427 static const tw_instr twinstr[] = {
Chris@82 428 {TW_FULL, 1, 8},
Chris@82 429 {TW_NEXT, 1, 0}
Chris@82 430 };
Chris@82 431
Chris@82 432 static const hc2c_desc desc = { 8, "hc2cfdft_8", twinstr, &GENUS, {68, 30, 14, 0} };
Chris@82 433
Chris@82 434 void X(codelet_hc2cfdft_8) (planner *p) {
Chris@82 435 X(khc2c_register) (p, hc2cfdft_8, &desc, HC2C_VIA_DFT);
Chris@82 436 }
Chris@82 437 #endif