Mercurial > hg > sv-dependency-builds

annotate src/fftw-3.3.8/rdft/scalar/r2cb/hc2cbdft2_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 08:08:00 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 -sign 1 -n 8 -dif -name hc2cbdft2_8 -include rdft/scalar/hc2cb.h */
Chris@82 29
Chris@82 30 /*
Chris@82 31 * This function contains 82 FP additions, 36 FP multiplications,
Chris@82 32 * (or, 60 additions, 14 multiplications, 22 fused multiply/add),
Chris@82 33 * 41 stack variables, 1 constants, and 32 memory accesses
Chris@82 34 */
Chris@82 35 #include "rdft/scalar/hc2cb.h"
Chris@82 36
Chris@82 37 static void hc2cbdft2_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 {
Chris@82 41 INT m;
Chris@82 42 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 43 E Tl, T1p, T1g, TM, T1k, TE, TP, T1f, T7, Te, TU, TH, T1l, Tw, T1q;
Chris@82 44 E T1c, T1y;
Chris@82 45 {
Chris@82 46 E T3, TA, Tk, TN, T6, Th, TD, TO, Ta, Tm, Tp, TK, Td, Tr, Tu;
Chris@82 47 E TL, TF, TG;
Chris@82 48 {
Chris@82 49 E T1, T2, Ti, Tj;
Chris@82 50 T1 = Rp[0];
Chris@82 51 T2 = Rm[WS(rs, 3)];
Chris@82 52 T3 = T1 + T2;
Chris@82 53 TA = T1 - T2;
Chris@82 54 Ti = Ip[0];
Chris@82 55 Tj = Im[WS(rs, 3)];
Chris@82 56 Tk = Ti + Tj;
Chris@82 57 TN = Ti - Tj;
Chris@82 58 }
Chris@82 59 {
Chris@82 60 E T4, T5, TB, TC;
Chris@82 61 T4 = Rp[WS(rs, 2)];
Chris@82 62 T5 = Rm[WS(rs, 1)];
Chris@82 63 T6 = T4 + T5;
Chris@82 64 Th = T4 - T5;
Chris@82 65 TB = Ip[WS(rs, 2)];
Chris@82 66 TC = Im[WS(rs, 1)];
Chris@82 67 TD = TB + TC;
Chris@82 68 TO = TB - TC;
Chris@82 69 }
Chris@82 70 {
Chris@82 71 E T8, T9, Tn, To;
Chris@82 72 T8 = Rp[WS(rs, 1)];
Chris@82 73 T9 = Rm[WS(rs, 2)];
Chris@82 74 Ta = T8 + T9;
Chris@82 75 Tm = T8 - T9;
Chris@82 76 Tn = Ip[WS(rs, 1)];
Chris@82 77 To = Im[WS(rs, 2)];
Chris@82 78 Tp = Tn + To;
Chris@82 79 TK = Tn - To;
Chris@82 80 }
Chris@82 81 {
Chris@82 82 E Tb, Tc, Ts, Tt;
Chris@82 83 Tb = Rm[0];
Chris@82 84 Tc = Rp[WS(rs, 3)];
Chris@82 85 Td = Tb + Tc;
Chris@82 86 Tr = Tb - Tc;
Chris@82 87 Ts = Im[0];
Chris@82 88 Tt = Ip[WS(rs, 3)];
Chris@82 89 Tu = Ts + Tt;
Chris@82 90 TL = Tt - Ts;
Chris@82 91 }
Chris@82 92 Tl = Th + Tk;
Chris@82 93 T1p = TA + TD;
Chris@82 94 T1g = TN - TO;
Chris@82 95 TM = TK + TL;
Chris@82 96 T1k = Tk - Th;
Chris@82 97 TE = TA - TD;
Chris@82 98 TP = TN + TO;
Chris@82 99 T1f = Ta - Td;
Chris@82 100 T7 = T3 + T6;
Chris@82 101 Te = Ta + Td;
Chris@82 102 TU = T7 - Te;
Chris@82 103 TF = Tm - Tp;
Chris@82 104 TG = Tr - Tu;
Chris@82 105 TH = TF + TG;
Chris@82 106 T1l = TF - TG;
Chris@82 107 {
Chris@82 108 E Tq, Tv, T1a, T1b;
Chris@82 109 Tq = Tm + Tp;
Chris@82 110 Tv = Tr + Tu;
Chris@82 111 Tw = Tq - Tv;
Chris@82 112 T1q = Tq + Tv;
Chris@82 113 T1a = T3 - T6;
Chris@82 114 T1b = TL - TK;
Chris@82 115 T1c = T1a + T1b;
Chris@82 116 T1y = T1a - T1b;
Chris@82 117 }
Chris@82 118 }
Chris@82 119 {
Chris@82 120 E Tf, TQ, Tx, TI, Ty, TR, Tg, TJ, TS, Tz;
Chris@82 121 Tf = T7 + Te;
Chris@82 122 TQ = TM + TP;
Chris@82 123 Tx = FMA(KP707106781, Tw, Tl);
Chris@82 124 TI = FMA(KP707106781, TH, TE);
Chris@82 125 Tg = W[0];
Chris@82 126 Ty = Tg * Tx;
Chris@82 127 TR = Tg * TI;
Chris@82 128 Tz = W[1];
Chris@82 129 TJ = FMA(Tz, TI, Ty);
Chris@82 130 TS = FNMS(Tz, Tx, TR);
Chris@82 131 Rp[0] = Tf - TJ;
Chris@82 132 Ip[0] = TQ + TS;
Chris@82 133 Rm[0] = Tf + TJ;
Chris@82 134 Im[0] = TS - TQ;
Chris@82 135 }
Chris@82 136 {
Chris@82 137 E T1B, T1A, T1J, T1x, T1z, T1E, T1H, T1F, T1L, T1D;
Chris@82 138 T1B = T1g - T1f;
Chris@82 139 T1A = W[11];
Chris@82 140 T1J = T1A * T1y;
Chris@82 141 T1x = W[10];
Chris@82 142 T1z = T1x * T1y;
Chris@82 143 T1E = FNMS(KP707106781, T1l, T1k);
Chris@82 144 T1H = FMA(KP707106781, T1q, T1p);
Chris@82 145 T1D = W[12];
Chris@82 146 T1F = T1D * T1E;
Chris@82 147 T1L = T1D * T1H;
Chris@82 148 {
Chris@82 149 E T1C, T1K, T1I, T1M, T1G;
Chris@82 150 T1C = FNMS(T1A, T1B, T1z);
Chris@82 151 T1K = FMA(T1x, T1B, T1J);
Chris@82 152 T1G = W[13];
Chris@82 153 T1I = FMA(T1G, T1H, T1F);
Chris@82 154 T1M = FNMS(T1G, T1E, T1L);
Chris@82 155 Rp[WS(rs, 3)] = T1C - T1I;
Chris@82 156 Ip[WS(rs, 3)] = T1K + T1M;
Chris@82 157 Rm[WS(rs, 3)] = T1C + T1I;
Chris@82 158 Im[WS(rs, 3)] = T1M - T1K;
Chris@82 159 }
Chris@82 160 }
Chris@82 161 {
Chris@82 162 E TX, TW, T15, TT, TV, T10, T13, T11, T17, TZ;
Chris@82 163 TX = TP - TM;
Chris@82 164 TW = W[7];
Chris@82 165 T15 = TW * TU;
Chris@82 166 TT = W[6];
Chris@82 167 TV = TT * TU;
Chris@82 168 T10 = FNMS(KP707106781, Tw, Tl);
Chris@82 169 T13 = FNMS(KP707106781, TH, TE);
Chris@82 170 TZ = W[8];
Chris@82 171 T11 = TZ * T10;
Chris@82 172 T17 = TZ * T13;
Chris@82 173 {
Chris@82 174 E TY, T16, T14, T18, T12;
Chris@82 175 TY = FNMS(TW, TX, TV);
Chris@82 176 T16 = FMA(TT, TX, T15);
Chris@82 177 T12 = W[9];
Chris@82 178 T14 = FMA(T12, T13, T11);
Chris@82 179 T18 = FNMS(T12, T10, T17);
Chris@82 180 Rp[WS(rs, 2)] = TY - T14;
Chris@82 181 Ip[WS(rs, 2)] = T16 + T18;
Chris@82 182 Rm[WS(rs, 2)] = TY + T14;
Chris@82 183 Im[WS(rs, 2)] = T18 - T16;
Chris@82 184 }
Chris@82 185 }
Chris@82 186 {
Chris@82 187 E T1h, T1e, T1t, T19, T1d, T1m, T1r, T1n, T1v, T1j;
Chris@82 188 T1h = T1f + T1g;
Chris@82 189 T1e = W[3];
Chris@82 190 T1t = T1e * T1c;
Chris@82 191 T19 = W[2];
Chris@82 192 T1d = T19 * T1c;
Chris@82 193 T1m = FMA(KP707106781, T1l, T1k);
Chris@82 194 T1r = FNMS(KP707106781, T1q, T1p);
Chris@82 195 T1j = W[4];
Chris@82 196 T1n = T1j * T1m;
Chris@82 197 T1v = T1j * T1r;
Chris@82 198 {
Chris@82 199 E T1i, T1u, T1s, T1w, T1o;
Chris@82 200 T1i = FNMS(T1e, T1h, T1d);
Chris@82 201 T1u = FMA(T19, T1h, T1t);
Chris@82 202 T1o = W[5];
Chris@82 203 T1s = FMA(T1o, T1r, T1n);
Chris@82 204 T1w = FNMS(T1o, T1m, T1v);
Chris@82 205 Rp[WS(rs, 1)] = T1i - T1s;
Chris@82 206 Ip[WS(rs, 1)] = T1u + T1w;
Chris@82 207 Rm[WS(rs, 1)] = T1i + T1s;
Chris@82 208 Im[WS(rs, 1)] = T1w - T1u;
Chris@82 209 }
Chris@82 210 }
Chris@82 211 }
Chris@82 212 }
Chris@82 213 }
Chris@82 214
Chris@82 215 static const tw_instr twinstr[] = {
Chris@82 216 {TW_FULL, 1, 8},
Chris@82 217 {TW_NEXT, 1, 0}
Chris@82 218 };
Chris@82 219
Chris@82 220 static const hc2c_desc desc = { 8, "hc2cbdft2_8", twinstr, &GENUS, {60, 14, 22, 0} };
Chris@82 221
Chris@82 222 void X(codelet_hc2cbdft2_8) (planner *p) {
Chris@82 223 X(khc2c_register) (p, hc2cbdft2_8, &desc, HC2C_VIA_DFT);
Chris@82 224 }
Chris@82 225 #else
Chris@82 226
Chris@82 227 /* Generated by: ../../../genfft/gen_hc2cdft.native -compact -variables 4 -pipeline-latency 4 -sign 1 -n 8 -dif -name hc2cbdft2_8 -include rdft/scalar/hc2cb.h */
Chris@82 228
Chris@82 229 /*
Chris@82 230 * This function contains 82 FP additions, 32 FP multiplications,
Chris@82 231 * (or, 68 additions, 18 multiplications, 14 fused multiply/add),
Chris@82 232 * 30 stack variables, 1 constants, and 32 memory accesses
Chris@82 233 */
Chris@82 234 #include "rdft/scalar/hc2cb.h"
Chris@82 235
Chris@82 236 static void hc2cbdft2_8(R *Rp, R *Ip, R *Rm, R *Im, const R *W, stride rs, INT mb, INT me, INT ms)
Chris@82 237 {
Chris@82 238 DK(KP707106781, +0.707106781186547524400844362104849039284835938);
Chris@82 239 {
Chris@82 240 INT m;
Chris@82 241 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@82 242 E T7, T1d, T1h, Tl, TG, T14, T19, TO, Te, TL, T18, T15, TB, T1e, Tw;
Chris@82 243 E T1i;
Chris@82 244 {
Chris@82 245 E T3, TC, Tk, TM, T6, Th, TF, TN;
Chris@82 246 {
Chris@82 247 E T1, T2, Ti, Tj;
Chris@82 248 T1 = Rp[0];
Chris@82 249 T2 = Rm[WS(rs, 3)];
Chris@82 250 T3 = T1 + T2;
Chris@82 251 TC = T1 - T2;
Chris@82 252 Ti = Ip[0];
Chris@82 253 Tj = Im[WS(rs, 3)];
Chris@82 254 Tk = Ti + Tj;
Chris@82 255 TM = Ti - Tj;
Chris@82 256 }
Chris@82 257 {
Chris@82 258 E T4, T5, TD, TE;
Chris@82 259 T4 = Rp[WS(rs, 2)];
Chris@82 260 T5 = Rm[WS(rs, 1)];
Chris@82 261 T6 = T4 + T5;
Chris@82 262 Th = T4 - T5;
Chris@82 263 TD = Ip[WS(rs, 2)];
Chris@82 264 TE = Im[WS(rs, 1)];
Chris@82 265 TF = TD + TE;
Chris@82 266 TN = TD - TE;
Chris@82 267 }
Chris@82 268 T7 = T3 + T6;
Chris@82 269 T1d = Tk - Th;
Chris@82 270 T1h = TC + TF;
Chris@82 271 Tl = Th + Tk;
Chris@82 272 TG = TC - TF;
Chris@82 273 T14 = T3 - T6;
Chris@82 274 T19 = TM - TN;
Chris@82 275 TO = TM + TN;
Chris@82 276 }
Chris@82 277 {
Chris@82 278 E Ta, Tm, Tp, TJ, Td, Tr, Tu, TK;
Chris@82 279 {
Chris@82 280 E T8, T9, Tn, To;
Chris@82 281 T8 = Rp[WS(rs, 1)];
Chris@82 282 T9 = Rm[WS(rs, 2)];
Chris@82 283 Ta = T8 + T9;
Chris@82 284 Tm = T8 - T9;
Chris@82 285 Tn = Ip[WS(rs, 1)];
Chris@82 286 To = Im[WS(rs, 2)];
Chris@82 287 Tp = Tn + To;
Chris@82 288 TJ = Tn - To;
Chris@82 289 }
Chris@82 290 {
Chris@82 291 E Tb, Tc, Ts, Tt;
Chris@82 292 Tb = Rm[0];
Chris@82 293 Tc = Rp[WS(rs, 3)];
Chris@82 294 Td = Tb + Tc;
Chris@82 295 Tr = Tb - Tc;
Chris@82 296 Ts = Im[0];
Chris@82 297 Tt = Ip[WS(rs, 3)];
Chris@82 298 Tu = Ts + Tt;
Chris@82 299 TK = Tt - Ts;
Chris@82 300 }
Chris@82 301 Te = Ta + Td;
Chris@82 302 TL = TJ + TK;
Chris@82 303 T18 = Ta - Td;
Chris@82 304 T15 = TK - TJ;
Chris@82 305 {
Chris@82 306 E Tz, TA, Tq, Tv;
Chris@82 307 Tz = Tm - Tp;
Chris@82 308 TA = Tr - Tu;
Chris@82 309 TB = KP707106781 * (Tz + TA);
Chris@82 310 T1e = KP707106781 * (Tz - TA);
Chris@82 311 Tq = Tm + Tp;
Chris@82 312 Tv = Tr + Tu;
Chris@82 313 Tw = KP707106781 * (Tq - Tv);
Chris@82 314 T1i = KP707106781 * (Tq + Tv);
Chris@82 315 }
Chris@82 316 }
Chris@82 317 {
Chris@82 318 E Tf, TP, TI, TQ;
Chris@82 319 Tf = T7 + Te;
Chris@82 320 TP = TL + TO;
Chris@82 321 {
Chris@82 322 E Tx, TH, Tg, Ty;
Chris@82 323 Tx = Tl + Tw;
Chris@82 324 TH = TB + TG;
Chris@82 325 Tg = W[0];
Chris@82 326 Ty = W[1];
Chris@82 327 TI = FMA(Tg, Tx, Ty * TH);
Chris@82 328 TQ = FNMS(Ty, Tx, Tg * TH);
Chris@82 329 }
Chris@82 330 Rp[0] = Tf - TI;
Chris@82 331 Ip[0] = TP + TQ;
Chris@82 332 Rm[0] = Tf + TI;
Chris@82 333 Im[0] = TQ - TP;
Chris@82 334 }
Chris@82 335 {
Chris@82 336 E T1r, T1x, T1w, T1y;
Chris@82 337 {
Chris@82 338 E T1o, T1q, T1n, T1p;
Chris@82 339 T1o = T14 - T15;
Chris@82 340 T1q = T19 - T18;
Chris@82 341 T1n = W[10];
Chris@82 342 T1p = W[11];
Chris@82 343 T1r = FNMS(T1p, T1q, T1n * T1o);
Chris@82 344 T1x = FMA(T1p, T1o, T1n * T1q);
Chris@82 345 }
Chris@82 346 {
Chris@82 347 E T1t, T1v, T1s, T1u;
Chris@82 348 T1t = T1d - T1e;
Chris@82 349 T1v = T1i + T1h;
Chris@82 350 T1s = W[12];
Chris@82 351 T1u = W[13];
Chris@82 352 T1w = FMA(T1s, T1t, T1u * T1v);
Chris@82 353 T1y = FNMS(T1u, T1t, T1s * T1v);
Chris@82 354 }
Chris@82 355 Rp[WS(rs, 3)] = T1r - T1w;
Chris@82 356 Ip[WS(rs, 3)] = T1x + T1y;
Chris@82 357 Rm[WS(rs, 3)] = T1r + T1w;
Chris@82 358 Im[WS(rs, 3)] = T1y - T1x;
Chris@82 359 }
Chris@82 360 {
Chris@82 361 E TV, T11, T10, T12;
Chris@82 362 {
Chris@82 363 E TS, TU, TR, TT;
Chris@82 364 TS = T7 - Te;
Chris@82 365 TU = TO - TL;
Chris@82 366 TR = W[6];
Chris@82 367 TT = W[7];
Chris@82 368 TV = FNMS(TT, TU, TR * TS);
Chris@82 369 T11 = FMA(TT, TS, TR * TU);
Chris@82 370 }
Chris@82 371 {
Chris@82 372 E TX, TZ, TW, TY;
Chris@82 373 TX = Tl - Tw;
Chris@82 374 TZ = TG - TB;
Chris@82 375 TW = W[8];
Chris@82 376 TY = W[9];
Chris@82 377 T10 = FMA(TW, TX, TY * TZ);
Chris@82 378 T12 = FNMS(TY, TX, TW * TZ);
Chris@82 379 }
Chris@82 380 Rp[WS(rs, 2)] = TV - T10;
Chris@82 381 Ip[WS(rs, 2)] = T11 + T12;
Chris@82 382 Rm[WS(rs, 2)] = TV + T10;
Chris@82 383 Im[WS(rs, 2)] = T12 - T11;
Chris@82 384 }
Chris@82 385 {
Chris@82 386 E T1b, T1l, T1k, T1m;
Chris@82 387 {
Chris@82 388 E T16, T1a, T13, T17;
Chris@82 389 T16 = T14 + T15;
Chris@82 390 T1a = T18 + T19;
Chris@82 391 T13 = W[2];
Chris@82 392 T17 = W[3];
Chris@82 393 T1b = FNMS(T17, T1a, T13 * T16);
Chris@82 394 T1l = FMA(T17, T16, T13 * T1a);
Chris@82 395 }
Chris@82 396 {
Chris@82 397 E T1f, T1j, T1c, T1g;
Chris@82 398 T1f = T1d + T1e;
Chris@82 399 T1j = T1h - T1i;
Chris@82 400 T1c = W[4];
Chris@82 401 T1g = W[5];
Chris@82 402 T1k = FMA(T1c, T1f, T1g * T1j);
Chris@82 403 T1m = FNMS(T1g, T1f, T1c * T1j);
Chris@82 404 }
Chris@82 405 Rp[WS(rs, 1)] = T1b - T1k;
Chris@82 406 Ip[WS(rs, 1)] = T1l + T1m;
Chris@82 407 Rm[WS(rs, 1)] = T1b + T1k;
Chris@82 408 Im[WS(rs, 1)] = T1m - T1l;
Chris@82 409 }
Chris@82 410 }
Chris@82 411 }
Chris@82 412 }
Chris@82 413
Chris@82 414 static const tw_instr twinstr[] = {
Chris@82 415 {TW_FULL, 1, 8},
Chris@82 416 {TW_NEXT, 1, 0}
Chris@82 417 };
Chris@82 418
Chris@82 419 static const hc2c_desc desc = { 8, "hc2cbdft2_8", twinstr, &GENUS, {68, 18, 14, 0} };
Chris@82 420
Chris@82 421 void X(codelet_hc2cbdft2_8) (planner *p) {
Chris@82 422 X(khc2c_register) (p, hc2cbdft2_8, &desc, HC2C_VIA_DFT);
Chris@82 423 }
Chris@82 424 #endif