Mercurial > hg > sv-dependency-builds

annotate src/fftw-3.3.8/dft/scalar/codelets/n1_12.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 08:04:10 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_notw.native -fma -compact -variables 4 -pipeline-latency 4 -n 12 -name n1_12 -include dft/scalar/n.h */
Chris@82 29
Chris@82 30 /*
Chris@82 31 * This function contains 96 FP additions, 24 FP multiplications,
Chris@82 32 * (or, 72 additions, 0 multiplications, 24 fused multiply/add),
Chris@82 33 * 43 stack variables, 2 constants, and 48 memory accesses
Chris@82 34 */
Chris@82 35 #include "dft/scalar/n.h"
Chris@82 36
Chris@82 37 static void n1_12(const R *ri, const R *ii, R *ro, R *io, stride is, stride os, INT v, INT ivs, INT ovs)
Chris@82 38 {
Chris@82 39 DK(KP866025403, +0.866025403784438646763723170752936183471402627);
Chris@82 40 DK(KP500000000, +0.500000000000000000000000000000000000000000000);
Chris@82 41 {
Chris@82 42 INT i;
Chris@82 43 for (i = v; i > 0; i = i - 1, ri = ri + ivs, ii = ii + ivs, ro = ro + ovs, io = io + ovs, MAKE_VOLATILE_STRIDE(48, is), MAKE_VOLATILE_STRIDE(48, os)) {
Chris@82 44 E T5, TR, TA, Ts, TS, Tz, Ta, TU, TD, Tx, TV, TC, Tg, T1d, TG;
Chris@82 45 E TJ, T1u, T1c, Tl, T1i, TL, TO, T1v, T1h;
Chris@82 46 {
Chris@82 47 E T1, T2, T3, T4;
Chris@82 48 T1 = ri[0];
Chris@82 49 T2 = ri[WS(is, 4)];
Chris@82 50 T3 = ri[WS(is, 8)];
Chris@82 51 T4 = T2 + T3;
Chris@82 52 T5 = T1 + T4;
Chris@82 53 TR = FNMS(KP500000000, T4, T1);
Chris@82 54 TA = T3 - T2;
Chris@82 55 }
Chris@82 56 {
Chris@82 57 E To, Tp, Tq, Tr;
Chris@82 58 To = ii[0];
Chris@82 59 Tp = ii[WS(is, 4)];
Chris@82 60 Tq = ii[WS(is, 8)];
Chris@82 61 Tr = Tp + Tq;
Chris@82 62 Ts = To + Tr;
Chris@82 63 TS = Tp - Tq;
Chris@82 64 Tz = FNMS(KP500000000, Tr, To);
Chris@82 65 }
Chris@82 66 {
Chris@82 67 E T6, T7, T8, T9;
Chris@82 68 T6 = ri[WS(is, 6)];
Chris@82 69 T7 = ri[WS(is, 10)];
Chris@82 70 T8 = ri[WS(is, 2)];
Chris@82 71 T9 = T7 + T8;
Chris@82 72 Ta = T6 + T9;
Chris@82 73 TU = FNMS(KP500000000, T9, T6);
Chris@82 74 TD = T8 - T7;
Chris@82 75 }
Chris@82 76 {
Chris@82 77 E Tt, Tu, Tv, Tw;
Chris@82 78 Tt = ii[WS(is, 6)];
Chris@82 79 Tu = ii[WS(is, 10)];
Chris@82 80 Tv = ii[WS(is, 2)];
Chris@82 81 Tw = Tu + Tv;
Chris@82 82 Tx = Tt + Tw;
Chris@82 83 TV = Tu - Tv;
Chris@82 84 TC = FNMS(KP500000000, Tw, Tt);
Chris@82 85 }
Chris@82 86 {
Chris@82 87 E Tc, Td, Te, Tf;
Chris@82 88 Tc = ri[WS(is, 3)];
Chris@82 89 Td = ri[WS(is, 7)];
Chris@82 90 Te = ri[WS(is, 11)];
Chris@82 91 Tf = Td + Te;
Chris@82 92 Tg = Tc + Tf;
Chris@82 93 T1d = Te - Td;
Chris@82 94 TG = FNMS(KP500000000, Tf, Tc);
Chris@82 95 }
Chris@82 96 {
Chris@82 97 E T1a, TH, TI, T1b;
Chris@82 98 T1a = ii[WS(is, 3)];
Chris@82 99 TH = ii[WS(is, 7)];
Chris@82 100 TI = ii[WS(is, 11)];
Chris@82 101 T1b = TH + TI;
Chris@82 102 TJ = TH - TI;
Chris@82 103 T1u = T1a + T1b;
Chris@82 104 T1c = FNMS(KP500000000, T1b, T1a);
Chris@82 105 }
Chris@82 106 {
Chris@82 107 E Th, Ti, Tj, Tk;
Chris@82 108 Th = ri[WS(is, 9)];
Chris@82 109 Ti = ri[WS(is, 1)];
Chris@82 110 Tj = ri[WS(is, 5)];
Chris@82 111 Tk = Ti + Tj;
Chris@82 112 Tl = Th + Tk;
Chris@82 113 T1i = Tj - Ti;
Chris@82 114 TL = FNMS(KP500000000, Tk, Th);
Chris@82 115 }
Chris@82 116 {
Chris@82 117 E T1f, TM, TN, T1g;
Chris@82 118 T1f = ii[WS(is, 9)];
Chris@82 119 TM = ii[WS(is, 1)];
Chris@82 120 TN = ii[WS(is, 5)];
Chris@82 121 T1g = TM + TN;
Chris@82 122 TO = TM - TN;
Chris@82 123 T1v = T1f + T1g;
Chris@82 124 T1h = FNMS(KP500000000, T1g, T1f);
Chris@82 125 }
Chris@82 126 {
Chris@82 127 E Tb, Tm, T1t, T1w;
Chris@82 128 Tb = T5 + Ta;
Chris@82 129 Tm = Tg + Tl;
Chris@82 130 ro[WS(os, 6)] = Tb - Tm;
Chris@82 131 ro[0] = Tb + Tm;
Chris@82 132 {
Chris@82 133 E T1x, T1y, Tn, Ty;
Chris@82 134 T1x = Ts + Tx;
Chris@82 135 T1y = T1u + T1v;
Chris@82 136 io[WS(os, 6)] = T1x - T1y;
Chris@82 137 io[0] = T1x + T1y;
Chris@82 138 Tn = Tg - Tl;
Chris@82 139 Ty = Ts - Tx;
Chris@82 140 io[WS(os, 3)] = Tn + Ty;
Chris@82 141 io[WS(os, 9)] = Ty - Tn;
Chris@82 142 }
Chris@82 143 T1t = T5 - Ta;
Chris@82 144 T1w = T1u - T1v;
Chris@82 145 ro[WS(os, 3)] = T1t - T1w;
Chris@82 146 ro[WS(os, 9)] = T1t + T1w;
Chris@82 147 {
Chris@82 148 E T11, T1l, T1k, T1m, T14, T18, T17, T19;
Chris@82 149 {
Chris@82 150 E TZ, T10, T1e, T1j;
Chris@82 151 TZ = FMA(KP866025403, TA, Tz);
Chris@82 152 T10 = FMA(KP866025403, TD, TC);
Chris@82 153 T11 = TZ - T10;
Chris@82 154 T1l = TZ + T10;
Chris@82 155 T1e = FMA(KP866025403, T1d, T1c);
Chris@82 156 T1j = FMA(KP866025403, T1i, T1h);
Chris@82 157 T1k = T1e - T1j;
Chris@82 158 T1m = T1e + T1j;
Chris@82 159 }
Chris@82 160 {
Chris@82 161 E T12, T13, T15, T16;
Chris@82 162 T12 = FMA(KP866025403, TJ, TG);
Chris@82 163 T13 = FMA(KP866025403, TO, TL);
Chris@82 164 T14 = T12 - T13;
Chris@82 165 T18 = T12 + T13;
Chris@82 166 T15 = FMA(KP866025403, TS, TR);
Chris@82 167 T16 = FMA(KP866025403, TV, TU);
Chris@82 168 T17 = T15 + T16;
Chris@82 169 T19 = T15 - T16;
Chris@82 170 }
Chris@82 171 io[WS(os, 1)] = T11 - T14;
Chris@82 172 ro[WS(os, 1)] = T19 + T1k;
Chris@82 173 io[WS(os, 7)] = T11 + T14;
Chris@82 174 ro[WS(os, 7)] = T19 - T1k;
Chris@82 175 ro[WS(os, 10)] = T17 - T18;
Chris@82 176 io[WS(os, 10)] = T1l - T1m;
Chris@82 177 ro[WS(os, 4)] = T17 + T18;
Chris@82 178 io[WS(os, 4)] = T1l + T1m;
Chris@82 179 }
Chris@82 180 {
Chris@82 181 E TF, T1r, T1q, T1s, TQ, TY, TX, T1n;
Chris@82 182 {
Chris@82 183 E TB, TE, T1o, T1p;
Chris@82 184 TB = FNMS(KP866025403, TA, Tz);
Chris@82 185 TE = FNMS(KP866025403, TD, TC);
Chris@82 186 TF = TB - TE;
Chris@82 187 T1r = TB + TE;
Chris@82 188 T1o = FNMS(KP866025403, T1d, T1c);
Chris@82 189 T1p = FNMS(KP866025403, T1i, T1h);
Chris@82 190 T1q = T1o - T1p;
Chris@82 191 T1s = T1o + T1p;
Chris@82 192 }
Chris@82 193 {
Chris@82 194 E TK, TP, TT, TW;
Chris@82 195 TK = FNMS(KP866025403, TJ, TG);
Chris@82 196 TP = FNMS(KP866025403, TO, TL);
Chris@82 197 TQ = TK - TP;
Chris@82 198 TY = TK + TP;
Chris@82 199 TT = FNMS(KP866025403, TS, TR);
Chris@82 200 TW = FNMS(KP866025403, TV, TU);
Chris@82 201 TX = TT + TW;
Chris@82 202 T1n = TT - TW;
Chris@82 203 }
Chris@82 204 io[WS(os, 5)] = TF - TQ;
Chris@82 205 ro[WS(os, 5)] = T1n + T1q;
Chris@82 206 io[WS(os, 11)] = TF + TQ;
Chris@82 207 ro[WS(os, 11)] = T1n - T1q;
Chris@82 208 ro[WS(os, 2)] = TX - TY;
Chris@82 209 io[WS(os, 2)] = T1r - T1s;
Chris@82 210 ro[WS(os, 8)] = TX + TY;
Chris@82 211 io[WS(os, 8)] = T1r + T1s;
Chris@82 212 }
Chris@82 213 }
Chris@82 214 }
Chris@82 215 }
Chris@82 216 }
Chris@82 217
Chris@82 218 static const kdft_desc desc = { 12, "n1_12", {72, 0, 24, 0}, &GENUS, 0, 0, 0, 0 };
Chris@82 219
Chris@82 220 void X(codelet_n1_12) (planner *p) {
Chris@82 221 X(kdft_register) (p, n1_12, &desc);
Chris@82 222 }
Chris@82 223
Chris@82 224 #else
Chris@82 225
Chris@82 226 /* Generated by: ../../../genfft/gen_notw.native -compact -variables 4 -pipeline-latency 4 -n 12 -name n1_12 -include dft/scalar/n.h */
Chris@82 227
Chris@82 228 /*
Chris@82 229 * This function contains 96 FP additions, 16 FP multiplications,
Chris@82 230 * (or, 88 additions, 8 multiplications, 8 fused multiply/add),
Chris@82 231 * 43 stack variables, 2 constants, and 48 memory accesses
Chris@82 232 */
Chris@82 233 #include "dft/scalar/n.h"
Chris@82 234
Chris@82 235 static void n1_12(const R *ri, const R *ii, R *ro, R *io, stride is, stride os, INT v, INT ivs, INT ovs)
Chris@82 236 {
Chris@82 237 DK(KP866025403, +0.866025403784438646763723170752936183471402627);
Chris@82 238 DK(KP500000000, +0.500000000000000000000000000000000000000000000);
Chris@82 239 {
Chris@82 240 INT i;
Chris@82 241 for (i = v; i > 0; i = i - 1, ri = ri + ivs, ii = ii + ivs, ro = ro + ovs, io = io + ovs, MAKE_VOLATILE_STRIDE(48, is), MAKE_VOLATILE_STRIDE(48, os)) {
Chris@82 242 E T5, TR, TA, Ts, TS, Tz, Ta, TU, TD, Tx, TV, TC, Tg, T1a, TG;
Chris@82 243 E TJ, T1u, T1d, Tl, T1f, TL, TO, T1v, T1i;
Chris@82 244 {
Chris@82 245 E T1, T2, T3, T4;
Chris@82 246 T1 = ri[0];
Chris@82 247 T2 = ri[WS(is, 4)];
Chris@82 248 T3 = ri[WS(is, 8)];
Chris@82 249 T4 = T2 + T3;
Chris@82 250 T5 = T1 + T4;
Chris@82 251 TR = FNMS(KP500000000, T4, T1);
Chris@82 252 TA = KP866025403 * (T3 - T2);
Chris@82 253 }
Chris@82 254 {
Chris@82 255 E To, Tp, Tq, Tr;
Chris@82 256 To = ii[0];
Chris@82 257 Tp = ii[WS(is, 4)];
Chris@82 258 Tq = ii[WS(is, 8)];
Chris@82 259 Tr = Tp + Tq;
Chris@82 260 Ts = To + Tr;
Chris@82 261 TS = KP866025403 * (Tp - Tq);
Chris@82 262 Tz = FNMS(KP500000000, Tr, To);
Chris@82 263 }
Chris@82 264 {
Chris@82 265 E T6, T7, T8, T9;
Chris@82 266 T6 = ri[WS(is, 6)];
Chris@82 267 T7 = ri[WS(is, 10)];
Chris@82 268 T8 = ri[WS(is, 2)];
Chris@82 269 T9 = T7 + T8;
Chris@82 270 Ta = T6 + T9;
Chris@82 271 TU = FNMS(KP500000000, T9, T6);
Chris@82 272 TD = KP866025403 * (T8 - T7);
Chris@82 273 }
Chris@82 274 {
Chris@82 275 E Tt, Tu, Tv, Tw;
Chris@82 276 Tt = ii[WS(is, 6)];
Chris@82 277 Tu = ii[WS(is, 10)];
Chris@82 278 Tv = ii[WS(is, 2)];
Chris@82 279 Tw = Tu + Tv;
Chris@82 280 Tx = Tt + Tw;
Chris@82 281 TV = KP866025403 * (Tu - Tv);
Chris@82 282 TC = FNMS(KP500000000, Tw, Tt);
Chris@82 283 }
Chris@82 284 {
Chris@82 285 E Tc, Td, Te, Tf;
Chris@82 286 Tc = ri[WS(is, 3)];
Chris@82 287 Td = ri[WS(is, 7)];
Chris@82 288 Te = ri[WS(is, 11)];
Chris@82 289 Tf = Td + Te;
Chris@82 290 Tg = Tc + Tf;
Chris@82 291 T1a = KP866025403 * (Te - Td);
Chris@82 292 TG = FNMS(KP500000000, Tf, Tc);
Chris@82 293 }
Chris@82 294 {
Chris@82 295 E T1b, TH, TI, T1c;
Chris@82 296 T1b = ii[WS(is, 3)];
Chris@82 297 TH = ii[WS(is, 7)];
Chris@82 298 TI = ii[WS(is, 11)];
Chris@82 299 T1c = TH + TI;
Chris@82 300 TJ = KP866025403 * (TH - TI);
Chris@82 301 T1u = T1b + T1c;
Chris@82 302 T1d = FNMS(KP500000000, T1c, T1b);
Chris@82 303 }
Chris@82 304 {
Chris@82 305 E Th, Ti, Tj, Tk;
Chris@82 306 Th = ri[WS(is, 9)];
Chris@82 307 Ti = ri[WS(is, 1)];
Chris@82 308 Tj = ri[WS(is, 5)];
Chris@82 309 Tk = Ti + Tj;
Chris@82 310 Tl = Th + Tk;
Chris@82 311 T1f = KP866025403 * (Tj - Ti);
Chris@82 312 TL = FNMS(KP500000000, Tk, Th);
Chris@82 313 }
Chris@82 314 {
Chris@82 315 E T1g, TM, TN, T1h;
Chris@82 316 T1g = ii[WS(is, 9)];
Chris@82 317 TM = ii[WS(is, 1)];
Chris@82 318 TN = ii[WS(is, 5)];
Chris@82 319 T1h = TM + TN;
Chris@82 320 TO = KP866025403 * (TM - TN);
Chris@82 321 T1v = T1g + T1h;
Chris@82 322 T1i = FNMS(KP500000000, T1h, T1g);
Chris@82 323 }
Chris@82 324 {
Chris@82 325 E Tb, Tm, T1t, T1w;
Chris@82 326 Tb = T5 + Ta;
Chris@82 327 Tm = Tg + Tl;
Chris@82 328 ro[WS(os, 6)] = Tb - Tm;
Chris@82 329 ro[0] = Tb + Tm;
Chris@82 330 {
Chris@82 331 E T1x, T1y, Tn, Ty;
Chris@82 332 T1x = Ts + Tx;
Chris@82 333 T1y = T1u + T1v;
Chris@82 334 io[WS(os, 6)] = T1x - T1y;
Chris@82 335 io[0] = T1x + T1y;
Chris@82 336 Tn = Tg - Tl;
Chris@82 337 Ty = Ts - Tx;
Chris@82 338 io[WS(os, 3)] = Tn + Ty;
Chris@82 339 io[WS(os, 9)] = Ty - Tn;
Chris@82 340 }
Chris@82 341 T1t = T5 - Ta;
Chris@82 342 T1w = T1u - T1v;
Chris@82 343 ro[WS(os, 3)] = T1t - T1w;
Chris@82 344 ro[WS(os, 9)] = T1t + T1w;
Chris@82 345 {
Chris@82 346 E T11, T1l, T1k, T1m, T14, T18, T17, T19;
Chris@82 347 {
Chris@82 348 E TZ, T10, T1e, T1j;
Chris@82 349 TZ = TA + Tz;
Chris@82 350 T10 = TD + TC;
Chris@82 351 T11 = TZ - T10;
Chris@82 352 T1l = TZ + T10;
Chris@82 353 T1e = T1a + T1d;
Chris@82 354 T1j = T1f + T1i;
Chris@82 355 T1k = T1e - T1j;
Chris@82 356 T1m = T1e + T1j;
Chris@82 357 }
Chris@82 358 {
Chris@82 359 E T12, T13, T15, T16;
Chris@82 360 T12 = TG + TJ;
Chris@82 361 T13 = TL + TO;
Chris@82 362 T14 = T12 - T13;
Chris@82 363 T18 = T12 + T13;
Chris@82 364 T15 = TR + TS;
Chris@82 365 T16 = TU + TV;
Chris@82 366 T17 = T15 + T16;
Chris@82 367 T19 = T15 - T16;
Chris@82 368 }
Chris@82 369 io[WS(os, 1)] = T11 - T14;
Chris@82 370 ro[WS(os, 1)] = T19 + T1k;
Chris@82 371 io[WS(os, 7)] = T11 + T14;
Chris@82 372 ro[WS(os, 7)] = T19 - T1k;
Chris@82 373 ro[WS(os, 10)] = T17 - T18;
Chris@82 374 io[WS(os, 10)] = T1l - T1m;
Chris@82 375 ro[WS(os, 4)] = T17 + T18;
Chris@82 376 io[WS(os, 4)] = T1l + T1m;
Chris@82 377 }
Chris@82 378 {
Chris@82 379 E TF, T1r, T1q, T1s, TQ, TY, TX, T1n;
Chris@82 380 {
Chris@82 381 E TB, TE, T1o, T1p;
Chris@82 382 TB = Tz - TA;
Chris@82 383 TE = TC - TD;
Chris@82 384 TF = TB - TE;
Chris@82 385 T1r = TB + TE;
Chris@82 386 T1o = T1d - T1a;
Chris@82 387 T1p = T1i - T1f;
Chris@82 388 T1q = T1o - T1p;
Chris@82 389 T1s = T1o + T1p;
Chris@82 390 }
Chris@82 391 {
Chris@82 392 E TK, TP, TT, TW;
Chris@82 393 TK = TG - TJ;
Chris@82 394 TP = TL - TO;
Chris@82 395 TQ = TK - TP;
Chris@82 396 TY = TK + TP;
Chris@82 397 TT = TR - TS;
Chris@82 398 TW = TU - TV;
Chris@82 399 TX = TT + TW;
Chris@82 400 T1n = TT - TW;
Chris@82 401 }
Chris@82 402 io[WS(os, 5)] = TF - TQ;
Chris@82 403 ro[WS(os, 5)] = T1n + T1q;
Chris@82 404 io[WS(os, 11)] = TF + TQ;
Chris@82 405 ro[WS(os, 11)] = T1n - T1q;
Chris@82 406 ro[WS(os, 2)] = TX - TY;
Chris@82 407 io[WS(os, 2)] = T1r - T1s;
Chris@82 408 ro[WS(os, 8)] = TX + TY;
Chris@82 409 io[WS(os, 8)] = T1r + T1s;
Chris@82 410 }
Chris@82 411 }
Chris@82 412 }
Chris@82 413 }
Chris@82 414 }
Chris@82 415
Chris@82 416 static const kdft_desc desc = { 12, "n1_12", {88, 8, 8, 0}, &GENUS, 0, 0, 0, 0 };
Chris@82 417
Chris@82 418 void X(codelet_n1_12) (planner *p) {
Chris@82 419 X(kdft_register) (p, n1_12, &desc);
Chris@82 420 }
Chris@82 421
Chris@82 422 #endif