Mercurial > hg > sv-dependency-builds

annotate src/fftw-3.3.5/dft/scalar/codelets/n1_12.c @ 84:08ae793730bd

Find changesets by keywords (author, files, the commit message), revision number or hash, or revset expression.
Add null config files
author Chris Cannam
date Mon, 02 Mar 2020 14:03:47 +0000
parents 2cd0e3b3e1fd
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 16:35:52 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_notw.native -fma -reorder-insns -schedule-for-pipeline -compact -variables 4 -pipeline-latency 4 -n 12 -name n1_12 -include n.h */
Chris@42 29
Chris@42 30 /*
Chris@42 31 * This function contains 96 FP additions, 24 FP multiplications,
Chris@42 32 * (or, 72 additions, 0 multiplications, 24 fused multiply/add),
Chris@42 33 * 63 stack variables, 2 constants, and 48 memory accesses
Chris@42 34 */
Chris@42 35 #include "n.h"
Chris@42 36
Chris@42 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@42 38 {
Chris@42 39 DK(KP866025403, +0.866025403784438646763723170752936183471402627);
Chris@42 40 DK(KP500000000, +0.500000000000000000000000000000000000000000000);
Chris@42 41 {
Chris@42 42 INT i;
Chris@42 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@42 44 E TT, TW, TF, T1q, TY, TQ, TX, T1n;
Chris@42 45 {
Chris@42 46 E TA, TS, TR, T5, Ts, Tz, TD, TV, TU, Ta, Tx, TC, T1d, Th, TJ;
Chris@42 47 E TG, Tg, T1u, T1c, T1f, TM, TN, Tk, T1i;
Chris@42 48 {
Chris@42 49 E T6, Tt, Tu, Tv, T9;
Chris@42 50 {
Chris@42 51 E T1, To, Tp, Tq, T4, T2, T3, T7, T8, Tr;
Chris@42 52 T1 = ri[0];
Chris@42 53 T2 = ri[WS(is, 4)];
Chris@42 54 T3 = ri[WS(is, 8)];
Chris@42 55 To = ii[0];
Chris@42 56 Tp = ii[WS(is, 4)];
Chris@42 57 Tq = ii[WS(is, 8)];
Chris@42 58 T4 = T2 + T3;
Chris@42 59 TA = T3 - T2;
Chris@42 60 T6 = ri[WS(is, 6)];
Chris@42 61 TS = Tp - Tq;
Chris@42 62 Tr = Tp + Tq;
Chris@42 63 TR = FNMS(KP500000000, T4, T1);
Chris@42 64 T5 = T1 + T4;
Chris@42 65 T7 = ri[WS(is, 10)];
Chris@42 66 Ts = To + Tr;
Chris@42 67 Tz = FNMS(KP500000000, Tr, To);
Chris@42 68 T8 = ri[WS(is, 2)];
Chris@42 69 Tt = ii[WS(is, 6)];
Chris@42 70 Tu = ii[WS(is, 10)];
Chris@42 71 Tv = ii[WS(is, 2)];
Chris@42 72 T9 = T7 + T8;
Chris@42 73 TD = T8 - T7;
Chris@42 74 }
Chris@42 75 {
Chris@42 76 E Tc, T1a, TH, TI, Tf, Td, Te, Tw, Ti, Tj, T1b;
Chris@42 77 Tc = ri[WS(is, 3)];
Chris@42 78 TV = Tu - Tv;
Chris@42 79 Tw = Tu + Tv;
Chris@42 80 TU = FNMS(KP500000000, T9, T6);
Chris@42 81 Ta = T6 + T9;
Chris@42 82 Td = ri[WS(is, 7)];
Chris@42 83 Tx = Tt + Tw;
Chris@42 84 TC = FNMS(KP500000000, Tw, Tt);
Chris@42 85 Te = ri[WS(is, 11)];
Chris@42 86 T1a = ii[WS(is, 3)];
Chris@42 87 TH = ii[WS(is, 7)];
Chris@42 88 TI = ii[WS(is, 11)];
Chris@42 89 Tf = Td + Te;
Chris@42 90 T1d = Te - Td;
Chris@42 91 Th = ri[WS(is, 9)];
Chris@42 92 TJ = TH - TI;
Chris@42 93 T1b = TH + TI;
Chris@42 94 TG = FNMS(KP500000000, Tf, Tc);
Chris@42 95 Tg = Tc + Tf;
Chris@42 96 Ti = ri[WS(is, 1)];
Chris@42 97 T1u = T1a + T1b;
Chris@42 98 T1c = FNMS(KP500000000, T1b, T1a);
Chris@42 99 Tj = ri[WS(is, 5)];
Chris@42 100 T1f = ii[WS(is, 9)];
Chris@42 101 TM = ii[WS(is, 1)];
Chris@42 102 TN = ii[WS(is, 5)];
Chris@42 103 Tk = Ti + Tj;
Chris@42 104 T1i = Tj - Ti;
Chris@42 105 }
Chris@42 106 }
Chris@42 107 {
Chris@42 108 E T1t, TO, TL, T1h, T1w, Tb, T1g, Tl;
Chris@42 109 T1t = T5 - Ta;
Chris@42 110 Tb = T5 + Ta;
Chris@42 111 TO = TM - TN;
Chris@42 112 T1g = TM + TN;
Chris@42 113 TL = FNMS(KP500000000, Tk, Th);
Chris@42 114 Tl = Th + Tk;
Chris@42 115 {
Chris@42 116 E T1x, Ty, T1v, Tn, Tm, T1y;
Chris@42 117 T1x = Ts + Tx;
Chris@42 118 Ty = Ts - Tx;
Chris@42 119 T1v = T1f + T1g;
Chris@42 120 T1h = FNMS(KP500000000, T1g, T1f);
Chris@42 121 Tn = Tg - Tl;
Chris@42 122 Tm = Tg + Tl;
Chris@42 123 T1y = T1u + T1v;
Chris@42 124 T1w = T1u - T1v;
Chris@42 125 ro[0] = Tb + Tm;
Chris@42 126 ro[WS(os, 6)] = Tb - Tm;
Chris@42 127 io[WS(os, 3)] = Tn + Ty;
Chris@42 128 io[0] = T1x + T1y;
Chris@42 129 io[WS(os, 6)] = T1x - T1y;
Chris@42 130 io[WS(os, 9)] = Ty - Tn;
Chris@42 131 }
Chris@42 132 {
Chris@42 133 E TB, TE, T1o, T11, T1p, TK, TP, T15, T1k, T18, T14, T16, T1l, T1m;
Chris@42 134 {
Chris@42 135 E T1e, T1j, TZ, T10, T12, T13;
Chris@42 136 TB = FNMS(KP866025403, TA, Tz);
Chris@42 137 TZ = FMA(KP866025403, TA, Tz);
Chris@42 138 T10 = FMA(KP866025403, TD, TC);
Chris@42 139 TE = FNMS(KP866025403, TD, TC);
Chris@42 140 T1o = FNMS(KP866025403, T1d, T1c);
Chris@42 141 T1e = FMA(KP866025403, T1d, T1c);
Chris@42 142 ro[WS(os, 9)] = T1t + T1w;
Chris@42 143 ro[WS(os, 3)] = T1t - T1w;
Chris@42 144 T1l = TZ + T10;
Chris@42 145 T11 = TZ - T10;
Chris@42 146 T1j = FMA(KP866025403, T1i, T1h);
Chris@42 147 T1p = FNMS(KP866025403, T1i, T1h);
Chris@42 148 TK = FNMS(KP866025403, TJ, TG);
Chris@42 149 T12 = FMA(KP866025403, TJ, TG);
Chris@42 150 T13 = FMA(KP866025403, TO, TL);
Chris@42 151 TP = FNMS(KP866025403, TO, TL);
Chris@42 152 TT = FNMS(KP866025403, TS, TR);
Chris@42 153 T15 = FMA(KP866025403, TS, TR);
Chris@42 154 T1m = T1e + T1j;
Chris@42 155 T1k = T1e - T1j;
Chris@42 156 T18 = T12 + T13;
Chris@42 157 T14 = T12 - T13;
Chris@42 158 T16 = FMA(KP866025403, TV, TU);
Chris@42 159 TW = FNMS(KP866025403, TV, TU);
Chris@42 160 }
Chris@42 161 io[WS(os, 10)] = T1l - T1m;
Chris@42 162 io[WS(os, 4)] = T1l + T1m;
Chris@42 163 io[WS(os, 7)] = T11 + T14;
Chris@42 164 io[WS(os, 1)] = T11 - T14;
Chris@42 165 {
Chris@42 166 E T17, T19, T1r, T1s;
Chris@42 167 T17 = T15 + T16;
Chris@42 168 T19 = T15 - T16;
Chris@42 169 ro[WS(os, 7)] = T19 - T1k;
Chris@42 170 ro[WS(os, 1)] = T19 + T1k;
Chris@42 171 ro[WS(os, 4)] = T17 + T18;
Chris@42 172 ro[WS(os, 10)] = T17 - T18;
Chris@42 173 T1r = TB + TE;
Chris@42 174 TF = TB - TE;
Chris@42 175 T1s = T1o + T1p;
Chris@42 176 T1q = T1o - T1p;
Chris@42 177 TY = TK + TP;
Chris@42 178 TQ = TK - TP;
Chris@42 179 io[WS(os, 2)] = T1r - T1s;
Chris@42 180 io[WS(os, 8)] = T1r + T1s;
Chris@42 181 }
Chris@42 182 }
Chris@42 183 }
Chris@42 184 }
Chris@42 185 io[WS(os, 11)] = TF + TQ;
Chris@42 186 io[WS(os, 5)] = TF - TQ;
Chris@42 187 TX = TT + TW;
Chris@42 188 T1n = TT - TW;
Chris@42 189 ro[WS(os, 11)] = T1n - T1q;
Chris@42 190 ro[WS(os, 5)] = T1n + T1q;
Chris@42 191 ro[WS(os, 8)] = TX + TY;
Chris@42 192 ro[WS(os, 2)] = TX - TY;
Chris@42 193 }
Chris@42 194 }
Chris@42 195 }
Chris@42 196
Chris@42 197 static const kdft_desc desc = { 12, "n1_12", {72, 0, 24, 0}, &GENUS, 0, 0, 0, 0 };
Chris@42 198
Chris@42 199 void X(codelet_n1_12) (planner *p) {
Chris@42 200 X(kdft_register) (p, n1_12, &desc);
Chris@42 201 }
Chris@42 202
Chris@42 203 #else /* HAVE_FMA */
Chris@42 204
Chris@42 205 /* Generated by: ../../../genfft/gen_notw.native -compact -variables 4 -pipeline-latency 4 -n 12 -name n1_12 -include n.h */
Chris@42 206
Chris@42 207 /*
Chris@42 208 * This function contains 96 FP additions, 16 FP multiplications,
Chris@42 209 * (or, 88 additions, 8 multiplications, 8 fused multiply/add),
Chris@42 210 * 43 stack variables, 2 constants, and 48 memory accesses
Chris@42 211 */
Chris@42 212 #include "n.h"
Chris@42 213
Chris@42 214 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@42 215 {
Chris@42 216 DK(KP866025403, +0.866025403784438646763723170752936183471402627);
Chris@42 217 DK(KP500000000, +0.500000000000000000000000000000000000000000000);
Chris@42 218 {
Chris@42 219 INT i;
Chris@42 220 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@42 221 E T5, TR, TA, Ts, TS, Tz, Ta, TU, TD, Tx, TV, TC, Tg, T1a, TG;
Chris@42 222 E TJ, T1u, T1d, Tl, T1f, TL, TO, T1v, T1i;
Chris@42 223 {
Chris@42 224 E T1, T2, T3, T4;
Chris@42 225 T1 = ri[0];
Chris@42 226 T2 = ri[WS(is, 4)];
Chris@42 227 T3 = ri[WS(is, 8)];
Chris@42 228 T4 = T2 + T3;
Chris@42 229 T5 = T1 + T4;
Chris@42 230 TR = FNMS(KP500000000, T4, T1);
Chris@42 231 TA = KP866025403 * (T3 - T2);
Chris@42 232 }
Chris@42 233 {
Chris@42 234 E To, Tp, Tq, Tr;
Chris@42 235 To = ii[0];
Chris@42 236 Tp = ii[WS(is, 4)];
Chris@42 237 Tq = ii[WS(is, 8)];
Chris@42 238 Tr = Tp + Tq;
Chris@42 239 Ts = To + Tr;
Chris@42 240 TS = KP866025403 * (Tp - Tq);
Chris@42 241 Tz = FNMS(KP500000000, Tr, To);
Chris@42 242 }
Chris@42 243 {
Chris@42 244 E T6, T7, T8, T9;
Chris@42 245 T6 = ri[WS(is, 6)];
Chris@42 246 T7 = ri[WS(is, 10)];
Chris@42 247 T8 = ri[WS(is, 2)];
Chris@42 248 T9 = T7 + T8;
Chris@42 249 Ta = T6 + T9;
Chris@42 250 TU = FNMS(KP500000000, T9, T6);
Chris@42 251 TD = KP866025403 * (T8 - T7);
Chris@42 252 }
Chris@42 253 {
Chris@42 254 E Tt, Tu, Tv, Tw;
Chris@42 255 Tt = ii[WS(is, 6)];
Chris@42 256 Tu = ii[WS(is, 10)];
Chris@42 257 Tv = ii[WS(is, 2)];
Chris@42 258 Tw = Tu + Tv;
Chris@42 259 Tx = Tt + Tw;
Chris@42 260 TV = KP866025403 * (Tu - Tv);
Chris@42 261 TC = FNMS(KP500000000, Tw, Tt);
Chris@42 262 }
Chris@42 263 {
Chris@42 264 E Tc, Td, Te, Tf;
Chris@42 265 Tc = ri[WS(is, 3)];
Chris@42 266 Td = ri[WS(is, 7)];
Chris@42 267 Te = ri[WS(is, 11)];
Chris@42 268 Tf = Td + Te;
Chris@42 269 Tg = Tc + Tf;
Chris@42 270 T1a = KP866025403 * (Te - Td);
Chris@42 271 TG = FNMS(KP500000000, Tf, Tc);
Chris@42 272 }
Chris@42 273 {
Chris@42 274 E T1b, TH, TI, T1c;
Chris@42 275 T1b = ii[WS(is, 3)];
Chris@42 276 TH = ii[WS(is, 7)];
Chris@42 277 TI = ii[WS(is, 11)];
Chris@42 278 T1c = TH + TI;
Chris@42 279 TJ = KP866025403 * (TH - TI);
Chris@42 280 T1u = T1b + T1c;
Chris@42 281 T1d = FNMS(KP500000000, T1c, T1b);
Chris@42 282 }
Chris@42 283 {
Chris@42 284 E Th, Ti, Tj, Tk;
Chris@42 285 Th = ri[WS(is, 9)];
Chris@42 286 Ti = ri[WS(is, 1)];
Chris@42 287 Tj = ri[WS(is, 5)];
Chris@42 288 Tk = Ti + Tj;
Chris@42 289 Tl = Th + Tk;
Chris@42 290 T1f = KP866025403 * (Tj - Ti);
Chris@42 291 TL = FNMS(KP500000000, Tk, Th);
Chris@42 292 }
Chris@42 293 {
Chris@42 294 E T1g, TM, TN, T1h;
Chris@42 295 T1g = ii[WS(is, 9)];
Chris@42 296 TM = ii[WS(is, 1)];
Chris@42 297 TN = ii[WS(is, 5)];
Chris@42 298 T1h = TM + TN;
Chris@42 299 TO = KP866025403 * (TM - TN);
Chris@42 300 T1v = T1g + T1h;
Chris@42 301 T1i = FNMS(KP500000000, T1h, T1g);
Chris@42 302 }
Chris@42 303 {
Chris@42 304 E Tb, Tm, T1t, T1w;
Chris@42 305 Tb = T5 + Ta;
Chris@42 306 Tm = Tg + Tl;
Chris@42 307 ro[WS(os, 6)] = Tb - Tm;
Chris@42 308 ro[0] = Tb + Tm;
Chris@42 309 {
Chris@42 310 E T1x, T1y, Tn, Ty;
Chris@42 311 T1x = Ts + Tx;
Chris@42 312 T1y = T1u + T1v;
Chris@42 313 io[WS(os, 6)] = T1x - T1y;
Chris@42 314 io[0] = T1x + T1y;
Chris@42 315 Tn = Tg - Tl;
Chris@42 316 Ty = Ts - Tx;
Chris@42 317 io[WS(os, 3)] = Tn + Ty;
Chris@42 318 io[WS(os, 9)] = Ty - Tn;
Chris@42 319 }
Chris@42 320 T1t = T5 - Ta;
Chris@42 321 T1w = T1u - T1v;
Chris@42 322 ro[WS(os, 3)] = T1t - T1w;
Chris@42 323 ro[WS(os, 9)] = T1t + T1w;
Chris@42 324 {
Chris@42 325 E T11, T1l, T1k, T1m, T14, T18, T17, T19;
Chris@42 326 {
Chris@42 327 E TZ, T10, T1e, T1j;
Chris@42 328 TZ = TA + Tz;
Chris@42 329 T10 = TD + TC;
Chris@42 330 T11 = TZ - T10;
Chris@42 331 T1l = TZ + T10;
Chris@42 332 T1e = T1a + T1d;
Chris@42 333 T1j = T1f + T1i;
Chris@42 334 T1k = T1e - T1j;
Chris@42 335 T1m = T1e + T1j;
Chris@42 336 }
Chris@42 337 {
Chris@42 338 E T12, T13, T15, T16;
Chris@42 339 T12 = TG + TJ;
Chris@42 340 T13 = TL + TO;
Chris@42 341 T14 = T12 - T13;
Chris@42 342 T18 = T12 + T13;
Chris@42 343 T15 = TR + TS;
Chris@42 344 T16 = TU + TV;
Chris@42 345 T17 = T15 + T16;
Chris@42 346 T19 = T15 - T16;
Chris@42 347 }
Chris@42 348 io[WS(os, 1)] = T11 - T14;
Chris@42 349 ro[WS(os, 1)] = T19 + T1k;
Chris@42 350 io[WS(os, 7)] = T11 + T14;
Chris@42 351 ro[WS(os, 7)] = T19 - T1k;
Chris@42 352 ro[WS(os, 10)] = T17 - T18;
Chris@42 353 io[WS(os, 10)] = T1l - T1m;
Chris@42 354 ro[WS(os, 4)] = T17 + T18;
Chris@42 355 io[WS(os, 4)] = T1l + T1m;
Chris@42 356 }
Chris@42 357 {
Chris@42 358 E TF, T1r, T1q, T1s, TQ, TY, TX, T1n;
Chris@42 359 {
Chris@42 360 E TB, TE, T1o, T1p;
Chris@42 361 TB = Tz - TA;
Chris@42 362 TE = TC - TD;
Chris@42 363 TF = TB - TE;
Chris@42 364 T1r = TB + TE;
Chris@42 365 T1o = T1d - T1a;
Chris@42 366 T1p = T1i - T1f;
Chris@42 367 T1q = T1o - T1p;
Chris@42 368 T1s = T1o + T1p;
Chris@42 369 }
Chris@42 370 {
Chris@42 371 E TK, TP, TT, TW;
Chris@42 372 TK = TG - TJ;
Chris@42 373 TP = TL - TO;
Chris@42 374 TQ = TK - TP;
Chris@42 375 TY = TK + TP;
Chris@42 376 TT = TR - TS;
Chris@42 377 TW = TU - TV;
Chris@42 378 TX = TT + TW;
Chris@42 379 T1n = TT - TW;
Chris@42 380 }
Chris@42 381 io[WS(os, 5)] = TF - TQ;
Chris@42 382 ro[WS(os, 5)] = T1n + T1q;
Chris@42 383 io[WS(os, 11)] = TF + TQ;
Chris@42 384 ro[WS(os, 11)] = T1n - T1q;
Chris@42 385 ro[WS(os, 2)] = TX - TY;
Chris@42 386 io[WS(os, 2)] = T1r - T1s;
Chris@42 387 ro[WS(os, 8)] = TX + TY;
Chris@42 388 io[WS(os, 8)] = T1r + T1s;
Chris@42 389 }
Chris@42 390 }
Chris@42 391 }
Chris@42 392 }
Chris@42 393 }
Chris@42 394
Chris@42 395 static const kdft_desc desc = { 12, "n1_12", {88, 8, 8, 0}, &GENUS, 0, 0, 0, 0 };
Chris@42 396
Chris@42 397 void X(codelet_n1_12) (planner *p) {
Chris@42 398 X(kdft_register) (p, n1_12, &desc);
Chris@42 399 }
Chris@42 400
Chris@42 401 #endif /* HAVE_FMA */