Mercurial > hg > sv-dependency-builds

annotate src/fftw-3.3.8/dft/scalar/codelets/n1_10.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 10 -name n1_10 -include dft/scalar/n.h */
Chris@82 29
Chris@82 30 /*
Chris@82 31 * This function contains 84 FP additions, 36 FP multiplications,
Chris@82 32 * (or, 48 additions, 0 multiplications, 36 fused multiply/add),
Chris@82 33 * 41 stack variables, 4 constants, and 40 memory accesses
Chris@82 34 */
Chris@82 35 #include "dft/scalar/n.h"
Chris@82 36
Chris@82 37 static void n1_10(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(KP951056516, +0.951056516295153572116439333379382143405698634);
Chris@82 40 DK(KP559016994, +0.559016994374947424102293417182819058860154590);
Chris@82 41 DK(KP250000000, +0.250000000000000000000000000000000000000000000);
Chris@82 42 DK(KP618033988, +0.618033988749894848204586834365638117720309180);
Chris@82 43 {
Chris@82 44 INT i;
Chris@82 45 for (i = v; i > 0; i = i - 1, ri = ri + ivs, ii = ii + ivs, ro = ro + ovs, io = io + ovs, MAKE_VOLATILE_STRIDE(40, is), MAKE_VOLATILE_STRIDE(40, os)) {
Chris@82 46 E T3, Tj, TN, T1b, TU, TV, T1j, T1i, Tm, Tp, Tq, Ta, Th, Ti, TA;
Chris@82 47 E TH, T17, T14, T1c, T1d, T1e, TO, TP, TQ;
Chris@82 48 {
Chris@82 49 E T1, T2, TL, TM;
Chris@82 50 T1 = ri[0];
Chris@82 51 T2 = ri[WS(is, 5)];
Chris@82 52 T3 = T1 - T2;
Chris@82 53 Tj = T1 + T2;
Chris@82 54 TL = ii[0];
Chris@82 55 TM = ii[WS(is, 5)];
Chris@82 56 TN = TL - TM;
Chris@82 57 T1b = TL + TM;
Chris@82 58 }
Chris@82 59 {
Chris@82 60 E T6, Tk, Tg, To, T9, Tl, Td, Tn;
Chris@82 61 {
Chris@82 62 E T4, T5, Te, Tf;
Chris@82 63 T4 = ri[WS(is, 2)];
Chris@82 64 T5 = ri[WS(is, 7)];
Chris@82 65 T6 = T4 - T5;
Chris@82 66 Tk = T4 + T5;
Chris@82 67 Te = ri[WS(is, 6)];
Chris@82 68 Tf = ri[WS(is, 1)];
Chris@82 69 Tg = Te - Tf;
Chris@82 70 To = Te + Tf;
Chris@82 71 }
Chris@82 72 {
Chris@82 73 E T7, T8, Tb, Tc;
Chris@82 74 T7 = ri[WS(is, 8)];
Chris@82 75 T8 = ri[WS(is, 3)];
Chris@82 76 T9 = T7 - T8;
Chris@82 77 Tl = T7 + T8;
Chris@82 78 Tb = ri[WS(is, 4)];
Chris@82 79 Tc = ri[WS(is, 9)];
Chris@82 80 Td = Tb - Tc;
Chris@82 81 Tn = Tb + Tc;
Chris@82 82 }
Chris@82 83 TU = T6 - T9;
Chris@82 84 TV = Td - Tg;
Chris@82 85 T1j = Tk - Tl;
Chris@82 86 T1i = Tn - To;
Chris@82 87 Tm = Tk + Tl;
Chris@82 88 Tp = Tn + To;
Chris@82 89 Tq = Tm + Tp;
Chris@82 90 Ta = T6 + T9;
Chris@82 91 Th = Td + Tg;
Chris@82 92 Ti = Ta + Th;
Chris@82 93 }
Chris@82 94 {
Chris@82 95 E Tw, T15, TG, T13, Tz, T16, TD, T12;
Chris@82 96 {
Chris@82 97 E Tu, Tv, TE, TF;
Chris@82 98 Tu = ii[WS(is, 2)];
Chris@82 99 Tv = ii[WS(is, 7)];
Chris@82 100 Tw = Tu - Tv;
Chris@82 101 T15 = Tu + Tv;
Chris@82 102 TE = ii[WS(is, 6)];
Chris@82 103 TF = ii[WS(is, 1)];
Chris@82 104 TG = TE - TF;
Chris@82 105 T13 = TE + TF;
Chris@82 106 }
Chris@82 107 {
Chris@82 108 E Tx, Ty, TB, TC;
Chris@82 109 Tx = ii[WS(is, 8)];
Chris@82 110 Ty = ii[WS(is, 3)];
Chris@82 111 Tz = Tx - Ty;
Chris@82 112 T16 = Tx + Ty;
Chris@82 113 TB = ii[WS(is, 4)];
Chris@82 114 TC = ii[WS(is, 9)];
Chris@82 115 TD = TB - TC;
Chris@82 116 T12 = TB + TC;
Chris@82 117 }
Chris@82 118 TA = Tw - Tz;
Chris@82 119 TH = TD - TG;
Chris@82 120 T17 = T15 - T16;
Chris@82 121 T14 = T12 - T13;
Chris@82 122 T1c = T15 + T16;
Chris@82 123 T1d = T12 + T13;
Chris@82 124 T1e = T1c + T1d;
Chris@82 125 TO = Tw + Tz;
Chris@82 126 TP = TD + TG;
Chris@82 127 TQ = TO + TP;
Chris@82 128 }
Chris@82 129 ro[WS(os, 5)] = T3 + Ti;
Chris@82 130 io[WS(os, 5)] = TN + TQ;
Chris@82 131 ro[0] = Tj + Tq;
Chris@82 132 io[0] = T1b + T1e;
Chris@82 133 {
Chris@82 134 E TI, TK, Tt, TJ, Tr, Ts;
Chris@82 135 TI = FMA(KP618033988, TH, TA);
Chris@82 136 TK = FNMS(KP618033988, TA, TH);
Chris@82 137 Tr = FNMS(KP250000000, Ti, T3);
Chris@82 138 Ts = Ta - Th;
Chris@82 139 Tt = FMA(KP559016994, Ts, Tr);
Chris@82 140 TJ = FNMS(KP559016994, Ts, Tr);
Chris@82 141 ro[WS(os, 9)] = FNMS(KP951056516, TI, Tt);
Chris@82 142 ro[WS(os, 3)] = FMA(KP951056516, TK, TJ);
Chris@82 143 ro[WS(os, 1)] = FMA(KP951056516, TI, Tt);
Chris@82 144 ro[WS(os, 7)] = FNMS(KP951056516, TK, TJ);
Chris@82 145 }
Chris@82 146 {
Chris@82 147 E TW, TY, TT, TX, TR, TS;
Chris@82 148 TW = FMA(KP618033988, TV, TU);
Chris@82 149 TY = FNMS(KP618033988, TU, TV);
Chris@82 150 TR = FNMS(KP250000000, TQ, TN);
Chris@82 151 TS = TO - TP;
Chris@82 152 TT = FMA(KP559016994, TS, TR);
Chris@82 153 TX = FNMS(KP559016994, TS, TR);
Chris@82 154 io[WS(os, 1)] = FNMS(KP951056516, TW, TT);
Chris@82 155 io[WS(os, 7)] = FMA(KP951056516, TY, TX);
Chris@82 156 io[WS(os, 9)] = FMA(KP951056516, TW, TT);
Chris@82 157 io[WS(os, 3)] = FNMS(KP951056516, TY, TX);
Chris@82 158 }
Chris@82 159 {
Chris@82 160 E T18, T1a, T11, T19, TZ, T10;
Chris@82 161 T18 = FNMS(KP618033988, T17, T14);
Chris@82 162 T1a = FMA(KP618033988, T14, T17);
Chris@82 163 TZ = FNMS(KP250000000, Tq, Tj);
Chris@82 164 T10 = Tm - Tp;
Chris@82 165 T11 = FNMS(KP559016994, T10, TZ);
Chris@82 166 T19 = FMA(KP559016994, T10, TZ);
Chris@82 167 ro[WS(os, 2)] = FNMS(KP951056516, T18, T11);
Chris@82 168 ro[WS(os, 6)] = FMA(KP951056516, T1a, T19);
Chris@82 169 ro[WS(os, 8)] = FMA(KP951056516, T18, T11);
Chris@82 170 ro[WS(os, 4)] = FNMS(KP951056516, T1a, T19);
Chris@82 171 }
Chris@82 172 {
Chris@82 173 E T1k, T1m, T1h, T1l, T1f, T1g;
Chris@82 174 T1k = FNMS(KP618033988, T1j, T1i);
Chris@82 175 T1m = FMA(KP618033988, T1i, T1j);
Chris@82 176 T1f = FNMS(KP250000000, T1e, T1b);
Chris@82 177 T1g = T1c - T1d;
Chris@82 178 T1h = FNMS(KP559016994, T1g, T1f);
Chris@82 179 T1l = FMA(KP559016994, T1g, T1f);
Chris@82 180 io[WS(os, 2)] = FMA(KP951056516, T1k, T1h);
Chris@82 181 io[WS(os, 6)] = FNMS(KP951056516, T1m, T1l);
Chris@82 182 io[WS(os, 8)] = FNMS(KP951056516, T1k, T1h);
Chris@82 183 io[WS(os, 4)] = FMA(KP951056516, T1m, T1l);
Chris@82 184 }
Chris@82 185 }
Chris@82 186 }
Chris@82 187 }
Chris@82 188
Chris@82 189 static const kdft_desc desc = { 10, "n1_10", {48, 0, 36, 0}, &GENUS, 0, 0, 0, 0 };
Chris@82 190
Chris@82 191 void X(codelet_n1_10) (planner *p) {
Chris@82 192 X(kdft_register) (p, n1_10, &desc);
Chris@82 193 }
Chris@82 194
Chris@82 195 #else
Chris@82 196
Chris@82 197 /* Generated by: ../../../genfft/gen_notw.native -compact -variables 4 -pipeline-latency 4 -n 10 -name n1_10 -include dft/scalar/n.h */
Chris@82 198
Chris@82 199 /*
Chris@82 200 * This function contains 84 FP additions, 24 FP multiplications,
Chris@82 201 * (or, 72 additions, 12 multiplications, 12 fused multiply/add),
Chris@82 202 * 41 stack variables, 4 constants, and 40 memory accesses
Chris@82 203 */
Chris@82 204 #include "dft/scalar/n.h"
Chris@82 205
Chris@82 206 static void n1_10(const R *ri, const R *ii, R *ro, R *io, stride is, stride os, INT v, INT ivs, INT ovs)
Chris@82 207 {
Chris@82 208 DK(KP250000000, +0.250000000000000000000000000000000000000000000);
Chris@82 209 DK(KP559016994, +0.559016994374947424102293417182819058860154590);
Chris@82 210 DK(KP587785252, +0.587785252292473129168705954639072768597652438);
Chris@82 211 DK(KP951056516, +0.951056516295153572116439333379382143405698634);
Chris@82 212 {
Chris@82 213 INT i;
Chris@82 214 for (i = v; i > 0; i = i - 1, ri = ri + ivs, ii = ii + ivs, ro = ro + ovs, io = io + ovs, MAKE_VOLATILE_STRIDE(40, is), MAKE_VOLATILE_STRIDE(40, os)) {
Chris@82 215 E T3, Tj, TQ, T1e, TU, TV, T1c, T1b, Tm, Tp, Tq, Ta, Th, Ti, TA;
Chris@82 216 E TH, T17, T14, T1f, T1g, T1h, TL, TM, TR;
Chris@82 217 {
Chris@82 218 E T1, T2, TO, TP;
Chris@82 219 T1 = ri[0];
Chris@82 220 T2 = ri[WS(is, 5)];
Chris@82 221 T3 = T1 - T2;
Chris@82 222 Tj = T1 + T2;
Chris@82 223 TO = ii[0];
Chris@82 224 TP = ii[WS(is, 5)];
Chris@82 225 TQ = TO - TP;
Chris@82 226 T1e = TO + TP;
Chris@82 227 }
Chris@82 228 {
Chris@82 229 E T6, Tk, Tg, To, T9, Tl, Td, Tn;
Chris@82 230 {
Chris@82 231 E T4, T5, Te, Tf;
Chris@82 232 T4 = ri[WS(is, 2)];
Chris@82 233 T5 = ri[WS(is, 7)];
Chris@82 234 T6 = T4 - T5;
Chris@82 235 Tk = T4 + T5;
Chris@82 236 Te = ri[WS(is, 6)];
Chris@82 237 Tf = ri[WS(is, 1)];
Chris@82 238 Tg = Te - Tf;
Chris@82 239 To = Te + Tf;
Chris@82 240 }
Chris@82 241 {
Chris@82 242 E T7, T8, Tb, Tc;
Chris@82 243 T7 = ri[WS(is, 8)];
Chris@82 244 T8 = ri[WS(is, 3)];
Chris@82 245 T9 = T7 - T8;
Chris@82 246 Tl = T7 + T8;
Chris@82 247 Tb = ri[WS(is, 4)];
Chris@82 248 Tc = ri[WS(is, 9)];
Chris@82 249 Td = Tb - Tc;
Chris@82 250 Tn = Tb + Tc;
Chris@82 251 }
Chris@82 252 TU = T6 - T9;
Chris@82 253 TV = Td - Tg;
Chris@82 254 T1c = Tk - Tl;
Chris@82 255 T1b = Tn - To;
Chris@82 256 Tm = Tk + Tl;
Chris@82 257 Tp = Tn + To;
Chris@82 258 Tq = Tm + Tp;
Chris@82 259 Ta = T6 + T9;
Chris@82 260 Th = Td + Tg;
Chris@82 261 Ti = Ta + Th;
Chris@82 262 }
Chris@82 263 {
Chris@82 264 E Tw, T15, TG, T13, Tz, T16, TD, T12;
Chris@82 265 {
Chris@82 266 E Tu, Tv, TE, TF;
Chris@82 267 Tu = ii[WS(is, 2)];
Chris@82 268 Tv = ii[WS(is, 7)];
Chris@82 269 Tw = Tu - Tv;
Chris@82 270 T15 = Tu + Tv;
Chris@82 271 TE = ii[WS(is, 6)];
Chris@82 272 TF = ii[WS(is, 1)];
Chris@82 273 TG = TE - TF;
Chris@82 274 T13 = TE + TF;
Chris@82 275 }
Chris@82 276 {
Chris@82 277 E Tx, Ty, TB, TC;
Chris@82 278 Tx = ii[WS(is, 8)];
Chris@82 279 Ty = ii[WS(is, 3)];
Chris@82 280 Tz = Tx - Ty;
Chris@82 281 T16 = Tx + Ty;
Chris@82 282 TB = ii[WS(is, 4)];
Chris@82 283 TC = ii[WS(is, 9)];
Chris@82 284 TD = TB - TC;
Chris@82 285 T12 = TB + TC;
Chris@82 286 }
Chris@82 287 TA = Tw - Tz;
Chris@82 288 TH = TD - TG;
Chris@82 289 T17 = T15 - T16;
Chris@82 290 T14 = T12 - T13;
Chris@82 291 T1f = T15 + T16;
Chris@82 292 T1g = T12 + T13;
Chris@82 293 T1h = T1f + T1g;
Chris@82 294 TL = Tw + Tz;
Chris@82 295 TM = TD + TG;
Chris@82 296 TR = TL + TM;
Chris@82 297 }
Chris@82 298 ro[WS(os, 5)] = T3 + Ti;
Chris@82 299 io[WS(os, 5)] = TQ + TR;
Chris@82 300 ro[0] = Tj + Tq;
Chris@82 301 io[0] = T1e + T1h;
Chris@82 302 {
Chris@82 303 E TI, TK, Tt, TJ, Tr, Ts;
Chris@82 304 TI = FMA(KP951056516, TA, KP587785252 * TH);
Chris@82 305 TK = FNMS(KP587785252, TA, KP951056516 * TH);
Chris@82 306 Tr = KP559016994 * (Ta - Th);
Chris@82 307 Ts = FNMS(KP250000000, Ti, T3);
Chris@82 308 Tt = Tr + Ts;
Chris@82 309 TJ = Ts - Tr;
Chris@82 310 ro[WS(os, 9)] = Tt - TI;
Chris@82 311 ro[WS(os, 3)] = TJ + TK;
Chris@82 312 ro[WS(os, 1)] = Tt + TI;
Chris@82 313 ro[WS(os, 7)] = TJ - TK;
Chris@82 314 }
Chris@82 315 {
Chris@82 316 E TW, TY, TT, TX, TN, TS;
Chris@82 317 TW = FMA(KP951056516, TU, KP587785252 * TV);
Chris@82 318 TY = FNMS(KP587785252, TU, KP951056516 * TV);
Chris@82 319 TN = KP559016994 * (TL - TM);
Chris@82 320 TS = FNMS(KP250000000, TR, TQ);
Chris@82 321 TT = TN + TS;
Chris@82 322 TX = TS - TN;
Chris@82 323 io[WS(os, 1)] = TT - TW;
Chris@82 324 io[WS(os, 7)] = TY + TX;
Chris@82 325 io[WS(os, 9)] = TW + TT;
Chris@82 326 io[WS(os, 3)] = TX - TY;
Chris@82 327 }
Chris@82 328 {
Chris@82 329 E T18, T1a, T11, T19, TZ, T10;
Chris@82 330 T18 = FNMS(KP587785252, T17, KP951056516 * T14);
Chris@82 331 T1a = FMA(KP951056516, T17, KP587785252 * T14);
Chris@82 332 TZ = FNMS(KP250000000, Tq, Tj);
Chris@82 333 T10 = KP559016994 * (Tm - Tp);
Chris@82 334 T11 = TZ - T10;
Chris@82 335 T19 = T10 + TZ;
Chris@82 336 ro[WS(os, 2)] = T11 - T18;
Chris@82 337 ro[WS(os, 6)] = T19 + T1a;
Chris@82 338 ro[WS(os, 8)] = T11 + T18;
Chris@82 339 ro[WS(os, 4)] = T19 - T1a;
Chris@82 340 }
Chris@82 341 {
Chris@82 342 E T1d, T1l, T1k, T1m, T1i, T1j;
Chris@82 343 T1d = FNMS(KP587785252, T1c, KP951056516 * T1b);
Chris@82 344 T1l = FMA(KP951056516, T1c, KP587785252 * T1b);
Chris@82 345 T1i = FNMS(KP250000000, T1h, T1e);
Chris@82 346 T1j = KP559016994 * (T1f - T1g);
Chris@82 347 T1k = T1i - T1j;
Chris@82 348 T1m = T1j + T1i;
Chris@82 349 io[WS(os, 2)] = T1d + T1k;
Chris@82 350 io[WS(os, 6)] = T1m - T1l;
Chris@82 351 io[WS(os, 8)] = T1k - T1d;
Chris@82 352 io[WS(os, 4)] = T1l + T1m;
Chris@82 353 }
Chris@82 354 }
Chris@82 355 }
Chris@82 356 }
Chris@82 357
Chris@82 358 static const kdft_desc desc = { 10, "n1_10", {72, 12, 12, 0}, &GENUS, 0, 0, 0, 0 };
Chris@82 359
Chris@82 360 void X(codelet_n1_10) (planner *p) {
Chris@82 361 X(kdft_register) (p, n1_10, &desc);
Chris@82 362 }
Chris@82 363
Chris@82 364 #endif