Mercurial > hg > sv-dependency-builds

annotate src/fftw-3.3.5/rdft/scalar/r2cb/hb_8.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 16:49:42 EDT 2016 */
Chris@42 23
Chris@42 24 #include "codelet-rdft.h"
Chris@42 25
Chris@42 26 #ifdef HAVE_FMA
Chris@42 27
Chris@42 28 /* Generated by: ../../../genfft/gen_hc2hc.native -fma -reorder-insns -schedule-for-pipeline -compact -variables 4 -pipeline-latency 4 -sign 1 -n 8 -dif -name hb_8 -include hb.h */
Chris@42 29
Chris@42 30 /*
Chris@42 31 * This function contains 66 FP additions, 36 FP multiplications,
Chris@42 32 * (or, 44 additions, 14 multiplications, 22 fused multiply/add),
Chris@42 33 * 52 stack variables, 1 constants, and 32 memory accesses
Chris@42 34 */
Chris@42 35 #include "hb.h"
Chris@42 36
Chris@42 37 static void hb_8(R *cr, R *ci, const R *W, stride rs, INT mb, INT me, INT ms)
Chris@42 38 {
Chris@42 39 DK(KP707106781, +0.707106781186547524400844362104849039284835938);
Chris@42 40 {
Chris@42 41 INT m;
Chris@42 42 for (m = mb, W = W + ((mb - 1) * 14); m < me; m = m + 1, cr = cr + ms, ci = ci - ms, W = W + 14, MAKE_VOLATILE_STRIDE(16, rs)) {
Chris@42 43 E Tw, TH, Tf, Ty, Tx, TI;
Chris@42 44 {
Chris@42 45 E TV, TD, T1i, T7, T1b, T1n, TQ, Tk, Tb, Tl, Ta, T1d, Tt, Tc, Tm;
Chris@42 46 E Tn;
Chris@42 47 {
Chris@42 48 E T4, Tg, T3, T19, TC, T5, Th, Ti;
Chris@42 49 {
Chris@42 50 E T1, T2, TA, TB;
Chris@42 51 T1 = cr[0];
Chris@42 52 T2 = ci[WS(rs, 3)];
Chris@42 53 TA = ci[WS(rs, 7)];
Chris@42 54 TB = cr[WS(rs, 4)];
Chris@42 55 T4 = cr[WS(rs, 2)];
Chris@42 56 Tg = T1 - T2;
Chris@42 57 T3 = T1 + T2;
Chris@42 58 T19 = TA - TB;
Chris@42 59 TC = TA + TB;
Chris@42 60 T5 = ci[WS(rs, 1)];
Chris@42 61 Th = ci[WS(rs, 5)];
Chris@42 62 Ti = cr[WS(rs, 6)];
Chris@42 63 }
Chris@42 64 {
Chris@42 65 E T8, T9, Tr, Ts;
Chris@42 66 T8 = cr[WS(rs, 1)];
Chris@42 67 {
Chris@42 68 E Tz, T6, T1a, Tj;
Chris@42 69 Tz = T4 - T5;
Chris@42 70 T6 = T4 + T5;
Chris@42 71 T1a = Th - Ti;
Chris@42 72 Tj = Th + Ti;
Chris@42 73 TV = TC - Tz;
Chris@42 74 TD = Tz + TC;
Chris@42 75 T1i = T3 - T6;
Chris@42 76 T7 = T3 + T6;
Chris@42 77 T1b = T19 + T1a;
Chris@42 78 T1n = T19 - T1a;
Chris@42 79 TQ = Tg + Tj;
Chris@42 80 Tk = Tg - Tj;
Chris@42 81 T9 = ci[WS(rs, 2)];
Chris@42 82 }
Chris@42 83 Tr = ci[WS(rs, 4)];
Chris@42 84 Ts = cr[WS(rs, 7)];
Chris@42 85 Tb = ci[0];
Chris@42 86 Tl = T8 - T9;
Chris@42 87 Ta = T8 + T9;
Chris@42 88 T1d = Tr - Ts;
Chris@42 89 Tt = Tr + Ts;
Chris@42 90 Tc = cr[WS(rs, 3)];
Chris@42 91 Tm = ci[WS(rs, 6)];
Chris@42 92 Tn = cr[WS(rs, 5)];
Chris@42 93 }
Chris@42 94 }
Chris@42 95 {
Chris@42 96 E Te, T1e, Tv, TG, T13, T1k, T1s, T10, T1p, T1v, T1u, T1w, T1t;
Chris@42 97 {
Chris@42 98 E TP, T1o, T1j, TR, TU, TX, TW;
Chris@42 99 TP = W[4];
Chris@42 100 {
Chris@42 101 E Tq, Td, T1c, To;
Chris@42 102 Tq = Tb - Tc;
Chris@42 103 Td = Tb + Tc;
Chris@42 104 T1c = Tm - Tn;
Chris@42 105 To = Tm + Tn;
Chris@42 106 {
Chris@42 107 E Tu, TF, Tp, TE;
Chris@42 108 Tu = Tq - Tt;
Chris@42 109 TF = Tq + Tt;
Chris@42 110 T1o = Ta - Td;
Chris@42 111 Te = Ta + Td;
Chris@42 112 T1j = T1d - T1c;
Chris@42 113 T1e = T1c + T1d;
Chris@42 114 Tp = Tl - To;
Chris@42 115 TE = Tl + To;
Chris@42 116 cr[0] = T7 + Te;
Chris@42 117 ci[0] = T1b + T1e;
Chris@42 118 TW = Tp - Tu;
Chris@42 119 Tv = Tp + Tu;
Chris@42 120 TR = TE + TF;
Chris@42 121 TG = TE - TF;
Chris@42 122 }
Chris@42 123 }
Chris@42 124 TU = W[5];
Chris@42 125 TX = FMA(KP707106781, TW, TV);
Chris@42 126 T13 = FNMS(KP707106781, TW, TV);
Chris@42 127 {
Chris@42 128 E TS, TY, T1r, TT;
Chris@42 129 T1k = T1i - T1j;
Chris@42 130 T1s = T1i + T1j;
Chris@42 131 TS = FNMS(KP707106781, TR, TQ);
Chris@42 132 T10 = FMA(KP707106781, TR, TQ);
Chris@42 133 T1p = T1n - T1o;
Chris@42 134 T1v = T1o + T1n;
Chris@42 135 TY = TP * TX;
Chris@42 136 T1r = W[2];
Chris@42 137 TT = TP * TS;
Chris@42 138 T1u = W[3];
Chris@42 139 ci[WS(rs, 3)] = FMA(TU, TS, TY);
Chris@42 140 T1w = T1r * T1v;
Chris@42 141 T1t = T1r * T1s;
Chris@42 142 cr[WS(rs, 3)] = FNMS(TU, TX, TT);
Chris@42 143 }
Chris@42 144 }
Chris@42 145 {
Chris@42 146 E T1f, T15, T18, T17, T1g, T1h, T1m;
Chris@42 147 {
Chris@42 148 E TZ, T12, T16, T14, T11;
Chris@42 149 ci[WS(rs, 2)] = FMA(T1u, T1s, T1w);
Chris@42 150 cr[WS(rs, 2)] = FNMS(T1u, T1v, T1t);
Chris@42 151 TZ = W[12];
Chris@42 152 T12 = W[13];
Chris@42 153 T1f = T1b - T1e;
Chris@42 154 T16 = T7 - Te;
Chris@42 155 T14 = TZ * T13;
Chris@42 156 T11 = TZ * T10;
Chris@42 157 T15 = W[6];
Chris@42 158 T18 = W[7];
Chris@42 159 ci[WS(rs, 7)] = FMA(T12, T10, T14);
Chris@42 160 cr[WS(rs, 7)] = FNMS(T12, T13, T11);
Chris@42 161 T17 = T15 * T16;
Chris@42 162 T1g = T18 * T16;
Chris@42 163 }
Chris@42 164 cr[WS(rs, 4)] = FNMS(T18, T1f, T17);
Chris@42 165 ci[WS(rs, 4)] = FMA(T15, T1f, T1g);
Chris@42 166 T1h = W[10];
Chris@42 167 T1m = W[11];
Chris@42 168 {
Chris@42 169 E TN, TJ, TM, TL, TO, TK, T1q, T1l;
Chris@42 170 Tw = FNMS(KP707106781, Tv, Tk);
Chris@42 171 TK = FMA(KP707106781, Tv, Tk);
Chris@42 172 T1q = T1h * T1p;
Chris@42 173 T1l = T1h * T1k;
Chris@42 174 TN = FMA(KP707106781, TG, TD);
Chris@42 175 TH = FNMS(KP707106781, TG, TD);
Chris@42 176 ci[WS(rs, 6)] = FMA(T1m, T1k, T1q);
Chris@42 177 cr[WS(rs, 6)] = FNMS(T1m, T1p, T1l);
Chris@42 178 TJ = W[0];
Chris@42 179 TM = W[1];
Chris@42 180 Tf = W[8];
Chris@42 181 TL = TJ * TK;
Chris@42 182 TO = TM * TK;
Chris@42 183 Ty = W[9];
Chris@42 184 Tx = Tf * Tw;
Chris@42 185 cr[WS(rs, 1)] = FNMS(TM, TN, TL);
Chris@42 186 ci[WS(rs, 1)] = FMA(TJ, TN, TO);
Chris@42 187 }
Chris@42 188 }
Chris@42 189 }
Chris@42 190 }
Chris@42 191 cr[WS(rs, 5)] = FNMS(Ty, TH, Tx);
Chris@42 192 TI = Ty * Tw;
Chris@42 193 ci[WS(rs, 5)] = FMA(Tf, TH, TI);
Chris@42 194 }
Chris@42 195 }
Chris@42 196 }
Chris@42 197
Chris@42 198 static const tw_instr twinstr[] = {
Chris@42 199 {TW_FULL, 1, 8},
Chris@42 200 {TW_NEXT, 1, 0}
Chris@42 201 };
Chris@42 202
Chris@42 203 static const hc2hc_desc desc = { 8, "hb_8", twinstr, &GENUS, {44, 14, 22, 0} };
Chris@42 204
Chris@42 205 void X(codelet_hb_8) (planner *p) {
Chris@42 206 X(khc2hc_register) (p, hb_8, &desc);
Chris@42 207 }
Chris@42 208 #else /* HAVE_FMA */
Chris@42 209
Chris@42 210 /* Generated by: ../../../genfft/gen_hc2hc.native -compact -variables 4 -pipeline-latency 4 -sign 1 -n 8 -dif -name hb_8 -include hb.h */
Chris@42 211
Chris@42 212 /*
Chris@42 213 * This function contains 66 FP additions, 32 FP multiplications,
Chris@42 214 * (or, 52 additions, 18 multiplications, 14 fused multiply/add),
Chris@42 215 * 30 stack variables, 1 constants, and 32 memory accesses
Chris@42 216 */
Chris@42 217 #include "hb.h"
Chris@42 218
Chris@42 219 static void hb_8(R *cr, R *ci, const R *W, stride rs, INT mb, INT me, INT ms)
Chris@42 220 {
Chris@42 221 DK(KP707106781, +0.707106781186547524400844362104849039284835938);
Chris@42 222 {
Chris@42 223 INT m;
Chris@42 224 for (m = mb, W = W + ((mb - 1) * 14); m < me; m = m + 1, cr = cr + ms, ci = ci - ms, W = W + 14, MAKE_VOLATILE_STRIDE(16, rs)) {
Chris@42 225 E T7, T18, T1c, To, Ty, TM, TY, TC, Te, TZ, T10, Tv, Tz, TP, TS;
Chris@42 226 E TD;
Chris@42 227 {
Chris@42 228 E T3, TK, Tn, TL, T6, TW, Tk, TX;
Chris@42 229 {
Chris@42 230 E T1, T2, Tl, Tm;
Chris@42 231 T1 = cr[0];
Chris@42 232 T2 = ci[WS(rs, 3)];
Chris@42 233 T3 = T1 + T2;
Chris@42 234 TK = T1 - T2;
Chris@42 235 Tl = ci[WS(rs, 5)];
Chris@42 236 Tm = cr[WS(rs, 6)];
Chris@42 237 Tn = Tl - Tm;
Chris@42 238 TL = Tl + Tm;
Chris@42 239 }
Chris@42 240 {
Chris@42 241 E T4, T5, Ti, Tj;
Chris@42 242 T4 = cr[WS(rs, 2)];
Chris@42 243 T5 = ci[WS(rs, 1)];
Chris@42 244 T6 = T4 + T5;
Chris@42 245 TW = T4 - T5;
Chris@42 246 Ti = ci[WS(rs, 7)];
Chris@42 247 Tj = cr[WS(rs, 4)];
Chris@42 248 Tk = Ti - Tj;
Chris@42 249 TX = Ti + Tj;
Chris@42 250 }
Chris@42 251 T7 = T3 + T6;
Chris@42 252 T18 = TK + TL;
Chris@42 253 T1c = TX - TW;
Chris@42 254 To = Tk + Tn;
Chris@42 255 Ty = T3 - T6;
Chris@42 256 TM = TK - TL;
Chris@42 257 TY = TW + TX;
Chris@42 258 TC = Tk - Tn;
Chris@42 259 }
Chris@42 260 {
Chris@42 261 E Ta, TN, Tu, TR, Td, TQ, Tr, TO;
Chris@42 262 {
Chris@42 263 E T8, T9, Ts, Tt;
Chris@42 264 T8 = cr[WS(rs, 1)];
Chris@42 265 T9 = ci[WS(rs, 2)];
Chris@42 266 Ta = T8 + T9;
Chris@42 267 TN = T8 - T9;
Chris@42 268 Ts = ci[WS(rs, 4)];
Chris@42 269 Tt = cr[WS(rs, 7)];
Chris@42 270 Tu = Ts - Tt;
Chris@42 271 TR = Ts + Tt;
Chris@42 272 }
Chris@42 273 {
Chris@42 274 E Tb, Tc, Tp, Tq;
Chris@42 275 Tb = ci[0];
Chris@42 276 Tc = cr[WS(rs, 3)];
Chris@42 277 Td = Tb + Tc;
Chris@42 278 TQ = Tb - Tc;
Chris@42 279 Tp = ci[WS(rs, 6)];
Chris@42 280 Tq = cr[WS(rs, 5)];
Chris@42 281 Tr = Tp - Tq;
Chris@42 282 TO = Tp + Tq;
Chris@42 283 }
Chris@42 284 Te = Ta + Td;
Chris@42 285 TZ = TN + TO;
Chris@42 286 T10 = TQ + TR;
Chris@42 287 Tv = Tr + Tu;
Chris@42 288 Tz = Tu - Tr;
Chris@42 289 TP = TN - TO;
Chris@42 290 TS = TQ - TR;
Chris@42 291 TD = Ta - Td;
Chris@42 292 }
Chris@42 293 cr[0] = T7 + Te;
Chris@42 294 ci[0] = To + Tv;
Chris@42 295 {
Chris@42 296 E Tg, Tw, Tf, Th;
Chris@42 297 Tg = T7 - Te;
Chris@42 298 Tw = To - Tv;
Chris@42 299 Tf = W[6];
Chris@42 300 Th = W[7];
Chris@42 301 cr[WS(rs, 4)] = FNMS(Th, Tw, Tf * Tg);
Chris@42 302 ci[WS(rs, 4)] = FMA(Th, Tg, Tf * Tw);
Chris@42 303 }
Chris@42 304 {
Chris@42 305 E TG, TI, TF, TH;
Chris@42 306 TG = Ty + Tz;
Chris@42 307 TI = TD + TC;
Chris@42 308 TF = W[2];
Chris@42 309 TH = W[3];
Chris@42 310 cr[WS(rs, 2)] = FNMS(TH, TI, TF * TG);
Chris@42 311 ci[WS(rs, 2)] = FMA(TF, TI, TH * TG);
Chris@42 312 }
Chris@42 313 {
Chris@42 314 E TA, TE, Tx, TB;
Chris@42 315 TA = Ty - Tz;
Chris@42 316 TE = TC - TD;
Chris@42 317 Tx = W[10];
Chris@42 318 TB = W[11];
Chris@42 319 cr[WS(rs, 6)] = FNMS(TB, TE, Tx * TA);
Chris@42 320 ci[WS(rs, 6)] = FMA(Tx, TE, TB * TA);
Chris@42 321 }
Chris@42 322 {
Chris@42 323 E T1a, T1g, T1e, T1i, T19, T1d;
Chris@42 324 T19 = KP707106781 * (TZ + T10);
Chris@42 325 T1a = T18 - T19;
Chris@42 326 T1g = T18 + T19;
Chris@42 327 T1d = KP707106781 * (TP - TS);
Chris@42 328 T1e = T1c + T1d;
Chris@42 329 T1i = T1c - T1d;
Chris@42 330 {
Chris@42 331 E T17, T1b, T1f, T1h;
Chris@42 332 T17 = W[4];
Chris@42 333 T1b = W[5];
Chris@42 334 cr[WS(rs, 3)] = FNMS(T1b, T1e, T17 * T1a);
Chris@42 335 ci[WS(rs, 3)] = FMA(T17, T1e, T1b * T1a);
Chris@42 336 T1f = W[12];
Chris@42 337 T1h = W[13];
Chris@42 338 cr[WS(rs, 7)] = FNMS(T1h, T1i, T1f * T1g);
Chris@42 339 ci[WS(rs, 7)] = FMA(T1f, T1i, T1h * T1g);
Chris@42 340 }
Chris@42 341 }
Chris@42 342 {
Chris@42 343 E TU, T14, T12, T16, TT, T11;
Chris@42 344 TT = KP707106781 * (TP + TS);
Chris@42 345 TU = TM - TT;
Chris@42 346 T14 = TM + TT;
Chris@42 347 T11 = KP707106781 * (TZ - T10);
Chris@42 348 T12 = TY - T11;
Chris@42 349 T16 = TY + T11;
Chris@42 350 {
Chris@42 351 E TJ, TV, T13, T15;
Chris@42 352 TJ = W[8];
Chris@42 353 TV = W[9];
Chris@42 354 cr[WS(rs, 5)] = FNMS(TV, T12, TJ * TU);
Chris@42 355 ci[WS(rs, 5)] = FMA(TV, TU, TJ * T12);
Chris@42 356 T13 = W[0];
Chris@42 357 T15 = W[1];
Chris@42 358 cr[WS(rs, 1)] = FNMS(T15, T16, T13 * T14);
Chris@42 359 ci[WS(rs, 1)] = FMA(T15, T14, T13 * T16);
Chris@42 360 }
Chris@42 361 }
Chris@42 362 }
Chris@42 363 }
Chris@42 364 }
Chris@42 365
Chris@42 366 static const tw_instr twinstr[] = {
Chris@42 367 {TW_FULL, 1, 8},
Chris@42 368 {TW_NEXT, 1, 0}
Chris@42 369 };
Chris@42 370
Chris@42 371 static const hc2hc_desc desc = { 8, "hb_8", twinstr, &GENUS, {52, 18, 14, 0} };
Chris@42 372
Chris@42 373 void X(codelet_hb_8) (planner *p) {
Chris@42 374 X(khc2hc_register) (p, hb_8, &desc);
Chris@42 375 }
Chris@42 376 #endif /* HAVE_FMA */