Mercurial > hg > sv-dependency-builds

annotate src/fftw-3.3.8/rdft/scalar/r2cb/hb2_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 08:07:37 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_hc2hc.native -fma -compact -variables 4 -pipeline-latency 4 -sign 1 -twiddle-log3 -precompute-twiddles -n 8 -dif -name hb2_8 -include rdft/scalar/hb.h */
Chris@82 29
Chris@82 30 /*
Chris@82 31 * This function contains 74 FP additions, 50 FP multiplications,
Chris@82 32 * (or, 44 additions, 20 multiplications, 30 fused multiply/add),
Chris@82 33 * 47 stack variables, 1 constants, and 32 memory accesses
Chris@82 34 */
Chris@82 35 #include "rdft/scalar/hb.h"
Chris@82 36
Chris@82 37 static void hb2_8(R *cr, R *ci, 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) * 6); m < me; m = m + 1, cr = cr + ms, ci = ci - ms, W = W + 6, MAKE_VOLATILE_STRIDE(16, rs)) {
Chris@82 43 E Tf, Tg, Tl, Tp, Ti, Tj, Tk, T1b, T1u, T1e, T1o, To, Tq, TK;
Chris@82 44 {
Chris@82 45 E Th, T1n, T1t, Tn, Tm, TJ;
Chris@82 46 Tf = W[0];
Chris@82 47 Tg = W[2];
Chris@82 48 Th = Tf * Tg;
Chris@82 49 Tl = W[4];
Chris@82 50 T1n = Tf * Tl;
Chris@82 51 Tp = W[5];
Chris@82 52 T1t = Tf * Tp;
Chris@82 53 Ti = W[1];
Chris@82 54 Tj = W[3];
Chris@82 55 Tn = Tf * Tj;
Chris@82 56 Tk = FMA(Ti, Tj, Th);
Chris@82 57 T1b = FNMS(Ti, Tj, Th);
Chris@82 58 T1u = FNMS(Ti, Tl, T1t);
Chris@82 59 T1e = FMA(Ti, Tg, Tn);
Chris@82 60 T1o = FMA(Ti, Tp, T1n);
Chris@82 61 Tm = Tk * Tl;
Chris@82 62 TJ = Tk * Tp;
Chris@82 63 To = FNMS(Ti, Tg, Tn);
Chris@82 64 Tq = FMA(To, Tp, Tm);
Chris@82 65 TK = FNMS(To, Tl, TJ);
Chris@82 66 }
Chris@82 67 {
Chris@82 68 E T7, T1p, T1v, Tv, TP, T13, T1h, TZ, Te, T1k, T1w, T1q, TQ, TR, T10;
Chris@82 69 E TG, T14;
Chris@82 70 {
Chris@82 71 E T3, Tr, TO, T1f, T6, TL, Tu, T1g;
Chris@82 72 {
Chris@82 73 E T1, T2, TM, TN;
Chris@82 74 T1 = cr[0];
Chris@82 75 T2 = ci[WS(rs, 3)];
Chris@82 76 T3 = T1 + T2;
Chris@82 77 Tr = T1 - T2;
Chris@82 78 TM = ci[WS(rs, 7)];
Chris@82 79 TN = cr[WS(rs, 4)];
Chris@82 80 TO = TM + TN;
Chris@82 81 T1f = TM - TN;
Chris@82 82 }
Chris@82 83 {
Chris@82 84 E T4, T5, Ts, Tt;
Chris@82 85 T4 = cr[WS(rs, 2)];
Chris@82 86 T5 = ci[WS(rs, 1)];
Chris@82 87 T6 = T4 + T5;
Chris@82 88 TL = T4 - T5;
Chris@82 89 Ts = ci[WS(rs, 5)];
Chris@82 90 Tt = cr[WS(rs, 6)];
Chris@82 91 Tu = Ts + Tt;
Chris@82 92 T1g = Ts - Tt;
Chris@82 93 }
Chris@82 94 T7 = T3 + T6;
Chris@82 95 T1p = T3 - T6;
Chris@82 96 T1v = T1f - T1g;
Chris@82 97 Tv = Tr - Tu;
Chris@82 98 TP = TL + TO;
Chris@82 99 T13 = TO - TL;
Chris@82 100 T1h = T1f + T1g;
Chris@82 101 TZ = Tr + Tu;
Chris@82 102 }
Chris@82 103 {
Chris@82 104 E Ta, Tw, TE, T1j, Td, TB, Tz, T1i, TA, TF;
Chris@82 105 {
Chris@82 106 E T8, T9, TC, TD;
Chris@82 107 T8 = cr[WS(rs, 1)];
Chris@82 108 T9 = ci[WS(rs, 2)];
Chris@82 109 Ta = T8 + T9;
Chris@82 110 Tw = T8 - T9;
Chris@82 111 TC = ci[WS(rs, 4)];
Chris@82 112 TD = cr[WS(rs, 7)];
Chris@82 113 TE = TC + TD;
Chris@82 114 T1j = TC - TD;
Chris@82 115 }
Chris@82 116 {
Chris@82 117 E Tb, Tc, Tx, Ty;
Chris@82 118 Tb = ci[0];
Chris@82 119 Tc = cr[WS(rs, 3)];
Chris@82 120 Td = Tb + Tc;
Chris@82 121 TB = Tb - Tc;
Chris@82 122 Tx = ci[WS(rs, 6)];
Chris@82 123 Ty = cr[WS(rs, 5)];
Chris@82 124 Tz = Tx + Ty;
Chris@82 125 T1i = Tx - Ty;
Chris@82 126 }
Chris@82 127 Te = Ta + Td;
Chris@82 128 T1k = T1i + T1j;
Chris@82 129 T1w = Ta - Td;
Chris@82 130 T1q = T1j - T1i;
Chris@82 131 TQ = Tw + Tz;
Chris@82 132 TR = TB + TE;
Chris@82 133 T10 = TQ + TR;
Chris@82 134 TA = Tw - Tz;
Chris@82 135 TF = TB - TE;
Chris@82 136 TG = TA + TF;
Chris@82 137 T14 = TA - TF;
Chris@82 138 }
Chris@82 139 cr[0] = T7 + Te;
Chris@82 140 ci[0] = T1h + T1k;
Chris@82 141 {
Chris@82 142 E T11, T12, T15, T16;
Chris@82 143 T11 = FNMS(KP707106781, T10, TZ);
Chris@82 144 T12 = Tg * T11;
Chris@82 145 T15 = FMA(KP707106781, T14, T13);
Chris@82 146 T16 = Tg * T15;
Chris@82 147 cr[WS(rs, 3)] = FNMS(Tj, T15, T12);
Chris@82 148 ci[WS(rs, 3)] = FMA(Tj, T11, T16);
Chris@82 149 }
Chris@82 150 {
Chris@82 151 E T1z, T1A, T1B, T1C;
Chris@82 152 T1z = T1p + T1q;
Chris@82 153 T1A = Tk * T1z;
Chris@82 154 T1B = T1w + T1v;
Chris@82 155 T1C = Tk * T1B;
Chris@82 156 cr[WS(rs, 2)] = FNMS(To, T1B, T1A);
Chris@82 157 ci[WS(rs, 2)] = FMA(To, T1z, T1C);
Chris@82 158 }
Chris@82 159 {
Chris@82 160 E T17, T18, T19, T1a;
Chris@82 161 T17 = FMA(KP707106781, T10, TZ);
Chris@82 162 T18 = Tl * T17;
Chris@82 163 T19 = FNMS(KP707106781, T14, T13);
Chris@82 164 T1a = Tl * T19;
Chris@82 165 cr[WS(rs, 7)] = FNMS(Tp, T19, T18);
Chris@82 166 ci[WS(rs, 7)] = FMA(Tp, T17, T1a);
Chris@82 167 }
Chris@82 168 {
Chris@82 169 E T1l, T1d, T1m, T1c;
Chris@82 170 T1l = T1h - T1k;
Chris@82 171 T1c = T7 - Te;
Chris@82 172 T1d = T1b * T1c;
Chris@82 173 T1m = T1e * T1c;
Chris@82 174 cr[WS(rs, 4)] = FNMS(T1e, T1l, T1d);
Chris@82 175 ci[WS(rs, 4)] = FMA(T1b, T1l, T1m);
Chris@82 176 }
Chris@82 177 {
Chris@82 178 E T1r, T1s, T1x, T1y;
Chris@82 179 T1r = T1p - T1q;
Chris@82 180 T1s = T1o * T1r;
Chris@82 181 T1x = T1v - T1w;
Chris@82 182 T1y = T1o * T1x;
Chris@82 183 cr[WS(rs, 6)] = FNMS(T1u, T1x, T1s);
Chris@82 184 ci[WS(rs, 6)] = FMA(T1u, T1r, T1y);
Chris@82 185 }
Chris@82 186 {
Chris@82 187 E TT, TX, TW, TY, TI, TU, TS, TV, TH;
Chris@82 188 TS = TQ - TR;
Chris@82 189 TT = FNMS(KP707106781, TS, TP);
Chris@82 190 TX = FMA(KP707106781, TS, TP);
Chris@82 191 TV = FMA(KP707106781, TG, Tv);
Chris@82 192 TW = Tf * TV;
Chris@82 193 TY = Ti * TV;
Chris@82 194 TH = FNMS(KP707106781, TG, Tv);
Chris@82 195 TI = Tq * TH;
Chris@82 196 TU = TK * TH;
Chris@82 197 cr[WS(rs, 5)] = FNMS(TK, TT, TI);
Chris@82 198 ci[WS(rs, 5)] = FMA(Tq, TT, TU);
Chris@82 199 cr[WS(rs, 1)] = FNMS(Ti, TX, TW);
Chris@82 200 ci[WS(rs, 1)] = FMA(Tf, TX, TY);
Chris@82 201 }
Chris@82 202 }
Chris@82 203 }
Chris@82 204 }
Chris@82 205 }
Chris@82 206
Chris@82 207 static const tw_instr twinstr[] = {
Chris@82 208 {TW_CEXP, 1, 1},
Chris@82 209 {TW_CEXP, 1, 3},
Chris@82 210 {TW_CEXP, 1, 7},
Chris@82 211 {TW_NEXT, 1, 0}
Chris@82 212 };
Chris@82 213
Chris@82 214 static const hc2hc_desc desc = { 8, "hb2_8", twinstr, &GENUS, {44, 20, 30, 0} };
Chris@82 215
Chris@82 216 void X(codelet_hb2_8) (planner *p) {
Chris@82 217 X(khc2hc_register) (p, hb2_8, &desc);
Chris@82 218 }
Chris@82 219 #else
Chris@82 220
Chris@82 221 /* Generated by: ../../../genfft/gen_hc2hc.native -compact -variables 4 -pipeline-latency 4 -sign 1 -twiddle-log3 -precompute-twiddles -n 8 -dif -name hb2_8 -include rdft/scalar/hb.h */
Chris@82 222
Chris@82 223 /*
Chris@82 224 * This function contains 74 FP additions, 44 FP multiplications,
Chris@82 225 * (or, 56 additions, 26 multiplications, 18 fused multiply/add),
Chris@82 226 * 46 stack variables, 1 constants, and 32 memory accesses
Chris@82 227 */
Chris@82 228 #include "rdft/scalar/hb.h"
Chris@82 229
Chris@82 230 static void hb2_8(R *cr, R *ci, const R *W, stride rs, INT mb, INT me, INT ms)
Chris@82 231 {
Chris@82 232 DK(KP707106781, +0.707106781186547524400844362104849039284835938);
Chris@82 233 {
Chris@82 234 INT m;
Chris@82 235 for (m = mb, W = W + ((mb - 1) * 6); m < me; m = m + 1, cr = cr + ms, ci = ci - ms, W = W + 6, MAKE_VOLATILE_STRIDE(16, rs)) {
Chris@82 236 E Tf, Ti, Tg, Tj, Tl, Tp, TP, TR, TF, TG, TH, T15, TL, TT;
Chris@82 237 {
Chris@82 238 E Th, To, Tk, Tn;
Chris@82 239 Tf = W[0];
Chris@82 240 Ti = W[1];
Chris@82 241 Tg = W[2];
Chris@82 242 Tj = W[3];
Chris@82 243 Th = Tf * Tg;
Chris@82 244 To = Ti * Tg;
Chris@82 245 Tk = Ti * Tj;
Chris@82 246 Tn = Tf * Tj;
Chris@82 247 Tl = Th - Tk;
Chris@82 248 Tp = Tn + To;
Chris@82 249 TP = Th + Tk;
Chris@82 250 TR = Tn - To;
Chris@82 251 TF = W[4];
Chris@82 252 TG = W[5];
Chris@82 253 TH = FMA(Tf, TF, Ti * TG);
Chris@82 254 T15 = FNMS(TR, TF, TP * TG);
Chris@82 255 TL = FNMS(Ti, TF, Tf * TG);
Chris@82 256 TT = FMA(TP, TF, TR * TG);
Chris@82 257 }
Chris@82 258 {
Chris@82 259 E T7, T1f, T1i, Tw, TI, TW, T18, TM, Te, T19, T1a, TD, TJ, TZ, T12;
Chris@82 260 E TN, Tm, TE;
Chris@82 261 {
Chris@82 262 E T3, TU, Tv, TV, T6, T16, Ts, T17;
Chris@82 263 {
Chris@82 264 E T1, T2, Tt, Tu;
Chris@82 265 T1 = cr[0];
Chris@82 266 T2 = ci[WS(rs, 3)];
Chris@82 267 T3 = T1 + T2;
Chris@82 268 TU = T1 - T2;
Chris@82 269 Tt = ci[WS(rs, 5)];
Chris@82 270 Tu = cr[WS(rs, 6)];
Chris@82 271 Tv = Tt - Tu;
Chris@82 272 TV = Tt + Tu;
Chris@82 273 }
Chris@82 274 {
Chris@82 275 E T4, T5, Tq, Tr;
Chris@82 276 T4 = cr[WS(rs, 2)];
Chris@82 277 T5 = ci[WS(rs, 1)];
Chris@82 278 T6 = T4 + T5;
Chris@82 279 T16 = T4 - T5;
Chris@82 280 Tq = ci[WS(rs, 7)];
Chris@82 281 Tr = cr[WS(rs, 4)];
Chris@82 282 Ts = Tq - Tr;
Chris@82 283 T17 = Tq + Tr;
Chris@82 284 }
Chris@82 285 T7 = T3 + T6;
Chris@82 286 T1f = TU + TV;
Chris@82 287 T1i = T17 - T16;
Chris@82 288 Tw = Ts + Tv;
Chris@82 289 TI = T3 - T6;
Chris@82 290 TW = TU - TV;
Chris@82 291 T18 = T16 + T17;
Chris@82 292 TM = Ts - Tv;
Chris@82 293 }
Chris@82 294 {
Chris@82 295 E Ta, TX, TC, T11, Td, T10, Tz, TY;
Chris@82 296 {
Chris@82 297 E T8, T9, TA, TB;
Chris@82 298 T8 = cr[WS(rs, 1)];
Chris@82 299 T9 = ci[WS(rs, 2)];
Chris@82 300 Ta = T8 + T9;
Chris@82 301 TX = T8 - T9;
Chris@82 302 TA = ci[WS(rs, 4)];
Chris@82 303 TB = cr[WS(rs, 7)];
Chris@82 304 TC = TA - TB;
Chris@82 305 T11 = TA + TB;
Chris@82 306 }
Chris@82 307 {
Chris@82 308 E Tb, Tc, Tx, Ty;
Chris@82 309 Tb = ci[0];
Chris@82 310 Tc = cr[WS(rs, 3)];
Chris@82 311 Td = Tb + Tc;
Chris@82 312 T10 = Tb - Tc;
Chris@82 313 Tx = ci[WS(rs, 6)];
Chris@82 314 Ty = cr[WS(rs, 5)];
Chris@82 315 Tz = Tx - Ty;
Chris@82 316 TY = Tx + Ty;
Chris@82 317 }
Chris@82 318 Te = Ta + Td;
Chris@82 319 T19 = TX + TY;
Chris@82 320 T1a = T10 + T11;
Chris@82 321 TD = Tz + TC;
Chris@82 322 TJ = TC - Tz;
Chris@82 323 TZ = TX - TY;
Chris@82 324 T12 = T10 - T11;
Chris@82 325 TN = Ta - Td;
Chris@82 326 }
Chris@82 327 cr[0] = T7 + Te;
Chris@82 328 ci[0] = Tw + TD;
Chris@82 329 Tm = T7 - Te;
Chris@82 330 TE = Tw - TD;
Chris@82 331 cr[WS(rs, 4)] = FNMS(Tp, TE, Tl * Tm);
Chris@82 332 ci[WS(rs, 4)] = FMA(Tp, Tm, Tl * TE);
Chris@82 333 {
Chris@82 334 E TQ, TS, TK, TO;
Chris@82 335 TQ = TI + TJ;
Chris@82 336 TS = TN + TM;
Chris@82 337 cr[WS(rs, 2)] = FNMS(TR, TS, TP * TQ);
Chris@82 338 ci[WS(rs, 2)] = FMA(TP, TS, TR * TQ);
Chris@82 339 TK = TI - TJ;
Chris@82 340 TO = TM - TN;
Chris@82 341 cr[WS(rs, 6)] = FNMS(TL, TO, TH * TK);
Chris@82 342 ci[WS(rs, 6)] = FMA(TH, TO, TL * TK);
Chris@82 343 }
Chris@82 344 {
Chris@82 345 E T1h, T1l, T1k, T1m, T1g, T1j;
Chris@82 346 T1g = KP707106781 * (T19 + T1a);
Chris@82 347 T1h = T1f - T1g;
Chris@82 348 T1l = T1f + T1g;
Chris@82 349 T1j = KP707106781 * (TZ - T12);
Chris@82 350 T1k = T1i + T1j;
Chris@82 351 T1m = T1i - T1j;
Chris@82 352 cr[WS(rs, 3)] = FNMS(Tj, T1k, Tg * T1h);
Chris@82 353 ci[WS(rs, 3)] = FMA(Tg, T1k, Tj * T1h);
Chris@82 354 cr[WS(rs, 7)] = FNMS(TG, T1m, TF * T1l);
Chris@82 355 ci[WS(rs, 7)] = FMA(TF, T1m, TG * T1l);
Chris@82 356 }
Chris@82 357 {
Chris@82 358 E T14, T1d, T1c, T1e, T13, T1b;
Chris@82 359 T13 = KP707106781 * (TZ + T12);
Chris@82 360 T14 = TW - T13;
Chris@82 361 T1d = TW + T13;
Chris@82 362 T1b = KP707106781 * (T19 - T1a);
Chris@82 363 T1c = T18 - T1b;
Chris@82 364 T1e = T18 + T1b;
Chris@82 365 cr[WS(rs, 5)] = FNMS(T15, T1c, TT * T14);
Chris@82 366 ci[WS(rs, 5)] = FMA(T15, T14, TT * T1c);
Chris@82 367 cr[WS(rs, 1)] = FNMS(Ti, T1e, Tf * T1d);
Chris@82 368 ci[WS(rs, 1)] = FMA(Ti, T1d, Tf * T1e);
Chris@82 369 }
Chris@82 370 }
Chris@82 371 }
Chris@82 372 }
Chris@82 373 }
Chris@82 374
Chris@82 375 static const tw_instr twinstr[] = {
Chris@82 376 {TW_CEXP, 1, 1},
Chris@82 377 {TW_CEXP, 1, 3},
Chris@82 378 {TW_CEXP, 1, 7},
Chris@82 379 {TW_NEXT, 1, 0}
Chris@82 380 };
Chris@82 381
Chris@82 382 static const hc2hc_desc desc = { 8, "hb2_8", twinstr, &GENUS, {56, 26, 18, 0} };
Chris@82 383
Chris@82 384 void X(codelet_hb2_8) (planner *p) {
Chris@82 385 X(khc2hc_register) (p, hb2_8, &desc);
Chris@82 386 }
Chris@82 387 #endif