Mercurial > hg > sv-dependency-builds

annotate src/fftw-3.3.5/rdft/scalar/r2cb/hb2_8.c @ 56:af97cad61ff0

Find changesets by keywords (author, files, the commit message), revision number or hash, or revset expression.
Add updated build of PortAudio for OSX
author Chris Cannam <cannam@all-day-breakfast.com>
date Tue, 03 Jan 2017 15:10:52 +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 16:50:10 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 -twiddle-log3 -precompute-twiddles -n 8 -dif -name hb2_8 -include hb.h */
Chris@42 29
Chris@42 30 /*
Chris@42 31 * This function contains 74 FP additions, 50 FP multiplications,
Chris@42 32 * (or, 44 additions, 20 multiplications, 30 fused multiply/add),
Chris@42 33 * 77 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 hb2_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) * 6); m < me; m = m + 1, cr = cr + ms, ci = ci - ms, W = W + 6, MAKE_VOLATILE_STRIDE(16, rs)) {
Chris@42 43 E Tf, Tg, Tl, Tp, Ti, Tj, T1o, T1u, Tk, T1b, To, T1e, TK, Tq, T13;
Chris@42 44 E TP, T1p, T7, T1h, T1v, TZ, Tv, Tw, Ta, Tx, T1j, TE, TB, Td, Ty;
Chris@42 45 E Th, T1n, T1t;
Chris@42 46 Tf = W[0];
Chris@42 47 Tg = W[2];
Chris@42 48 Tl = W[4];
Chris@42 49 Tp = W[5];
Chris@42 50 Ti = W[1];
Chris@42 51 Th = Tf * Tg;
Chris@42 52 T1n = Tf * Tl;
Chris@42 53 T1t = Tf * Tp;
Chris@42 54 Tj = W[3];
Chris@42 55 {
Chris@42 56 E Tr, T3, Ts, T1f, TO, TL, T6, Tt;
Chris@42 57 {
Chris@42 58 E TM, TN, T4, T5;
Chris@42 59 {
Chris@42 60 E T1, Tn, T2, TJ, Tm;
Chris@42 61 T1 = cr[0];
Chris@42 62 T1o = FMA(Ti, Tp, T1n);
Chris@42 63 T1u = FNMS(Ti, Tl, T1t);
Chris@42 64 Tk = FMA(Ti, Tj, Th);
Chris@42 65 T1b = FNMS(Ti, Tj, Th);
Chris@42 66 Tn = Tf * Tj;
Chris@42 67 T2 = ci[WS(rs, 3)];
Chris@42 68 TM = ci[WS(rs, 7)];
Chris@42 69 TJ = Tk * Tp;
Chris@42 70 Tm = Tk * Tl;
Chris@42 71 To = FNMS(Ti, Tg, Tn);
Chris@42 72 T1e = FMA(Ti, Tg, Tn);
Chris@42 73 Tr = T1 - T2;
Chris@42 74 T3 = T1 + T2;
Chris@42 75 TK = FNMS(To, Tl, TJ);
Chris@42 76 Tq = FMA(To, Tp, Tm);
Chris@42 77 TN = cr[WS(rs, 4)];
Chris@42 78 }
Chris@42 79 T4 = cr[WS(rs, 2)];
Chris@42 80 T5 = ci[WS(rs, 1)];
Chris@42 81 Ts = ci[WS(rs, 5)];
Chris@42 82 T1f = TM - TN;
Chris@42 83 TO = TM + TN;
Chris@42 84 TL = T4 - T5;
Chris@42 85 T6 = T4 + T5;
Chris@42 86 Tt = cr[WS(rs, 6)];
Chris@42 87 }
Chris@42 88 {
Chris@42 89 E TC, TD, Tb, Tc;
Chris@42 90 {
Chris@42 91 E T8, T1g, Tu, T9;
Chris@42 92 T8 = cr[WS(rs, 1)];
Chris@42 93 T13 = TO - TL;
Chris@42 94 TP = TL + TO;
Chris@42 95 T1p = T3 - T6;
Chris@42 96 T7 = T3 + T6;
Chris@42 97 T1g = Ts - Tt;
Chris@42 98 Tu = Ts + Tt;
Chris@42 99 T9 = ci[WS(rs, 2)];
Chris@42 100 TC = ci[WS(rs, 4)];
Chris@42 101 T1h = T1f + T1g;
Chris@42 102 T1v = T1f - T1g;
Chris@42 103 TZ = Tr + Tu;
Chris@42 104 Tv = Tr - Tu;
Chris@42 105 Tw = T8 - T9;
Chris@42 106 Ta = T8 + T9;
Chris@42 107 TD = cr[WS(rs, 7)];
Chris@42 108 }
Chris@42 109 Tb = ci[0];
Chris@42 110 Tc = cr[WS(rs, 3)];
Chris@42 111 Tx = ci[WS(rs, 6)];
Chris@42 112 T1j = TC - TD;
Chris@42 113 TE = TC + TD;
Chris@42 114 TB = Tb - Tc;
Chris@42 115 Td = Tb + Tc;
Chris@42 116 Ty = cr[WS(rs, 5)];
Chris@42 117 }
Chris@42 118 }
Chris@42 119 {
Chris@42 120 E TR, TF, Te, T1w;
Chris@42 121 TR = TB + TE;
Chris@42 122 TF = TB - TE;
Chris@42 123 Te = Ta + Td;
Chris@42 124 T1w = Ta - Td;
Chris@42 125 {
Chris@42 126 E Tz, T1i, T1B, T1x, T1c;
Chris@42 127 Tz = Tx + Ty;
Chris@42 128 T1i = Tx - Ty;
Chris@42 129 T1B = T1w + T1v;
Chris@42 130 T1x = T1v - T1w;
Chris@42 131 T1c = T7 - Te;
Chris@42 132 cr[0] = T7 + Te;
Chris@42 133 {
Chris@42 134 E T1k, T1q, TQ, TA;
Chris@42 135 T1k = T1i + T1j;
Chris@42 136 T1q = T1j - T1i;
Chris@42 137 TQ = Tw + Tz;
Chris@42 138 TA = Tw - Tz;
Chris@42 139 {
Chris@42 140 E T1y, T1C, T1m, T1d;
Chris@42 141 T1y = T1o * T1x;
Chris@42 142 T1C = Tk * T1B;
Chris@42 143 T1m = T1e * T1c;
Chris@42 144 T1d = T1b * T1c;
Chris@42 145 {
Chris@42 146 E T1z, T1r, T1l, TG, T14;
Chris@42 147 T1z = T1p + T1q;
Chris@42 148 T1r = T1p - T1q;
Chris@42 149 T1l = T1h - T1k;
Chris@42 150 ci[0] = T1h + T1k;
Chris@42 151 TG = TA + TF;
Chris@42 152 T14 = TA - TF;
Chris@42 153 {
Chris@42 154 E T10, TS, T1s, T1A;
Chris@42 155 T10 = TQ + TR;
Chris@42 156 TS = TQ - TR;
Chris@42 157 ci[WS(rs, 6)] = FMA(T1u, T1r, T1y);
Chris@42 158 T1s = T1o * T1r;
Chris@42 159 ci[WS(rs, 2)] = FMA(To, T1z, T1C);
Chris@42 160 T1A = Tk * T1z;
Chris@42 161 ci[WS(rs, 4)] = FMA(T1b, T1l, T1m);
Chris@42 162 cr[WS(rs, 4)] = FNMS(T1e, T1l, T1d);
Chris@42 163 {
Chris@42 164 E T15, T19, TV, TH;
Chris@42 165 T15 = FMA(KP707106781, T14, T13);
Chris@42 166 T19 = FNMS(KP707106781, T14, T13);
Chris@42 167 TV = FMA(KP707106781, TG, Tv);
Chris@42 168 TH = FNMS(KP707106781, TG, Tv);
Chris@42 169 {
Chris@42 170 E TT, TX, T11, T17;
Chris@42 171 TT = FNMS(KP707106781, TS, TP);
Chris@42 172 TX = FMA(KP707106781, TS, TP);
Chris@42 173 T11 = FNMS(KP707106781, T10, TZ);
Chris@42 174 T17 = FMA(KP707106781, T10, TZ);
Chris@42 175 cr[WS(rs, 6)] = FNMS(T1u, T1x, T1s);
Chris@42 176 cr[WS(rs, 2)] = FNMS(To, T1B, T1A);
Chris@42 177 {
Chris@42 178 E T1a, T16, TU, TI;
Chris@42 179 T1a = Tl * T19;
Chris@42 180 T16 = Tg * T15;
Chris@42 181 TU = TK * TH;
Chris@42 182 TI = Tq * TH;
Chris@42 183 {
Chris@42 184 E TY, TW, T18, T12;
Chris@42 185 TY = Ti * TV;
Chris@42 186 TW = Tf * TV;
Chris@42 187 T18 = Tl * T17;
Chris@42 188 T12 = Tg * T11;
Chris@42 189 ci[WS(rs, 7)] = FMA(Tp, T17, T1a);
Chris@42 190 ci[WS(rs, 3)] = FMA(Tj, T11, T16);
Chris@42 191 ci[WS(rs, 5)] = FMA(Tq, TT, TU);
Chris@42 192 cr[WS(rs, 5)] = FNMS(TK, TT, TI);
Chris@42 193 ci[WS(rs, 1)] = FMA(Tf, TX, TY);
Chris@42 194 cr[WS(rs, 1)] = FNMS(Ti, TX, TW);
Chris@42 195 cr[WS(rs, 7)] = FNMS(Tp, T19, T18);
Chris@42 196 cr[WS(rs, 3)] = FNMS(Tj, T15, T12);
Chris@42 197 }
Chris@42 198 }
Chris@42 199 }
Chris@42 200 }
Chris@42 201 }
Chris@42 202 }
Chris@42 203 }
Chris@42 204 }
Chris@42 205 }
Chris@42 206 }
Chris@42 207 }
Chris@42 208 }
Chris@42 209 }
Chris@42 210
Chris@42 211 static const tw_instr twinstr[] = {
Chris@42 212 {TW_CEXP, 1, 1},
Chris@42 213 {TW_CEXP, 1, 3},
Chris@42 214 {TW_CEXP, 1, 7},
Chris@42 215 {TW_NEXT, 1, 0}
Chris@42 216 };
Chris@42 217
Chris@42 218 static const hc2hc_desc desc = { 8, "hb2_8", twinstr, &GENUS, {44, 20, 30, 0} };
Chris@42 219
Chris@42 220 void X(codelet_hb2_8) (planner *p) {
Chris@42 221 X(khc2hc_register) (p, hb2_8, &desc);
Chris@42 222 }
Chris@42 223 #else /* HAVE_FMA */
Chris@42 224
Chris@42 225 /* 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 hb.h */
Chris@42 226
Chris@42 227 /*
Chris@42 228 * This function contains 74 FP additions, 44 FP multiplications,
Chris@42 229 * (or, 56 additions, 26 multiplications, 18 fused multiply/add),
Chris@42 230 * 46 stack variables, 1 constants, and 32 memory accesses
Chris@42 231 */
Chris@42 232 #include "hb.h"
Chris@42 233
Chris@42 234 static void hb2_8(R *cr, R *ci, const R *W, stride rs, INT mb, INT me, INT ms)
Chris@42 235 {
Chris@42 236 DK(KP707106781, +0.707106781186547524400844362104849039284835938);
Chris@42 237 {
Chris@42 238 INT m;
Chris@42 239 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@42 240 E Tf, Ti, Tg, Tj, Tl, Tp, TP, TR, TF, TG, TH, T15, TL, TT;
Chris@42 241 {
Chris@42 242 E Th, To, Tk, Tn;
Chris@42 243 Tf = W[0];
Chris@42 244 Ti = W[1];
Chris@42 245 Tg = W[2];
Chris@42 246 Tj = W[3];
Chris@42 247 Th = Tf * Tg;
Chris@42 248 To = Ti * Tg;
Chris@42 249 Tk = Ti * Tj;
Chris@42 250 Tn = Tf * Tj;
Chris@42 251 Tl = Th - Tk;
Chris@42 252 Tp = Tn + To;
Chris@42 253 TP = Th + Tk;
Chris@42 254 TR = Tn - To;
Chris@42 255 TF = W[4];
Chris@42 256 TG = W[5];
Chris@42 257 TH = FMA(Tf, TF, Ti * TG);
Chris@42 258 T15 = FNMS(TR, TF, TP * TG);
Chris@42 259 TL = FNMS(Ti, TF, Tf * TG);
Chris@42 260 TT = FMA(TP, TF, TR * TG);
Chris@42 261 }
Chris@42 262 {
Chris@42 263 E T7, T1f, T1i, Tw, TI, TW, T18, TM, Te, T19, T1a, TD, TJ, TZ, T12;
Chris@42 264 E TN, Tm, TE;
Chris@42 265 {
Chris@42 266 E T3, TU, Tv, TV, T6, T16, Ts, T17;
Chris@42 267 {
Chris@42 268 E T1, T2, Tt, Tu;
Chris@42 269 T1 = cr[0];
Chris@42 270 T2 = ci[WS(rs, 3)];
Chris@42 271 T3 = T1 + T2;
Chris@42 272 TU = T1 - T2;
Chris@42 273 Tt = ci[WS(rs, 5)];
Chris@42 274 Tu = cr[WS(rs, 6)];
Chris@42 275 Tv = Tt - Tu;
Chris@42 276 TV = Tt + Tu;
Chris@42 277 }
Chris@42 278 {
Chris@42 279 E T4, T5, Tq, Tr;
Chris@42 280 T4 = cr[WS(rs, 2)];
Chris@42 281 T5 = ci[WS(rs, 1)];
Chris@42 282 T6 = T4 + T5;
Chris@42 283 T16 = T4 - T5;
Chris@42 284 Tq = ci[WS(rs, 7)];
Chris@42 285 Tr = cr[WS(rs, 4)];
Chris@42 286 Ts = Tq - Tr;
Chris@42 287 T17 = Tq + Tr;
Chris@42 288 }
Chris@42 289 T7 = T3 + T6;
Chris@42 290 T1f = TU + TV;
Chris@42 291 T1i = T17 - T16;
Chris@42 292 Tw = Ts + Tv;
Chris@42 293 TI = T3 - T6;
Chris@42 294 TW = TU - TV;
Chris@42 295 T18 = T16 + T17;
Chris@42 296 TM = Ts - Tv;
Chris@42 297 }
Chris@42 298 {
Chris@42 299 E Ta, TX, TC, T11, Td, T10, Tz, TY;
Chris@42 300 {
Chris@42 301 E T8, T9, TA, TB;
Chris@42 302 T8 = cr[WS(rs, 1)];
Chris@42 303 T9 = ci[WS(rs, 2)];
Chris@42 304 Ta = T8 + T9;
Chris@42 305 TX = T8 - T9;
Chris@42 306 TA = ci[WS(rs, 4)];
Chris@42 307 TB = cr[WS(rs, 7)];
Chris@42 308 TC = TA - TB;
Chris@42 309 T11 = TA + TB;
Chris@42 310 }
Chris@42 311 {
Chris@42 312 E Tb, Tc, Tx, Ty;
Chris@42 313 Tb = ci[0];
Chris@42 314 Tc = cr[WS(rs, 3)];
Chris@42 315 Td = Tb + Tc;
Chris@42 316 T10 = Tb - Tc;
Chris@42 317 Tx = ci[WS(rs, 6)];
Chris@42 318 Ty = cr[WS(rs, 5)];
Chris@42 319 Tz = Tx - Ty;
Chris@42 320 TY = Tx + Ty;
Chris@42 321 }
Chris@42 322 Te = Ta + Td;
Chris@42 323 T19 = TX + TY;
Chris@42 324 T1a = T10 + T11;
Chris@42 325 TD = Tz + TC;
Chris@42 326 TJ = TC - Tz;
Chris@42 327 TZ = TX - TY;
Chris@42 328 T12 = T10 - T11;
Chris@42 329 TN = Ta - Td;
Chris@42 330 }
Chris@42 331 cr[0] = T7 + Te;
Chris@42 332 ci[0] = Tw + TD;
Chris@42 333 Tm = T7 - Te;
Chris@42 334 TE = Tw - TD;
Chris@42 335 cr[WS(rs, 4)] = FNMS(Tp, TE, Tl * Tm);
Chris@42 336 ci[WS(rs, 4)] = FMA(Tp, Tm, Tl * TE);
Chris@42 337 {
Chris@42 338 E TQ, TS, TK, TO;
Chris@42 339 TQ = TI + TJ;
Chris@42 340 TS = TN + TM;
Chris@42 341 cr[WS(rs, 2)] = FNMS(TR, TS, TP * TQ);
Chris@42 342 ci[WS(rs, 2)] = FMA(TP, TS, TR * TQ);
Chris@42 343 TK = TI - TJ;
Chris@42 344 TO = TM - TN;
Chris@42 345 cr[WS(rs, 6)] = FNMS(TL, TO, TH * TK);
Chris@42 346 ci[WS(rs, 6)] = FMA(TH, TO, TL * TK);
Chris@42 347 }
Chris@42 348 {
Chris@42 349 E T1h, T1l, T1k, T1m, T1g, T1j;
Chris@42 350 T1g = KP707106781 * (T19 + T1a);
Chris@42 351 T1h = T1f - T1g;
Chris@42 352 T1l = T1f + T1g;
Chris@42 353 T1j = KP707106781 * (TZ - T12);
Chris@42 354 T1k = T1i + T1j;
Chris@42 355 T1m = T1i - T1j;
Chris@42 356 cr[WS(rs, 3)] = FNMS(Tj, T1k, Tg * T1h);
Chris@42 357 ci[WS(rs, 3)] = FMA(Tg, T1k, Tj * T1h);
Chris@42 358 cr[WS(rs, 7)] = FNMS(TG, T1m, TF * T1l);
Chris@42 359 ci[WS(rs, 7)] = FMA(TF, T1m, TG * T1l);
Chris@42 360 }
Chris@42 361 {
Chris@42 362 E T14, T1d, T1c, T1e, T13, T1b;
Chris@42 363 T13 = KP707106781 * (TZ + T12);
Chris@42 364 T14 = TW - T13;
Chris@42 365 T1d = TW + T13;
Chris@42 366 T1b = KP707106781 * (T19 - T1a);
Chris@42 367 T1c = T18 - T1b;
Chris@42 368 T1e = T18 + T1b;
Chris@42 369 cr[WS(rs, 5)] = FNMS(T15, T1c, TT * T14);
Chris@42 370 ci[WS(rs, 5)] = FMA(T15, T14, TT * T1c);
Chris@42 371 cr[WS(rs, 1)] = FNMS(Ti, T1e, Tf * T1d);
Chris@42 372 ci[WS(rs, 1)] = FMA(Ti, T1d, Tf * T1e);
Chris@42 373 }
Chris@42 374 }
Chris@42 375 }
Chris@42 376 }
Chris@42 377 }
Chris@42 378
Chris@42 379 static const tw_instr twinstr[] = {
Chris@42 380 {TW_CEXP, 1, 1},
Chris@42 381 {TW_CEXP, 1, 3},
Chris@42 382 {TW_CEXP, 1, 7},
Chris@42 383 {TW_NEXT, 1, 0}
Chris@42 384 };
Chris@42 385
Chris@42 386 static const hc2hc_desc desc = { 8, "hb2_8", twinstr, &GENUS, {56, 26, 18, 0} };
Chris@42 387
Chris@42 388 void X(codelet_hb2_8) (planner *p) {
Chris@42 389 X(khc2hc_register) (p, hb2_8, &desc);
Chris@42 390 }
Chris@42 391 #endif /* HAVE_FMA */