Mercurial > hg > sv-dependency-builds

annotate src/fftw-3.3.8/dft/scalar/codelets/t2_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 08:04:19 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_twiddle.native -fma -compact -variables 4 -pipeline-latency 4 -twiddle-log3 -precompute-twiddles -n 8 -name t2_8 -include dft/scalar/t.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 * 48 stack variables, 1 constants, and 32 memory accesses
Chris@82 34 */
Chris@82 35 #include "dft/scalar/t.h"
Chris@82 36
Chris@82 37 static void t2_8(R *ri, R *ii, 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 * 6); m < me; m = m + 1, ri = ri + ms, ii = ii + ms, W = W + 6, MAKE_VOLATILE_STRIDE(16, rs)) {
Chris@82 43 E T2, T3, Tl, Tn, T5, T6, Tf, T7, Ts, Tb, To, Ti, TC, TG;
Chris@82 44 {
Chris@82 45 E T4, Tm, Tr, Ta, TB, TF;
Chris@82 46 T2 = W[0];
Chris@82 47 T3 = W[2];
Chris@82 48 T4 = T2 * T3;
Chris@82 49 Tl = W[4];
Chris@82 50 Tm = T2 * Tl;
Chris@82 51 Tn = W[5];
Chris@82 52 Tr = T2 * Tn;
Chris@82 53 T5 = W[1];
Chris@82 54 T6 = W[3];
Chris@82 55 Ta = T2 * T6;
Chris@82 56 Tf = FMA(T5, T6, T4);
Chris@82 57 T7 = FNMS(T5, T6, T4);
Chris@82 58 Ts = FNMS(T5, Tl, Tr);
Chris@82 59 Tb = FMA(T5, T3, Ta);
Chris@82 60 To = FMA(T5, Tn, Tm);
Chris@82 61 TB = Tf * Tl;
Chris@82 62 TF = Tf * Tn;
Chris@82 63 Ti = FNMS(T5, T3, Ta);
Chris@82 64 TC = FMA(Ti, Tn, TB);
Chris@82 65 TG = FNMS(Ti, Tl, TF);
Chris@82 66 }
Chris@82 67 {
Chris@82 68 E T1, T1s, Td, T1r, Tu, TY, Tk, TW, TN, TR, T18, T1a, T1c, T1d, TA;
Chris@82 69 E TI, T11, T13, T15, T16;
Chris@82 70 T1 = ri[0];
Chris@82 71 T1s = ii[0];
Chris@82 72 {
Chris@82 73 E T8, T9, Tc, T1q;
Chris@82 74 T8 = ri[WS(rs, 4)];
Chris@82 75 T9 = T7 * T8;
Chris@82 76 Tc = ii[WS(rs, 4)];
Chris@82 77 T1q = T7 * Tc;
Chris@82 78 Td = FMA(Tb, Tc, T9);
Chris@82 79 T1r = FNMS(Tb, T8, T1q);
Chris@82 80 }
Chris@82 81 {
Chris@82 82 E Tp, Tq, Tt, TX;
Chris@82 83 Tp = ri[WS(rs, 6)];
Chris@82 84 Tq = To * Tp;
Chris@82 85 Tt = ii[WS(rs, 6)];
Chris@82 86 TX = To * Tt;
Chris@82 87 Tu = FMA(Ts, Tt, Tq);
Chris@82 88 TY = FNMS(Ts, Tp, TX);
Chris@82 89 }
Chris@82 90 {
Chris@82 91 E Tg, Th, Tj, TV;
Chris@82 92 Tg = ri[WS(rs, 2)];
Chris@82 93 Th = Tf * Tg;
Chris@82 94 Tj = ii[WS(rs, 2)];
Chris@82 95 TV = Tf * Tj;
Chris@82 96 Tk = FMA(Ti, Tj, Th);
Chris@82 97 TW = FNMS(Ti, Tg, TV);
Chris@82 98 }
Chris@82 99 {
Chris@82 100 E TK, TL, TM, T19, TO, TP, TQ, T1b;
Chris@82 101 TK = ri[WS(rs, 7)];
Chris@82 102 TL = Tl * TK;
Chris@82 103 TM = ii[WS(rs, 7)];
Chris@82 104 T19 = Tl * TM;
Chris@82 105 TO = ri[WS(rs, 3)];
Chris@82 106 TP = T3 * TO;
Chris@82 107 TQ = ii[WS(rs, 3)];
Chris@82 108 T1b = T3 * TQ;
Chris@82 109 TN = FMA(Tn, TM, TL);
Chris@82 110 TR = FMA(T6, TQ, TP);
Chris@82 111 T18 = TN - TR;
Chris@82 112 T1a = FNMS(Tn, TK, T19);
Chris@82 113 T1c = FNMS(T6, TO, T1b);
Chris@82 114 T1d = T1a - T1c;
Chris@82 115 }
Chris@82 116 {
Chris@82 117 E Tx, Ty, Tz, T12, TD, TE, TH, T14;
Chris@82 118 Tx = ri[WS(rs, 1)];
Chris@82 119 Ty = T2 * Tx;
Chris@82 120 Tz = ii[WS(rs, 1)];
Chris@82 121 T12 = T2 * Tz;
Chris@82 122 TD = ri[WS(rs, 5)];
Chris@82 123 TE = TC * TD;
Chris@82 124 TH = ii[WS(rs, 5)];
Chris@82 125 T14 = TC * TH;
Chris@82 126 TA = FMA(T5, Tz, Ty);
Chris@82 127 TI = FMA(TG, TH, TE);
Chris@82 128 T11 = TA - TI;
Chris@82 129 T13 = FNMS(T5, Tx, T12);
Chris@82 130 T15 = FNMS(TG, TD, T14);
Chris@82 131 T16 = T13 - T15;
Chris@82 132 }
Chris@82 133 {
Chris@82 134 E T10, T1g, T1z, T1B, T1f, T1C, T1j, T1A;
Chris@82 135 {
Chris@82 136 E TU, TZ, T1x, T1y;
Chris@82 137 TU = T1 - Td;
Chris@82 138 TZ = TW - TY;
Chris@82 139 T10 = TU + TZ;
Chris@82 140 T1g = TU - TZ;
Chris@82 141 T1x = T1s - T1r;
Chris@82 142 T1y = Tk - Tu;
Chris@82 143 T1z = T1x - T1y;
Chris@82 144 T1B = T1y + T1x;
Chris@82 145 }
Chris@82 146 {
Chris@82 147 E T17, T1e, T1h, T1i;
Chris@82 148 T17 = T11 + T16;
Chris@82 149 T1e = T18 - T1d;
Chris@82 150 T1f = T17 + T1e;
Chris@82 151 T1C = T1e - T17;
Chris@82 152 T1h = T16 - T11;
Chris@82 153 T1i = T18 + T1d;
Chris@82 154 T1j = T1h - T1i;
Chris@82 155 T1A = T1h + T1i;
Chris@82 156 }
Chris@82 157 ri[WS(rs, 5)] = FNMS(KP707106781, T1f, T10);
Chris@82 158 ii[WS(rs, 5)] = FNMS(KP707106781, T1A, T1z);
Chris@82 159 ri[WS(rs, 1)] = FMA(KP707106781, T1f, T10);
Chris@82 160 ii[WS(rs, 1)] = FMA(KP707106781, T1A, T1z);
Chris@82 161 ri[WS(rs, 7)] = FNMS(KP707106781, T1j, T1g);
Chris@82 162 ii[WS(rs, 7)] = FNMS(KP707106781, T1C, T1B);
Chris@82 163 ri[WS(rs, 3)] = FMA(KP707106781, T1j, T1g);
Chris@82 164 ii[WS(rs, 3)] = FMA(KP707106781, T1C, T1B);
Chris@82 165 }
Chris@82 166 {
Chris@82 167 E Tw, T1k, T1u, T1w, TT, T1v, T1n, T1o;
Chris@82 168 {
Chris@82 169 E Te, Tv, T1p, T1t;
Chris@82 170 Te = T1 + Td;
Chris@82 171 Tv = Tk + Tu;
Chris@82 172 Tw = Te + Tv;
Chris@82 173 T1k = Te - Tv;
Chris@82 174 T1p = TW + TY;
Chris@82 175 T1t = T1r + T1s;
Chris@82 176 T1u = T1p + T1t;
Chris@82 177 T1w = T1t - T1p;
Chris@82 178 }
Chris@82 179 {
Chris@82 180 E TJ, TS, T1l, T1m;
Chris@82 181 TJ = TA + TI;
Chris@82 182 TS = TN + TR;
Chris@82 183 TT = TJ + TS;
Chris@82 184 T1v = TS - TJ;
Chris@82 185 T1l = T13 + T15;
Chris@82 186 T1m = T1a + T1c;
Chris@82 187 T1n = T1l - T1m;
Chris@82 188 T1o = T1l + T1m;
Chris@82 189 }
Chris@82 190 ri[WS(rs, 4)] = Tw - TT;
Chris@82 191 ii[WS(rs, 4)] = T1u - T1o;
Chris@82 192 ri[0] = Tw + TT;
Chris@82 193 ii[0] = T1o + T1u;
Chris@82 194 ri[WS(rs, 6)] = T1k - T1n;
Chris@82 195 ii[WS(rs, 6)] = T1w - T1v;
Chris@82 196 ri[WS(rs, 2)] = T1k + T1n;
Chris@82 197 ii[WS(rs, 2)] = T1v + T1w;
Chris@82 198 }
Chris@82 199 }
Chris@82 200 }
Chris@82 201 }
Chris@82 202 }
Chris@82 203
Chris@82 204 static const tw_instr twinstr[] = {
Chris@82 205 {TW_CEXP, 0, 1},
Chris@82 206 {TW_CEXP, 0, 3},
Chris@82 207 {TW_CEXP, 0, 7},
Chris@82 208 {TW_NEXT, 1, 0}
Chris@82 209 };
Chris@82 210
Chris@82 211 static const ct_desc desc = { 8, "t2_8", twinstr, &GENUS, {44, 20, 30, 0}, 0, 0, 0 };
Chris@82 212
Chris@82 213 void X(codelet_t2_8) (planner *p) {
Chris@82 214 X(kdft_dit_register) (p, t2_8, &desc);
Chris@82 215 }
Chris@82 216 #else
Chris@82 217
Chris@82 218 /* Generated by: ../../../genfft/gen_twiddle.native -compact -variables 4 -pipeline-latency 4 -twiddle-log3 -precompute-twiddles -n 8 -name t2_8 -include dft/scalar/t.h */
Chris@82 219
Chris@82 220 /*
Chris@82 221 * This function contains 74 FP additions, 44 FP multiplications,
Chris@82 222 * (or, 56 additions, 26 multiplications, 18 fused multiply/add),
Chris@82 223 * 42 stack variables, 1 constants, and 32 memory accesses
Chris@82 224 */
Chris@82 225 #include "dft/scalar/t.h"
Chris@82 226
Chris@82 227 static void t2_8(R *ri, R *ii, const R *W, stride rs, INT mb, INT me, INT ms)
Chris@82 228 {
Chris@82 229 DK(KP707106781, +0.707106781186547524400844362104849039284835938);
Chris@82 230 {
Chris@82 231 INT m;
Chris@82 232 for (m = mb, W = W + (mb * 6); m < me; m = m + 1, ri = ri + ms, ii = ii + ms, W = W + 6, MAKE_VOLATILE_STRIDE(16, rs)) {
Chris@82 233 E T2, T5, T3, T6, T8, Tc, Tg, Ti, Tl, Tm, Tn, Tz, Tp, Tx;
Chris@82 234 {
Chris@82 235 E T4, Tb, T7, Ta;
Chris@82 236 T2 = W[0];
Chris@82 237 T5 = W[1];
Chris@82 238 T3 = W[2];
Chris@82 239 T6 = W[3];
Chris@82 240 T4 = T2 * T3;
Chris@82 241 Tb = T5 * T3;
Chris@82 242 T7 = T5 * T6;
Chris@82 243 Ta = T2 * T6;
Chris@82 244 T8 = T4 - T7;
Chris@82 245 Tc = Ta + Tb;
Chris@82 246 Tg = T4 + T7;
Chris@82 247 Ti = Ta - Tb;
Chris@82 248 Tl = W[4];
Chris@82 249 Tm = W[5];
Chris@82 250 Tn = FMA(T2, Tl, T5 * Tm);
Chris@82 251 Tz = FNMS(Ti, Tl, Tg * Tm);
Chris@82 252 Tp = FNMS(T5, Tl, T2 * Tm);
Chris@82 253 Tx = FMA(Tg, Tl, Ti * Tm);
Chris@82 254 }
Chris@82 255 {
Chris@82 256 E Tf, T1i, TL, T1d, TJ, T17, TV, TY, Ts, T1j, TO, T1a, TC, T16, TQ;
Chris@82 257 E TT;
Chris@82 258 {
Chris@82 259 E T1, T1c, Te, T1b, T9, Td;
Chris@82 260 T1 = ri[0];
Chris@82 261 T1c = ii[0];
Chris@82 262 T9 = ri[WS(rs, 4)];
Chris@82 263 Td = ii[WS(rs, 4)];
Chris@82 264 Te = FMA(T8, T9, Tc * Td);
Chris@82 265 T1b = FNMS(Tc, T9, T8 * Td);
Chris@82 266 Tf = T1 + Te;
Chris@82 267 T1i = T1c - T1b;
Chris@82 268 TL = T1 - Te;
Chris@82 269 T1d = T1b + T1c;
Chris@82 270 }
Chris@82 271 {
Chris@82 272 E TF, TW, TI, TX;
Chris@82 273 {
Chris@82 274 E TD, TE, TG, TH;
Chris@82 275 TD = ri[WS(rs, 7)];
Chris@82 276 TE = ii[WS(rs, 7)];
Chris@82 277 TF = FMA(Tl, TD, Tm * TE);
Chris@82 278 TW = FNMS(Tm, TD, Tl * TE);
Chris@82 279 TG = ri[WS(rs, 3)];
Chris@82 280 TH = ii[WS(rs, 3)];
Chris@82 281 TI = FMA(T3, TG, T6 * TH);
Chris@82 282 TX = FNMS(T6, TG, T3 * TH);
Chris@82 283 }
Chris@82 284 TJ = TF + TI;
Chris@82 285 T17 = TW + TX;
Chris@82 286 TV = TF - TI;
Chris@82 287 TY = TW - TX;
Chris@82 288 }
Chris@82 289 {
Chris@82 290 E Tk, TM, Tr, TN;
Chris@82 291 {
Chris@82 292 E Th, Tj, To, Tq;
Chris@82 293 Th = ri[WS(rs, 2)];
Chris@82 294 Tj = ii[WS(rs, 2)];
Chris@82 295 Tk = FMA(Tg, Th, Ti * Tj);
Chris@82 296 TM = FNMS(Ti, Th, Tg * Tj);
Chris@82 297 To = ri[WS(rs, 6)];
Chris@82 298 Tq = ii[WS(rs, 6)];
Chris@82 299 Tr = FMA(Tn, To, Tp * Tq);
Chris@82 300 TN = FNMS(Tp, To, Tn * Tq);
Chris@82 301 }
Chris@82 302 Ts = Tk + Tr;
Chris@82 303 T1j = Tk - Tr;
Chris@82 304 TO = TM - TN;
Chris@82 305 T1a = TM + TN;
Chris@82 306 }
Chris@82 307 {
Chris@82 308 E Tw, TR, TB, TS;
Chris@82 309 {
Chris@82 310 E Tu, Tv, Ty, TA;
Chris@82 311 Tu = ri[WS(rs, 1)];
Chris@82 312 Tv = ii[WS(rs, 1)];
Chris@82 313 Tw = FMA(T2, Tu, T5 * Tv);
Chris@82 314 TR = FNMS(T5, Tu, T2 * Tv);
Chris@82 315 Ty = ri[WS(rs, 5)];
Chris@82 316 TA = ii[WS(rs, 5)];
Chris@82 317 TB = FMA(Tx, Ty, Tz * TA);
Chris@82 318 TS = FNMS(Tz, Ty, Tx * TA);
Chris@82 319 }
Chris@82 320 TC = Tw + TB;
Chris@82 321 T16 = TR + TS;
Chris@82 322 TQ = Tw - TB;
Chris@82 323 TT = TR - TS;
Chris@82 324 }
Chris@82 325 {
Chris@82 326 E Tt, TK, T1f, T1g;
Chris@82 327 Tt = Tf + Ts;
Chris@82 328 TK = TC + TJ;
Chris@82 329 ri[WS(rs, 4)] = Tt - TK;
Chris@82 330 ri[0] = Tt + TK;
Chris@82 331 {
Chris@82 332 E T19, T1e, T15, T18;
Chris@82 333 T19 = T16 + T17;
Chris@82 334 T1e = T1a + T1d;
Chris@82 335 ii[0] = T19 + T1e;
Chris@82 336 ii[WS(rs, 4)] = T1e - T19;
Chris@82 337 T15 = Tf - Ts;
Chris@82 338 T18 = T16 - T17;
Chris@82 339 ri[WS(rs, 6)] = T15 - T18;
Chris@82 340 ri[WS(rs, 2)] = T15 + T18;
Chris@82 341 }
Chris@82 342 T1f = TJ - TC;
Chris@82 343 T1g = T1d - T1a;
Chris@82 344 ii[WS(rs, 2)] = T1f + T1g;
Chris@82 345 ii[WS(rs, 6)] = T1g - T1f;
Chris@82 346 {
Chris@82 347 E T11, T1k, T14, T1h, T12, T13;
Chris@82 348 T11 = TL - TO;
Chris@82 349 T1k = T1i - T1j;
Chris@82 350 T12 = TT - TQ;
Chris@82 351 T13 = TV + TY;
Chris@82 352 T14 = KP707106781 * (T12 - T13);
Chris@82 353 T1h = KP707106781 * (T12 + T13);
Chris@82 354 ri[WS(rs, 7)] = T11 - T14;
Chris@82 355 ii[WS(rs, 5)] = T1k - T1h;
Chris@82 356 ri[WS(rs, 3)] = T11 + T14;
Chris@82 357 ii[WS(rs, 1)] = T1h + T1k;
Chris@82 358 }
Chris@82 359 {
Chris@82 360 E TP, T1m, T10, T1l, TU, TZ;
Chris@82 361 TP = TL + TO;
Chris@82 362 T1m = T1j + T1i;
Chris@82 363 TU = TQ + TT;
Chris@82 364 TZ = TV - TY;
Chris@82 365 T10 = KP707106781 * (TU + TZ);
Chris@82 366 T1l = KP707106781 * (TZ - TU);
Chris@82 367 ri[WS(rs, 5)] = TP - T10;
Chris@82 368 ii[WS(rs, 7)] = T1m - T1l;
Chris@82 369 ri[WS(rs, 1)] = TP + T10;
Chris@82 370 ii[WS(rs, 3)] = T1l + T1m;
Chris@82 371 }
Chris@82 372 }
Chris@82 373 }
Chris@82 374 }
Chris@82 375 }
Chris@82 376 }
Chris@82 377
Chris@82 378 static const tw_instr twinstr[] = {
Chris@82 379 {TW_CEXP, 0, 1},
Chris@82 380 {TW_CEXP, 0, 3},
Chris@82 381 {TW_CEXP, 0, 7},
Chris@82 382 {TW_NEXT, 1, 0}
Chris@82 383 };
Chris@82 384
Chris@82 385 static const ct_desc desc = { 8, "t2_8", twinstr, &GENUS, {56, 26, 18, 0}, 0, 0, 0 };
Chris@82 386
Chris@82 387 void X(codelet_t2_8) (planner *p) {
Chris@82 388 X(kdft_dit_register) (p, t2_8, &desc);
Chris@82 389 }
Chris@82 390 #endif