Mercurial > hg > sv-dependency-builds

annotate src/fftw-3.3.8/rdft/scalar/r2cf/hc2cfdft_12.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 08:07:11 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_hc2cdft.native -fma -compact -variables 4 -pipeline-latency 4 -n 12 -dit -name hc2cfdft_12 -include rdft/scalar/hc2cf.h */
Chris@82 29
Chris@82 30 /*
Chris@82 31 * This function contains 142 FP additions, 92 FP multiplications,
Chris@82 32 * (or, 96 additions, 46 multiplications, 46 fused multiply/add),
Chris@82 33 * 65 stack variables, 2 constants, and 48 memory accesses
Chris@82 34 */
Chris@82 35 #include "rdft/scalar/hc2cf.h"
Chris@82 36
Chris@82 37 static void hc2cfdft_12(R *Rp, R *Ip, R *Rm, R *Im, const R *W, stride rs, INT mb, INT me, INT ms)
Chris@82 38 {
Chris@82 39 DK(KP866025403, +0.866025403784438646763723170752936183471402627);
Chris@82 40 DK(KP500000000, +0.500000000000000000000000000000000000000000000);
Chris@82 41 {
Chris@82 42 INT m;
Chris@82 43 for (m = mb, W = W + ((mb - 1) * 22); m < me; m = m + 1, Rp = Rp + ms, Ip = Ip + ms, Rm = Rm - ms, Im = Im - ms, W = W + 22, MAKE_VOLATILE_STRIDE(48, rs)) {
Chris@82 44 E To, T1E, T1m, T2H, Ta, T1G, Tk, T1I, Tl, T1J, T1s, T2b, T1A, T2d, T1B;
Chris@82 45 E T2I, T12, T18, T19, T24, T26, T2C, Tz, T1M, T1f, T2B, TJ, T1O, TT, T1Q;
Chris@82 46 E TU, T1R;
Chris@82 47 {
Chris@82 48 E Tm, Tn, T1u, T1x, T1y, T1z, T1v, T2c, Te, Tj, T1i, T1l, Tf, T1H, T4;
Chris@82 49 E T1o, T9, T1r, T5, T1F, T1p, T2a, T1t, T1, T1n;
Chris@82 50 Tm = Ip[0];
Chris@82 51 Tn = Im[0];
Chris@82 52 T1u = Tm + Tn;
Chris@82 53 T1x = Rp[0];
Chris@82 54 T1y = Rm[0];
Chris@82 55 T1z = T1x - T1y;
Chris@82 56 T1t = W[0];
Chris@82 57 T1v = T1t * T1u;
Chris@82 58 T2c = T1t * T1z;
Chris@82 59 {
Chris@82 60 E Tc, Td, Th, Ti, Tb;
Chris@82 61 Tc = Ip[WS(rs, 4)];
Chris@82 62 Td = Im[WS(rs, 4)];
Chris@82 63 Te = Tc - Td;
Chris@82 64 Th = Rp[WS(rs, 4)];
Chris@82 65 Ti = Rm[WS(rs, 4)];
Chris@82 66 Tj = Th + Ti;
Chris@82 67 T1i = Tc + Td;
Chris@82 68 T1l = Th - Ti;
Chris@82 69 Tb = W[14];
Chris@82 70 Tf = Tb * Te;
Chris@82 71 T1H = Tb * Tj;
Chris@82 72 }
Chris@82 73 {
Chris@82 74 E T2, T3, T7, T8;
Chris@82 75 T2 = Ip[WS(rs, 2)];
Chris@82 76 T3 = Im[WS(rs, 2)];
Chris@82 77 T4 = T2 - T3;
Chris@82 78 T1o = T2 + T3;
Chris@82 79 T7 = Rp[WS(rs, 2)];
Chris@82 80 T8 = Rm[WS(rs, 2)];
Chris@82 81 T9 = T7 + T8;
Chris@82 82 T1r = T7 - T8;
Chris@82 83 }
Chris@82 84 T1 = W[6];
Chris@82 85 T5 = T1 * T4;
Chris@82 86 T1F = T1 * T9;
Chris@82 87 T1n = W[8];
Chris@82 88 T1p = T1n * T1o;
Chris@82 89 T2a = T1n * T1r;
Chris@82 90 To = Tm - Tn;
Chris@82 91 T1E = T1x + T1y;
Chris@82 92 {
Chris@82 93 E T1j, T2G, T1h, T1k;
Chris@82 94 T1h = W[16];
Chris@82 95 T1j = T1h * T1i;
Chris@82 96 T2G = T1h * T1l;
Chris@82 97 T1k = W[17];
Chris@82 98 T1m = FNMS(T1k, T1l, T1j);
Chris@82 99 T2H = FMA(T1k, T1i, T2G);
Chris@82 100 }
Chris@82 101 {
Chris@82 102 E T6, Tg, T1q, T1w;
Chris@82 103 T6 = W[7];
Chris@82 104 Ta = FNMS(T6, T9, T5);
Chris@82 105 T1G = FMA(T6, T4, T1F);
Chris@82 106 Tg = W[15];
Chris@82 107 Tk = FNMS(Tg, Tj, Tf);
Chris@82 108 T1I = FMA(Tg, Te, T1H);
Chris@82 109 Tl = Ta + Tk;
Chris@82 110 T1J = T1G + T1I;
Chris@82 111 T1q = W[9];
Chris@82 112 T1s = FNMS(T1q, T1r, T1p);
Chris@82 113 T2b = FMA(T1q, T1o, T2a);
Chris@82 114 T1w = W[1];
Chris@82 115 T1A = FNMS(T1w, T1z, T1v);
Chris@82 116 T2d = FMA(T1w, T1u, T2c);
Chris@82 117 T1B = T1s + T1A;
Chris@82 118 T2I = T2b + T2d;
Chris@82 119 }
Chris@82 120 }
Chris@82 121 {
Chris@82 122 E Tt, T11, Ty, T10, T23, TX, TZ, TN, TS, T1b, T1e, TO, T1P, TD, TI;
Chris@82 123 E T17, T16, T25, T13, T15, TE, T1N, TF, TP;
Chris@82 124 {
Chris@82 125 E Tr, Ts, Tw, Tx, TY;
Chris@82 126 Tr = Ip[WS(rs, 3)];
Chris@82 127 Ts = Im[WS(rs, 3)];
Chris@82 128 Tt = Tr - Ts;
Chris@82 129 T11 = Tr + Ts;
Chris@82 130 Tw = Rp[WS(rs, 3)];
Chris@82 131 Tx = Rm[WS(rs, 3)];
Chris@82 132 TY = Tx - Tw;
Chris@82 133 Ty = Tw + Tx;
Chris@82 134 T10 = W[12];
Chris@82 135 T23 = T10 * TY;
Chris@82 136 TX = W[13];
Chris@82 137 TZ = TX * TY;
Chris@82 138 }
Chris@82 139 {
Chris@82 140 E TL, TM, TQ, TR, TK;
Chris@82 141 TL = Ip[WS(rs, 1)];
Chris@82 142 TM = Im[WS(rs, 1)];
Chris@82 143 TN = TL - TM;
Chris@82 144 TQ = Rp[WS(rs, 1)];
Chris@82 145 TR = Rm[WS(rs, 1)];
Chris@82 146 TS = TQ + TR;
Chris@82 147 T1b = TL + TM;
Chris@82 148 T1e = TQ - TR;
Chris@82 149 TK = W[2];
Chris@82 150 TO = TK * TN;
Chris@82 151 T1P = TK * TS;
Chris@82 152 }
Chris@82 153 {
Chris@82 154 E TB, TC, T14, TG, TH, TA;
Chris@82 155 TB = Ip[WS(rs, 5)];
Chris@82 156 TC = Im[WS(rs, 5)];
Chris@82 157 TD = TB - TC;
Chris@82 158 TG = Rp[WS(rs, 5)];
Chris@82 159 TH = Rm[WS(rs, 5)];
Chris@82 160 TI = TG + TH;
Chris@82 161 T14 = TH - TG;
Chris@82 162 T17 = TB + TC;
Chris@82 163 T16 = W[20];
Chris@82 164 T25 = T16 * T14;
Chris@82 165 T13 = W[21];
Chris@82 166 T15 = T13 * T14;
Chris@82 167 TA = W[18];
Chris@82 168 TE = TA * TD;
Chris@82 169 T1N = TA * TI;
Chris@82 170 }
Chris@82 171 T12 = FMA(T10, T11, TZ);
Chris@82 172 T18 = FMA(T16, T17, T15);
Chris@82 173 T19 = T12 + T18;
Chris@82 174 T24 = FNMS(TX, T11, T23);
Chris@82 175 T26 = FNMS(T13, T17, T25);
Chris@82 176 T2C = T24 + T26;
Chris@82 177 {
Chris@82 178 E Tu, T1L, Tq, Tv;
Chris@82 179 Tq = W[10];
Chris@82 180 Tu = Tq * Tt;
Chris@82 181 T1L = Tq * Ty;
Chris@82 182 Tv = W[11];
Chris@82 183 Tz = FNMS(Tv, Ty, Tu);
Chris@82 184 T1M = FMA(Tv, Tt, T1L);
Chris@82 185 }
Chris@82 186 {
Chris@82 187 E T1c, T2A, T1a, T1d;
Chris@82 188 T1a = W[4];
Chris@82 189 T1c = T1a * T1b;
Chris@82 190 T2A = T1a * T1e;
Chris@82 191 T1d = W[5];
Chris@82 192 T1f = FNMS(T1d, T1e, T1c);
Chris@82 193 T2B = FMA(T1d, T1b, T2A);
Chris@82 194 }
Chris@82 195 TF = W[19];
Chris@82 196 TJ = FNMS(TF, TI, TE);
Chris@82 197 T1O = FMA(TF, TD, T1N);
Chris@82 198 TP = W[3];
Chris@82 199 TT = FNMS(TP, TS, TO);
Chris@82 200 T1Q = FMA(TP, TN, T1P);
Chris@82 201 TU = TJ + TT;
Chris@82 202 T1R = T1O + T1Q;
Chris@82 203 }
Chris@82 204 {
Chris@82 205 E TW, T2V, T2Y, T30, T1D, T1U, T1T, T2Z;
Chris@82 206 {
Chris@82 207 E Tp, TV, T2W, T2X;
Chris@82 208 Tp = Tl + To;
Chris@82 209 TV = Tz + TU;
Chris@82 210 TW = Tp - TV;
Chris@82 211 T2V = TV + Tp;
Chris@82 212 T2W = T2C - T2B;
Chris@82 213 T2X = T2H + T2I;
Chris@82 214 T2Y = T2W - T2X;
Chris@82 215 T30 = T2W + T2X;
Chris@82 216 }
Chris@82 217 {
Chris@82 218 E T1g, T1C, T1K, T1S;
Chris@82 219 T1g = T19 + T1f;
Chris@82 220 T1C = T1m + T1B;
Chris@82 221 T1D = T1g - T1C;
Chris@82 222 T1U = T1g + T1C;
Chris@82 223 T1K = T1E + T1J;
Chris@82 224 T1S = T1M + T1R;
Chris@82 225 T1T = T1K + T1S;
Chris@82 226 T2Z = T1K - T1S;
Chris@82 227 }
Chris@82 228 Ip[WS(rs, 3)] = KP500000000 * (TW + T1D);
Chris@82 229 Rp[WS(rs, 3)] = KP500000000 * (T2Z - T30);
Chris@82 230 Im[WS(rs, 2)] = KP500000000 * (T1D - TW);
Chris@82 231 Rm[WS(rs, 2)] = KP500000000 * (T2Z + T30);
Chris@82 232 Rm[WS(rs, 5)] = KP500000000 * (T1T - T1U);
Chris@82 233 Im[WS(rs, 5)] = KP500000000 * (T2Y - T2V);
Chris@82 234 Rp[0] = KP500000000 * (T1T + T1U);
Chris@82 235 Ip[0] = KP500000000 * (T2V + T2Y);
Chris@82 236 }
Chris@82 237 {
Chris@82 238 E T1X, T2v, T2F, T2Q, T2L, T2R, T20, T2w, T28, T2t, T2j, T2p, T2m, T2q, T2f;
Chris@82 239 E T2s;
Chris@82 240 {
Chris@82 241 E T1V, T1W, T2D, T2E;
Chris@82 242 T1V = FNMS(KP500000000, T1J, T1E);
Chris@82 243 T1W = Ta - Tk;
Chris@82 244 T1X = FNMS(KP866025403, T1W, T1V);
Chris@82 245 T2v = FMA(KP866025403, T1W, T1V);
Chris@82 246 T2D = FMA(KP500000000, T2C, T2B);
Chris@82 247 T2E = T18 - T12;
Chris@82 248 T2F = FNMS(KP866025403, T2E, T2D);
Chris@82 249 T2Q = FMA(KP866025403, T2E, T2D);
Chris@82 250 }
Chris@82 251 {
Chris@82 252 E T2J, T2K, T1Y, T1Z;
Chris@82 253 T2J = FNMS(KP500000000, T2I, T2H);
Chris@82 254 T2K = T1s - T1A;
Chris@82 255 T2L = FNMS(KP866025403, T2K, T2J);
Chris@82 256 T2R = FMA(KP866025403, T2K, T2J);
Chris@82 257 T1Y = FNMS(KP500000000, T1R, T1M);
Chris@82 258 T1Z = TJ - TT;
Chris@82 259 T20 = FNMS(KP866025403, T1Z, T1Y);
Chris@82 260 T2w = FMA(KP866025403, T1Z, T1Y);
Chris@82 261 }
Chris@82 262 {
Chris@82 263 E T22, T27, T2h, T2i;
Chris@82 264 T22 = FNMS(KP500000000, T19, T1f);
Chris@82 265 T27 = T24 - T26;
Chris@82 266 T28 = FNMS(KP866025403, T27, T22);
Chris@82 267 T2t = FMA(KP866025403, T27, T22);
Chris@82 268 T2h = FNMS(KP500000000, Tl, To);
Chris@82 269 T2i = T1I - T1G;
Chris@82 270 T2j = FNMS(KP866025403, T2i, T2h);
Chris@82 271 T2p = FMA(KP866025403, T2i, T2h);
Chris@82 272 }
Chris@82 273 {
Chris@82 274 E T2k, T2l, T29, T2e;
Chris@82 275 T2k = FNMS(KP500000000, TU, Tz);
Chris@82 276 T2l = T1Q - T1O;
Chris@82 277 T2m = FNMS(KP866025403, T2l, T2k);
Chris@82 278 T2q = FMA(KP866025403, T2l, T2k);
Chris@82 279 T29 = FNMS(KP500000000, T1B, T1m);
Chris@82 280 T2e = T2b - T2d;
Chris@82 281 T2f = FNMS(KP866025403, T2e, T29);
Chris@82 282 T2s = FMA(KP866025403, T2e, T29);
Chris@82 283 }
Chris@82 284 {
Chris@82 285 E T21, T2g, T2P, T2S;
Chris@82 286 T21 = T1X + T20;
Chris@82 287 T2g = T28 + T2f;
Chris@82 288 Rp[WS(rs, 2)] = KP500000000 * (T21 - T2g);
Chris@82 289 Rm[WS(rs, 3)] = KP500000000 * (T21 + T2g);
Chris@82 290 T2P = T2m + T2j;
Chris@82 291 T2S = T2Q + T2R;
Chris@82 292 Ip[WS(rs, 2)] = KP500000000 * (T2P + T2S);
Chris@82 293 Im[WS(rs, 3)] = KP500000000 * (T2S - T2P);
Chris@82 294 }
Chris@82 295 {
Chris@82 296 E T2n, T2o, T2T, T2U;
Chris@82 297 T2n = T2j - T2m;
Chris@82 298 T2o = T2f - T28;
Chris@82 299 Ip[WS(rs, 5)] = KP500000000 * (T2n + T2o);
Chris@82 300 Im[0] = KP500000000 * (T2o - T2n);
Chris@82 301 T2T = T1X - T20;
Chris@82 302 T2U = T2R - T2Q;
Chris@82 303 Rm[0] = KP500000000 * (T2T - T2U);
Chris@82 304 Rp[WS(rs, 5)] = KP500000000 * (T2T + T2U);
Chris@82 305 }
Chris@82 306 {
Chris@82 307 E T2r, T2u, T2N, T2O;
Chris@82 308 T2r = T2p - T2q;
Chris@82 309 T2u = T2s - T2t;
Chris@82 310 Ip[WS(rs, 1)] = KP500000000 * (T2r + T2u);
Chris@82 311 Im[WS(rs, 4)] = KP500000000 * (T2u - T2r);
Chris@82 312 T2N = T2v - T2w;
Chris@82 313 T2O = T2L - T2F;
Chris@82 314 Rm[WS(rs, 4)] = KP500000000 * (T2N - T2O);
Chris@82 315 Rp[WS(rs, 1)] = KP500000000 * (T2N + T2O);
Chris@82 316 }
Chris@82 317 {
Chris@82 318 E T2x, T2y, T2z, T2M;
Chris@82 319 T2x = T2v + T2w;
Chris@82 320 T2y = T2t + T2s;
Chris@82 321 Rm[WS(rs, 1)] = KP500000000 * (T2x - T2y);
Chris@82 322 Rp[WS(rs, 4)] = KP500000000 * (T2x + T2y);
Chris@82 323 T2z = T2q + T2p;
Chris@82 324 T2M = T2F + T2L;
Chris@82 325 Ip[WS(rs, 4)] = KP500000000 * (T2z - T2M);
Chris@82 326 Im[WS(rs, 1)] = -(KP500000000 * (T2z + T2M));
Chris@82 327 }
Chris@82 328 }
Chris@82 329 }
Chris@82 330 }
Chris@82 331 }
Chris@82 332
Chris@82 333 static const tw_instr twinstr[] = {
Chris@82 334 {TW_FULL, 1, 12},
Chris@82 335 {TW_NEXT, 1, 0}
Chris@82 336 };
Chris@82 337
Chris@82 338 static const hc2c_desc desc = { 12, "hc2cfdft_12", twinstr, &GENUS, {96, 46, 46, 0} };
Chris@82 339
Chris@82 340 void X(codelet_hc2cfdft_12) (planner *p) {
Chris@82 341 X(khc2c_register) (p, hc2cfdft_12, &desc, HC2C_VIA_DFT);
Chris@82 342 }
Chris@82 343 #else
Chris@82 344
Chris@82 345 /* Generated by: ../../../genfft/gen_hc2cdft.native -compact -variables 4 -pipeline-latency 4 -n 12 -dit -name hc2cfdft_12 -include rdft/scalar/hc2cf.h */
Chris@82 346
Chris@82 347 /*
Chris@82 348 * This function contains 142 FP additions, 76 FP multiplications,
Chris@82 349 * (or, 112 additions, 46 multiplications, 30 fused multiply/add),
Chris@82 350 * 52 stack variables, 3 constants, and 48 memory accesses
Chris@82 351 */
Chris@82 352 #include "rdft/scalar/hc2cf.h"
Chris@82 353
Chris@82 354 static void hc2cfdft_12(R *Rp, R *Ip, R *Rm, R *Im, const R *W, stride rs, INT mb, INT me, INT ms)
Chris@82 355 {
Chris@82 356 DK(KP250000000, +0.250000000000000000000000000000000000000000000);
Chris@82 357 DK(KP500000000, +0.500000000000000000000000000000000000000000000);
Chris@82 358 DK(KP433012701, +0.433012701892219323381861585376468091735701313);
Chris@82 359 {
Chris@82 360 INT m;
Chris@82 361 for (m = mb, W = W + ((mb - 1) * 22); m < me; m = m + 1, Rp = Rp + ms, Ip = Ip + ms, Rm = Rm - ms, Im = Im - ms, W = W + 22, MAKE_VOLATILE_STRIDE(48, rs)) {
Chris@82 362 E Tm, T1t, T1d, T2j, Tj, T1Y, T1w, T1G, T1q, T2q, T1U, T2k, Tw, T1y, T17;
Chris@82 363 E T2g, TP, T21, T1B, T1J, T12, T2u, T1P, T2h;
Chris@82 364 {
Chris@82 365 E Tk, Tl, T1k, T1m, T1n, T1o, T4, T1f, T8, T1h, Th, T1c, Td, T1a, T19;
Chris@82 366 E T1b;
Chris@82 367 {
Chris@82 368 E T2, T3, T6, T7;
Chris@82 369 Tk = Ip[0];
Chris@82 370 Tl = Im[0];
Chris@82 371 T1k = Tk + Tl;
Chris@82 372 T1m = Rp[0];
Chris@82 373 T1n = Rm[0];
Chris@82 374 T1o = T1m - T1n;
Chris@82 375 T2 = Ip[WS(rs, 2)];
Chris@82 376 T3 = Im[WS(rs, 2)];
Chris@82 377 T4 = T2 - T3;
Chris@82 378 T1f = T2 + T3;
Chris@82 379 T6 = Rp[WS(rs, 2)];
Chris@82 380 T7 = Rm[WS(rs, 2)];
Chris@82 381 T8 = T6 + T7;
Chris@82 382 T1h = T6 - T7;
Chris@82 383 {
Chris@82 384 E Tf, Tg, Tb, Tc;
Chris@82 385 Tf = Rp[WS(rs, 4)];
Chris@82 386 Tg = Rm[WS(rs, 4)];
Chris@82 387 Th = Tf + Tg;
Chris@82 388 T1c = Tf - Tg;
Chris@82 389 Tb = Ip[WS(rs, 4)];
Chris@82 390 Tc = Im[WS(rs, 4)];
Chris@82 391 Td = Tb - Tc;
Chris@82 392 T1a = Tb + Tc;
Chris@82 393 }
Chris@82 394 }
Chris@82 395 Tm = Tk - Tl;
Chris@82 396 T1t = T1m + T1n;
Chris@82 397 T19 = W[16];
Chris@82 398 T1b = W[17];
Chris@82 399 T1d = FNMS(T1b, T1c, T19 * T1a);
Chris@82 400 T2j = FMA(T19, T1c, T1b * T1a);
Chris@82 401 {
Chris@82 402 E T9, T1u, Ti, T1v;
Chris@82 403 {
Chris@82 404 E T1, T5, Ta, Te;
Chris@82 405 T1 = W[6];
Chris@82 406 T5 = W[7];
Chris@82 407 T9 = FNMS(T5, T8, T1 * T4);
Chris@82 408 T1u = FMA(T1, T8, T5 * T4);
Chris@82 409 Ta = W[14];
Chris@82 410 Te = W[15];
Chris@82 411 Ti = FNMS(Te, Th, Ta * Td);
Chris@82 412 T1v = FMA(Ta, Th, Te * Td);
Chris@82 413 }
Chris@82 414 Tj = T9 + Ti;
Chris@82 415 T1Y = KP433012701 * (T1v - T1u);
Chris@82 416 T1w = T1u + T1v;
Chris@82 417 T1G = KP433012701 * (T9 - Ti);
Chris@82 418 }
Chris@82 419 {
Chris@82 420 E T1i, T1S, T1p, T1T;
Chris@82 421 {
Chris@82 422 E T1e, T1g, T1j, T1l;
Chris@82 423 T1e = W[8];
Chris@82 424 T1g = W[9];
Chris@82 425 T1i = FNMS(T1g, T1h, T1e * T1f);
Chris@82 426 T1S = FMA(T1e, T1h, T1g * T1f);
Chris@82 427 T1j = W[0];
Chris@82 428 T1l = W[1];
Chris@82 429 T1p = FNMS(T1l, T1o, T1j * T1k);
Chris@82 430 T1T = FMA(T1j, T1o, T1l * T1k);
Chris@82 431 }
Chris@82 432 T1q = T1i + T1p;
Chris@82 433 T2q = KP433012701 * (T1i - T1p);
Chris@82 434 T1U = KP433012701 * (T1S - T1T);
Chris@82 435 T2k = T1S + T1T;
Chris@82 436 }
Chris@82 437 }
Chris@82 438 {
Chris@82 439 E Tr, TT, Tv, TV, TA, TY, TE, T10, TN, T14, TJ, T16;
Chris@82 440 {
Chris@82 441 E Tp, Tq, TC, TD;
Chris@82 442 Tp = Ip[WS(rs, 3)];
Chris@82 443 Tq = Im[WS(rs, 3)];
Chris@82 444 Tr = Tp - Tq;
Chris@82 445 TT = Tp + Tq;
Chris@82 446 {
Chris@82 447 E Tt, Tu, Ty, Tz;
Chris@82 448 Tt = Rp[WS(rs, 3)];
Chris@82 449 Tu = Rm[WS(rs, 3)];
Chris@82 450 Tv = Tt + Tu;
Chris@82 451 TV = Tt - Tu;
Chris@82 452 Ty = Ip[WS(rs, 5)];
Chris@82 453 Tz = Im[WS(rs, 5)];
Chris@82 454 TA = Ty - Tz;
Chris@82 455 TY = Ty + Tz;
Chris@82 456 }
Chris@82 457 TC = Rp[WS(rs, 5)];
Chris@82 458 TD = Rm[WS(rs, 5)];
Chris@82 459 TE = TC + TD;
Chris@82 460 T10 = TC - TD;
Chris@82 461 {
Chris@82 462 E TL, TM, TH, TI;
Chris@82 463 TL = Rp[WS(rs, 1)];
Chris@82 464 TM = Rm[WS(rs, 1)];
Chris@82 465 TN = TL + TM;
Chris@82 466 T14 = TM - TL;
Chris@82 467 TH = Ip[WS(rs, 1)];
Chris@82 468 TI = Im[WS(rs, 1)];
Chris@82 469 TJ = TH - TI;
Chris@82 470 T16 = TH + TI;
Chris@82 471 }
Chris@82 472 }
Chris@82 473 {
Chris@82 474 E To, Ts, T13, T15;
Chris@82 475 To = W[10];
Chris@82 476 Ts = W[11];
Chris@82 477 Tw = FNMS(Ts, Tv, To * Tr);
Chris@82 478 T1y = FMA(To, Tv, Ts * Tr);
Chris@82 479 T13 = W[5];
Chris@82 480 T15 = W[4];
Chris@82 481 T17 = FMA(T13, T14, T15 * T16);
Chris@82 482 T2g = FNMS(T13, T16, T15 * T14);
Chris@82 483 }
Chris@82 484 {
Chris@82 485 E TF, T1z, TO, T1A;
Chris@82 486 {
Chris@82 487 E Tx, TB, TG, TK;
Chris@82 488 Tx = W[18];
Chris@82 489 TB = W[19];
Chris@82 490 TF = FNMS(TB, TE, Tx * TA);
Chris@82 491 T1z = FMA(Tx, TE, TB * TA);
Chris@82 492 TG = W[2];
Chris@82 493 TK = W[3];
Chris@82 494 TO = FNMS(TK, TN, TG * TJ);
Chris@82 495 T1A = FMA(TG, TN, TK * TJ);
Chris@82 496 }
Chris@82 497 TP = TF + TO;
Chris@82 498 T21 = KP433012701 * (T1A - T1z);
Chris@82 499 T1B = T1z + T1A;
Chris@82 500 T1J = KP433012701 * (TF - TO);
Chris@82 501 }
Chris@82 502 {
Chris@82 503 E TW, T1O, T11, T1N;
Chris@82 504 {
Chris@82 505 E TS, TU, TX, TZ;
Chris@82 506 TS = W[12];
Chris@82 507 TU = W[13];
Chris@82 508 TW = FNMS(TU, TV, TS * TT);
Chris@82 509 T1O = FMA(TS, TV, TU * TT);
Chris@82 510 TX = W[20];
Chris@82 511 TZ = W[21];
Chris@82 512 T11 = FNMS(TZ, T10, TX * TY);
Chris@82 513 T1N = FMA(TX, T10, TZ * TY);
Chris@82 514 }
Chris@82 515 T12 = TW + T11;
Chris@82 516 T2u = KP433012701 * (T11 - TW);
Chris@82 517 T1P = KP433012701 * (T1N - T1O);
Chris@82 518 T2h = T1O + T1N;
Chris@82 519 }
Chris@82 520 }
Chris@82 521 {
Chris@82 522 E TR, T2f, T2m, T2o, T1s, T1E, T1D, T2n;
Chris@82 523 {
Chris@82 524 E Tn, TQ, T2i, T2l;
Chris@82 525 Tn = Tj + Tm;
Chris@82 526 TQ = Tw + TP;
Chris@82 527 TR = Tn - TQ;
Chris@82 528 T2f = TQ + Tn;
Chris@82 529 T2i = T2g - T2h;
Chris@82 530 T2l = T2j + T2k;
Chris@82 531 T2m = T2i - T2l;
Chris@82 532 T2o = T2i + T2l;
Chris@82 533 }
Chris@82 534 {
Chris@82 535 E T18, T1r, T1x, T1C;
Chris@82 536 T18 = T12 + T17;
Chris@82 537 T1r = T1d + T1q;
Chris@82 538 T1s = T18 - T1r;
Chris@82 539 T1E = T18 + T1r;
Chris@82 540 T1x = T1t + T1w;
Chris@82 541 T1C = T1y + T1B;
Chris@82 542 T1D = T1x + T1C;
Chris@82 543 T2n = T1x - T1C;
Chris@82 544 }
Chris@82 545 Ip[WS(rs, 3)] = KP500000000 * (TR + T1s);
Chris@82 546 Rp[WS(rs, 3)] = KP500000000 * (T2n - T2o);
Chris@82 547 Im[WS(rs, 2)] = KP500000000 * (T1s - TR);
Chris@82 548 Rm[WS(rs, 2)] = KP500000000 * (T2n + T2o);
Chris@82 549 Rm[WS(rs, 5)] = KP500000000 * (T1D - T1E);
Chris@82 550 Im[WS(rs, 5)] = KP500000000 * (T2m - T2f);
Chris@82 551 Rp[0] = KP500000000 * (T1D + T1E);
Chris@82 552 Ip[0] = KP500000000 * (T2f + T2m);
Chris@82 553 }
Chris@82 554 {
Chris@82 555 E T1H, T2b, T2s, T2B, T2v, T2A, T1K, T2c, T1Q, T29, T1Z, T25, T22, T26, T1V;
Chris@82 556 E T28;
Chris@82 557 {
Chris@82 558 E T1F, T2r, T2t, T1I;
Chris@82 559 T1F = FNMS(KP250000000, T1w, KP500000000 * T1t);
Chris@82 560 T1H = T1F - T1G;
Chris@82 561 T2b = T1F + T1G;
Chris@82 562 T2r = FNMS(KP500000000, T2j, KP250000000 * T2k);
Chris@82 563 T2s = T2q - T2r;
Chris@82 564 T2B = T2q + T2r;
Chris@82 565 T2t = FMA(KP250000000, T2h, KP500000000 * T2g);
Chris@82 566 T2v = T2t - T2u;
Chris@82 567 T2A = T2u + T2t;
Chris@82 568 T1I = FNMS(KP250000000, T1B, KP500000000 * T1y);
Chris@82 569 T1K = T1I - T1J;
Chris@82 570 T2c = T1I + T1J;
Chris@82 571 }
Chris@82 572 {
Chris@82 573 E T1M, T1X, T20, T1R;
Chris@82 574 T1M = FNMS(KP250000000, T12, KP500000000 * T17);
Chris@82 575 T1Q = T1M - T1P;
Chris@82 576 T29 = T1P + T1M;
Chris@82 577 T1X = FNMS(KP250000000, Tj, KP500000000 * Tm);
Chris@82 578 T1Z = T1X - T1Y;
Chris@82 579 T25 = T1Y + T1X;
Chris@82 580 T20 = FNMS(KP250000000, TP, KP500000000 * Tw);
Chris@82 581 T22 = T20 - T21;
Chris@82 582 T26 = T21 + T20;
Chris@82 583 T1R = FNMS(KP250000000, T1q, KP500000000 * T1d);
Chris@82 584 T1V = T1R - T1U;
Chris@82 585 T28 = T1R + T1U;
Chris@82 586 }
Chris@82 587 {
Chris@82 588 E T1L, T1W, T2p, T2w;
Chris@82 589 T1L = T1H + T1K;
Chris@82 590 T1W = T1Q + T1V;
Chris@82 591 Rp[WS(rs, 2)] = T1L - T1W;
Chris@82 592 Rm[WS(rs, 3)] = T1L + T1W;
Chris@82 593 T2p = T22 + T1Z;
Chris@82 594 T2w = T2s - T2v;
Chris@82 595 Ip[WS(rs, 2)] = T2p + T2w;
Chris@82 596 Im[WS(rs, 3)] = T2w - T2p;
Chris@82 597 }
Chris@82 598 {
Chris@82 599 E T23, T24, T2x, T2y;
Chris@82 600 T23 = T1Z - T22;
Chris@82 601 T24 = T1V - T1Q;
Chris@82 602 Ip[WS(rs, 5)] = T23 + T24;
Chris@82 603 Im[0] = T24 - T23;
Chris@82 604 T2x = T1H - T1K;
Chris@82 605 T2y = T2v + T2s;
Chris@82 606 Rm[0] = T2x - T2y;
Chris@82 607 Rp[WS(rs, 5)] = T2x + T2y;
Chris@82 608 }
Chris@82 609 {
Chris@82 610 E T27, T2a, T2z, T2C;
Chris@82 611 T27 = T25 - T26;
Chris@82 612 T2a = T28 - T29;
Chris@82 613 Ip[WS(rs, 1)] = T27 + T2a;
Chris@82 614 Im[WS(rs, 4)] = T2a - T27;
Chris@82 615 T2z = T2b - T2c;
Chris@82 616 T2C = T2A - T2B;
Chris@82 617 Rm[WS(rs, 4)] = T2z - T2C;
Chris@82 618 Rp[WS(rs, 1)] = T2z + T2C;
Chris@82 619 }
Chris@82 620 {
Chris@82 621 E T2d, T2e, T2D, T2E;
Chris@82 622 T2d = T2b + T2c;
Chris@82 623 T2e = T29 + T28;
Chris@82 624 Rm[WS(rs, 1)] = T2d - T2e;
Chris@82 625 Rp[WS(rs, 4)] = T2d + T2e;
Chris@82 626 T2D = T26 + T25;
Chris@82 627 T2E = T2A + T2B;
Chris@82 628 Ip[WS(rs, 4)] = T2D + T2E;
Chris@82 629 Im[WS(rs, 1)] = T2E - T2D;
Chris@82 630 }
Chris@82 631 }
Chris@82 632 }
Chris@82 633 }
Chris@82 634 }
Chris@82 635
Chris@82 636 static const tw_instr twinstr[] = {
Chris@82 637 {TW_FULL, 1, 12},
Chris@82 638 {TW_NEXT, 1, 0}
Chris@82 639 };
Chris@82 640
Chris@82 641 static const hc2c_desc desc = { 12, "hc2cfdft_12", twinstr, &GENUS, {112, 46, 30, 0} };
Chris@82 642
Chris@82 643 void X(codelet_hc2cfdft_12) (planner *p) {
Chris@82 644 X(khc2c_register) (p, hc2cfdft_12, &desc, HC2C_VIA_DFT);
Chris@82 645 }
Chris@82 646 #endif