Mercurial > hg > sv-dependency-builds

annotate src/fftw-3.3.8/rdft/scalar/r2cb/hc2cbdft_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 08:07:58 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 -sign 1 -n 12 -dif -name hc2cbdft_12 -include rdft/scalar/hc2cb.h */
Chris@82 29
Chris@82 30 /*
Chris@82 31 * This function contains 142 FP additions, 68 FP multiplications,
Chris@82 32 * (or, 96 additions, 22 multiplications, 46 fused multiply/add),
Chris@82 33 * 55 stack variables, 2 constants, and 48 memory accesses
Chris@82 34 */
Chris@82 35 #include "rdft/scalar/hc2cb.h"
Chris@82 36
Chris@82 37 static void hc2cbdft_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 Tv, TC, TD, T1L, T1M, T2y, Tb, T1Z, T1E, T2D, T1e, T1U, TY, T2o, T13;
Chris@82 45 E T18, T19, T1O, T1P, T2E, Tm, T1V, T1H, T2z, T1h, T20, TO, T2p;
Chris@82 46 {
Chris@82 47 E T1, T4, Tu, TS, Tp, Ts, Tt, TT, T6, T9, TB, TV, Tw, Tz, TA;
Chris@82 48 E TW;
Chris@82 49 {
Chris@82 50 E T2, T3, Tq, Tr;
Chris@82 51 T1 = Rp[0];
Chris@82 52 T2 = Rp[WS(rs, 4)];
Chris@82 53 T3 = Rm[WS(rs, 3)];
Chris@82 54 T4 = T2 + T3;
Chris@82 55 Tu = T2 - T3;
Chris@82 56 TS = FNMS(KP500000000, T4, T1);
Chris@82 57 Tp = Ip[0];
Chris@82 58 Tq = Ip[WS(rs, 4)];
Chris@82 59 Tr = Im[WS(rs, 3)];
Chris@82 60 Ts = Tq - Tr;
Chris@82 61 Tt = FNMS(KP500000000, Ts, Tp);
Chris@82 62 TT = Tr + Tq;
Chris@82 63 }
Chris@82 64 {
Chris@82 65 E T7, T8, Tx, Ty;
Chris@82 66 T6 = Rm[WS(rs, 5)];
Chris@82 67 T7 = Rm[WS(rs, 1)];
Chris@82 68 T8 = Rp[WS(rs, 2)];
Chris@82 69 T9 = T7 + T8;
Chris@82 70 TB = T7 - T8;
Chris@82 71 TV = FNMS(KP500000000, T9, T6);
Chris@82 72 Tw = Im[WS(rs, 5)];
Chris@82 73 Tx = Im[WS(rs, 1)];
Chris@82 74 Ty = Ip[WS(rs, 2)];
Chris@82 75 Tz = Tx - Ty;
Chris@82 76 TA = FNMS(KP500000000, Tz, Tw);
Chris@82 77 TW = Tx + Ty;
Chris@82 78 }
Chris@82 79 {
Chris@82 80 E T5, Ta, T1C, T1D;
Chris@82 81 Tv = FMA(KP866025403, Tu, Tt);
Chris@82 82 TC = FNMS(KP866025403, TB, TA);
Chris@82 83 TD = Tv + TC;
Chris@82 84 T1L = FNMS(KP866025403, Tu, Tt);
Chris@82 85 T1M = FMA(KP866025403, TB, TA);
Chris@82 86 T2y = T1L + T1M;
Chris@82 87 T5 = T1 + T4;
Chris@82 88 Ta = T6 + T9;
Chris@82 89 Tb = T5 + Ta;
Chris@82 90 T1Z = T5 - Ta;
Chris@82 91 T1C = FMA(KP866025403, TT, TS);
Chris@82 92 T1D = FNMS(KP866025403, TW, TV);
Chris@82 93 T1E = T1C + T1D;
Chris@82 94 T2D = T1C - T1D;
Chris@82 95 {
Chris@82 96 E T1c, T1d, TU, TX;
Chris@82 97 T1c = Tp + Ts;
Chris@82 98 T1d = Tw + Tz;
Chris@82 99 T1e = T1c - T1d;
Chris@82 100 T1U = T1c + T1d;
Chris@82 101 TU = FNMS(KP866025403, TT, TS);
Chris@82 102 TX = FMA(KP866025403, TW, TV);
Chris@82 103 TY = TU - TX;
Chris@82 104 T2o = TU + TX;
Chris@82 105 }
Chris@82 106 }
Chris@82 107 }
Chris@82 108 {
Chris@82 109 E Tc, Tf, TE, T12, TZ, T10, TH, T11, Th, Tk, TJ, T17, T14, T15, TM;
Chris@82 110 E T16;
Chris@82 111 {
Chris@82 112 E Td, Te, TF, TG;
Chris@82 113 Tc = Rp[WS(rs, 3)];
Chris@82 114 Td = Rm[WS(rs, 4)];
Chris@82 115 Te = Rm[0];
Chris@82 116 Tf = Td + Te;
Chris@82 117 TE = FNMS(KP500000000, Tf, Tc);
Chris@82 118 T12 = Td - Te;
Chris@82 119 TZ = Ip[WS(rs, 3)];
Chris@82 120 TF = Im[WS(rs, 4)];
Chris@82 121 TG = Im[0];
Chris@82 122 T10 = TF + TG;
Chris@82 123 TH = TF - TG;
Chris@82 124 T11 = FMA(KP500000000, T10, TZ);
Chris@82 125 }
Chris@82 126 {
Chris@82 127 E Ti, Tj, TK, TL;
Chris@82 128 Th = Rm[WS(rs, 2)];
Chris@82 129 Ti = Rp[WS(rs, 1)];
Chris@82 130 Tj = Rp[WS(rs, 5)];
Chris@82 131 Tk = Ti + Tj;
Chris@82 132 TJ = FNMS(KP500000000, Tk, Th);
Chris@82 133 T17 = Ti - Tj;
Chris@82 134 T14 = Im[WS(rs, 2)];
Chris@82 135 TK = Ip[WS(rs, 5)];
Chris@82 136 TL = Ip[WS(rs, 1)];
Chris@82 137 T15 = TK + TL;
Chris@82 138 TM = TK - TL;
Chris@82 139 T16 = FMA(KP500000000, T15, T14);
Chris@82 140 }
Chris@82 141 {
Chris@82 142 E Tg, Tl, T1F, T1G;
Chris@82 143 T13 = FMA(KP866025403, T12, T11);
Chris@82 144 T18 = FNMS(KP866025403, T17, T16);
Chris@82 145 T19 = T13 + T18;
Chris@82 146 T1O = FNMS(KP866025403, T12, T11);
Chris@82 147 T1P = FMA(KP866025403, T17, T16);
Chris@82 148 T2E = T1O + T1P;
Chris@82 149 Tg = Tc + Tf;
Chris@82 150 Tl = Th + Tk;
Chris@82 151 Tm = Tg + Tl;
Chris@82 152 T1V = Tg - Tl;
Chris@82 153 T1F = FNMS(KP866025403, TH, TE);
Chris@82 154 T1G = FNMS(KP866025403, TM, TJ);
Chris@82 155 T1H = T1F + T1G;
Chris@82 156 T2z = T1F - T1G;
Chris@82 157 {
Chris@82 158 E T1f, T1g, TI, TN;
Chris@82 159 T1f = TZ - T10;
Chris@82 160 T1g = T15 - T14;
Chris@82 161 T1h = T1f + T1g;
Chris@82 162 T20 = T1f - T1g;
Chris@82 163 TI = FMA(KP866025403, TH, TE);
Chris@82 164 TN = FMA(KP866025403, TM, TJ);
Chris@82 165 TO = TI - TN;
Chris@82 166 T2p = TI + TN;
Chris@82 167 }
Chris@82 168 }
Chris@82 169 }
Chris@82 170 {
Chris@82 171 E Tn, T1i, TP, T1a, TQ, T1j, To, T1b, T1k, TR;
Chris@82 172 Tn = Tb + Tm;
Chris@82 173 T1i = T1e + T1h;
Chris@82 174 TP = TD + TO;
Chris@82 175 T1a = TY - T19;
Chris@82 176 To = W[0];
Chris@82 177 TQ = To * TP;
Chris@82 178 T1j = To * T1a;
Chris@82 179 TR = W[1];
Chris@82 180 T1b = FMA(TR, T1a, TQ);
Chris@82 181 T1k = FNMS(TR, TP, T1j);
Chris@82 182 Rp[0] = Tn - T1b;
Chris@82 183 Ip[0] = T1i + T1k;
Chris@82 184 Rm[0] = Tn + T1b;
Chris@82 185 Im[0] = T1k - T1i;
Chris@82 186 }
Chris@82 187 {
Chris@82 188 E T1p, T1l, T1n, T1o, T1x, T1s, T1v, T1t, T1z, T1m, T1r;
Chris@82 189 T1p = T1e - T1h;
Chris@82 190 T1m = Tb - Tm;
Chris@82 191 T1l = W[10];
Chris@82 192 T1n = T1l * T1m;
Chris@82 193 T1o = W[11];
Chris@82 194 T1x = T1o * T1m;
Chris@82 195 T1s = TD - TO;
Chris@82 196 T1v = TY + T19;
Chris@82 197 T1r = W[12];
Chris@82 198 T1t = T1r * T1s;
Chris@82 199 T1z = T1r * T1v;
Chris@82 200 {
Chris@82 201 E T1q, T1y, T1w, T1A, T1u;
Chris@82 202 T1q = FNMS(T1o, T1p, T1n);
Chris@82 203 T1y = FMA(T1l, T1p, T1x);
Chris@82 204 T1u = W[13];
Chris@82 205 T1w = FMA(T1u, T1v, T1t);
Chris@82 206 T1A = FNMS(T1u, T1s, T1z);
Chris@82 207 Rp[WS(rs, 3)] = T1q - T1w;
Chris@82 208 Ip[WS(rs, 3)] = T1y + T1A;
Chris@82 209 Rm[WS(rs, 3)] = T1q + T1w;
Chris@82 210 Im[WS(rs, 3)] = T1A - T1y;
Chris@82 211 }
Chris@82 212 }
Chris@82 213 {
Chris@82 214 E T1R, T2b, T27, T29, T2a, T2l, T1B, T1J, T1K, T25, T1W, T21, T1X, T23, T2e;
Chris@82 215 E T2h, T2f, T2j;
Chris@82 216 {
Chris@82 217 E T1N, T1Q, T28, T1I, T1T, T2d;
Chris@82 218 T1N = T1L - T1M;
Chris@82 219 T1Q = T1O - T1P;
Chris@82 220 T1R = T1N - T1Q;
Chris@82 221 T2b = T1N + T1Q;
Chris@82 222 T28 = T1E + T1H;
Chris@82 223 T27 = W[14];
Chris@82 224 T29 = T27 * T28;
Chris@82 225 T2a = W[15];
Chris@82 226 T2l = T2a * T28;
Chris@82 227 T1I = T1E - T1H;
Chris@82 228 T1B = W[2];
Chris@82 229 T1J = T1B * T1I;
Chris@82 230 T1K = W[3];
Chris@82 231 T25 = T1K * T1I;
Chris@82 232 T1W = T1U - T1V;
Chris@82 233 T21 = T1Z + T20;
Chris@82 234 T1T = W[4];
Chris@82 235 T1X = T1T * T1W;
Chris@82 236 T23 = T1T * T21;
Chris@82 237 T2e = T1V + T1U;
Chris@82 238 T2h = T1Z - T20;
Chris@82 239 T2d = W[16];
Chris@82 240 T2f = T2d * T2e;
Chris@82 241 T2j = T2d * T2h;
Chris@82 242 }
Chris@82 243 {
Chris@82 244 E T1S, T26, T22, T24, T1Y;
Chris@82 245 T1S = FNMS(T1K, T1R, T1J);
Chris@82 246 T26 = FMA(T1B, T1R, T25);
Chris@82 247 T1Y = W[5];
Chris@82 248 T22 = FMA(T1Y, T21, T1X);
Chris@82 249 T24 = FNMS(T1Y, T1W, T23);
Chris@82 250 Rp[WS(rs, 1)] = T1S - T22;
Chris@82 251 Ip[WS(rs, 1)] = T24 + T26;
Chris@82 252 Rm[WS(rs, 1)] = T22 + T1S;
Chris@82 253 Im[WS(rs, 1)] = T24 - T26;
Chris@82 254 }
Chris@82 255 {
Chris@82 256 E T2c, T2m, T2i, T2k, T2g;
Chris@82 257 T2c = FNMS(T2a, T2b, T29);
Chris@82 258 T2m = FMA(T27, T2b, T2l);
Chris@82 259 T2g = W[17];
Chris@82 260 T2i = FMA(T2g, T2h, T2f);
Chris@82 261 T2k = FNMS(T2g, T2e, T2j);
Chris@82 262 Rp[WS(rs, 4)] = T2c - T2i;
Chris@82 263 Ip[WS(rs, 4)] = T2k + T2m;
Chris@82 264 Rm[WS(rs, 4)] = T2i + T2c;
Chris@82 265 Im[WS(rs, 4)] = T2k - T2m;
Chris@82 266 }
Chris@82 267 }
Chris@82 268 {
Chris@82 269 E T2v, T2P, T2L, T2N, T2O, T2X, T2n, T2r, T2s, T2H, T2A, T2F, T2B, T2J, T2S;
Chris@82 270 E T2V, T2T, T2Z;
Chris@82 271 {
Chris@82 272 E T2t, T2u, T2M, T2q, T2x, T2R;
Chris@82 273 T2t = Tv - TC;
Chris@82 274 T2u = T13 - T18;
Chris@82 275 T2v = T2t + T2u;
Chris@82 276 T2P = T2t - T2u;
Chris@82 277 T2M = T2o - T2p;
Chris@82 278 T2L = W[18];
Chris@82 279 T2N = T2L * T2M;
Chris@82 280 T2O = W[19];
Chris@82 281 T2X = T2O * T2M;
Chris@82 282 T2q = T2o + T2p;
Chris@82 283 T2n = W[6];
Chris@82 284 T2r = T2n * T2q;
Chris@82 285 T2s = W[7];
Chris@82 286 T2H = T2s * T2q;
Chris@82 287 T2A = T2y + T2z;
Chris@82 288 T2F = T2D - T2E;
Chris@82 289 T2x = W[8];
Chris@82 290 T2B = T2x * T2A;
Chris@82 291 T2J = T2x * T2F;
Chris@82 292 T2S = T2y - T2z;
Chris@82 293 T2V = T2D + T2E;
Chris@82 294 T2R = W[20];
Chris@82 295 T2T = T2R * T2S;
Chris@82 296 T2Z = T2R * T2V;
Chris@82 297 }
Chris@82 298 {
Chris@82 299 E T2w, T2I, T2G, T2K, T2C;
Chris@82 300 T2w = FNMS(T2s, T2v, T2r);
Chris@82 301 T2I = FMA(T2n, T2v, T2H);
Chris@82 302 T2C = W[9];
Chris@82 303 T2G = FMA(T2C, T2F, T2B);
Chris@82 304 T2K = FNMS(T2C, T2A, T2J);
Chris@82 305 Rp[WS(rs, 2)] = T2w - T2G;
Chris@82 306 Ip[WS(rs, 2)] = T2I + T2K;
Chris@82 307 Rm[WS(rs, 2)] = T2w + T2G;
Chris@82 308 Im[WS(rs, 2)] = T2K - T2I;
Chris@82 309 }
Chris@82 310 {
Chris@82 311 E T2Q, T2Y, T2W, T30, T2U;
Chris@82 312 T2Q = FNMS(T2O, T2P, T2N);
Chris@82 313 T2Y = FMA(T2L, T2P, T2X);
Chris@82 314 T2U = W[21];
Chris@82 315 T2W = FMA(T2U, T2V, T2T);
Chris@82 316 T30 = FNMS(T2U, T2S, T2Z);
Chris@82 317 Rp[WS(rs, 5)] = T2Q - T2W;
Chris@82 318 Ip[WS(rs, 5)] = T2Y + T30;
Chris@82 319 Rm[WS(rs, 5)] = T2Q + T2W;
Chris@82 320 Im[WS(rs, 5)] = T30 - T2Y;
Chris@82 321 }
Chris@82 322 }
Chris@82 323 }
Chris@82 324 }
Chris@82 325 }
Chris@82 326
Chris@82 327 static const tw_instr twinstr[] = {
Chris@82 328 {TW_FULL, 1, 12},
Chris@82 329 {TW_NEXT, 1, 0}
Chris@82 330 };
Chris@82 331
Chris@82 332 static const hc2c_desc desc = { 12, "hc2cbdft_12", twinstr, &GENUS, {96, 22, 46, 0} };
Chris@82 333
Chris@82 334 void X(codelet_hc2cbdft_12) (planner *p) {
Chris@82 335 X(khc2c_register) (p, hc2cbdft_12, &desc, HC2C_VIA_DFT);
Chris@82 336 }
Chris@82 337 #else
Chris@82 338
Chris@82 339 /* Generated by: ../../../genfft/gen_hc2cdft.native -compact -variables 4 -pipeline-latency 4 -sign 1 -n 12 -dif -name hc2cbdft_12 -include rdft/scalar/hc2cb.h */
Chris@82 340
Chris@82 341 /*
Chris@82 342 * This function contains 142 FP additions, 60 FP multiplications,
Chris@82 343 * (or, 112 additions, 30 multiplications, 30 fused multiply/add),
Chris@82 344 * 47 stack variables, 2 constants, and 48 memory accesses
Chris@82 345 */
Chris@82 346 #include "rdft/scalar/hc2cb.h"
Chris@82 347
Chris@82 348 static void hc2cbdft_12(R *Rp, R *Ip, R *Rm, R *Im, const R *W, stride rs, INT mb, INT me, INT ms)
Chris@82 349 {
Chris@82 350 DK(KP500000000, +0.500000000000000000000000000000000000000000000);
Chris@82 351 DK(KP866025403, +0.866025403784438646763723170752936183471402627);
Chris@82 352 {
Chris@82 353 INT m;
Chris@82 354 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@82 355 E Tv, T1E, TC, T1F, TW, T1x, TT, T1w, T1d, T1N, Tb, T1R, TI, T1z, TN;
Chris@82 356 E T1A, T17, T1I, T12, T1H, T1g, T1S, Tm, T1O;
Chris@82 357 {
Chris@82 358 E T1, Tq, T6, TA, T4, Tp, Tt, TS, T9, Tw, Tz, TV;
Chris@82 359 T1 = Rp[0];
Chris@82 360 Tq = Ip[0];
Chris@82 361 T6 = Rm[WS(rs, 5)];
Chris@82 362 TA = Im[WS(rs, 5)];
Chris@82 363 {
Chris@82 364 E T2, T3, Tr, Ts;
Chris@82 365 T2 = Rp[WS(rs, 4)];
Chris@82 366 T3 = Rm[WS(rs, 3)];
Chris@82 367 T4 = T2 + T3;
Chris@82 368 Tp = KP866025403 * (T2 - T3);
Chris@82 369 Tr = Im[WS(rs, 3)];
Chris@82 370 Ts = Ip[WS(rs, 4)];
Chris@82 371 Tt = Tr - Ts;
Chris@82 372 TS = KP866025403 * (Tr + Ts);
Chris@82 373 }
Chris@82 374 {
Chris@82 375 E T7, T8, Tx, Ty;
Chris@82 376 T7 = Rm[WS(rs, 1)];
Chris@82 377 T8 = Rp[WS(rs, 2)];
Chris@82 378 T9 = T7 + T8;
Chris@82 379 Tw = KP866025403 * (T7 - T8);
Chris@82 380 Tx = Im[WS(rs, 1)];
Chris@82 381 Ty = Ip[WS(rs, 2)];
Chris@82 382 Tz = Tx - Ty;
Chris@82 383 TV = KP866025403 * (Tx + Ty);
Chris@82 384 }
Chris@82 385 {
Chris@82 386 E Tu, TB, TU, TR;
Chris@82 387 Tu = FMA(KP500000000, Tt, Tq);
Chris@82 388 Tv = Tp + Tu;
Chris@82 389 T1E = Tu - Tp;
Chris@82 390 TB = FMS(KP500000000, Tz, TA);
Chris@82 391 TC = Tw + TB;
Chris@82 392 T1F = TB - Tw;
Chris@82 393 TU = FNMS(KP500000000, T9, T6);
Chris@82 394 TW = TU + TV;
Chris@82 395 T1x = TU - TV;
Chris@82 396 TR = FNMS(KP500000000, T4, T1);
Chris@82 397 TT = TR - TS;
Chris@82 398 T1w = TR + TS;
Chris@82 399 {
Chris@82 400 E T1b, T1c, T5, Ta;
Chris@82 401 T1b = Tq - Tt;
Chris@82 402 T1c = Tz + TA;
Chris@82 403 T1d = T1b - T1c;
Chris@82 404 T1N = T1b + T1c;
Chris@82 405 T5 = T1 + T4;
Chris@82 406 Ta = T6 + T9;
Chris@82 407 Tb = T5 + Ta;
Chris@82 408 T1R = T5 - Ta;
Chris@82 409 }
Chris@82 410 }
Chris@82 411 }
Chris@82 412 {
Chris@82 413 E Tc, T10, Th, T15, Tf, TY, TH, TZ, Tk, T13, TM, T14;
Chris@82 414 Tc = Rp[WS(rs, 3)];
Chris@82 415 T10 = Ip[WS(rs, 3)];
Chris@82 416 Th = Rm[WS(rs, 2)];
Chris@82 417 T15 = Im[WS(rs, 2)];
Chris@82 418 {
Chris@82 419 E Td, Te, TF, TG;
Chris@82 420 Td = Rm[WS(rs, 4)];
Chris@82 421 Te = Rm[0];
Chris@82 422 Tf = Td + Te;
Chris@82 423 TY = KP866025403 * (Td - Te);
Chris@82 424 TF = Im[WS(rs, 4)];
Chris@82 425 TG = Im[0];
Chris@82 426 TH = KP866025403 * (TF - TG);
Chris@82 427 TZ = TF + TG;
Chris@82 428 }
Chris@82 429 {
Chris@82 430 E Ti, Tj, TK, TL;
Chris@82 431 Ti = Rp[WS(rs, 1)];
Chris@82 432 Tj = Rp[WS(rs, 5)];
Chris@82 433 Tk = Ti + Tj;
Chris@82 434 T13 = KP866025403 * (Ti - Tj);
Chris@82 435 TK = Ip[WS(rs, 5)];
Chris@82 436 TL = Ip[WS(rs, 1)];
Chris@82 437 TM = KP866025403 * (TK - TL);
Chris@82 438 T14 = TK + TL;
Chris@82 439 }
Chris@82 440 {
Chris@82 441 E TE, TJ, T16, T11;
Chris@82 442 TE = FNMS(KP500000000, Tf, Tc);
Chris@82 443 TI = TE + TH;
Chris@82 444 T1z = TE - TH;
Chris@82 445 TJ = FNMS(KP500000000, Tk, Th);
Chris@82 446 TN = TJ + TM;
Chris@82 447 T1A = TJ - TM;
Chris@82 448 T16 = FMA(KP500000000, T14, T15);
Chris@82 449 T17 = T13 - T16;
Chris@82 450 T1I = T13 + T16;
Chris@82 451 T11 = FMA(KP500000000, TZ, T10);
Chris@82 452 T12 = TY + T11;
Chris@82 453 T1H = T11 - TY;
Chris@82 454 {
Chris@82 455 E T1e, T1f, Tg, Tl;
Chris@82 456 T1e = T10 - TZ;
Chris@82 457 T1f = T14 - T15;
Chris@82 458 T1g = T1e + T1f;
Chris@82 459 T1S = T1e - T1f;
Chris@82 460 Tg = Tc + Tf;
Chris@82 461 Tl = Th + Tk;
Chris@82 462 Tm = Tg + Tl;
Chris@82 463 T1O = Tg - Tl;
Chris@82 464 }
Chris@82 465 }
Chris@82 466 }
Chris@82 467 {
Chris@82 468 E Tn, T1h, TP, T1p, T19, T1r, T1n, T1t;
Chris@82 469 Tn = Tb + Tm;
Chris@82 470 T1h = T1d + T1g;
Chris@82 471 {
Chris@82 472 E TD, TO, TX, T18;
Chris@82 473 TD = Tv - TC;
Chris@82 474 TO = TI - TN;
Chris@82 475 TP = TD + TO;
Chris@82 476 T1p = TD - TO;
Chris@82 477 TX = TT - TW;
Chris@82 478 T18 = T12 - T17;
Chris@82 479 T19 = TX - T18;
Chris@82 480 T1r = TX + T18;
Chris@82 481 {
Chris@82 482 E T1k, T1m, T1j, T1l;
Chris@82 483 T1k = Tb - Tm;
Chris@82 484 T1m = T1d - T1g;
Chris@82 485 T1j = W[10];
Chris@82 486 T1l = W[11];
Chris@82 487 T1n = FNMS(T1l, T1m, T1j * T1k);
Chris@82 488 T1t = FMA(T1l, T1k, T1j * T1m);
Chris@82 489 }
Chris@82 490 }
Chris@82 491 {
Chris@82 492 E T1a, T1i, To, TQ;
Chris@82 493 To = W[0];
Chris@82 494 TQ = W[1];
Chris@82 495 T1a = FMA(To, TP, TQ * T19);
Chris@82 496 T1i = FNMS(TQ, TP, To * T19);
Chris@82 497 Rp[0] = Tn - T1a;
Chris@82 498 Ip[0] = T1h + T1i;
Chris@82 499 Rm[0] = Tn + T1a;
Chris@82 500 Im[0] = T1i - T1h;
Chris@82 501 }
Chris@82 502 {
Chris@82 503 E T1s, T1u, T1o, T1q;
Chris@82 504 T1o = W[12];
Chris@82 505 T1q = W[13];
Chris@82 506 T1s = FMA(T1o, T1p, T1q * T1r);
Chris@82 507 T1u = FNMS(T1q, T1p, T1o * T1r);
Chris@82 508 Rp[WS(rs, 3)] = T1n - T1s;
Chris@82 509 Ip[WS(rs, 3)] = T1t + T1u;
Chris@82 510 Rm[WS(rs, 3)] = T1n + T1s;
Chris@82 511 Im[WS(rs, 3)] = T1u - T1t;
Chris@82 512 }
Chris@82 513 }
Chris@82 514 {
Chris@82 515 E T1C, T1Y, T1K, T20, T1U, T1V, T26, T27;
Chris@82 516 {
Chris@82 517 E T1y, T1B, T1G, T1J;
Chris@82 518 T1y = T1w + T1x;
Chris@82 519 T1B = T1z + T1A;
Chris@82 520 T1C = T1y - T1B;
Chris@82 521 T1Y = T1y + T1B;
Chris@82 522 T1G = T1E + T1F;
Chris@82 523 T1J = T1H - T1I;
Chris@82 524 T1K = T1G - T1J;
Chris@82 525 T20 = T1G + T1J;
Chris@82 526 }
Chris@82 527 {
Chris@82 528 E T1P, T1T, T1M, T1Q;
Chris@82 529 T1P = T1N - T1O;
Chris@82 530 T1T = T1R + T1S;
Chris@82 531 T1M = W[4];
Chris@82 532 T1Q = W[5];
Chris@82 533 T1U = FMA(T1M, T1P, T1Q * T1T);
Chris@82 534 T1V = FNMS(T1Q, T1P, T1M * T1T);
Chris@82 535 }
Chris@82 536 {
Chris@82 537 E T23, T25, T22, T24;
Chris@82 538 T23 = T1O + T1N;
Chris@82 539 T25 = T1R - T1S;
Chris@82 540 T22 = W[16];
Chris@82 541 T24 = W[17];
Chris@82 542 T26 = FMA(T22, T23, T24 * T25);
Chris@82 543 T27 = FNMS(T24, T23, T22 * T25);
Chris@82 544 }
Chris@82 545 {
Chris@82 546 E T1L, T1W, T1v, T1D;
Chris@82 547 T1v = W[2];
Chris@82 548 T1D = W[3];
Chris@82 549 T1L = FNMS(T1D, T1K, T1v * T1C);
Chris@82 550 T1W = FMA(T1D, T1C, T1v * T1K);
Chris@82 551 Rp[WS(rs, 1)] = T1L - T1U;
Chris@82 552 Ip[WS(rs, 1)] = T1V + T1W;
Chris@82 553 Rm[WS(rs, 1)] = T1U + T1L;
Chris@82 554 Im[WS(rs, 1)] = T1V - T1W;
Chris@82 555 }
Chris@82 556 {
Chris@82 557 E T21, T28, T1X, T1Z;
Chris@82 558 T1X = W[14];
Chris@82 559 T1Z = W[15];
Chris@82 560 T21 = FNMS(T1Z, T20, T1X * T1Y);
Chris@82 561 T28 = FMA(T1Z, T1Y, T1X * T20);
Chris@82 562 Rp[WS(rs, 4)] = T21 - T26;
Chris@82 563 Ip[WS(rs, 4)] = T27 + T28;
Chris@82 564 Rm[WS(rs, 4)] = T26 + T21;
Chris@82 565 Im[WS(rs, 4)] = T27 - T28;
Chris@82 566 }
Chris@82 567 }
Chris@82 568 {
Chris@82 569 E T2c, T2u, T2p, T2B, T2g, T2w, T2l, T2z;
Chris@82 570 {
Chris@82 571 E T2a, T2b, T2n, T2o;
Chris@82 572 T2a = TT + TW;
Chris@82 573 T2b = TI + TN;
Chris@82 574 T2c = T2a + T2b;
Chris@82 575 T2u = T2a - T2b;
Chris@82 576 T2n = T1w - T1x;
Chris@82 577 T2o = T1H + T1I;
Chris@82 578 T2p = T2n - T2o;
Chris@82 579 T2B = T2n + T2o;
Chris@82 580 }
Chris@82 581 {
Chris@82 582 E T2e, T2f, T2j, T2k;
Chris@82 583 T2e = Tv + TC;
Chris@82 584 T2f = T12 + T17;
Chris@82 585 T2g = T2e + T2f;
Chris@82 586 T2w = T2e - T2f;
Chris@82 587 T2j = T1E - T1F;
Chris@82 588 T2k = T1z - T1A;
Chris@82 589 T2l = T2j + T2k;
Chris@82 590 T2z = T2j - T2k;
Chris@82 591 }
Chris@82 592 {
Chris@82 593 E T2h, T2r, T2q, T2s;
Chris@82 594 {
Chris@82 595 E T29, T2d, T2i, T2m;
Chris@82 596 T29 = W[6];
Chris@82 597 T2d = W[7];
Chris@82 598 T2h = FNMS(T2d, T2g, T29 * T2c);
Chris@82 599 T2r = FMA(T2d, T2c, T29 * T2g);
Chris@82 600 T2i = W[8];
Chris@82 601 T2m = W[9];
Chris@82 602 T2q = FMA(T2i, T2l, T2m * T2p);
Chris@82 603 T2s = FNMS(T2m, T2l, T2i * T2p);
Chris@82 604 }
Chris@82 605 Rp[WS(rs, 2)] = T2h - T2q;
Chris@82 606 Ip[WS(rs, 2)] = T2r + T2s;
Chris@82 607 Rm[WS(rs, 2)] = T2h + T2q;
Chris@82 608 Im[WS(rs, 2)] = T2s - T2r;
Chris@82 609 }
Chris@82 610 {
Chris@82 611 E T2x, T2D, T2C, T2E;
Chris@82 612 {
Chris@82 613 E T2t, T2v, T2y, T2A;
Chris@82 614 T2t = W[18];
Chris@82 615 T2v = W[19];
Chris@82 616 T2x = FNMS(T2v, T2w, T2t * T2u);
Chris@82 617 T2D = FMA(T2v, T2u, T2t * T2w);
Chris@82 618 T2y = W[20];
Chris@82 619 T2A = W[21];
Chris@82 620 T2C = FMA(T2y, T2z, T2A * T2B);
Chris@82 621 T2E = FNMS(T2A, T2z, T2y * T2B);
Chris@82 622 }
Chris@82 623 Rp[WS(rs, 5)] = T2x - T2C;
Chris@82 624 Ip[WS(rs, 5)] = T2D + T2E;
Chris@82 625 Rm[WS(rs, 5)] = T2x + T2C;
Chris@82 626 Im[WS(rs, 5)] = T2E - T2D;
Chris@82 627 }
Chris@82 628 }
Chris@82 629 }
Chris@82 630 }
Chris@82 631 }
Chris@82 632
Chris@82 633 static const tw_instr twinstr[] = {
Chris@82 634 {TW_FULL, 1, 12},
Chris@82 635 {TW_NEXT, 1, 0}
Chris@82 636 };
Chris@82 637
Chris@82 638 static const hc2c_desc desc = { 12, "hc2cbdft_12", twinstr, &GENUS, {112, 30, 30, 0} };
Chris@82 639
Chris@82 640 void X(codelet_hc2cbdft_12) (planner *p) {
Chris@82 641 X(khc2c_register) (p, hc2cbdft_12, &desc, HC2C_VIA_DFT);
Chris@82 642 }
Chris@82 643 #endif