Mercurial > hg > sv-dependency-builds

annotate src/fftw-3.3.8/rdft/scalar/r2cb/hc2cb_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 08:07:51 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_hc2c.native -fma -compact -variables 4 -pipeline-latency 4 -sign 1 -n 8 -dif -name hc2cb_8 -include rdft/scalar/hc2cb.h */
Chris@82 29
Chris@82 30 /*
Chris@82 31 * This function contains 66 FP additions, 36 FP multiplications,
Chris@82 32 * (or, 44 additions, 14 multiplications, 22 fused multiply/add),
Chris@82 33 * 33 stack variables, 1 constants, and 32 memory accesses
Chris@82 34 */
Chris@82 35 #include "rdft/scalar/hc2cb.h"
Chris@82 36
Chris@82 37 static void hc2cb_8(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(KP707106781, +0.707106781186547524400844362104849039284835938);
Chris@82 40 {
Chris@82 41 INT m;
Chris@82 42 for (m = mb, W = W + ((mb - 1) * 14); m < me; m = m + 1, Rp = Rp + ms, Ip = Ip + ms, Rm = Rm - ms, Im = Im - ms, W = W + 14, MAKE_VOLATILE_STRIDE(32, rs)) {
Chris@82 43 E T7, T1i, T1n, Tk, TD, TV, T1b, TQ, Te, T1e, T1o, T1j, TE, TF, TR;
Chris@82 44 E Tv, TW;
Chris@82 45 {
Chris@82 46 E T3, Tg, TC, T19, T6, Tz, Tj, T1a;
Chris@82 47 {
Chris@82 48 E T1, T2, TA, TB;
Chris@82 49 T1 = Rp[0];
Chris@82 50 T2 = Rm[WS(rs, 3)];
Chris@82 51 T3 = T1 + T2;
Chris@82 52 Tg = T1 - T2;
Chris@82 53 TA = Ip[0];
Chris@82 54 TB = Im[WS(rs, 3)];
Chris@82 55 TC = TA + TB;
Chris@82 56 T19 = TA - TB;
Chris@82 57 }
Chris@82 58 {
Chris@82 59 E T4, T5, Th, Ti;
Chris@82 60 T4 = Rp[WS(rs, 2)];
Chris@82 61 T5 = Rm[WS(rs, 1)];
Chris@82 62 T6 = T4 + T5;
Chris@82 63 Tz = T4 - T5;
Chris@82 64 Th = Ip[WS(rs, 2)];
Chris@82 65 Ti = Im[WS(rs, 1)];
Chris@82 66 Tj = Th + Ti;
Chris@82 67 T1a = Th - Ti;
Chris@82 68 }
Chris@82 69 T7 = T3 + T6;
Chris@82 70 T1i = T3 - T6;
Chris@82 71 T1n = T19 - T1a;
Chris@82 72 Tk = Tg - Tj;
Chris@82 73 TD = Tz + TC;
Chris@82 74 TV = TC - Tz;
Chris@82 75 T1b = T19 + T1a;
Chris@82 76 TQ = Tg + Tj;
Chris@82 77 }
Chris@82 78 {
Chris@82 79 E Ta, Tl, To, T1c, Td, Tq, Tt, T1d, Tp, Tu;
Chris@82 80 {
Chris@82 81 E T8, T9, Tm, Tn;
Chris@82 82 T8 = Rp[WS(rs, 1)];
Chris@82 83 T9 = Rm[WS(rs, 2)];
Chris@82 84 Ta = T8 + T9;
Chris@82 85 Tl = T8 - T9;
Chris@82 86 Tm = Ip[WS(rs, 1)];
Chris@82 87 Tn = Im[WS(rs, 2)];
Chris@82 88 To = Tm + Tn;
Chris@82 89 T1c = Tm - Tn;
Chris@82 90 }
Chris@82 91 {
Chris@82 92 E Tb, Tc, Tr, Ts;
Chris@82 93 Tb = Rm[0];
Chris@82 94 Tc = Rp[WS(rs, 3)];
Chris@82 95 Td = Tb + Tc;
Chris@82 96 Tq = Tb - Tc;
Chris@82 97 Tr = Ip[WS(rs, 3)];
Chris@82 98 Ts = Im[0];
Chris@82 99 Tt = Tr + Ts;
Chris@82 100 T1d = Tr - Ts;
Chris@82 101 }
Chris@82 102 Te = Ta + Td;
Chris@82 103 T1e = T1c + T1d;
Chris@82 104 T1o = Ta - Td;
Chris@82 105 T1j = T1d - T1c;
Chris@82 106 TE = Tl + To;
Chris@82 107 TF = Tq + Tt;
Chris@82 108 TR = TE + TF;
Chris@82 109 Tp = Tl - To;
Chris@82 110 Tu = Tq - Tt;
Chris@82 111 Tv = Tp + Tu;
Chris@82 112 TW = Tp - Tu;
Chris@82 113 }
Chris@82 114 Rp[0] = T7 + Te;
Chris@82 115 Rm[0] = T1b + T1e;
Chris@82 116 {
Chris@82 117 E TS, TX, TT, TY, TP, TU;
Chris@82 118 TS = FNMS(KP707106781, TR, TQ);
Chris@82 119 TX = FMA(KP707106781, TW, TV);
Chris@82 120 TP = W[4];
Chris@82 121 TT = TP * TS;
Chris@82 122 TY = TP * TX;
Chris@82 123 TU = W[5];
Chris@82 124 Ip[WS(rs, 1)] = FNMS(TU, TX, TT);
Chris@82 125 Im[WS(rs, 1)] = FMA(TU, TS, TY);
Chris@82 126 }
Chris@82 127 {
Chris@82 128 E T1s, T1v, T1t, T1w, T1r, T1u;
Chris@82 129 T1s = T1i + T1j;
Chris@82 130 T1v = T1o + T1n;
Chris@82 131 T1r = W[2];
Chris@82 132 T1t = T1r * T1s;
Chris@82 133 T1w = T1r * T1v;
Chris@82 134 T1u = W[3];
Chris@82 135 Rp[WS(rs, 1)] = FNMS(T1u, T1v, T1t);
Chris@82 136 Rm[WS(rs, 1)] = FMA(T1u, T1s, T1w);
Chris@82 137 }
Chris@82 138 {
Chris@82 139 E T10, T13, T11, T14, TZ, T12;
Chris@82 140 T10 = FMA(KP707106781, TR, TQ);
Chris@82 141 T13 = FNMS(KP707106781, TW, TV);
Chris@82 142 TZ = W[12];
Chris@82 143 T11 = TZ * T10;
Chris@82 144 T14 = TZ * T13;
Chris@82 145 T12 = W[13];
Chris@82 146 Ip[WS(rs, 3)] = FNMS(T12, T13, T11);
Chris@82 147 Im[WS(rs, 3)] = FMA(T12, T10, T14);
Chris@82 148 }
Chris@82 149 {
Chris@82 150 E T1f, T15, T17, T18, T1g, T16;
Chris@82 151 T1f = T1b - T1e;
Chris@82 152 T16 = T7 - Te;
Chris@82 153 T15 = W[6];
Chris@82 154 T17 = T15 * T16;
Chris@82 155 T18 = W[7];
Chris@82 156 T1g = T18 * T16;
Chris@82 157 Rp[WS(rs, 2)] = FNMS(T18, T1f, T17);
Chris@82 158 Rm[WS(rs, 2)] = FMA(T15, T1f, T1g);
Chris@82 159 }
Chris@82 160 {
Chris@82 161 E T1k, T1p, T1l, T1q, T1h, T1m;
Chris@82 162 T1k = T1i - T1j;
Chris@82 163 T1p = T1n - T1o;
Chris@82 164 T1h = W[10];
Chris@82 165 T1l = T1h * T1k;
Chris@82 166 T1q = T1h * T1p;
Chris@82 167 T1m = W[11];
Chris@82 168 Rp[WS(rs, 3)] = FNMS(T1m, T1p, T1l);
Chris@82 169 Rm[WS(rs, 3)] = FMA(T1m, T1k, T1q);
Chris@82 170 }
Chris@82 171 {
Chris@82 172 E TH, TN, TJ, TL, TM, TO, Tf, Tx, Ty, TI, TG, TK, Tw;
Chris@82 173 TG = TE - TF;
Chris@82 174 TH = FNMS(KP707106781, TG, TD);
Chris@82 175 TN = FMA(KP707106781, TG, TD);
Chris@82 176 TK = FMA(KP707106781, Tv, Tk);
Chris@82 177 TJ = W[0];
Chris@82 178 TL = TJ * TK;
Chris@82 179 TM = W[1];
Chris@82 180 TO = TM * TK;
Chris@82 181 Tw = FNMS(KP707106781, Tv, Tk);
Chris@82 182 Tf = W[8];
Chris@82 183 Tx = Tf * Tw;
Chris@82 184 Ty = W[9];
Chris@82 185 TI = Ty * Tw;
Chris@82 186 Ip[WS(rs, 2)] = FNMS(Ty, TH, Tx);
Chris@82 187 Im[WS(rs, 2)] = FMA(Tf, TH, TI);
Chris@82 188 Ip[0] = FNMS(TM, TN, TL);
Chris@82 189 Im[0] = FMA(TJ, TN, TO);
Chris@82 190 }
Chris@82 191 }
Chris@82 192 }
Chris@82 193 }
Chris@82 194
Chris@82 195 static const tw_instr twinstr[] = {
Chris@82 196 {TW_FULL, 1, 8},
Chris@82 197 {TW_NEXT, 1, 0}
Chris@82 198 };
Chris@82 199
Chris@82 200 static const hc2c_desc desc = { 8, "hc2cb_8", twinstr, &GENUS, {44, 14, 22, 0} };
Chris@82 201
Chris@82 202 void X(codelet_hc2cb_8) (planner *p) {
Chris@82 203 X(khc2c_register) (p, hc2cb_8, &desc, HC2C_VIA_RDFT);
Chris@82 204 }
Chris@82 205 #else
Chris@82 206
Chris@82 207 /* Generated by: ../../../genfft/gen_hc2c.native -compact -variables 4 -pipeline-latency 4 -sign 1 -n 8 -dif -name hc2cb_8 -include rdft/scalar/hc2cb.h */
Chris@82 208
Chris@82 209 /*
Chris@82 210 * This function contains 66 FP additions, 32 FP multiplications,
Chris@82 211 * (or, 52 additions, 18 multiplications, 14 fused multiply/add),
Chris@82 212 * 30 stack variables, 1 constants, and 32 memory accesses
Chris@82 213 */
Chris@82 214 #include "rdft/scalar/hc2cb.h"
Chris@82 215
Chris@82 216 static void hc2cb_8(R *Rp, R *Ip, R *Rm, R *Im, const R *W, stride rs, INT mb, INT me, INT ms)
Chris@82 217 {
Chris@82 218 DK(KP707106781, +0.707106781186547524400844362104849039284835938);
Chris@82 219 {
Chris@82 220 INT m;
Chris@82 221 for (m = mb, W = W + ((mb - 1) * 14); m < me; m = m + 1, Rp = Rp + ms, Ip = Ip + ms, Rm = Rm - ms, Im = Im - ms, W = W + 14, MAKE_VOLATILE_STRIDE(32, rs)) {
Chris@82 222 E T7, T18, T1c, To, Ty, TM, TY, TC, Te, TZ, T10, Tv, Tz, TP, TS;
Chris@82 223 E TD;
Chris@82 224 {
Chris@82 225 E T3, TK, Tk, TX, T6, TW, Tn, TL;
Chris@82 226 {
Chris@82 227 E T1, T2, Ti, Tj;
Chris@82 228 T1 = Rp[0];
Chris@82 229 T2 = Rm[WS(rs, 3)];
Chris@82 230 T3 = T1 + T2;
Chris@82 231 TK = T1 - T2;
Chris@82 232 Ti = Ip[0];
Chris@82 233 Tj = Im[WS(rs, 3)];
Chris@82 234 Tk = Ti - Tj;
Chris@82 235 TX = Ti + Tj;
Chris@82 236 }
Chris@82 237 {
Chris@82 238 E T4, T5, Tl, Tm;
Chris@82 239 T4 = Rp[WS(rs, 2)];
Chris@82 240 T5 = Rm[WS(rs, 1)];
Chris@82 241 T6 = T4 + T5;
Chris@82 242 TW = T4 - T5;
Chris@82 243 Tl = Ip[WS(rs, 2)];
Chris@82 244 Tm = Im[WS(rs, 1)];
Chris@82 245 Tn = Tl - Tm;
Chris@82 246 TL = Tl + Tm;
Chris@82 247 }
Chris@82 248 T7 = T3 + T6;
Chris@82 249 T18 = TK + TL;
Chris@82 250 T1c = TX - TW;
Chris@82 251 To = Tk + Tn;
Chris@82 252 Ty = T3 - T6;
Chris@82 253 TM = TK - TL;
Chris@82 254 TY = TW + TX;
Chris@82 255 TC = Tk - Tn;
Chris@82 256 }
Chris@82 257 {
Chris@82 258 E Ta, TN, Tr, TO, Td, TQ, Tu, TR;
Chris@82 259 {
Chris@82 260 E T8, T9, Tp, Tq;
Chris@82 261 T8 = Rp[WS(rs, 1)];
Chris@82 262 T9 = Rm[WS(rs, 2)];
Chris@82 263 Ta = T8 + T9;
Chris@82 264 TN = T8 - T9;
Chris@82 265 Tp = Ip[WS(rs, 1)];
Chris@82 266 Tq = Im[WS(rs, 2)];
Chris@82 267 Tr = Tp - Tq;
Chris@82 268 TO = Tp + Tq;
Chris@82 269 }
Chris@82 270 {
Chris@82 271 E Tb, Tc, Ts, Tt;
Chris@82 272 Tb = Rm[0];
Chris@82 273 Tc = Rp[WS(rs, 3)];
Chris@82 274 Td = Tb + Tc;
Chris@82 275 TQ = Tb - Tc;
Chris@82 276 Ts = Ip[WS(rs, 3)];
Chris@82 277 Tt = Im[0];
Chris@82 278 Tu = Ts - Tt;
Chris@82 279 TR = Ts + Tt;
Chris@82 280 }
Chris@82 281 Te = Ta + Td;
Chris@82 282 TZ = TN + TO;
Chris@82 283 T10 = TQ + TR;
Chris@82 284 Tv = Tr + Tu;
Chris@82 285 Tz = Tu - Tr;
Chris@82 286 TP = TN - TO;
Chris@82 287 TS = TQ - TR;
Chris@82 288 TD = Ta - Td;
Chris@82 289 }
Chris@82 290 Rp[0] = T7 + Te;
Chris@82 291 Rm[0] = To + Tv;
Chris@82 292 {
Chris@82 293 E Tg, Tw, Tf, Th;
Chris@82 294 Tg = T7 - Te;
Chris@82 295 Tw = To - Tv;
Chris@82 296 Tf = W[6];
Chris@82 297 Th = W[7];
Chris@82 298 Rp[WS(rs, 2)] = FNMS(Th, Tw, Tf * Tg);
Chris@82 299 Rm[WS(rs, 2)] = FMA(Th, Tg, Tf * Tw);
Chris@82 300 }
Chris@82 301 {
Chris@82 302 E TG, TI, TF, TH;
Chris@82 303 TG = Ty + Tz;
Chris@82 304 TI = TD + TC;
Chris@82 305 TF = W[2];
Chris@82 306 TH = W[3];
Chris@82 307 Rp[WS(rs, 1)] = FNMS(TH, TI, TF * TG);
Chris@82 308 Rm[WS(rs, 1)] = FMA(TF, TI, TH * TG);
Chris@82 309 }
Chris@82 310 {
Chris@82 311 E TA, TE, Tx, TB;
Chris@82 312 TA = Ty - Tz;
Chris@82 313 TE = TC - TD;
Chris@82 314 Tx = W[10];
Chris@82 315 TB = W[11];
Chris@82 316 Rp[WS(rs, 3)] = FNMS(TB, TE, Tx * TA);
Chris@82 317 Rm[WS(rs, 3)] = FMA(Tx, TE, TB * TA);
Chris@82 318 }
Chris@82 319 {
Chris@82 320 E T1a, T1g, T1e, T1i, T19, T1d;
Chris@82 321 T19 = KP707106781 * (TZ + T10);
Chris@82 322 T1a = T18 - T19;
Chris@82 323 T1g = T18 + T19;
Chris@82 324 T1d = KP707106781 * (TP - TS);
Chris@82 325 T1e = T1c + T1d;
Chris@82 326 T1i = T1c - T1d;
Chris@82 327 {
Chris@82 328 E T17, T1b, T1f, T1h;
Chris@82 329 T17 = W[4];
Chris@82 330 T1b = W[5];
Chris@82 331 Ip[WS(rs, 1)] = FNMS(T1b, T1e, T17 * T1a);
Chris@82 332 Im[WS(rs, 1)] = FMA(T17, T1e, T1b * T1a);
Chris@82 333 T1f = W[12];
Chris@82 334 T1h = W[13];
Chris@82 335 Ip[WS(rs, 3)] = FNMS(T1h, T1i, T1f * T1g);
Chris@82 336 Im[WS(rs, 3)] = FMA(T1f, T1i, T1h * T1g);
Chris@82 337 }
Chris@82 338 }
Chris@82 339 {
Chris@82 340 E TU, T14, T12, T16, TT, T11;
Chris@82 341 TT = KP707106781 * (TP + TS);
Chris@82 342 TU = TM - TT;
Chris@82 343 T14 = TM + TT;
Chris@82 344 T11 = KP707106781 * (TZ - T10);
Chris@82 345 T12 = TY - T11;
Chris@82 346 T16 = TY + T11;
Chris@82 347 {
Chris@82 348 E TJ, TV, T13, T15;
Chris@82 349 TJ = W[8];
Chris@82 350 TV = W[9];
Chris@82 351 Ip[WS(rs, 2)] = FNMS(TV, T12, TJ * TU);
Chris@82 352 Im[WS(rs, 2)] = FMA(TV, TU, TJ * T12);
Chris@82 353 T13 = W[0];
Chris@82 354 T15 = W[1];
Chris@82 355 Ip[0] = FNMS(T15, T16, T13 * T14);
Chris@82 356 Im[0] = FMA(T15, T14, T13 * T16);
Chris@82 357 }
Chris@82 358 }
Chris@82 359 }
Chris@82 360 }
Chris@82 361 }
Chris@82 362
Chris@82 363 static const tw_instr twinstr[] = {
Chris@82 364 {TW_FULL, 1, 8},
Chris@82 365 {TW_NEXT, 1, 0}
Chris@82 366 };
Chris@82 367
Chris@82 368 static const hc2c_desc desc = { 8, "hc2cb_8", twinstr, &GENUS, {52, 18, 14, 0} };
Chris@82 369
Chris@82 370 void X(codelet_hc2cb_8) (planner *p) {
Chris@82 371 X(khc2c_register) (p, hc2cb_8, &desc, HC2C_VIA_RDFT);
Chris@82 372 }
Chris@82 373 #endif