Mercurial > hg > sv-dependency-builds

annotate src/fftw-3.3.8/dft/scalar/codelets/q1_3.c @ 84:08ae793730bd

Find changesets by keywords (author, files, the commit message), revision number or hash, or revset expression.
Add null config files
author Chris Cannam
date Mon, 02 Mar 2020 14:03:47 +0000
parents d0c2a83c1364
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 08:04:30 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_twidsq.native -fma -compact -variables 4 -pipeline-latency 4 -reload-twiddle -dif -n 3 -name q1_3 -include dft/scalar/q.h */
Chris@82 29
Chris@82 30 /*
Chris@82 31 * This function contains 48 FP additions, 42 FP multiplications,
Chris@82 32 * (or, 18 additions, 12 multiplications, 30 fused multiply/add),
Chris@82 33 * 35 stack variables, 2 constants, and 36 memory accesses
Chris@82 34 */
Chris@82 35 #include "dft/scalar/q.h"
Chris@82 36
Chris@82 37 static void q1_3(R *rio, R *iio, const R *W, stride rs, stride vs, 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 * 4); m < me; m = m + 1, rio = rio + ms, iio = iio + ms, W = W + 4, MAKE_VOLATILE_STRIDE(6, rs), MAKE_VOLATILE_STRIDE(0, vs)) {
Chris@82 44 E T1, T4, T6, Tg, Td, Te, T9, Tf, Tp, Ts, Tu, TE, TB, TC, Tx;
Chris@82 45 E TD, TZ, T10, TV, T11, TN, TQ, TS, T12;
Chris@82 46 {
Chris@82 47 E T2, T3, Tv, Tw;
Chris@82 48 T1 = rio[0];
Chris@82 49 T2 = rio[WS(rs, 1)];
Chris@82 50 T3 = rio[WS(rs, 2)];
Chris@82 51 T4 = T2 + T3;
Chris@82 52 T6 = FNMS(KP500000000, T4, T1);
Chris@82 53 Tg = T3 - T2;
Chris@82 54 {
Chris@82 55 E T7, T8, Tq, Tr;
Chris@82 56 Td = iio[0];
Chris@82 57 T7 = iio[WS(rs, 1)];
Chris@82 58 T8 = iio[WS(rs, 2)];
Chris@82 59 Te = T7 + T8;
Chris@82 60 T9 = T7 - T8;
Chris@82 61 Tf = FNMS(KP500000000, Te, Td);
Chris@82 62 Tp = rio[WS(vs, 1)];
Chris@82 63 Tq = rio[WS(vs, 1) + WS(rs, 1)];
Chris@82 64 Tr = rio[WS(vs, 1) + WS(rs, 2)];
Chris@82 65 Ts = Tq + Tr;
Chris@82 66 Tu = FNMS(KP500000000, Ts, Tp);
Chris@82 67 TE = Tr - Tq;
Chris@82 68 }
Chris@82 69 TB = iio[WS(vs, 1)];
Chris@82 70 Tv = iio[WS(vs, 1) + WS(rs, 1)];
Chris@82 71 Tw = iio[WS(vs, 1) + WS(rs, 2)];
Chris@82 72 TC = Tv + Tw;
Chris@82 73 Tx = Tv - Tw;
Chris@82 74 TD = FNMS(KP500000000, TC, TB);
Chris@82 75 {
Chris@82 76 E TT, TU, TO, TP;
Chris@82 77 TZ = iio[WS(vs, 2)];
Chris@82 78 TT = iio[WS(vs, 2) + WS(rs, 1)];
Chris@82 79 TU = iio[WS(vs, 2) + WS(rs, 2)];
Chris@82 80 T10 = TT + TU;
Chris@82 81 TV = TT - TU;
Chris@82 82 T11 = FNMS(KP500000000, T10, TZ);
Chris@82 83 TN = rio[WS(vs, 2)];
Chris@82 84 TO = rio[WS(vs, 2) + WS(rs, 1)];
Chris@82 85 TP = rio[WS(vs, 2) + WS(rs, 2)];
Chris@82 86 TQ = TO + TP;
Chris@82 87 TS = FNMS(KP500000000, TQ, TN);
Chris@82 88 T12 = TP - TO;
Chris@82 89 }
Chris@82 90 }
Chris@82 91 rio[0] = T1 + T4;
Chris@82 92 iio[0] = Td + Te;
Chris@82 93 rio[WS(rs, 1)] = Tp + Ts;
Chris@82 94 iio[WS(rs, 1)] = TB + TC;
Chris@82 95 iio[WS(rs, 2)] = TZ + T10;
Chris@82 96 rio[WS(rs, 2)] = TN + TQ;
Chris@82 97 {
Chris@82 98 E Ta, Th, Tb, Ti, T5, Tc;
Chris@82 99 Ta = FMA(KP866025403, T9, T6);
Chris@82 100 Th = FMA(KP866025403, Tg, Tf);
Chris@82 101 T5 = W[0];
Chris@82 102 Tb = T5 * Ta;
Chris@82 103 Ti = T5 * Th;
Chris@82 104 Tc = W[1];
Chris@82 105 rio[WS(vs, 1)] = FMA(Tc, Th, Tb);
Chris@82 106 iio[WS(vs, 1)] = FNMS(Tc, Ta, Ti);
Chris@82 107 }
Chris@82 108 {
Chris@82 109 E T16, T19, T17, T1a, T15, T18;
Chris@82 110 T16 = FNMS(KP866025403, TV, TS);
Chris@82 111 T19 = FNMS(KP866025403, T12, T11);
Chris@82 112 T15 = W[2];
Chris@82 113 T17 = T15 * T16;
Chris@82 114 T1a = T15 * T19;
Chris@82 115 T18 = W[3];
Chris@82 116 rio[WS(vs, 2) + WS(rs, 2)] = FMA(T18, T19, T17);
Chris@82 117 iio[WS(vs, 2) + WS(rs, 2)] = FNMS(T18, T16, T1a);
Chris@82 118 }
Chris@82 119 {
Chris@82 120 E TI, TL, TJ, TM, TH, TK;
Chris@82 121 TI = FNMS(KP866025403, Tx, Tu);
Chris@82 122 TL = FNMS(KP866025403, TE, TD);
Chris@82 123 TH = W[2];
Chris@82 124 TJ = TH * TI;
Chris@82 125 TM = TH * TL;
Chris@82 126 TK = W[3];
Chris@82 127 rio[WS(vs, 2) + WS(rs, 1)] = FMA(TK, TL, TJ);
Chris@82 128 iio[WS(vs, 2) + WS(rs, 1)] = FNMS(TK, TI, TM);
Chris@82 129 }
Chris@82 130 {
Chris@82 131 E Ty, TF, Tz, TG, Tt, TA;
Chris@82 132 Ty = FMA(KP866025403, Tx, Tu);
Chris@82 133 TF = FMA(KP866025403, TE, TD);
Chris@82 134 Tt = W[0];
Chris@82 135 Tz = Tt * Ty;
Chris@82 136 TG = Tt * TF;
Chris@82 137 TA = W[1];
Chris@82 138 rio[WS(vs, 1) + WS(rs, 1)] = FMA(TA, TF, Tz);
Chris@82 139 iio[WS(vs, 1) + WS(rs, 1)] = FNMS(TA, Ty, TG);
Chris@82 140 }
Chris@82 141 {
Chris@82 142 E TW, T13, TX, T14, TR, TY;
Chris@82 143 TW = FMA(KP866025403, TV, TS);
Chris@82 144 T13 = FMA(KP866025403, T12, T11);
Chris@82 145 TR = W[0];
Chris@82 146 TX = TR * TW;
Chris@82 147 T14 = TR * T13;
Chris@82 148 TY = W[1];
Chris@82 149 rio[WS(vs, 1) + WS(rs, 2)] = FMA(TY, T13, TX);
Chris@82 150 iio[WS(vs, 1) + WS(rs, 2)] = FNMS(TY, TW, T14);
Chris@82 151 }
Chris@82 152 {
Chris@82 153 E Tk, Tn, Tl, To, Tj, Tm;
Chris@82 154 Tk = FNMS(KP866025403, T9, T6);
Chris@82 155 Tn = FNMS(KP866025403, Tg, Tf);
Chris@82 156 Tj = W[2];
Chris@82 157 Tl = Tj * Tk;
Chris@82 158 To = Tj * Tn;
Chris@82 159 Tm = W[3];
Chris@82 160 rio[WS(vs, 2)] = FMA(Tm, Tn, Tl);
Chris@82 161 iio[WS(vs, 2)] = FNMS(Tm, Tk, To);
Chris@82 162 }
Chris@82 163 }
Chris@82 164 }
Chris@82 165 }
Chris@82 166
Chris@82 167 static const tw_instr twinstr[] = {
Chris@82 168 {TW_FULL, 0, 3},
Chris@82 169 {TW_NEXT, 1, 0}
Chris@82 170 };
Chris@82 171
Chris@82 172 static const ct_desc desc = { 3, "q1_3", twinstr, &GENUS, {18, 12, 30, 0}, 0, 0, 0 };
Chris@82 173
Chris@82 174 void X(codelet_q1_3) (planner *p) {
Chris@82 175 X(kdft_difsq_register) (p, q1_3, &desc);
Chris@82 176 }
Chris@82 177 #else
Chris@82 178
Chris@82 179 /* Generated by: ../../../genfft/gen_twidsq.native -compact -variables 4 -pipeline-latency 4 -reload-twiddle -dif -n 3 -name q1_3 -include dft/scalar/q.h */
Chris@82 180
Chris@82 181 /*
Chris@82 182 * This function contains 48 FP additions, 36 FP multiplications,
Chris@82 183 * (or, 30 additions, 18 multiplications, 18 fused multiply/add),
Chris@82 184 * 35 stack variables, 2 constants, and 36 memory accesses
Chris@82 185 */
Chris@82 186 #include "dft/scalar/q.h"
Chris@82 187
Chris@82 188 static void q1_3(R *rio, R *iio, const R *W, stride rs, stride vs, INT mb, INT me, INT ms)
Chris@82 189 {
Chris@82 190 DK(KP866025403, +0.866025403784438646763723170752936183471402627);
Chris@82 191 DK(KP500000000, +0.500000000000000000000000000000000000000000000);
Chris@82 192 {
Chris@82 193 INT m;
Chris@82 194 for (m = mb, W = W + (mb * 4); m < me; m = m + 1, rio = rio + ms, iio = iio + ms, W = W + 4, MAKE_VOLATILE_STRIDE(6, rs), MAKE_VOLATILE_STRIDE(0, vs)) {
Chris@82 195 E T1, T4, T6, Tc, Td, Te, T9, Tf, Tl, To, Tq, Tw, Tx, Ty, Tt;
Chris@82 196 E Tz, TR, TS, TN, TT, TF, TI, TK, TQ;
Chris@82 197 {
Chris@82 198 E T2, T3, Tr, Ts;
Chris@82 199 T1 = rio[0];
Chris@82 200 T2 = rio[WS(rs, 1)];
Chris@82 201 T3 = rio[WS(rs, 2)];
Chris@82 202 T4 = T2 + T3;
Chris@82 203 T6 = FNMS(KP500000000, T4, T1);
Chris@82 204 Tc = KP866025403 * (T3 - T2);
Chris@82 205 {
Chris@82 206 E T7, T8, Tm, Tn;
Chris@82 207 Td = iio[0];
Chris@82 208 T7 = iio[WS(rs, 1)];
Chris@82 209 T8 = iio[WS(rs, 2)];
Chris@82 210 Te = T7 + T8;
Chris@82 211 T9 = KP866025403 * (T7 - T8);
Chris@82 212 Tf = FNMS(KP500000000, Te, Td);
Chris@82 213 Tl = rio[WS(vs, 1)];
Chris@82 214 Tm = rio[WS(vs, 1) + WS(rs, 1)];
Chris@82 215 Tn = rio[WS(vs, 1) + WS(rs, 2)];
Chris@82 216 To = Tm + Tn;
Chris@82 217 Tq = FNMS(KP500000000, To, Tl);
Chris@82 218 Tw = KP866025403 * (Tn - Tm);
Chris@82 219 }
Chris@82 220 Tx = iio[WS(vs, 1)];
Chris@82 221 Tr = iio[WS(vs, 1) + WS(rs, 1)];
Chris@82 222 Ts = iio[WS(vs, 1) + WS(rs, 2)];
Chris@82 223 Ty = Tr + Ts;
Chris@82 224 Tt = KP866025403 * (Tr - Ts);
Chris@82 225 Tz = FNMS(KP500000000, Ty, Tx);
Chris@82 226 {
Chris@82 227 E TL, TM, TG, TH;
Chris@82 228 TR = iio[WS(vs, 2)];
Chris@82 229 TL = iio[WS(vs, 2) + WS(rs, 1)];
Chris@82 230 TM = iio[WS(vs, 2) + WS(rs, 2)];
Chris@82 231 TS = TL + TM;
Chris@82 232 TN = KP866025403 * (TL - TM);
Chris@82 233 TT = FNMS(KP500000000, TS, TR);
Chris@82 234 TF = rio[WS(vs, 2)];
Chris@82 235 TG = rio[WS(vs, 2) + WS(rs, 1)];
Chris@82 236 TH = rio[WS(vs, 2) + WS(rs, 2)];
Chris@82 237 TI = TG + TH;
Chris@82 238 TK = FNMS(KP500000000, TI, TF);
Chris@82 239 TQ = KP866025403 * (TH - TG);
Chris@82 240 }
Chris@82 241 }
Chris@82 242 rio[0] = T1 + T4;
Chris@82 243 iio[0] = Td + Te;
Chris@82 244 rio[WS(rs, 1)] = Tl + To;
Chris@82 245 iio[WS(rs, 1)] = Tx + Ty;
Chris@82 246 iio[WS(rs, 2)] = TR + TS;
Chris@82 247 rio[WS(rs, 2)] = TF + TI;
Chris@82 248 {
Chris@82 249 E Ta, Tg, T5, Tb;
Chris@82 250 Ta = T6 + T9;
Chris@82 251 Tg = Tc + Tf;
Chris@82 252 T5 = W[0];
Chris@82 253 Tb = W[1];
Chris@82 254 rio[WS(vs, 1)] = FMA(T5, Ta, Tb * Tg);
Chris@82 255 iio[WS(vs, 1)] = FNMS(Tb, Ta, T5 * Tg);
Chris@82 256 }
Chris@82 257 {
Chris@82 258 E TW, TY, TV, TX;
Chris@82 259 TW = TK - TN;
Chris@82 260 TY = TT - TQ;
Chris@82 261 TV = W[2];
Chris@82 262 TX = W[3];
Chris@82 263 rio[WS(vs, 2) + WS(rs, 2)] = FMA(TV, TW, TX * TY);
Chris@82 264 iio[WS(vs, 2) + WS(rs, 2)] = FNMS(TX, TW, TV * TY);
Chris@82 265 }
Chris@82 266 {
Chris@82 267 E TC, TE, TB, TD;
Chris@82 268 TC = Tq - Tt;
Chris@82 269 TE = Tz - Tw;
Chris@82 270 TB = W[2];
Chris@82 271 TD = W[3];
Chris@82 272 rio[WS(vs, 2) + WS(rs, 1)] = FMA(TB, TC, TD * TE);
Chris@82 273 iio[WS(vs, 2) + WS(rs, 1)] = FNMS(TD, TC, TB * TE);
Chris@82 274 }
Chris@82 275 {
Chris@82 276 E Tu, TA, Tp, Tv;
Chris@82 277 Tu = Tq + Tt;
Chris@82 278 TA = Tw + Tz;
Chris@82 279 Tp = W[0];
Chris@82 280 Tv = W[1];
Chris@82 281 rio[WS(vs, 1) + WS(rs, 1)] = FMA(Tp, Tu, Tv * TA);
Chris@82 282 iio[WS(vs, 1) + WS(rs, 1)] = FNMS(Tv, Tu, Tp * TA);
Chris@82 283 }
Chris@82 284 {
Chris@82 285 E TO, TU, TJ, TP;
Chris@82 286 TO = TK + TN;
Chris@82 287 TU = TQ + TT;
Chris@82 288 TJ = W[0];
Chris@82 289 TP = W[1];
Chris@82 290 rio[WS(vs, 1) + WS(rs, 2)] = FMA(TJ, TO, TP * TU);
Chris@82 291 iio[WS(vs, 1) + WS(rs, 2)] = FNMS(TP, TO, TJ * TU);
Chris@82 292 }
Chris@82 293 {
Chris@82 294 E Ti, Tk, Th, Tj;
Chris@82 295 Ti = T6 - T9;
Chris@82 296 Tk = Tf - Tc;
Chris@82 297 Th = W[2];
Chris@82 298 Tj = W[3];
Chris@82 299 rio[WS(vs, 2)] = FMA(Th, Ti, Tj * Tk);
Chris@82 300 iio[WS(vs, 2)] = FNMS(Tj, Ti, Th * Tk);
Chris@82 301 }
Chris@82 302 }
Chris@82 303 }
Chris@82 304 }
Chris@82 305
Chris@82 306 static const tw_instr twinstr[] = {
Chris@82 307 {TW_FULL, 0, 3},
Chris@82 308 {TW_NEXT, 1, 0}
Chris@82 309 };
Chris@82 310
Chris@82 311 static const ct_desc desc = { 3, "q1_3", twinstr, &GENUS, {30, 18, 18, 0}, 0, 0, 0 };
Chris@82 312
Chris@82 313 void X(codelet_q1_3) (planner *p) {
Chris@82 314 X(kdft_difsq_register) (p, q1_3, &desc);
Chris@82 315 }
Chris@82 316 #endif