Mercurial > hg > sv-dependency-builds

annotate src/fftw-3.3.5/dft/simd/common/t3bv_20.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 2cd0e3b3e1fd
children
rev   line source
Chris@42 1 /*
Chris@42 2 * Copyright (c) 2003, 2007-14 Matteo Frigo
Chris@42 3 * Copyright (c) 2003, 2007-14 Massachusetts Institute of Technology
Chris@42 4 *
Chris@42 5 * This program is free software; you can redistribute it and/or modify
Chris@42 6 * it under the terms of the GNU General Public License as published by
Chris@42 7 * the Free Software Foundation; either version 2 of the License, or
Chris@42 8 * (at your option) any later version.
Chris@42 9 *
Chris@42 10 * This program is distributed in the hope that it will be useful,
Chris@42 11 * but WITHOUT ANY WARRANTY; without even the implied warranty of
Chris@42 12 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
Chris@42 13 * GNU General Public License for more details.
Chris@42 14 *
Chris@42 15 * You should have received a copy of the GNU General Public License
Chris@42 16 * along with this program; if not, write to the Free Software
Chris@42 17 * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
Chris@42 18 *
Chris@42 19 */
Chris@42 20
Chris@42 21 /* This file was automatically generated --- DO NOT EDIT */
Chris@42 22 /* Generated on Sat Jul 30 16:45:02 EDT 2016 */
Chris@42 23
Chris@42 24 #include "codelet-dft.h"
Chris@42 25
Chris@42 26 #ifdef HAVE_FMA
Chris@42 27
Chris@42 28 /* Generated by: ../../../genfft/gen_twiddle_c.native -fma -reorder-insns -schedule-for-pipeline -simd -compact -variables 4 -pipeline-latency 8 -twiddle-log3 -precompute-twiddles -no-generate-bytw -n 20 -name t3bv_20 -include t3b.h -sign 1 */
Chris@42 29
Chris@42 30 /*
Chris@42 31 * This function contains 138 FP additions, 118 FP multiplications,
Chris@42 32 * (or, 92 additions, 72 multiplications, 46 fused multiply/add),
Chris@42 33 * 90 stack variables, 4 constants, and 40 memory accesses
Chris@42 34 */
Chris@42 35 #include "t3b.h"
Chris@42 36
Chris@42 37 static void t3bv_20(R *ri, R *ii, const R *W, stride rs, INT mb, INT me, INT ms)
Chris@42 38 {
Chris@42 39 DVK(KP559016994, +0.559016994374947424102293417182819058860154590);
Chris@42 40 DVK(KP951056516, +0.951056516295153572116439333379382143405698634);
Chris@42 41 DVK(KP618033988, +0.618033988749894848204586834365638117720309180);
Chris@42 42 DVK(KP250000000, +0.250000000000000000000000000000000000000000000);
Chris@42 43 {
Chris@42 44 INT m;
Chris@42 45 R *x;
Chris@42 46 x = ii;
Chris@42 47 for (m = mb, W = W + (mb * ((TWVL / VL) * 8)); m < me; m = m + VL, x = x + (VL * ms), W = W + (TWVL * 8), MAKE_VOLATILE_STRIDE(20, rs)) {
Chris@42 48 V T19, T1u, T1p, T1x, T1m, T1w, T1t, TI;
Chris@42 49 {
Chris@42 50 V T2, T8, T3, Td;
Chris@42 51 T2 = LDW(&(W[0]));
Chris@42 52 T8 = LDW(&(W[TWVL * 2]));
Chris@42 53 T3 = LDW(&(W[TWVL * 4]));
Chris@42 54 Td = LDW(&(W[TWVL * 6]));
Chris@42 55 {
Chris@42 56 V T7, T1g, T1F, T23, T1n, Tp, T18, T27, T1P, T1I, TU, T1L, T28, T1S, T1o;
Chris@42 57 V TE, T1l, T1j, T26, T2e;
Chris@42 58 {
Chris@42 59 V T1, T1e, T5, T1b;
Chris@42 60 T1 = LD(&(x[0]), ms, &(x[0]));
Chris@42 61 T1e = LD(&(x[WS(rs, 15)]), ms, &(x[WS(rs, 1)]));
Chris@42 62 T5 = LD(&(x[WS(rs, 10)]), ms, &(x[0]));
Chris@42 63 T1b = LD(&(x[WS(rs, 5)]), ms, &(x[WS(rs, 1)]));
Chris@42 64 {
Chris@42 65 V TA, Tx, TQ, T1O, T10, Th, T1G, T1R, T17, T1J, To, Ts, TR, Tv, TK;
Chris@42 66 V TM, TP, Ty, TB;
Chris@42 67 {
Chris@42 68 V Tq, Tt, T13, T16, Tk, Tn;
Chris@42 69 {
Chris@42 70 V Tl, Ti, T11, T14, TV, Tc, T6, Tb, Tf, TW, TY, T1f;
Chris@42 71 {
Chris@42 72 V T1d, Ta, T9, T4;
Chris@42 73 Ta = LD(&(x[WS(rs, 4)]), ms, &(x[0]));
Chris@42 74 TA = VZMULJ(T2, T8);
Chris@42 75 T9 = VZMUL(T2, T8);
Chris@42 76 Tx = VZMUL(T8, T3);
Chris@42 77 Tl = VZMULJ(T8, T3);
Chris@42 78 T4 = VZMUL(T2, T3);
Chris@42 79 Tq = VZMULJ(T2, T3);
Chris@42 80 Tt = VZMULJ(T2, Td);
Chris@42 81 Ti = VZMULJ(T8, Td);
Chris@42 82 T11 = VZMULJ(TA, Td);
Chris@42 83 T14 = VZMULJ(TA, T3);
Chris@42 84 TQ = VZMUL(TA, T3);
Chris@42 85 T1d = VZMULJ(T9, Td);
Chris@42 86 TV = VZMUL(T9, T3);
Chris@42 87 Tc = VZMULJ(T9, T3);
Chris@42 88 T6 = VZMUL(T4, T5);
Chris@42 89 Tb = VZMUL(T9, Ta);
Chris@42 90 Tf = LD(&(x[WS(rs, 14)]), ms, &(x[0]));
Chris@42 91 TW = LD(&(x[WS(rs, 13)]), ms, &(x[WS(rs, 1)]));
Chris@42 92 TY = LD(&(x[WS(rs, 3)]), ms, &(x[WS(rs, 1)]));
Chris@42 93 T1f = VZMUL(T1d, T1e);
Chris@42 94 }
Chris@42 95 {
Chris@42 96 V T1D, TX, TZ, T15, T1E, Tg, T12, T1c, Te, Tj, Tm;
Chris@42 97 T12 = LD(&(x[WS(rs, 17)]), ms, &(x[WS(rs, 1)]));
Chris@42 98 T1c = VZMUL(Tc, T1b);
Chris@42 99 Te = VZMULJ(Tc, Td);
Chris@42 100 T7 = VSUB(T1, T6);
Chris@42 101 T1D = VADD(T1, T6);
Chris@42 102 TX = VZMUL(TV, TW);
Chris@42 103 TZ = VZMUL(T8, TY);
Chris@42 104 T15 = LD(&(x[WS(rs, 7)]), ms, &(x[WS(rs, 1)]));
Chris@42 105 T13 = VZMUL(T11, T12);
Chris@42 106 T1g = VSUB(T1c, T1f);
Chris@42 107 T1E = VADD(T1c, T1f);
Chris@42 108 Tg = VZMUL(Te, Tf);
Chris@42 109 Tj = LD(&(x[WS(rs, 16)]), ms, &(x[0]));
Chris@42 110 Tm = LD(&(x[WS(rs, 6)]), ms, &(x[0]));
Chris@42 111 T1O = VADD(TX, TZ);
Chris@42 112 T10 = VSUB(TX, TZ);
Chris@42 113 T16 = VZMUL(T14, T15);
Chris@42 114 T1F = VSUB(T1D, T1E);
Chris@42 115 T23 = VADD(T1D, T1E);
Chris@42 116 Th = VSUB(Tb, Tg);
Chris@42 117 T1G = VADD(Tb, Tg);
Chris@42 118 Tk = VZMUL(Ti, Tj);
Chris@42 119 Tn = VZMUL(Tl, Tm);
Chris@42 120 }
Chris@42 121 }
Chris@42 122 {
Chris@42 123 V Tr, Tu, TJ, TL, TO;
Chris@42 124 Tr = LD(&(x[WS(rs, 8)]), ms, &(x[0]));
Chris@42 125 T1R = VADD(T13, T16);
Chris@42 126 T17 = VSUB(T13, T16);
Chris@42 127 Tu = LD(&(x[WS(rs, 18)]), ms, &(x[0]));
Chris@42 128 TJ = LD(&(x[WS(rs, 9)]), ms, &(x[WS(rs, 1)]));
Chris@42 129 TL = LD(&(x[WS(rs, 19)]), ms, &(x[WS(rs, 1)]));
Chris@42 130 TO = LD(&(x[WS(rs, 1)]), ms, &(x[WS(rs, 1)]));
Chris@42 131 T1J = VADD(Tk, Tn);
Chris@42 132 To = VSUB(Tk, Tn);
Chris@42 133 Ts = VZMUL(Tq, Tr);
Chris@42 134 TR = LD(&(x[WS(rs, 11)]), ms, &(x[WS(rs, 1)]));
Chris@42 135 Tv = VZMUL(Tt, Tu);
Chris@42 136 TK = VZMUL(T3, TJ);
Chris@42 137 TM = VZMUL(Td, TL);
Chris@42 138 TP = VZMUL(T2, TO);
Chris@42 139 Ty = LD(&(x[WS(rs, 12)]), ms, &(x[0]));
Chris@42 140 TB = LD(&(x[WS(rs, 2)]), ms, &(x[0]));
Chris@42 141 }
Chris@42 142 }
Chris@42 143 {
Chris@42 144 V T1N, Tw, T1H, TN, Tz, TC, T1i, TT, T1K, TS;
Chris@42 145 T1n = VSUB(Th, To);
Chris@42 146 Tp = VADD(Th, To);
Chris@42 147 TS = VZMUL(TQ, TR);
Chris@42 148 T1N = VADD(Ts, Tv);
Chris@42 149 Tw = VSUB(Ts, Tv);
Chris@42 150 T1H = VADD(TK, TM);
Chris@42 151 TN = VSUB(TK, TM);
Chris@42 152 Tz = VZMUL(Tx, Ty);
Chris@42 153 TC = VZMUL(TA, TB);
Chris@42 154 T18 = VSUB(T10, T17);
Chris@42 155 T1i = VADD(T10, T17);
Chris@42 156 TT = VSUB(TP, TS);
Chris@42 157 T1K = VADD(TP, TS);
Chris@42 158 T27 = VADD(T1N, T1O);
Chris@42 159 T1P = VSUB(T1N, T1O);
Chris@42 160 {
Chris@42 161 V TD, T1Q, T24, T1h, T25;
Chris@42 162 TD = VSUB(Tz, TC);
Chris@42 163 T1Q = VADD(Tz, TC);
Chris@42 164 T1I = VSUB(T1G, T1H);
Chris@42 165 T24 = VADD(T1G, T1H);
Chris@42 166 T1h = VADD(TN, TT);
Chris@42 167 TU = VSUB(TN, TT);
Chris@42 168 T25 = VADD(T1J, T1K);
Chris@42 169 T1L = VSUB(T1J, T1K);
Chris@42 170 T28 = VADD(T1Q, T1R);
Chris@42 171 T1S = VSUB(T1Q, T1R);
Chris@42 172 T1o = VSUB(Tw, TD);
Chris@42 173 TE = VADD(Tw, TD);
Chris@42 174 T1l = VSUB(T1h, T1i);
Chris@42 175 T1j = VADD(T1h, T1i);
Chris@42 176 T26 = VADD(T24, T25);
Chris@42 177 T2e = VSUB(T24, T25);
Chris@42 178 }
Chris@42 179 }
Chris@42 180 }
Chris@42 181 }
Chris@42 182 {
Chris@42 183 V T1M, T1Z, T1Y, T1T, T29, T2f, TH, TF, T1k, T1C;
Chris@42 184 T1M = VADD(T1I, T1L);
Chris@42 185 T1Z = VSUB(T1I, T1L);
Chris@42 186 T1Y = VSUB(T1P, T1S);
Chris@42 187 T1T = VADD(T1P, T1S);
Chris@42 188 T29 = VADD(T27, T28);
Chris@42 189 T2f = VSUB(T27, T28);
Chris@42 190 TH = VSUB(Tp, TE);
Chris@42 191 TF = VADD(Tp, TE);
Chris@42 192 T1k = VFNMS(LDK(KP250000000), T1j, T1g);
Chris@42 193 T1C = VADD(T1g, T1j);
Chris@42 194 {
Chris@42 195 V T1W, T2c, TG, T2i, T2g, T22, T20, T1V, T2b, T1U, T2a, T1B;
Chris@42 196 T19 = VFMA(LDK(KP618033988), T18, TU);
Chris@42 197 T1u = VFNMS(LDK(KP618033988), TU, T18);
Chris@42 198 T1W = VSUB(T1M, T1T);
Chris@42 199 T1U = VADD(T1M, T1T);
Chris@42 200 T2c = VSUB(T26, T29);
Chris@42 201 T2a = VADD(T26, T29);
Chris@42 202 TG = VFNMS(LDK(KP250000000), TF, T7);
Chris@42 203 T1B = VADD(T7, TF);
Chris@42 204 T2i = VMUL(LDK(KP951056516), VFNMS(LDK(KP618033988), T2e, T2f));
Chris@42 205 T2g = VMUL(LDK(KP951056516), VFMA(LDK(KP618033988), T2f, T2e));
Chris@42 206 T22 = VMUL(LDK(KP951056516), VFMA(LDK(KP618033988), T1Y, T1Z));
Chris@42 207 T20 = VMUL(LDK(KP951056516), VFNMS(LDK(KP618033988), T1Z, T1Y));
Chris@42 208 ST(&(x[WS(rs, 10)]), VADD(T1F, T1U), ms, &(x[0]));
Chris@42 209 T1V = VFNMS(LDK(KP250000000), T1U, T1F);
Chris@42 210 ST(&(x[0]), VADD(T23, T2a), ms, &(x[0]));
Chris@42 211 T2b = VFNMS(LDK(KP250000000), T2a, T23);
Chris@42 212 ST(&(x[WS(rs, 5)]), VFMAI(T1C, T1B), ms, &(x[WS(rs, 1)]));
Chris@42 213 ST(&(x[WS(rs, 15)]), VFNMSI(T1C, T1B), ms, &(x[WS(rs, 1)]));
Chris@42 214 T1p = VFMA(LDK(KP618033988), T1o, T1n);
Chris@42 215 T1x = VFNMS(LDK(KP618033988), T1n, T1o);
Chris@42 216 {
Chris@42 217 V T21, T1X, T2h, T2d;
Chris@42 218 T21 = VFMA(LDK(KP559016994), T1W, T1V);
Chris@42 219 T1X = VFNMS(LDK(KP559016994), T1W, T1V);
Chris@42 220 T2h = VFNMS(LDK(KP559016994), T2c, T2b);
Chris@42 221 T2d = VFMA(LDK(KP559016994), T2c, T2b);
Chris@42 222 ST(&(x[WS(rs, 18)]), VFMAI(T20, T1X), ms, &(x[0]));
Chris@42 223 ST(&(x[WS(rs, 2)]), VFNMSI(T20, T1X), ms, &(x[0]));
Chris@42 224 ST(&(x[WS(rs, 14)]), VFNMSI(T22, T21), ms, &(x[0]));
Chris@42 225 ST(&(x[WS(rs, 6)]), VFMAI(T22, T21), ms, &(x[0]));
Chris@42 226 ST(&(x[WS(rs, 16)]), VFMAI(T2g, T2d), ms, &(x[0]));
Chris@42 227 ST(&(x[WS(rs, 4)]), VFNMSI(T2g, T2d), ms, &(x[0]));
Chris@42 228 ST(&(x[WS(rs, 12)]), VFNMSI(T2i, T2h), ms, &(x[0]));
Chris@42 229 ST(&(x[WS(rs, 8)]), VFMAI(T2i, T2h), ms, &(x[0]));
Chris@42 230 T1m = VFMA(LDK(KP559016994), T1l, T1k);
Chris@42 231 T1w = VFNMS(LDK(KP559016994), T1l, T1k);
Chris@42 232 T1t = VFNMS(LDK(KP559016994), TH, TG);
Chris@42 233 TI = VFMA(LDK(KP559016994), TH, TG);
Chris@42 234 }
Chris@42 235 }
Chris@42 236 }
Chris@42 237 }
Chris@42 238 }
Chris@42 239 {
Chris@42 240 V T1A, T1y, T1q, T1s, T1a, T1r, T1z, T1v;
Chris@42 241 T1A = VFMA(LDK(KP951056516), T1x, T1w);
Chris@42 242 T1y = VFNMS(LDK(KP951056516), T1x, T1w);
Chris@42 243 T1q = VFMA(LDK(KP951056516), T1p, T1m);
Chris@42 244 T1s = VFNMS(LDK(KP951056516), T1p, T1m);
Chris@42 245 T1a = VFNMS(LDK(KP951056516), T19, TI);
Chris@42 246 T1r = VFMA(LDK(KP951056516), T19, TI);
Chris@42 247 T1z = VFNMS(LDK(KP951056516), T1u, T1t);
Chris@42 248 T1v = VFMA(LDK(KP951056516), T1u, T1t);
Chris@42 249 ST(&(x[WS(rs, 9)]), VFMAI(T1s, T1r), ms, &(x[WS(rs, 1)]));
Chris@42 250 ST(&(x[WS(rs, 11)]), VFNMSI(T1s, T1r), ms, &(x[WS(rs, 1)]));
Chris@42 251 ST(&(x[WS(rs, 1)]), VFMAI(T1q, T1a), ms, &(x[WS(rs, 1)]));
Chris@42 252 ST(&(x[WS(rs, 19)]), VFNMSI(T1q, T1a), ms, &(x[WS(rs, 1)]));
Chris@42 253 ST(&(x[WS(rs, 17)]), VFMAI(T1y, T1v), ms, &(x[WS(rs, 1)]));
Chris@42 254 ST(&(x[WS(rs, 3)]), VFNMSI(T1y, T1v), ms, &(x[WS(rs, 1)]));
Chris@42 255 ST(&(x[WS(rs, 13)]), VFMAI(T1A, T1z), ms, &(x[WS(rs, 1)]));
Chris@42 256 ST(&(x[WS(rs, 7)]), VFNMSI(T1A, T1z), ms, &(x[WS(rs, 1)]));
Chris@42 257 }
Chris@42 258 }
Chris@42 259 }
Chris@42 260 VLEAVE();
Chris@42 261 }
Chris@42 262
Chris@42 263 static const tw_instr twinstr[] = {
Chris@42 264 VTW(0, 1),
Chris@42 265 VTW(0, 3),
Chris@42 266 VTW(0, 9),
Chris@42 267 VTW(0, 19),
Chris@42 268 {TW_NEXT, VL, 0}
Chris@42 269 };
Chris@42 270
Chris@42 271 static const ct_desc desc = { 20, XSIMD_STRING("t3bv_20"), twinstr, &GENUS, {92, 72, 46, 0}, 0, 0, 0 };
Chris@42 272
Chris@42 273 void XSIMD(codelet_t3bv_20) (planner *p) {
Chris@42 274 X(kdft_dit_register) (p, t3bv_20, &desc);
Chris@42 275 }
Chris@42 276 #else /* HAVE_FMA */
Chris@42 277
Chris@42 278 /* Generated by: ../../../genfft/gen_twiddle_c.native -simd -compact -variables 4 -pipeline-latency 8 -twiddle-log3 -precompute-twiddles -no-generate-bytw -n 20 -name t3bv_20 -include t3b.h -sign 1 */
Chris@42 279
Chris@42 280 /*
Chris@42 281 * This function contains 138 FP additions, 92 FP multiplications,
Chris@42 282 * (or, 126 additions, 80 multiplications, 12 fused multiply/add),
Chris@42 283 * 73 stack variables, 4 constants, and 40 memory accesses
Chris@42 284 */
Chris@42 285 #include "t3b.h"
Chris@42 286
Chris@42 287 static void t3bv_20(R *ri, R *ii, const R *W, stride rs, INT mb, INT me, INT ms)
Chris@42 288 {
Chris@42 289 DVK(KP587785252, +0.587785252292473129168705954639072768597652438);
Chris@42 290 DVK(KP951056516, +0.951056516295153572116439333379382143405698634);
Chris@42 291 DVK(KP250000000, +0.250000000000000000000000000000000000000000000);
Chris@42 292 DVK(KP559016994, +0.559016994374947424102293417182819058860154590);
Chris@42 293 {
Chris@42 294 INT m;
Chris@42 295 R *x;
Chris@42 296 x = ii;
Chris@42 297 for (m = mb, W = W + (mb * ((TWVL / VL) * 8)); m < me; m = m + VL, x = x + (VL * ms), W = W + (TWVL * 8), MAKE_VOLATILE_STRIDE(20, rs)) {
Chris@42 298 V T2, T8, T9, TA, T3, Tc, T4, TV, T14, Tl, Tq, Tx, TQ, Td, Te;
Chris@42 299 V T1g, Ti, Tt, T11;
Chris@42 300 T2 = LDW(&(W[0]));
Chris@42 301 T8 = LDW(&(W[TWVL * 2]));
Chris@42 302 T9 = VZMUL(T2, T8);
Chris@42 303 TA = VZMULJ(T2, T8);
Chris@42 304 T3 = LDW(&(W[TWVL * 4]));
Chris@42 305 Tc = VZMULJ(T9, T3);
Chris@42 306 T4 = VZMUL(T2, T3);
Chris@42 307 TV = VZMUL(T9, T3);
Chris@42 308 T14 = VZMULJ(TA, T3);
Chris@42 309 Tl = VZMULJ(T8, T3);
Chris@42 310 Tq = VZMULJ(T2, T3);
Chris@42 311 Tx = VZMUL(T8, T3);
Chris@42 312 TQ = VZMUL(TA, T3);
Chris@42 313 Td = LDW(&(W[TWVL * 6]));
Chris@42 314 Te = VZMULJ(Tc, Td);
Chris@42 315 T1g = VZMULJ(T9, Td);
Chris@42 316 Ti = VZMULJ(T8, Td);
Chris@42 317 Tt = VZMULJ(T2, Td);
Chris@42 318 T11 = VZMULJ(TA, Td);
Chris@42 319 {
Chris@42 320 V T7, T1j, T1U, T2a, TU, T1n, T1o, T18, Tp, TE, TF, T26, T27, T28, T1M;
Chris@42 321 V T1P, T1W, T1b, T1c, T1k, T23, T24, T25, T1F, T1I, T1V, T1B, T1C;
Chris@42 322 {
Chris@42 323 V T1, T1i, T6, T1f, T1h, T5, T1e, T1S, T1T;
Chris@42 324 T1 = LD(&(x[0]), ms, &(x[0]));
Chris@42 325 T1h = LD(&(x[WS(rs, 15)]), ms, &(x[WS(rs, 1)]));
Chris@42 326 T1i = VZMUL(T1g, T1h);
Chris@42 327 T5 = LD(&(x[WS(rs, 10)]), ms, &(x[0]));
Chris@42 328 T6 = VZMUL(T4, T5);
Chris@42 329 T1e = LD(&(x[WS(rs, 5)]), ms, &(x[WS(rs, 1)]));
Chris@42 330 T1f = VZMUL(Tc, T1e);
Chris@42 331 T7 = VSUB(T1, T6);
Chris@42 332 T1j = VSUB(T1f, T1i);
Chris@42 333 T1S = VADD(T1, T6);
Chris@42 334 T1T = VADD(T1f, T1i);
Chris@42 335 T1U = VSUB(T1S, T1T);
Chris@42 336 T2a = VADD(T1S, T1T);
Chris@42 337 }
Chris@42 338 {
Chris@42 339 V Th, T1D, T10, T1L, T17, T1O, To, T1G, Tw, T1K, TN, T1E, TT, T1H, TD;
Chris@42 340 V T1N;
Chris@42 341 {
Chris@42 342 V Tb, Tg, Ta, Tf;
Chris@42 343 Ta = LD(&(x[WS(rs, 4)]), ms, &(x[0]));
Chris@42 344 Tb = VZMUL(T9, Ta);
Chris@42 345 Tf = LD(&(x[WS(rs, 14)]), ms, &(x[0]));
Chris@42 346 Tg = VZMUL(Te, Tf);
Chris@42 347 Th = VSUB(Tb, Tg);
Chris@42 348 T1D = VADD(Tb, Tg);
Chris@42 349 }
Chris@42 350 {
Chris@42 351 V TX, TZ, TW, TY;
Chris@42 352 TW = LD(&(x[WS(rs, 13)]), ms, &(x[WS(rs, 1)]));
Chris@42 353 TX = VZMUL(TV, TW);
Chris@42 354 TY = LD(&(x[WS(rs, 3)]), ms, &(x[WS(rs, 1)]));
Chris@42 355 TZ = VZMUL(T8, TY);
Chris@42 356 T10 = VSUB(TX, TZ);
Chris@42 357 T1L = VADD(TX, TZ);
Chris@42 358 }
Chris@42 359 {
Chris@42 360 V T13, T16, T12, T15;
Chris@42 361 T12 = LD(&(x[WS(rs, 17)]), ms, &(x[WS(rs, 1)]));
Chris@42 362 T13 = VZMUL(T11, T12);
Chris@42 363 T15 = LD(&(x[WS(rs, 7)]), ms, &(x[WS(rs, 1)]));
Chris@42 364 T16 = VZMUL(T14, T15);
Chris@42 365 T17 = VSUB(T13, T16);
Chris@42 366 T1O = VADD(T13, T16);
Chris@42 367 }
Chris@42 368 {
Chris@42 369 V Tk, Tn, Tj, Tm;
Chris@42 370 Tj = LD(&(x[WS(rs, 16)]), ms, &(x[0]));
Chris@42 371 Tk = VZMUL(Ti, Tj);
Chris@42 372 Tm = LD(&(x[WS(rs, 6)]), ms, &(x[0]));
Chris@42 373 Tn = VZMUL(Tl, Tm);
Chris@42 374 To = VSUB(Tk, Tn);
Chris@42 375 T1G = VADD(Tk, Tn);
Chris@42 376 }
Chris@42 377 {
Chris@42 378 V Ts, Tv, Tr, Tu;
Chris@42 379 Tr = LD(&(x[WS(rs, 8)]), ms, &(x[0]));
Chris@42 380 Ts = VZMUL(Tq, Tr);
Chris@42 381 Tu = LD(&(x[WS(rs, 18)]), ms, &(x[0]));
Chris@42 382 Tv = VZMUL(Tt, Tu);
Chris@42 383 Tw = VSUB(Ts, Tv);
Chris@42 384 T1K = VADD(Ts, Tv);
Chris@42 385 }
Chris@42 386 {
Chris@42 387 V TK, TM, TJ, TL;
Chris@42 388 TJ = LD(&(x[WS(rs, 9)]), ms, &(x[WS(rs, 1)]));
Chris@42 389 TK = VZMUL(T3, TJ);
Chris@42 390 TL = LD(&(x[WS(rs, 19)]), ms, &(x[WS(rs, 1)]));
Chris@42 391 TM = VZMUL(Td, TL);
Chris@42 392 TN = VSUB(TK, TM);
Chris@42 393 T1E = VADD(TK, TM);
Chris@42 394 }
Chris@42 395 {
Chris@42 396 V TP, TS, TO, TR;
Chris@42 397 TO = LD(&(x[WS(rs, 1)]), ms, &(x[WS(rs, 1)]));
Chris@42 398 TP = VZMUL(T2, TO);
Chris@42 399 TR = LD(&(x[WS(rs, 11)]), ms, &(x[WS(rs, 1)]));
Chris@42 400 TS = VZMUL(TQ, TR);
Chris@42 401 TT = VSUB(TP, TS);
Chris@42 402 T1H = VADD(TP, TS);
Chris@42 403 }
Chris@42 404 {
Chris@42 405 V Tz, TC, Ty, TB;
Chris@42 406 Ty = LD(&(x[WS(rs, 12)]), ms, &(x[0]));
Chris@42 407 Tz = VZMUL(Tx, Ty);
Chris@42 408 TB = LD(&(x[WS(rs, 2)]), ms, &(x[0]));
Chris@42 409 TC = VZMUL(TA, TB);
Chris@42 410 TD = VSUB(Tz, TC);
Chris@42 411 T1N = VADD(Tz, TC);
Chris@42 412 }
Chris@42 413 TU = VSUB(TN, TT);
Chris@42 414 T1n = VSUB(Th, To);
Chris@42 415 T1o = VSUB(Tw, TD);
Chris@42 416 T18 = VSUB(T10, T17);
Chris@42 417 Tp = VADD(Th, To);
Chris@42 418 TE = VADD(Tw, TD);
Chris@42 419 TF = VADD(Tp, TE);
Chris@42 420 T26 = VADD(T1K, T1L);
Chris@42 421 T27 = VADD(T1N, T1O);
Chris@42 422 T28 = VADD(T26, T27);
Chris@42 423 T1M = VSUB(T1K, T1L);
Chris@42 424 T1P = VSUB(T1N, T1O);
Chris@42 425 T1W = VADD(T1M, T1P);
Chris@42 426 T1b = VADD(TN, TT);
Chris@42 427 T1c = VADD(T10, T17);
Chris@42 428 T1k = VADD(T1b, T1c);
Chris@42 429 T23 = VADD(T1D, T1E);
Chris@42 430 T24 = VADD(T1G, T1H);
Chris@42 431 T25 = VADD(T23, T24);
Chris@42 432 T1F = VSUB(T1D, T1E);
Chris@42 433 T1I = VSUB(T1G, T1H);
Chris@42 434 T1V = VADD(T1F, T1I);
Chris@42 435 }
Chris@42 436 T1B = VADD(T7, TF);
Chris@42 437 T1C = VBYI(VADD(T1j, T1k));
Chris@42 438 ST(&(x[WS(rs, 15)]), VSUB(T1B, T1C), ms, &(x[WS(rs, 1)]));
Chris@42 439 ST(&(x[WS(rs, 5)]), VADD(T1B, T1C), ms, &(x[WS(rs, 1)]));
Chris@42 440 {
Chris@42 441 V T29, T2b, T2c, T2g, T2i, T2e, T2f, T2h, T2d;
Chris@42 442 T29 = VMUL(LDK(KP559016994), VSUB(T25, T28));
Chris@42 443 T2b = VADD(T25, T28);
Chris@42 444 T2c = VFNMS(LDK(KP250000000), T2b, T2a);
Chris@42 445 T2e = VSUB(T23, T24);
Chris@42 446 T2f = VSUB(T26, T27);
Chris@42 447 T2g = VBYI(VFMA(LDK(KP951056516), T2e, VMUL(LDK(KP587785252), T2f)));
Chris@42 448 T2i = VBYI(VFNMS(LDK(KP951056516), T2f, VMUL(LDK(KP587785252), T2e)));
Chris@42 449 ST(&(x[0]), VADD(T2a, T2b), ms, &(x[0]));
Chris@42 450 T2h = VSUB(T2c, T29);
Chris@42 451 ST(&(x[WS(rs, 8)]), VSUB(T2h, T2i), ms, &(x[0]));
Chris@42 452 ST(&(x[WS(rs, 12)]), VADD(T2i, T2h), ms, &(x[0]));
Chris@42 453 T2d = VADD(T29, T2c);
Chris@42 454 ST(&(x[WS(rs, 4)]), VSUB(T2d, T2g), ms, &(x[0]));
Chris@42 455 ST(&(x[WS(rs, 16)]), VADD(T2g, T2d), ms, &(x[0]));
Chris@42 456 }
Chris@42 457 {
Chris@42 458 V T1Z, T1X, T1Y, T1R, T21, T1J, T1Q, T22, T20;
Chris@42 459 T1Z = VMUL(LDK(KP559016994), VSUB(T1V, T1W));
Chris@42 460 T1X = VADD(T1V, T1W);
Chris@42 461 T1Y = VFNMS(LDK(KP250000000), T1X, T1U);
Chris@42 462 T1J = VSUB(T1F, T1I);
Chris@42 463 T1Q = VSUB(T1M, T1P);
Chris@42 464 T1R = VBYI(VFNMS(LDK(KP951056516), T1Q, VMUL(LDK(KP587785252), T1J)));
Chris@42 465 T21 = VBYI(VFMA(LDK(KP951056516), T1J, VMUL(LDK(KP587785252), T1Q)));
Chris@42 466 ST(&(x[WS(rs, 10)]), VADD(T1U, T1X), ms, &(x[0]));
Chris@42 467 T22 = VADD(T1Z, T1Y);
Chris@42 468 ST(&(x[WS(rs, 6)]), VADD(T21, T22), ms, &(x[0]));
Chris@42 469 ST(&(x[WS(rs, 14)]), VSUB(T22, T21), ms, &(x[0]));
Chris@42 470 T20 = VSUB(T1Y, T1Z);
Chris@42 471 ST(&(x[WS(rs, 2)]), VADD(T1R, T20), ms, &(x[0]));
Chris@42 472 ST(&(x[WS(rs, 18)]), VSUB(T20, T1R), ms, &(x[0]));
Chris@42 473 }
Chris@42 474 {
Chris@42 475 V T19, T1p, T1w, T1u, T1m, T1x, TI, T1t;
Chris@42 476 T19 = VFNMS(LDK(KP951056516), T18, VMUL(LDK(KP587785252), TU));
Chris@42 477 T1p = VFNMS(LDK(KP951056516), T1o, VMUL(LDK(KP587785252), T1n));
Chris@42 478 T1w = VFMA(LDK(KP951056516), T1n, VMUL(LDK(KP587785252), T1o));
Chris@42 479 T1u = VFMA(LDK(KP951056516), TU, VMUL(LDK(KP587785252), T18));
Chris@42 480 {
Chris@42 481 V T1d, T1l, TG, TH;
Chris@42 482 T1d = VMUL(LDK(KP559016994), VSUB(T1b, T1c));
Chris@42 483 T1l = VFNMS(LDK(KP250000000), T1k, T1j);
Chris@42 484 T1m = VSUB(T1d, T1l);
Chris@42 485 T1x = VADD(T1d, T1l);
Chris@42 486 TG = VFNMS(LDK(KP250000000), TF, T7);
Chris@42 487 TH = VMUL(LDK(KP559016994), VSUB(Tp, TE));
Chris@42 488 TI = VSUB(TG, TH);
Chris@42 489 T1t = VADD(TH, TG);
Chris@42 490 }
Chris@42 491 {
Chris@42 492 V T1a, T1q, T1z, T1A;
Chris@42 493 T1a = VSUB(TI, T19);
Chris@42 494 T1q = VBYI(VSUB(T1m, T1p));
Chris@42 495 ST(&(x[WS(rs, 17)]), VSUB(T1a, T1q), ms, &(x[WS(rs, 1)]));
Chris@42 496 ST(&(x[WS(rs, 3)]), VADD(T1a, T1q), ms, &(x[WS(rs, 1)]));
Chris@42 497 T1z = VADD(T1t, T1u);
Chris@42 498 T1A = VBYI(VSUB(T1x, T1w));
Chris@42 499 ST(&(x[WS(rs, 11)]), VSUB(T1z, T1A), ms, &(x[WS(rs, 1)]));
Chris@42 500 ST(&(x[WS(rs, 9)]), VADD(T1z, T1A), ms, &(x[WS(rs, 1)]));
Chris@42 501 }
Chris@42 502 {
Chris@42 503 V T1r, T1s, T1v, T1y;
Chris@42 504 T1r = VADD(TI, T19);
Chris@42 505 T1s = VBYI(VADD(T1p, T1m));
Chris@42 506 ST(&(x[WS(rs, 13)]), VSUB(T1r, T1s), ms, &(x[WS(rs, 1)]));
Chris@42 507 ST(&(x[WS(rs, 7)]), VADD(T1r, T1s), ms, &(x[WS(rs, 1)]));
Chris@42 508 T1v = VSUB(T1t, T1u);
Chris@42 509 T1y = VBYI(VADD(T1w, T1x));
Chris@42 510 ST(&(x[WS(rs, 19)]), VSUB(T1v, T1y), ms, &(x[WS(rs, 1)]));
Chris@42 511 ST(&(x[WS(rs, 1)]), VADD(T1v, T1y), ms, &(x[WS(rs, 1)]));
Chris@42 512 }
Chris@42 513 }
Chris@42 514 }
Chris@42 515 }
Chris@42 516 }
Chris@42 517 VLEAVE();
Chris@42 518 }
Chris@42 519
Chris@42 520 static const tw_instr twinstr[] = {
Chris@42 521 VTW(0, 1),
Chris@42 522 VTW(0, 3),
Chris@42 523 VTW(0, 9),
Chris@42 524 VTW(0, 19),
Chris@42 525 {TW_NEXT, VL, 0}
Chris@42 526 };
Chris@42 527
Chris@42 528 static const ct_desc desc = { 20, XSIMD_STRING("t3bv_20"), twinstr, &GENUS, {126, 80, 12, 0}, 0, 0, 0 };
Chris@42 529
Chris@42 530 void XSIMD(codelet_t3bv_20) (planner *p) {
Chris@42 531 X(kdft_dit_register) (p, t3bv_20, &desc);
Chris@42 532 }
Chris@42 533 #endif /* HAVE_FMA */