Mercurial > hg > sv-dependency-builds

annotate src/fftw-3.3.8/dft/simd/common/t2bv_20.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 08:06:05 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_twiddle_c.native -fma -simd -compact -variables 4 -pipeline-latency 8 -n 20 -name t2bv_20 -include dft/simd/t2b.h -sign 1 */
Chris@82 29
Chris@82 30 /*
Chris@82 31 * This function contains 123 FP additions, 88 FP multiplications,
Chris@82 32 * (or, 77 additions, 42 multiplications, 46 fused multiply/add),
Chris@82 33 * 54 stack variables, 4 constants, and 40 memory accesses
Chris@82 34 */
Chris@82 35 #include "dft/simd/t2b.h"
Chris@82 36
Chris@82 37 static void t2bv_20(R *ri, R *ii, const R *W, stride rs, INT mb, INT me, INT ms)
Chris@82 38 {
Chris@82 39 DVK(KP559016994, +0.559016994374947424102293417182819058860154590);
Chris@82 40 DVK(KP618033988, +0.618033988749894848204586834365638117720309180);
Chris@82 41 DVK(KP951056516, +0.951056516295153572116439333379382143405698634);
Chris@82 42 DVK(KP250000000, +0.250000000000000000000000000000000000000000000);
Chris@82 43 {
Chris@82 44 INT m;
Chris@82 45 R *x;
Chris@82 46 x = ii;
Chris@82 47 for (m = mb, W = W + (mb * ((TWVL / VL) * 38)); m < me; m = m + VL, x = x + (VL * ms), W = W + (TWVL * 38), MAKE_VOLATILE_STRIDE(20, rs)) {
Chris@82 48 V T4, TX, T1m, T1K, TF, T14, T15, TQ, Tf, Tq, Tr, T1O, T1P, T1Q, T1w;
Chris@82 49 V T1z, T1A, TY, TZ, T10, T1L, T1M, T1N, T1p, T1s, T1t, T1i, T1j;
Chris@82 50 {
Chris@82 51 V T1, TW, T3, TU, TV, T2, TT, T1k, T1l;
Chris@82 52 T1 = LD(&(x[0]), ms, &(x[0]));
Chris@82 53 TV = LD(&(x[WS(rs, 15)]), ms, &(x[WS(rs, 1)]));
Chris@82 54 TW = BYTW(&(W[TWVL * 28]), TV);
Chris@82 55 T2 = LD(&(x[WS(rs, 10)]), ms, &(x[0]));
Chris@82 56 T3 = BYTW(&(W[TWVL * 18]), T2);
Chris@82 57 TT = LD(&(x[WS(rs, 5)]), ms, &(x[WS(rs, 1)]));
Chris@82 58 TU = BYTW(&(W[TWVL * 8]), TT);
Chris@82 59 T4 = VSUB(T1, T3);
Chris@82 60 TX = VSUB(TU, TW);
Chris@82 61 T1k = VADD(T1, T3);
Chris@82 62 T1l = VADD(TU, TW);
Chris@82 63 T1m = VSUB(T1k, T1l);
Chris@82 64 T1K = VADD(T1k, T1l);
Chris@82 65 }
Chris@82 66 {
Chris@82 67 V T9, T1n, TK, T1v, TP, T1y, Te, T1q, Tk, T1u, Tz, T1o, TE, T1r, Tp;
Chris@82 68 V T1x;
Chris@82 69 {
Chris@82 70 V T6, T8, T5, T7;
Chris@82 71 T5 = LD(&(x[WS(rs, 4)]), ms, &(x[0]));
Chris@82 72 T6 = BYTW(&(W[TWVL * 6]), T5);
Chris@82 73 T7 = LD(&(x[WS(rs, 14)]), ms, &(x[0]));
Chris@82 74 T8 = BYTW(&(W[TWVL * 26]), T7);
Chris@82 75 T9 = VSUB(T6, T8);
Chris@82 76 T1n = VADD(T6, T8);
Chris@82 77 }
Chris@82 78 {
Chris@82 79 V TH, TJ, TG, TI;
Chris@82 80 TG = LD(&(x[WS(rs, 13)]), ms, &(x[WS(rs, 1)]));
Chris@82 81 TH = BYTW(&(W[TWVL * 24]), TG);
Chris@82 82 TI = LD(&(x[WS(rs, 3)]), ms, &(x[WS(rs, 1)]));
Chris@82 83 TJ = BYTW(&(W[TWVL * 4]), TI);
Chris@82 84 TK = VSUB(TH, TJ);
Chris@82 85 T1v = VADD(TH, TJ);
Chris@82 86 }
Chris@82 87 {
Chris@82 88 V TM, TO, TL, TN;
Chris@82 89 TL = LD(&(x[WS(rs, 17)]), ms, &(x[WS(rs, 1)]));
Chris@82 90 TM = BYTW(&(W[TWVL * 32]), TL);
Chris@82 91 TN = LD(&(x[WS(rs, 7)]), ms, &(x[WS(rs, 1)]));
Chris@82 92 TO = BYTW(&(W[TWVL * 12]), TN);
Chris@82 93 TP = VSUB(TM, TO);
Chris@82 94 T1y = VADD(TM, TO);
Chris@82 95 }
Chris@82 96 {
Chris@82 97 V Tb, Td, Ta, Tc;
Chris@82 98 Ta = LD(&(x[WS(rs, 16)]), ms, &(x[0]));
Chris@82 99 Tb = BYTW(&(W[TWVL * 30]), Ta);
Chris@82 100 Tc = LD(&(x[WS(rs, 6)]), ms, &(x[0]));
Chris@82 101 Td = BYTW(&(W[TWVL * 10]), Tc);
Chris@82 102 Te = VSUB(Tb, Td);
Chris@82 103 T1q = VADD(Tb, Td);
Chris@82 104 }
Chris@82 105 {
Chris@82 106 V Th, Tj, Tg, Ti;
Chris@82 107 Tg = LD(&(x[WS(rs, 8)]), ms, &(x[0]));
Chris@82 108 Th = BYTW(&(W[TWVL * 14]), Tg);
Chris@82 109 Ti = LD(&(x[WS(rs, 18)]), ms, &(x[0]));
Chris@82 110 Tj = BYTW(&(W[TWVL * 34]), Ti);
Chris@82 111 Tk = VSUB(Th, Tj);
Chris@82 112 T1u = VADD(Th, Tj);
Chris@82 113 }
Chris@82 114 {
Chris@82 115 V Tw, Ty, Tv, Tx;
Chris@82 116 Tv = LD(&(x[WS(rs, 9)]), ms, &(x[WS(rs, 1)]));
Chris@82 117 Tw = BYTW(&(W[TWVL * 16]), Tv);
Chris@82 118 Tx = LD(&(x[WS(rs, 19)]), ms, &(x[WS(rs, 1)]));
Chris@82 119 Ty = BYTW(&(W[TWVL * 36]), Tx);
Chris@82 120 Tz = VSUB(Tw, Ty);
Chris@82 121 T1o = VADD(Tw, Ty);
Chris@82 122 }
Chris@82 123 {
Chris@82 124 V TB, TD, TA, TC;
Chris@82 125 TA = LD(&(x[WS(rs, 1)]), ms, &(x[WS(rs, 1)]));
Chris@82 126 TB = BYTW(&(W[0]), TA);
Chris@82 127 TC = LD(&(x[WS(rs, 11)]), ms, &(x[WS(rs, 1)]));
Chris@82 128 TD = BYTW(&(W[TWVL * 20]), TC);
Chris@82 129 TE = VSUB(TB, TD);
Chris@82 130 T1r = VADD(TB, TD);
Chris@82 131 }
Chris@82 132 {
Chris@82 133 V Tm, To, Tl, Tn;
Chris@82 134 Tl = LD(&(x[WS(rs, 12)]), ms, &(x[0]));
Chris@82 135 Tm = BYTW(&(W[TWVL * 22]), Tl);
Chris@82 136 Tn = LD(&(x[WS(rs, 2)]), ms, &(x[0]));
Chris@82 137 To = BYTW(&(W[TWVL * 2]), Tn);
Chris@82 138 Tp = VSUB(Tm, To);
Chris@82 139 T1x = VADD(Tm, To);
Chris@82 140 }
Chris@82 141 TF = VSUB(Tz, TE);
Chris@82 142 T14 = VSUB(T9, Te);
Chris@82 143 T15 = VSUB(Tk, Tp);
Chris@82 144 TQ = VSUB(TK, TP);
Chris@82 145 Tf = VADD(T9, Te);
Chris@82 146 Tq = VADD(Tk, Tp);
Chris@82 147 Tr = VADD(Tf, Tq);
Chris@82 148 T1O = VADD(T1u, T1v);
Chris@82 149 T1P = VADD(T1x, T1y);
Chris@82 150 T1Q = VADD(T1O, T1P);
Chris@82 151 T1w = VSUB(T1u, T1v);
Chris@82 152 T1z = VSUB(T1x, T1y);
Chris@82 153 T1A = VADD(T1w, T1z);
Chris@82 154 TY = VADD(Tz, TE);
Chris@82 155 TZ = VADD(TK, TP);
Chris@82 156 T10 = VADD(TY, TZ);
Chris@82 157 T1L = VADD(T1n, T1o);
Chris@82 158 T1M = VADD(T1q, T1r);
Chris@82 159 T1N = VADD(T1L, T1M);
Chris@82 160 T1p = VSUB(T1n, T1o);
Chris@82 161 T1s = VSUB(T1q, T1r);
Chris@82 162 T1t = VADD(T1p, T1s);
Chris@82 163 }
Chris@82 164 T1i = VADD(T4, Tr);
Chris@82 165 T1j = VADD(TX, T10);
Chris@82 166 ST(&(x[WS(rs, 15)]), VFNMSI(T1j, T1i), ms, &(x[WS(rs, 1)]));
Chris@82 167 ST(&(x[WS(rs, 5)]), VFMAI(T1j, T1i), ms, &(x[WS(rs, 1)]));
Chris@82 168 {
Chris@82 169 V T1T, T1R, T1S, T1X, T1Z, T1V, T1W, T1Y, T1U;
Chris@82 170 T1T = VSUB(T1N, T1Q);
Chris@82 171 T1R = VADD(T1N, T1Q);
Chris@82 172 T1S = VFNMS(LDK(KP250000000), T1R, T1K);
Chris@82 173 T1V = VSUB(T1L, T1M);
Chris@82 174 T1W = VSUB(T1O, T1P);
Chris@82 175 T1X = VMUL(LDK(KP951056516), VFMA(LDK(KP618033988), T1W, T1V));
Chris@82 176 T1Z = VMUL(LDK(KP951056516), VFNMS(LDK(KP618033988), T1V, T1W));
Chris@82 177 ST(&(x[0]), VADD(T1K, T1R), ms, &(x[0]));
Chris@82 178 T1Y = VFNMS(LDK(KP559016994), T1T, T1S);
Chris@82 179 ST(&(x[WS(rs, 8)]), VFMAI(T1Z, T1Y), ms, &(x[0]));
Chris@82 180 ST(&(x[WS(rs, 12)]), VFNMSI(T1Z, T1Y), ms, &(x[0]));
Chris@82 181 T1U = VFMA(LDK(KP559016994), T1T, T1S);
Chris@82 182 ST(&(x[WS(rs, 4)]), VFNMSI(T1X, T1U), ms, &(x[0]));
Chris@82 183 ST(&(x[WS(rs, 16)]), VFMAI(T1X, T1U), ms, &(x[0]));
Chris@82 184 }
Chris@82 185 {
Chris@82 186 V T1D, T1B, T1C, T1H, T1J, T1F, T1G, T1I, T1E;
Chris@82 187 T1D = VSUB(T1t, T1A);
Chris@82 188 T1B = VADD(T1t, T1A);
Chris@82 189 T1C = VFNMS(LDK(KP250000000), T1B, T1m);
Chris@82 190 T1F = VSUB(T1w, T1z);
Chris@82 191 T1G = VSUB(T1p, T1s);
Chris@82 192 T1H = VMUL(LDK(KP951056516), VFNMS(LDK(KP618033988), T1G, T1F));
Chris@82 193 T1J = VMUL(LDK(KP951056516), VFMA(LDK(KP618033988), T1F, T1G));
Chris@82 194 ST(&(x[WS(rs, 10)]), VADD(T1m, T1B), ms, &(x[0]));
Chris@82 195 T1I = VFMA(LDK(KP559016994), T1D, T1C);
Chris@82 196 ST(&(x[WS(rs, 6)]), VFMAI(T1J, T1I), ms, &(x[0]));
Chris@82 197 ST(&(x[WS(rs, 14)]), VFNMSI(T1J, T1I), ms, &(x[0]));
Chris@82 198 T1E = VFNMS(LDK(KP559016994), T1D, T1C);
Chris@82 199 ST(&(x[WS(rs, 2)]), VFNMSI(T1H, T1E), ms, &(x[0]));
Chris@82 200 ST(&(x[WS(rs, 18)]), VFMAI(T1H, T1E), ms, &(x[0]));
Chris@82 201 }
Chris@82 202 {
Chris@82 203 V TR, T16, T1e, T1b, T13, T1d, Tu, T1a;
Chris@82 204 TR = VFMA(LDK(KP618033988), TQ, TF);
Chris@82 205 T16 = VFMA(LDK(KP618033988), T15, T14);
Chris@82 206 T1e = VFNMS(LDK(KP618033988), T14, T15);
Chris@82 207 T1b = VFNMS(LDK(KP618033988), TF, TQ);
Chris@82 208 {
Chris@82 209 V T11, T12, Ts, Tt;
Chris@82 210 T11 = VFNMS(LDK(KP250000000), T10, TX);
Chris@82 211 T12 = VSUB(TY, TZ);
Chris@82 212 T13 = VFMA(LDK(KP559016994), T12, T11);
Chris@82 213 T1d = VFNMS(LDK(KP559016994), T12, T11);
Chris@82 214 Ts = VFNMS(LDK(KP250000000), Tr, T4);
Chris@82 215 Tt = VSUB(Tf, Tq);
Chris@82 216 Tu = VFMA(LDK(KP559016994), Tt, Ts);
Chris@82 217 T1a = VFNMS(LDK(KP559016994), Tt, Ts);
Chris@82 218 }
Chris@82 219 {
Chris@82 220 V TS, T17, T1g, T1h;
Chris@82 221 TS = VFNMS(LDK(KP951056516), TR, Tu);
Chris@82 222 T17 = VFMA(LDK(KP951056516), T16, T13);
Chris@82 223 ST(&(x[WS(rs, 19)]), VFNMSI(T17, TS), ms, &(x[WS(rs, 1)]));
Chris@82 224 ST(&(x[WS(rs, 1)]), VFMAI(T17, TS), ms, &(x[WS(rs, 1)]));
Chris@82 225 T1g = VFNMS(LDK(KP951056516), T1b, T1a);
Chris@82 226 T1h = VFMA(LDK(KP951056516), T1e, T1d);
Chris@82 227 ST(&(x[WS(rs, 7)]), VFNMSI(T1h, T1g), ms, &(x[WS(rs, 1)]));
Chris@82 228 ST(&(x[WS(rs, 13)]), VFMAI(T1h, T1g), ms, &(x[WS(rs, 1)]));
Chris@82 229 }
Chris@82 230 {
Chris@82 231 V T18, T19, T1c, T1f;
Chris@82 232 T18 = VFMA(LDK(KP951056516), TR, Tu);
Chris@82 233 T19 = VFNMS(LDK(KP951056516), T16, T13);
Chris@82 234 ST(&(x[WS(rs, 11)]), VFNMSI(T19, T18), ms, &(x[WS(rs, 1)]));
Chris@82 235 ST(&(x[WS(rs, 9)]), VFMAI(T19, T18), ms, &(x[WS(rs, 1)]));
Chris@82 236 T1c = VFMA(LDK(KP951056516), T1b, T1a);
Chris@82 237 T1f = VFNMS(LDK(KP951056516), T1e, T1d);
Chris@82 238 ST(&(x[WS(rs, 3)]), VFNMSI(T1f, T1c), ms, &(x[WS(rs, 1)]));
Chris@82 239 ST(&(x[WS(rs, 17)]), VFMAI(T1f, T1c), ms, &(x[WS(rs, 1)]));
Chris@82 240 }
Chris@82 241 }
Chris@82 242 }
Chris@82 243 }
Chris@82 244 VLEAVE();
Chris@82 245 }
Chris@82 246
Chris@82 247 static const tw_instr twinstr[] = {
Chris@82 248 VTW(0, 1),
Chris@82 249 VTW(0, 2),
Chris@82 250 VTW(0, 3),
Chris@82 251 VTW(0, 4),
Chris@82 252 VTW(0, 5),
Chris@82 253 VTW(0, 6),
Chris@82 254 VTW(0, 7),
Chris@82 255 VTW(0, 8),
Chris@82 256 VTW(0, 9),
Chris@82 257 VTW(0, 10),
Chris@82 258 VTW(0, 11),
Chris@82 259 VTW(0, 12),
Chris@82 260 VTW(0, 13),
Chris@82 261 VTW(0, 14),
Chris@82 262 VTW(0, 15),
Chris@82 263 VTW(0, 16),
Chris@82 264 VTW(0, 17),
Chris@82 265 VTW(0, 18),
Chris@82 266 VTW(0, 19),
Chris@82 267 {TW_NEXT, VL, 0}
Chris@82 268 };
Chris@82 269
Chris@82 270 static const ct_desc desc = { 20, XSIMD_STRING("t2bv_20"), twinstr, &GENUS, {77, 42, 46, 0}, 0, 0, 0 };
Chris@82 271
Chris@82 272 void XSIMD(codelet_t2bv_20) (planner *p) {
Chris@82 273 X(kdft_dit_register) (p, t2bv_20, &desc);
Chris@82 274 }
Chris@82 275 #else
Chris@82 276
Chris@82 277 /* Generated by: ../../../genfft/gen_twiddle_c.native -simd -compact -variables 4 -pipeline-latency 8 -n 20 -name t2bv_20 -include dft/simd/t2b.h -sign 1 */
Chris@82 278
Chris@82 279 /*
Chris@82 280 * This function contains 123 FP additions, 62 FP multiplications,
Chris@82 281 * (or, 111 additions, 50 multiplications, 12 fused multiply/add),
Chris@82 282 * 54 stack variables, 4 constants, and 40 memory accesses
Chris@82 283 */
Chris@82 284 #include "dft/simd/t2b.h"
Chris@82 285
Chris@82 286 static void t2bv_20(R *ri, R *ii, const R *W, stride rs, INT mb, INT me, INT ms)
Chris@82 287 {
Chris@82 288 DVK(KP587785252, +0.587785252292473129168705954639072768597652438);
Chris@82 289 DVK(KP951056516, +0.951056516295153572116439333379382143405698634);
Chris@82 290 DVK(KP250000000, +0.250000000000000000000000000000000000000000000);
Chris@82 291 DVK(KP559016994, +0.559016994374947424102293417182819058860154590);
Chris@82 292 {
Chris@82 293 INT m;
Chris@82 294 R *x;
Chris@82 295 x = ii;
Chris@82 296 for (m = mb, W = W + (mb * ((TWVL / VL) * 38)); m < me; m = m + VL, x = x + (VL * ms), W = W + (TWVL * 38), MAKE_VOLATILE_STRIDE(20, rs)) {
Chris@82 297 V T4, T10, T1B, T1R, TF, T14, T15, TQ, Tf, Tq, Tr, T1N, T1O, T1P, T1t;
Chris@82 298 V T1w, T1D, TT, TU, T11, T1K, T1L, T1M, T1m, T1p, T1C, T1i, T1j;
Chris@82 299 {
Chris@82 300 V T1, TZ, T3, TX, TY, T2, TW, T1z, T1A;
Chris@82 301 T1 = LD(&(x[0]), ms, &(x[0]));
Chris@82 302 TY = LD(&(x[WS(rs, 15)]), ms, &(x[WS(rs, 1)]));
Chris@82 303 TZ = BYTW(&(W[TWVL * 28]), TY);
Chris@82 304 T2 = LD(&(x[WS(rs, 10)]), ms, &(x[0]));
Chris@82 305 T3 = BYTW(&(W[TWVL * 18]), T2);
Chris@82 306 TW = LD(&(x[WS(rs, 5)]), ms, &(x[WS(rs, 1)]));
Chris@82 307 TX = BYTW(&(W[TWVL * 8]), TW);
Chris@82 308 T4 = VSUB(T1, T3);
Chris@82 309 T10 = VSUB(TX, TZ);
Chris@82 310 T1z = VADD(T1, T3);
Chris@82 311 T1A = VADD(TX, TZ);
Chris@82 312 T1B = VSUB(T1z, T1A);
Chris@82 313 T1R = VADD(T1z, T1A);
Chris@82 314 }
Chris@82 315 {
Chris@82 316 V T9, T1k, TK, T1s, TP, T1v, Te, T1n, Tk, T1r, Tz, T1l, TE, T1o, Tp;
Chris@82 317 V T1u;
Chris@82 318 {
Chris@82 319 V T6, T8, T5, T7;
Chris@82 320 T5 = LD(&(x[WS(rs, 4)]), ms, &(x[0]));
Chris@82 321 T6 = BYTW(&(W[TWVL * 6]), T5);
Chris@82 322 T7 = LD(&(x[WS(rs, 14)]), ms, &(x[0]));
Chris@82 323 T8 = BYTW(&(W[TWVL * 26]), T7);
Chris@82 324 T9 = VSUB(T6, T8);
Chris@82 325 T1k = VADD(T6, T8);
Chris@82 326 }
Chris@82 327 {
Chris@82 328 V TH, TJ, TG, TI;
Chris@82 329 TG = LD(&(x[WS(rs, 13)]), ms, &(x[WS(rs, 1)]));
Chris@82 330 TH = BYTW(&(W[TWVL * 24]), TG);
Chris@82 331 TI = LD(&(x[WS(rs, 3)]), ms, &(x[WS(rs, 1)]));
Chris@82 332 TJ = BYTW(&(W[TWVL * 4]), TI);
Chris@82 333 TK = VSUB(TH, TJ);
Chris@82 334 T1s = VADD(TH, TJ);
Chris@82 335 }
Chris@82 336 {
Chris@82 337 V TM, TO, TL, TN;
Chris@82 338 TL = LD(&(x[WS(rs, 17)]), ms, &(x[WS(rs, 1)]));
Chris@82 339 TM = BYTW(&(W[TWVL * 32]), TL);
Chris@82 340 TN = LD(&(x[WS(rs, 7)]), ms, &(x[WS(rs, 1)]));
Chris@82 341 TO = BYTW(&(W[TWVL * 12]), TN);
Chris@82 342 TP = VSUB(TM, TO);
Chris@82 343 T1v = VADD(TM, TO);
Chris@82 344 }
Chris@82 345 {
Chris@82 346 V Tb, Td, Ta, Tc;
Chris@82 347 Ta = LD(&(x[WS(rs, 16)]), ms, &(x[0]));
Chris@82 348 Tb = BYTW(&(W[TWVL * 30]), Ta);
Chris@82 349 Tc = LD(&(x[WS(rs, 6)]), ms, &(x[0]));
Chris@82 350 Td = BYTW(&(W[TWVL * 10]), Tc);
Chris@82 351 Te = VSUB(Tb, Td);
Chris@82 352 T1n = VADD(Tb, Td);
Chris@82 353 }
Chris@82 354 {
Chris@82 355 V Th, Tj, Tg, Ti;
Chris@82 356 Tg = LD(&(x[WS(rs, 8)]), ms, &(x[0]));
Chris@82 357 Th = BYTW(&(W[TWVL * 14]), Tg);
Chris@82 358 Ti = LD(&(x[WS(rs, 18)]), ms, &(x[0]));
Chris@82 359 Tj = BYTW(&(W[TWVL * 34]), Ti);
Chris@82 360 Tk = VSUB(Th, Tj);
Chris@82 361 T1r = VADD(Th, Tj);
Chris@82 362 }
Chris@82 363 {
Chris@82 364 V Tw, Ty, Tv, Tx;
Chris@82 365 Tv = LD(&(x[WS(rs, 9)]), ms, &(x[WS(rs, 1)]));
Chris@82 366 Tw = BYTW(&(W[TWVL * 16]), Tv);
Chris@82 367 Tx = LD(&(x[WS(rs, 19)]), ms, &(x[WS(rs, 1)]));
Chris@82 368 Ty = BYTW(&(W[TWVL * 36]), Tx);
Chris@82 369 Tz = VSUB(Tw, Ty);
Chris@82 370 T1l = VADD(Tw, Ty);
Chris@82 371 }
Chris@82 372 {
Chris@82 373 V TB, TD, TA, TC;
Chris@82 374 TA = LD(&(x[WS(rs, 1)]), ms, &(x[WS(rs, 1)]));
Chris@82 375 TB = BYTW(&(W[0]), TA);
Chris@82 376 TC = LD(&(x[WS(rs, 11)]), ms, &(x[WS(rs, 1)]));
Chris@82 377 TD = BYTW(&(W[TWVL * 20]), TC);
Chris@82 378 TE = VSUB(TB, TD);
Chris@82 379 T1o = VADD(TB, TD);
Chris@82 380 }
Chris@82 381 {
Chris@82 382 V Tm, To, Tl, Tn;
Chris@82 383 Tl = LD(&(x[WS(rs, 12)]), ms, &(x[0]));
Chris@82 384 Tm = BYTW(&(W[TWVL * 22]), Tl);
Chris@82 385 Tn = LD(&(x[WS(rs, 2)]), ms, &(x[0]));
Chris@82 386 To = BYTW(&(W[TWVL * 2]), Tn);
Chris@82 387 Tp = VSUB(Tm, To);
Chris@82 388 T1u = VADD(Tm, To);
Chris@82 389 }
Chris@82 390 TF = VSUB(Tz, TE);
Chris@82 391 T14 = VSUB(T9, Te);
Chris@82 392 T15 = VSUB(Tk, Tp);
Chris@82 393 TQ = VSUB(TK, TP);
Chris@82 394 Tf = VADD(T9, Te);
Chris@82 395 Tq = VADD(Tk, Tp);
Chris@82 396 Tr = VADD(Tf, Tq);
Chris@82 397 T1N = VADD(T1r, T1s);
Chris@82 398 T1O = VADD(T1u, T1v);
Chris@82 399 T1P = VADD(T1N, T1O);
Chris@82 400 T1t = VSUB(T1r, T1s);
Chris@82 401 T1w = VSUB(T1u, T1v);
Chris@82 402 T1D = VADD(T1t, T1w);
Chris@82 403 TT = VADD(Tz, TE);
Chris@82 404 TU = VADD(TK, TP);
Chris@82 405 T11 = VADD(TT, TU);
Chris@82 406 T1K = VADD(T1k, T1l);
Chris@82 407 T1L = VADD(T1n, T1o);
Chris@82 408 T1M = VADD(T1K, T1L);
Chris@82 409 T1m = VSUB(T1k, T1l);
Chris@82 410 T1p = VSUB(T1n, T1o);
Chris@82 411 T1C = VADD(T1m, T1p);
Chris@82 412 }
Chris@82 413 T1i = VADD(T4, Tr);
Chris@82 414 T1j = VBYI(VADD(T10, T11));
Chris@82 415 ST(&(x[WS(rs, 15)]), VSUB(T1i, T1j), ms, &(x[WS(rs, 1)]));
Chris@82 416 ST(&(x[WS(rs, 5)]), VADD(T1i, T1j), ms, &(x[WS(rs, 1)]));
Chris@82 417 {
Chris@82 418 V T1Q, T1S, T1T, T1X, T1Z, T1V, T1W, T1Y, T1U;
Chris@82 419 T1Q = VMUL(LDK(KP559016994), VSUB(T1M, T1P));
Chris@82 420 T1S = VADD(T1M, T1P);
Chris@82 421 T1T = VFNMS(LDK(KP250000000), T1S, T1R);
Chris@82 422 T1V = VSUB(T1K, T1L);
Chris@82 423 T1W = VSUB(T1N, T1O);
Chris@82 424 T1X = VBYI(VFMA(LDK(KP951056516), T1V, VMUL(LDK(KP587785252), T1W)));
Chris@82 425 T1Z = VBYI(VFNMS(LDK(KP951056516), T1W, VMUL(LDK(KP587785252), T1V)));
Chris@82 426 ST(&(x[0]), VADD(T1R, T1S), ms, &(x[0]));
Chris@82 427 T1Y = VSUB(T1T, T1Q);
Chris@82 428 ST(&(x[WS(rs, 8)]), VSUB(T1Y, T1Z), ms, &(x[0]));
Chris@82 429 ST(&(x[WS(rs, 12)]), VADD(T1Z, T1Y), ms, &(x[0]));
Chris@82 430 T1U = VADD(T1Q, T1T);
Chris@82 431 ST(&(x[WS(rs, 4)]), VSUB(T1U, T1X), ms, &(x[0]));
Chris@82 432 ST(&(x[WS(rs, 16)]), VADD(T1X, T1U), ms, &(x[0]));
Chris@82 433 }
Chris@82 434 {
Chris@82 435 V T1G, T1E, T1F, T1y, T1I, T1q, T1x, T1J, T1H;
Chris@82 436 T1G = VMUL(LDK(KP559016994), VSUB(T1C, T1D));
Chris@82 437 T1E = VADD(T1C, T1D);
Chris@82 438 T1F = VFNMS(LDK(KP250000000), T1E, T1B);
Chris@82 439 T1q = VSUB(T1m, T1p);
Chris@82 440 T1x = VSUB(T1t, T1w);
Chris@82 441 T1y = VBYI(VFNMS(LDK(KP951056516), T1x, VMUL(LDK(KP587785252), T1q)));
Chris@82 442 T1I = VBYI(VFMA(LDK(KP951056516), T1q, VMUL(LDK(KP587785252), T1x)));
Chris@82 443 ST(&(x[WS(rs, 10)]), VADD(T1B, T1E), ms, &(x[0]));
Chris@82 444 T1J = VADD(T1G, T1F);
Chris@82 445 ST(&(x[WS(rs, 6)]), VADD(T1I, T1J), ms, &(x[0]));
Chris@82 446 ST(&(x[WS(rs, 14)]), VSUB(T1J, T1I), ms, &(x[0]));
Chris@82 447 T1H = VSUB(T1F, T1G);
Chris@82 448 ST(&(x[WS(rs, 2)]), VADD(T1y, T1H), ms, &(x[0]));
Chris@82 449 ST(&(x[WS(rs, 18)]), VSUB(T1H, T1y), ms, &(x[0]));
Chris@82 450 }
Chris@82 451 {
Chris@82 452 V TR, T16, T1d, T1b, T13, T1e, Tu, T1a;
Chris@82 453 TR = VFNMS(LDK(KP951056516), TQ, VMUL(LDK(KP587785252), TF));
Chris@82 454 T16 = VFNMS(LDK(KP951056516), T15, VMUL(LDK(KP587785252), T14));
Chris@82 455 T1d = VFMA(LDK(KP951056516), T14, VMUL(LDK(KP587785252), T15));
Chris@82 456 T1b = VFMA(LDK(KP951056516), TF, VMUL(LDK(KP587785252), TQ));
Chris@82 457 {
Chris@82 458 V TV, T12, Ts, Tt;
Chris@82 459 TV = VMUL(LDK(KP559016994), VSUB(TT, TU));
Chris@82 460 T12 = VFNMS(LDK(KP250000000), T11, T10);
Chris@82 461 T13 = VSUB(TV, T12);
Chris@82 462 T1e = VADD(TV, T12);
Chris@82 463 Ts = VFNMS(LDK(KP250000000), Tr, T4);
Chris@82 464 Tt = VMUL(LDK(KP559016994), VSUB(Tf, Tq));
Chris@82 465 Tu = VSUB(Ts, Tt);
Chris@82 466 T1a = VADD(Tt, Ts);
Chris@82 467 }
Chris@82 468 {
Chris@82 469 V TS, T17, T1g, T1h;
Chris@82 470 TS = VSUB(Tu, TR);
Chris@82 471 T17 = VBYI(VSUB(T13, T16));
Chris@82 472 ST(&(x[WS(rs, 17)]), VSUB(TS, T17), ms, &(x[WS(rs, 1)]));
Chris@82 473 ST(&(x[WS(rs, 3)]), VADD(TS, T17), ms, &(x[WS(rs, 1)]));
Chris@82 474 T1g = VADD(T1a, T1b);
Chris@82 475 T1h = VBYI(VSUB(T1e, T1d));
Chris@82 476 ST(&(x[WS(rs, 11)]), VSUB(T1g, T1h), ms, &(x[WS(rs, 1)]));
Chris@82 477 ST(&(x[WS(rs, 9)]), VADD(T1g, T1h), ms, &(x[WS(rs, 1)]));
Chris@82 478 }
Chris@82 479 {
Chris@82 480 V T18, T19, T1c, T1f;
Chris@82 481 T18 = VADD(Tu, TR);
Chris@82 482 T19 = VBYI(VADD(T16, T13));
Chris@82 483 ST(&(x[WS(rs, 13)]), VSUB(T18, T19), ms, &(x[WS(rs, 1)]));
Chris@82 484 ST(&(x[WS(rs, 7)]), VADD(T18, T19), ms, &(x[WS(rs, 1)]));
Chris@82 485 T1c = VSUB(T1a, T1b);
Chris@82 486 T1f = VBYI(VADD(T1d, T1e));
Chris@82 487 ST(&(x[WS(rs, 19)]), VSUB(T1c, T1f), ms, &(x[WS(rs, 1)]));
Chris@82 488 ST(&(x[WS(rs, 1)]), VADD(T1c, T1f), ms, &(x[WS(rs, 1)]));
Chris@82 489 }
Chris@82 490 }
Chris@82 491 }
Chris@82 492 }
Chris@82 493 VLEAVE();
Chris@82 494 }
Chris@82 495
Chris@82 496 static const tw_instr twinstr[] = {
Chris@82 497 VTW(0, 1),
Chris@82 498 VTW(0, 2),
Chris@82 499 VTW(0, 3),
Chris@82 500 VTW(0, 4),
Chris@82 501 VTW(0, 5),
Chris@82 502 VTW(0, 6),
Chris@82 503 VTW(0, 7),
Chris@82 504 VTW(0, 8),
Chris@82 505 VTW(0, 9),
Chris@82 506 VTW(0, 10),
Chris@82 507 VTW(0, 11),
Chris@82 508 VTW(0, 12),
Chris@82 509 VTW(0, 13),
Chris@82 510 VTW(0, 14),
Chris@82 511 VTW(0, 15),
Chris@82 512 VTW(0, 16),
Chris@82 513 VTW(0, 17),
Chris@82 514 VTW(0, 18),
Chris@82 515 VTW(0, 19),
Chris@82 516 {TW_NEXT, VL, 0}
Chris@82 517 };
Chris@82 518
Chris@82 519 static const ct_desc desc = { 20, XSIMD_STRING("t2bv_20"), twinstr, &GENUS, {111, 50, 12, 0}, 0, 0, 0 };
Chris@82 520
Chris@82 521 void XSIMD(codelet_t2bv_20) (planner *p) {
Chris@82 522 X(kdft_dit_register) (p, t2bv_20, &desc);
Chris@82 523 }
Chris@82 524 #endif