Mercurial > hg > sv-dependency-builds

annotate src/fftw-3.3.5/rdft/simd/common/hc2cfdftv_20.c @ 42:2cd0e3b3e1fd

Find changesets by keywords (author, files, the commit message), revision number or hash, or revset expression.
Current fftw source
author Chris Cannam
date Tue, 18 Oct 2016 13:40:26 +0100
parents
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 16:52:41 EDT 2016 */
Chris@42 23
Chris@42 24 #include "codelet-rdft.h"
Chris@42 25
Chris@42 26 #ifdef HAVE_FMA
Chris@42 27
Chris@42 28 /* Generated by: ../../../genfft/gen_hc2cdft_c.native -fma -reorder-insns -schedule-for-pipeline -simd -compact -variables 4 -pipeline-latency 8 -trivial-stores -variables 32 -no-generate-bytw -n 20 -dit -name hc2cfdftv_20 -include hc2cfv.h */
Chris@42 29
Chris@42 30 /*
Chris@42 31 * This function contains 143 FP additions, 128 FP multiplications,
Chris@42 32 * (or, 77 additions, 62 multiplications, 66 fused multiply/add),
Chris@42 33 * 130 stack variables, 5 constants, and 40 memory accesses
Chris@42 34 */
Chris@42 35 #include "hc2cfv.h"
Chris@42 36
Chris@42 37 static void hc2cfdftv_20(R *Rp, R *Ip, R *Rm, R *Im, 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(KP500000000, +0.500000000000000000000000000000000000000000000);
Chris@42 41 DVK(KP951056516, +0.951056516295153572116439333379382143405698634);
Chris@42 42 DVK(KP250000000, +0.250000000000000000000000000000000000000000000);
Chris@42 43 DVK(KP618033988, +0.618033988749894848204586834365638117720309180);
Chris@42 44 {
Chris@42 45 INT m;
Chris@42 46 for (m = mb, W = W + ((mb - 1) * ((TWVL / VL) * 38)); m < me; m = m + VL, Rp = Rp + (VL * ms), Ip = Ip + (VL * ms), Rm = Rm - (VL * ms), Im = Im - (VL * ms), W = W + (TWVL * 38), MAKE_VOLATILE_STRIDE(80, rs)) {
Chris@42 47 V T2g, T2f, T2w, T2k, T2A, T2u, T2e, T2o, T1O, T2b, T2i, T1R, T1X, T1k, TN;
Chris@42 48 V T1w, T1G, T1t, Ti, T2c, T12, T1x, T2j, T1U, T1y, T1d, T24, T2v, T2h, T2x;
Chris@42 49 V T2B, T2p, T2l, T2z, T2y, T2D, T2C, T2r, T2q, T2n, T2m;
Chris@42 50 {
Chris@42 51 V T3, T7, TC, T1Y, Tc, Tg, Tn, T1P, T1Z, Tw, T1S, TS, TY, TZ, T1Q;
Chris@42 52 V TL, T17, T21, TW, T19, TX, T1a, T8, T20, Th, Tx, T1u, T1v, TM, T10;
Chris@42 53 V T1b, T22, T11, T1T, T1c, T23;
Chris@42 54 {
Chris@42 55 V Ta, Tb, Tz, Te, TB, Tf, Tl, T9, Td, Tk, T1, T2, Ty, T5, T6;
Chris@42 56 V TA, T4, Tj, Tt, Tu, Ts, TQ, Tr, TP, Tp, Tq, Tm, To, TO, TG;
Chris@42 57 V T14, TK, T16, TE, TF, Tv, TD, T13, TR, TI, TJ, TH, T15, TU, TV;
Chris@42 58 V TT, T18;
Chris@42 59 T1 = LD(&(Rp[0]), ms, &(Rp[0]));
Chris@42 60 T2 = LD(&(Rm[0]), -ms, &(Rm[0]));
Chris@42 61 Ty = LDW(&(W[0]));
Chris@42 62 T5 = LD(&(Rp[WS(rs, 5)]), ms, &(Rp[WS(rs, 1)]));
Chris@42 63 T6 = LD(&(Rm[WS(rs, 5)]), -ms, &(Rm[WS(rs, 1)]));
Chris@42 64 TA = LDW(&(W[TWVL * 20]));
Chris@42 65 T4 = LDW(&(W[TWVL * 18]));
Chris@42 66 Ta = LD(&(Rp[WS(rs, 2)]), ms, &(Rp[0]));
Chris@42 67 Tb = LD(&(Rm[WS(rs, 2)]), -ms, &(Rm[0]));
Chris@42 68 T3 = VFMACONJ(T2, T1);
Chris@42 69 Tz = VZMULIJ(Ty, VFNMSCONJ(T2, T1));
Chris@42 70 Tj = LDW(&(W[TWVL * 6]));
Chris@42 71 Te = LD(&(Rp[WS(rs, 7)]), ms, &(Rp[WS(rs, 1)]));
Chris@42 72 TB = VZMULIJ(TA, VFNMSCONJ(T6, T5));
Chris@42 73 T7 = VZMULJ(T4, VFMACONJ(T6, T5));
Chris@42 74 Tf = LD(&(Rm[WS(rs, 7)]), -ms, &(Rm[WS(rs, 1)]));
Chris@42 75 Tl = LDW(&(W[TWVL * 26]));
Chris@42 76 T9 = LDW(&(W[TWVL * 8]));
Chris@42 77 Td = LDW(&(W[TWVL * 28]));
Chris@42 78 Tk = VZMULJ(Tj, VFMACONJ(Tb, Ta));
Chris@42 79 Tp = LD(&(Rp[WS(rs, 4)]), ms, &(Rp[0]));
Chris@42 80 TC = VADD(Tz, TB);
Chris@42 81 T1Y = VSUB(TB, Tz);
Chris@42 82 Tq = LD(&(Rm[WS(rs, 4)]), -ms, &(Rm[0]));
Chris@42 83 Tm = VZMULJ(Tl, VFMACONJ(Tf, Te));
Chris@42 84 Tc = VZMULIJ(T9, VFNMSCONJ(Tb, Ta));
Chris@42 85 Tg = VZMULIJ(Td, VFNMSCONJ(Tf, Te));
Chris@42 86 To = LDW(&(W[TWVL * 16]));
Chris@42 87 TO = LDW(&(W[TWVL * 14]));
Chris@42 88 Tt = LD(&(Rp[WS(rs, 9)]), ms, &(Rp[WS(rs, 1)]));
Chris@42 89 Tu = LD(&(Rm[WS(rs, 9)]), -ms, &(Rm[WS(rs, 1)]));
Chris@42 90 Ts = LDW(&(W[TWVL * 36]));
Chris@42 91 Tn = VADD(Tk, Tm);
Chris@42 92 T1P = VSUB(Tk, Tm);
Chris@42 93 TQ = LDW(&(W[TWVL * 34]));
Chris@42 94 Tr = VZMULIJ(To, VFNMSCONJ(Tq, Tp));
Chris@42 95 TP = VZMULJ(TO, VFMACONJ(Tq, Tp));
Chris@42 96 TE = LD(&(Rp[WS(rs, 8)]), ms, &(Rp[0]));
Chris@42 97 TF = LD(&(Rm[WS(rs, 8)]), -ms, &(Rm[0]));
Chris@42 98 Tv = VZMULIJ(Ts, VFNMSCONJ(Tu, Tt));
Chris@42 99 TD = LDW(&(W[TWVL * 30]));
Chris@42 100 T13 = LDW(&(W[TWVL * 32]));
Chris@42 101 TR = VZMULJ(TQ, VFMACONJ(Tu, Tt));
Chris@42 102 TI = LD(&(Rp[WS(rs, 3)]), ms, &(Rp[WS(rs, 1)]));
Chris@42 103 TJ = LD(&(Rm[WS(rs, 3)]), -ms, &(Rm[WS(rs, 1)]));
Chris@42 104 TH = LDW(&(W[TWVL * 10]));
Chris@42 105 T15 = LDW(&(W[TWVL * 12]));
Chris@42 106 T1Z = VSUB(Tv, Tr);
Chris@42 107 Tw = VADD(Tr, Tv);
Chris@42 108 TG = VZMULJ(TD, VFMACONJ(TF, TE));
Chris@42 109 T14 = VZMULIJ(T13, VFNMSCONJ(TF, TE));
Chris@42 110 T1S = VSUB(TP, TR);
Chris@42 111 TS = VADD(TP, TR);
Chris@42 112 TK = VZMULJ(TH, VFMACONJ(TJ, TI));
Chris@42 113 T16 = VZMULIJ(T15, VFNMSCONJ(TJ, TI));
Chris@42 114 TU = LD(&(Rp[WS(rs, 6)]), ms, &(Rp[0]));
Chris@42 115 TV = LD(&(Rm[WS(rs, 6)]), -ms, &(Rm[0]));
Chris@42 116 TT = LDW(&(W[TWVL * 24]));
Chris@42 117 T18 = LDW(&(W[TWVL * 22]));
Chris@42 118 TY = LD(&(Rp[WS(rs, 1)]), ms, &(Rp[WS(rs, 1)]));
Chris@42 119 TZ = LD(&(Rm[WS(rs, 1)]), -ms, &(Rm[WS(rs, 1)]));
Chris@42 120 T1Q = VSUB(TK, TG);
Chris@42 121 TL = VADD(TG, TK);
Chris@42 122 T17 = VADD(T14, T16);
Chris@42 123 T21 = VSUB(T16, T14);
Chris@42 124 TW = VZMULIJ(TT, VFNMSCONJ(TV, TU));
Chris@42 125 T19 = VZMULJ(T18, VFMACONJ(TV, TU));
Chris@42 126 TX = LDW(&(W[TWVL * 4]));
Chris@42 127 T1a = LDW(&(W[TWVL * 2]));
Chris@42 128 }
Chris@42 129 T1O = VSUB(T3, T7);
Chris@42 130 T8 = VADD(T3, T7);
Chris@42 131 T20 = VADD(T1Y, T1Z);
Chris@42 132 T2b = VSUB(T1Y, T1Z);
Chris@42 133 T2i = VADD(T1P, T1Q);
Chris@42 134 T1R = VSUB(T1P, T1Q);
Chris@42 135 Th = VADD(Tc, Tg);
Chris@42 136 T1X = VSUB(Tg, Tc);
Chris@42 137 Tx = VSUB(Tn, Tw);
Chris@42 138 T1u = VADD(Tn, Tw);
Chris@42 139 T1v = VADD(TC, TL);
Chris@42 140 TM = VSUB(TC, TL);
Chris@42 141 T10 = VZMULIJ(TX, VFNMSCONJ(TZ, TY));
Chris@42 142 T1b = VZMULJ(T1a, VFMACONJ(TZ, TY));
Chris@42 143 T1k = VADD(Tx, TM);
Chris@42 144 TN = VSUB(Tx, TM);
Chris@42 145 T22 = VSUB(T10, TW);
Chris@42 146 T11 = VADD(TW, T10);
Chris@42 147 T1T = VSUB(T1b, T19);
Chris@42 148 T1c = VADD(T19, T1b);
Chris@42 149 T1w = VADD(T1u, T1v);
Chris@42 150 T1G = VSUB(T1u, T1v);
Chris@42 151 T1t = VADD(T8, Th);
Chris@42 152 Ti = VSUB(T8, Th);
Chris@42 153 T23 = VADD(T21, T22);
Chris@42 154 T2c = VSUB(T21, T22);
Chris@42 155 T12 = VSUB(TS, T11);
Chris@42 156 T1x = VADD(TS, T11);
Chris@42 157 T2j = VADD(T1S, T1T);
Chris@42 158 T1U = VSUB(T1S, T1T);
Chris@42 159 T1y = VADD(T17, T1c);
Chris@42 160 T1d = VSUB(T17, T1c);
Chris@42 161 T2g = VSUB(T23, T20);
Chris@42 162 T24 = VADD(T20, T23);
Chris@42 163 }
Chris@42 164 {
Chris@42 165 V T2d, T2t, T29, T25, T1m, T1q, T1i, T1H, T1L, T1D, T1A, T28, T1W, T1h, T1g;
Chris@42 166 V T1e, T1l, T1z, T1F, T1V, T1f, T1C, T1B, T26, T27, T2a, T2s, T1j, T1p, T1K;
Chris@42 167 V T1E, T1n, T1o, T1s, T1r, T1I, T1J, T1N, T1M;
Chris@42 168 T2d = VFMA(LDK(KP618033988), T2c, T2b);
Chris@42 169 T2t = VFNMS(LDK(KP618033988), T2b, T2c);
Chris@42 170 T1e = VSUB(T12, T1d);
Chris@42 171 T1l = VADD(T12, T1d);
Chris@42 172 T1z = VADD(T1x, T1y);
Chris@42 173 T1F = VSUB(T1x, T1y);
Chris@42 174 T1V = VADD(T1R, T1U);
Chris@42 175 T29 = VSUB(T1R, T1U);
Chris@42 176 T2f = VFNMS(LDK(KP250000000), T24, T1X);
Chris@42 177 T25 = VADD(T1X, T24);
Chris@42 178 T1m = VMUL(LDK(KP951056516), VFMA(LDK(KP618033988), T1l, T1k));
Chris@42 179 T1q = VMUL(LDK(KP951056516), VFNMS(LDK(KP618033988), T1k, T1l));
Chris@42 180 T1i = VSUB(TN, T1e);
Chris@42 181 T1f = VADD(TN, T1e);
Chris@42 182 T1H = VMUL(LDK(KP951056516), VFNMS(LDK(KP618033988), T1G, T1F));
Chris@42 183 T1L = VMUL(LDK(KP951056516), VFMA(LDK(KP618033988), T1F, T1G));
Chris@42 184 T1D = VSUB(T1w, T1z);
Chris@42 185 T1A = VADD(T1w, T1z);
Chris@42 186 T28 = VFNMS(LDK(KP250000000), T1V, T1O);
Chris@42 187 T1W = VADD(T1O, T1V);
Chris@42 188 T1h = VFNMS(LDK(KP250000000), T1f, Ti);
Chris@42 189 T1g = VMUL(LDK(KP500000000), VADD(Ti, T1f));
Chris@42 190 T2w = VFNMS(LDK(KP618033988), T2i, T2j);
Chris@42 191 T2k = VFMA(LDK(KP618033988), T2j, T2i);
Chris@42 192 T1C = VFNMS(LDK(KP250000000), T1A, T1t);
Chris@42 193 T1B = VCONJ(VMUL(LDK(KP500000000), VADD(T1t, T1A)));
Chris@42 194 T26 = VMUL(LDK(KP500000000), VFNMSI(T25, T1W));
Chris@42 195 T27 = VCONJ(VMUL(LDK(KP500000000), VFMAI(T25, T1W)));
Chris@42 196 T2a = VFMA(LDK(KP559016994), T29, T28);
Chris@42 197 T2s = VFNMS(LDK(KP559016994), T29, T28);
Chris@42 198 ST(&(Rp[0]), T1g, ms, &(Rp[0]));
Chris@42 199 T1j = VFMA(LDK(KP559016994), T1i, T1h);
Chris@42 200 T1p = VFNMS(LDK(KP559016994), T1i, T1h);
Chris@42 201 ST(&(Rm[WS(rs, 9)]), T1B, -ms, &(Rm[WS(rs, 1)]));
Chris@42 202 T1K = VFMA(LDK(KP559016994), T1D, T1C);
Chris@42 203 T1E = VFNMS(LDK(KP559016994), T1D, T1C);
Chris@42 204 ST(&(Rm[WS(rs, 4)]), T27, -ms, &(Rm[0]));
Chris@42 205 ST(&(Rp[WS(rs, 5)]), T26, ms, &(Rp[WS(rs, 1)]));
Chris@42 206 T2A = VFMA(LDK(KP951056516), T2t, T2s);
Chris@42 207 T2u = VFNMS(LDK(KP951056516), T2t, T2s);
Chris@42 208 T2e = VFNMS(LDK(KP951056516), T2d, T2a);
Chris@42 209 T2o = VFMA(LDK(KP951056516), T2d, T2a);
Chris@42 210 T1n = VCONJ(VMUL(LDK(KP500000000), VFNMSI(T1m, T1j)));
Chris@42 211 T1o = VMUL(LDK(KP500000000), VFMAI(T1m, T1j));
Chris@42 212 T1s = VCONJ(VMUL(LDK(KP500000000), VFMAI(T1q, T1p)));
Chris@42 213 T1r = VMUL(LDK(KP500000000), VFNMSI(T1q, T1p));
Chris@42 214 T1I = VCONJ(VMUL(LDK(KP500000000), VFNMSI(T1H, T1E)));
Chris@42 215 T1J = VMUL(LDK(KP500000000), VFMAI(T1H, T1E));
Chris@42 216 T1N = VCONJ(VMUL(LDK(KP500000000), VFMAI(T1L, T1K)));
Chris@42 217 T1M = VMUL(LDK(KP500000000), VFNMSI(T1L, T1K));
Chris@42 218 ST(&(Rp[WS(rs, 4)]), T1o, ms, &(Rp[0]));
Chris@42 219 ST(&(Rm[WS(rs, 3)]), T1n, -ms, &(Rm[WS(rs, 1)]));
Chris@42 220 ST(&(Rp[WS(rs, 8)]), T1r, ms, &(Rp[0]));
Chris@42 221 ST(&(Rm[WS(rs, 7)]), T1s, -ms, &(Rm[WS(rs, 1)]));
Chris@42 222 ST(&(Rp[WS(rs, 2)]), T1J, ms, &(Rp[0]));
Chris@42 223 ST(&(Rm[WS(rs, 1)]), T1I, -ms, &(Rm[WS(rs, 1)]));
Chris@42 224 ST(&(Rp[WS(rs, 6)]), T1M, ms, &(Rp[0]));
Chris@42 225 ST(&(Rm[WS(rs, 5)]), T1N, -ms, &(Rm[WS(rs, 1)]));
Chris@42 226 }
Chris@42 227 T2v = VFMA(LDK(KP559016994), T2g, T2f);
Chris@42 228 T2h = VFNMS(LDK(KP559016994), T2g, T2f);
Chris@42 229 T2x = VFNMS(LDK(KP951056516), T2w, T2v);
Chris@42 230 T2B = VFMA(LDK(KP951056516), T2w, T2v);
Chris@42 231 T2p = VFMA(LDK(KP951056516), T2k, T2h);
Chris@42 232 T2l = VFNMS(LDK(KP951056516), T2k, T2h);
Chris@42 233 T2z = VMUL(LDK(KP500000000), VFMAI(T2x, T2u));
Chris@42 234 T2y = VCONJ(VMUL(LDK(KP500000000), VFNMSI(T2x, T2u)));
Chris@42 235 T2D = VMUL(LDK(KP500000000), VFMAI(T2B, T2A));
Chris@42 236 T2C = VCONJ(VMUL(LDK(KP500000000), VFNMSI(T2B, T2A)));
Chris@42 237 T2r = VCONJ(VMUL(LDK(KP500000000), VFMAI(T2p, T2o)));
Chris@42 238 T2q = VMUL(LDK(KP500000000), VFNMSI(T2p, T2o));
Chris@42 239 T2n = VCONJ(VMUL(LDK(KP500000000), VFMAI(T2l, T2e)));
Chris@42 240 T2m = VMUL(LDK(KP500000000), VFNMSI(T2l, T2e));
Chris@42 241 ST(&(Rp[WS(rs, 3)]), T2z, ms, &(Rp[WS(rs, 1)]));
Chris@42 242 ST(&(Rm[WS(rs, 2)]), T2y, -ms, &(Rm[0]));
Chris@42 243 ST(&(Rp[WS(rs, 7)]), T2D, ms, &(Rp[WS(rs, 1)]));
Chris@42 244 ST(&(Rm[WS(rs, 6)]), T2C, -ms, &(Rm[0]));
Chris@42 245 ST(&(Rm[0]), T2r, -ms, &(Rm[0]));
Chris@42 246 ST(&(Rp[WS(rs, 1)]), T2q, ms, &(Rp[WS(rs, 1)]));
Chris@42 247 ST(&(Rm[WS(rs, 8)]), T2n, -ms, &(Rm[0]));
Chris@42 248 ST(&(Rp[WS(rs, 9)]), T2m, ms, &(Rp[WS(rs, 1)]));
Chris@42 249 }
Chris@42 250 }
Chris@42 251 VLEAVE();
Chris@42 252 }
Chris@42 253
Chris@42 254 static const tw_instr twinstr[] = {
Chris@42 255 VTW(1, 1),
Chris@42 256 VTW(1, 2),
Chris@42 257 VTW(1, 3),
Chris@42 258 VTW(1, 4),
Chris@42 259 VTW(1, 5),
Chris@42 260 VTW(1, 6),
Chris@42 261 VTW(1, 7),
Chris@42 262 VTW(1, 8),
Chris@42 263 VTW(1, 9),
Chris@42 264 VTW(1, 10),
Chris@42 265 VTW(1, 11),
Chris@42 266 VTW(1, 12),
Chris@42 267 VTW(1, 13),
Chris@42 268 VTW(1, 14),
Chris@42 269 VTW(1, 15),
Chris@42 270 VTW(1, 16),
Chris@42 271 VTW(1, 17),
Chris@42 272 VTW(1, 18),
Chris@42 273 VTW(1, 19),
Chris@42 274 {TW_NEXT, VL, 0}
Chris@42 275 };
Chris@42 276
Chris@42 277 static const hc2c_desc desc = { 20, XSIMD_STRING("hc2cfdftv_20"), twinstr, &GENUS, {77, 62, 66, 0} };
Chris@42 278
Chris@42 279 void XSIMD(codelet_hc2cfdftv_20) (planner *p) {
Chris@42 280 X(khc2c_register) (p, hc2cfdftv_20, &desc, HC2C_VIA_DFT);
Chris@42 281 }
Chris@42 282 #else /* HAVE_FMA */
Chris@42 283
Chris@42 284 /* Generated by: ../../../genfft/gen_hc2cdft_c.native -simd -compact -variables 4 -pipeline-latency 8 -trivial-stores -variables 32 -no-generate-bytw -n 20 -dit -name hc2cfdftv_20 -include hc2cfv.h */
Chris@42 285
Chris@42 286 /*
Chris@42 287 * This function contains 143 FP additions, 77 FP multiplications,
Chris@42 288 * (or, 131 additions, 65 multiplications, 12 fused multiply/add),
Chris@42 289 * 141 stack variables, 9 constants, and 40 memory accesses
Chris@42 290 */
Chris@42 291 #include "hc2cfv.h"
Chris@42 292
Chris@42 293 static void hc2cfdftv_20(R *Rp, R *Ip, R *Rm, R *Im, const R *W, stride rs, INT mb, INT me, INT ms)
Chris@42 294 {
Chris@42 295 DVK(KP293892626, +0.293892626146236564584352977319536384298826219);
Chris@42 296 DVK(KP475528258, +0.475528258147576786058219666689691071702849317);
Chris@42 297 DVK(KP559016994, +0.559016994374947424102293417182819058860154590);
Chris@42 298 DVK(KP250000000, +0.250000000000000000000000000000000000000000000);
Chris@42 299 DVK(KP125000000, +0.125000000000000000000000000000000000000000000);
Chris@42 300 DVK(KP279508497, +0.279508497187473712051146708591409529430077295);
Chris@42 301 DVK(KP587785252, +0.587785252292473129168705954639072768597652438);
Chris@42 302 DVK(KP951056516, +0.951056516295153572116439333379382143405698634);
Chris@42 303 DVK(KP500000000, +0.500000000000000000000000000000000000000000000);
Chris@42 304 {
Chris@42 305 INT m;
Chris@42 306 for (m = mb, W = W + ((mb - 1) * ((TWVL / VL) * 38)); m < me; m = m + VL, Rp = Rp + (VL * ms), Ip = Ip + (VL * ms), Rm = Rm - (VL * ms), Im = Im - (VL * ms), W = W + (TWVL * 38), MAKE_VOLATILE_STRIDE(80, rs)) {
Chris@42 307 V TW, T1x, T2i, T2A, T1r, T1s, T1a, T1y, T1l, Tn, TK, TL, T1p, T1o, T27;
Chris@42 308 V T2t, T2a, T2u, T2e, T2C, T20, T2w, T23, T2x, T2d, T2B, T1W, T1X, T1U, T1V;
Chris@42 309 V T2z, T2K, T2G, T2N, T2J, T2v, T2y, T2F, T2D, T2E, T2M, T2H, T2I, T2L;
Chris@42 310 {
Chris@42 311 V T1u, T5, Tg, T1c, TV, T13, Ta, T1w, TQ, T11, TI, T1j, Tx, T18, Tl;
Chris@42 312 V T1e, TD, T1h, Ts, T16, T2g, T2h, T14, T19, T1f, T1k, Tb, Tm, Ty, TJ;
Chris@42 313 V T25, T26, T28, T29, T1Y, T1Z, T21, T22;
Chris@42 314 {
Chris@42 315 V T4, T3, T2, T1, Tf, Te, Td, Tc, T1b, TU, TT, TS, TR, T12, T9;
Chris@42 316 V T8, T7, T6, T1v, TP, TO, TN, TM, T10, TH, TG, TF, TE, T1i, Tw;
Chris@42 317 V Tv, Tu, Tt, T17, Tk, Tj, Ti, Th, T1d, TC, TB, TA, Tz, T1g, Tr;
Chris@42 318 V Tq, Tp, To, T15;
Chris@42 319 T4 = LD(&(Rp[0]), ms, &(Rp[0]));
Chris@42 320 T2 = LD(&(Rm[0]), -ms, &(Rm[0]));
Chris@42 321 T3 = VCONJ(T2);
Chris@42 322 T1u = VADD(T4, T3);
Chris@42 323 T1 = LDW(&(W[0]));
Chris@42 324 T5 = VZMULIJ(T1, VSUB(T3, T4));
Chris@42 325 Tf = LD(&(Rp[WS(rs, 4)]), ms, &(Rp[0]));
Chris@42 326 Td = LD(&(Rm[WS(rs, 4)]), -ms, &(Rm[0]));
Chris@42 327 Te = VCONJ(Td);
Chris@42 328 Tc = LDW(&(W[TWVL * 16]));
Chris@42 329 Tg = VZMULIJ(Tc, VSUB(Te, Tf));
Chris@42 330 T1b = LDW(&(W[TWVL * 14]));
Chris@42 331 T1c = VZMULJ(T1b, VADD(Te, Tf));
Chris@42 332 TU = LD(&(Rp[WS(rs, 7)]), ms, &(Rp[WS(rs, 1)]));
Chris@42 333 TS = LD(&(Rm[WS(rs, 7)]), -ms, &(Rm[WS(rs, 1)]));
Chris@42 334 TT = VCONJ(TS);
Chris@42 335 TR = LDW(&(W[TWVL * 28]));
Chris@42 336 TV = VZMULIJ(TR, VSUB(TT, TU));
Chris@42 337 T12 = LDW(&(W[TWVL * 26]));
Chris@42 338 T13 = VZMULJ(T12, VADD(TT, TU));
Chris@42 339 T9 = LD(&(Rp[WS(rs, 5)]), ms, &(Rp[WS(rs, 1)]));
Chris@42 340 T7 = LD(&(Rm[WS(rs, 5)]), -ms, &(Rm[WS(rs, 1)]));
Chris@42 341 T8 = VCONJ(T7);
Chris@42 342 T6 = LDW(&(W[TWVL * 20]));
Chris@42 343 Ta = VZMULIJ(T6, VSUB(T8, T9));
Chris@42 344 T1v = LDW(&(W[TWVL * 18]));
Chris@42 345 T1w = VZMULJ(T1v, VADD(T9, T8));
Chris@42 346 TP = LD(&(Rp[WS(rs, 2)]), ms, &(Rp[0]));
Chris@42 347 TN = LD(&(Rm[WS(rs, 2)]), -ms, &(Rm[0]));
Chris@42 348 TO = VCONJ(TN);
Chris@42 349 TM = LDW(&(W[TWVL * 8]));
Chris@42 350 TQ = VZMULIJ(TM, VSUB(TO, TP));
Chris@42 351 T10 = LDW(&(W[TWVL * 6]));
Chris@42 352 T11 = VZMULJ(T10, VADD(TO, TP));
Chris@42 353 TH = LD(&(Rp[WS(rs, 1)]), ms, &(Rp[WS(rs, 1)]));
Chris@42 354 TF = LD(&(Rm[WS(rs, 1)]), -ms, &(Rm[WS(rs, 1)]));
Chris@42 355 TG = VCONJ(TF);
Chris@42 356 TE = LDW(&(W[TWVL * 4]));
Chris@42 357 TI = VZMULIJ(TE, VSUB(TG, TH));
Chris@42 358 T1i = LDW(&(W[TWVL * 2]));
Chris@42 359 T1j = VZMULJ(T1i, VADD(TG, TH));
Chris@42 360 Tw = LD(&(Rp[WS(rs, 3)]), ms, &(Rp[WS(rs, 1)]));
Chris@42 361 Tu = LD(&(Rm[WS(rs, 3)]), -ms, &(Rm[WS(rs, 1)]));
Chris@42 362 Tv = VCONJ(Tu);
Chris@42 363 Tt = LDW(&(W[TWVL * 12]));
Chris@42 364 Tx = VZMULIJ(Tt, VSUB(Tv, Tw));
Chris@42 365 T17 = LDW(&(W[TWVL * 10]));
Chris@42 366 T18 = VZMULJ(T17, VADD(Tw, Tv));
Chris@42 367 Tk = LD(&(Rp[WS(rs, 9)]), ms, &(Rp[WS(rs, 1)]));
Chris@42 368 Ti = LD(&(Rm[WS(rs, 9)]), -ms, &(Rm[WS(rs, 1)]));
Chris@42 369 Tj = VCONJ(Ti);
Chris@42 370 Th = LDW(&(W[TWVL * 36]));
Chris@42 371 Tl = VZMULIJ(Th, VSUB(Tj, Tk));
Chris@42 372 T1d = LDW(&(W[TWVL * 34]));
Chris@42 373 T1e = VZMULJ(T1d, VADD(Tj, Tk));
Chris@42 374 TC = LD(&(Rp[WS(rs, 6)]), ms, &(Rp[0]));
Chris@42 375 TA = LD(&(Rm[WS(rs, 6)]), -ms, &(Rm[0]));
Chris@42 376 TB = VCONJ(TA);
Chris@42 377 Tz = LDW(&(W[TWVL * 24]));
Chris@42 378 TD = VZMULIJ(Tz, VSUB(TB, TC));
Chris@42 379 T1g = LDW(&(W[TWVL * 22]));
Chris@42 380 T1h = VZMULJ(T1g, VADD(TB, TC));
Chris@42 381 Tr = LD(&(Rp[WS(rs, 8)]), ms, &(Rp[0]));
Chris@42 382 Tp = LD(&(Rm[WS(rs, 8)]), -ms, &(Rm[0]));
Chris@42 383 Tq = VCONJ(Tp);
Chris@42 384 To = LDW(&(W[TWVL * 32]));
Chris@42 385 Ts = VZMULIJ(To, VSUB(Tq, Tr));
Chris@42 386 T15 = LDW(&(W[TWVL * 30]));
Chris@42 387 T16 = VZMULJ(T15, VADD(Tr, Tq));
Chris@42 388 }
Chris@42 389 TW = VSUB(TQ, TV);
Chris@42 390 T1x = VSUB(T1u, T1w);
Chris@42 391 T2g = VADD(T1u, T1w);
Chris@42 392 T2h = VADD(TQ, TV);
Chris@42 393 T2i = VADD(T2g, T2h);
Chris@42 394 T2A = VSUB(T2g, T2h);
Chris@42 395 T14 = VSUB(T11, T13);
Chris@42 396 T19 = VSUB(T16, T18);
Chris@42 397 T1r = VADD(T14, T19);
Chris@42 398 T1f = VSUB(T1c, T1e);
Chris@42 399 T1k = VSUB(T1h, T1j);
Chris@42 400 T1s = VADD(T1f, T1k);
Chris@42 401 T1a = VSUB(T14, T19);
Chris@42 402 T1y = VADD(T1r, T1s);
Chris@42 403 T1l = VSUB(T1f, T1k);
Chris@42 404 Tb = VSUB(T5, Ta);
Chris@42 405 Tm = VSUB(Tg, Tl);
Chris@42 406 Tn = VADD(Tb, Tm);
Chris@42 407 Ty = VSUB(Ts, Tx);
Chris@42 408 TJ = VSUB(TD, TI);
Chris@42 409 TK = VADD(Ty, TJ);
Chris@42 410 TL = VADD(Tn, TK);
Chris@42 411 T1p = VSUB(Ty, TJ);
Chris@42 412 T1o = VSUB(Tb, Tm);
Chris@42 413 T25 = VADD(T1c, T1e);
Chris@42 414 T26 = VADD(TD, TI);
Chris@42 415 T27 = VADD(T25, T26);
Chris@42 416 T2t = VSUB(T25, T26);
Chris@42 417 T28 = VADD(Ts, Tx);
Chris@42 418 T29 = VADD(T1h, T1j);
Chris@42 419 T2a = VADD(T28, T29);
Chris@42 420 T2u = VSUB(T29, T28);
Chris@42 421 T2e = VADD(T27, T2a);
Chris@42 422 T2C = VADD(T2t, T2u);
Chris@42 423 T1Y = VADD(T11, T13);
Chris@42 424 T1Z = VADD(Tg, Tl);
Chris@42 425 T20 = VADD(T1Y, T1Z);
Chris@42 426 T2w = VSUB(T1Y, T1Z);
Chris@42 427 T21 = VADD(T5, Ta);
Chris@42 428 T22 = VADD(T16, T18);
Chris@42 429 T23 = VADD(T21, T22);
Chris@42 430 T2x = VSUB(T22, T21);
Chris@42 431 T2d = VADD(T20, T23);
Chris@42 432 T2B = VADD(T2w, T2x);
Chris@42 433 }
Chris@42 434 T1U = VADD(T1x, T1y);
Chris@42 435 T1V = VBYI(VADD(TW, TL));
Chris@42 436 T1W = VMUL(LDK(KP500000000), VSUB(T1U, T1V));
Chris@42 437 T1X = VCONJ(VMUL(LDK(KP500000000), VADD(T1V, T1U)));
Chris@42 438 ST(&(Rp[WS(rs, 5)]), T1W, ms, &(Rp[WS(rs, 1)]));
Chris@42 439 ST(&(Rm[WS(rs, 4)]), T1X, -ms, &(Rm[0]));
Chris@42 440 T2v = VSUB(T2t, T2u);
Chris@42 441 T2y = VSUB(T2w, T2x);
Chris@42 442 T2z = VMUL(LDK(KP500000000), VBYI(VFNMS(LDK(KP587785252), T2y, VMUL(LDK(KP951056516), T2v))));
Chris@42 443 T2K = VMUL(LDK(KP500000000), VBYI(VFMA(LDK(KP951056516), T2y, VMUL(LDK(KP587785252), T2v))));
Chris@42 444 T2F = VMUL(LDK(KP279508497), VSUB(T2B, T2C));
Chris@42 445 T2D = VADD(T2B, T2C);
Chris@42 446 T2E = VFNMS(LDK(KP125000000), T2D, VMUL(LDK(KP500000000), T2A));
Chris@42 447 T2G = VSUB(T2E, T2F);
Chris@42 448 T2N = VCONJ(VMUL(LDK(KP500000000), VADD(T2A, T2D)));
Chris@42 449 T2J = VADD(T2F, T2E);
Chris@42 450 ST(&(Rm[WS(rs, 9)]), T2N, -ms, &(Rm[WS(rs, 1)]));
Chris@42 451 T2M = VCONJ(VADD(T2K, T2J));
Chris@42 452 ST(&(Rm[WS(rs, 5)]), T2M, -ms, &(Rm[WS(rs, 1)]));
Chris@42 453 T2H = VADD(T2z, T2G);
Chris@42 454 ST(&(Rp[WS(rs, 2)]), T2H, ms, &(Rp[0]));
Chris@42 455 T2I = VCONJ(VSUB(T2G, T2z));
Chris@42 456 ST(&(Rm[WS(rs, 1)]), T2I, -ms, &(Rm[WS(rs, 1)]));
Chris@42 457 T2L = VSUB(T2J, T2K);
Chris@42 458 ST(&(Rp[WS(rs, 6)]), T2L, ms, &(Rp[0]));
Chris@42 459 {
Chris@42 460 V T2c, T2p, T2l, T2s, T2o, T24, T2b, T2f, T2j, T2k, T2r, T2m, T2n, T2q, T1n;
Chris@42 461 V T1Q, T1E, T1K, T1B, T1R, T1F, T1N, T1m, T1J, TZ, T1I, TX, TY, T1q, T1M;
Chris@42 462 V T1A, T1L, T1t, T1z, T1C, T1S, T1T, T1D, T1G, T1O, T1P, T1H;
Chris@42 463 T24 = VSUB(T20, T23);
Chris@42 464 T2b = VSUB(T27, T2a);
Chris@42 465 T2c = VMUL(LDK(KP500000000), VBYI(VFMA(LDK(KP951056516), T24, VMUL(LDK(KP587785252), T2b))));
Chris@42 466 T2p = VMUL(LDK(KP500000000), VBYI(VFNMS(LDK(KP587785252), T24, VMUL(LDK(KP951056516), T2b))));
Chris@42 467 T2f = VMUL(LDK(KP279508497), VSUB(T2d, T2e));
Chris@42 468 T2j = VADD(T2d, T2e);
Chris@42 469 T2k = VFNMS(LDK(KP125000000), T2j, VMUL(LDK(KP500000000), T2i));
Chris@42 470 T2l = VADD(T2f, T2k);
Chris@42 471 T2s = VMUL(LDK(KP500000000), VADD(T2i, T2j));
Chris@42 472 T2o = VSUB(T2k, T2f);
Chris@42 473 ST(&(Rp[0]), T2s, ms, &(Rp[0]));
Chris@42 474 T2r = VCONJ(VADD(T2p, T2o));
Chris@42 475 ST(&(Rm[WS(rs, 7)]), T2r, -ms, &(Rm[WS(rs, 1)]));
Chris@42 476 T2m = VADD(T2c, T2l);
Chris@42 477 ST(&(Rp[WS(rs, 4)]), T2m, ms, &(Rp[0]));
Chris@42 478 T2n = VCONJ(VSUB(T2l, T2c));
Chris@42 479 ST(&(Rm[WS(rs, 3)]), T2n, -ms, &(Rm[WS(rs, 1)]));
Chris@42 480 T2q = VSUB(T2o, T2p);
Chris@42 481 ST(&(Rp[WS(rs, 8)]), T2q, ms, &(Rp[0]));
Chris@42 482 T1m = VFMA(LDK(KP951056516), T1a, VMUL(LDK(KP587785252), T1l));
Chris@42 483 T1J = VFNMS(LDK(KP587785252), T1a, VMUL(LDK(KP951056516), T1l));
Chris@42 484 TX = VFMS(LDK(KP250000000), TL, TW);
Chris@42 485 TY = VMUL(LDK(KP559016994), VSUB(TK, Tn));
Chris@42 486 TZ = VADD(TX, TY);
Chris@42 487 T1I = VSUB(TY, TX);
Chris@42 488 T1n = VMUL(LDK(KP500000000), VBYI(VSUB(TZ, T1m)));
Chris@42 489 T1Q = VMUL(LDK(KP500000000), VBYI(VADD(T1I, T1J)));
Chris@42 490 T1E = VMUL(LDK(KP500000000), VBYI(VADD(TZ, T1m)));
Chris@42 491 T1K = VMUL(LDK(KP500000000), VBYI(VSUB(T1I, T1J)));
Chris@42 492 T1q = VFMA(LDK(KP475528258), T1o, VMUL(LDK(KP293892626), T1p));
Chris@42 493 T1M = VFNMS(LDK(KP293892626), T1o, VMUL(LDK(KP475528258), T1p));
Chris@42 494 T1t = VMUL(LDK(KP279508497), VSUB(T1r, T1s));
Chris@42 495 T1z = VFNMS(LDK(KP125000000), T1y, VMUL(LDK(KP500000000), T1x));
Chris@42 496 T1A = VADD(T1t, T1z);
Chris@42 497 T1L = VSUB(T1z, T1t);
Chris@42 498 T1B = VADD(T1q, T1A);
Chris@42 499 T1R = VADD(T1M, T1L);
Chris@42 500 T1F = VSUB(T1A, T1q);
Chris@42 501 T1N = VSUB(T1L, T1M);
Chris@42 502 T1C = VADD(T1n, T1B);
Chris@42 503 ST(&(Rp[WS(rs, 1)]), T1C, ms, &(Rp[WS(rs, 1)]));
Chris@42 504 T1S = VADD(T1Q, T1R);
Chris@42 505 ST(&(Rp[WS(rs, 7)]), T1S, ms, &(Rp[WS(rs, 1)]));
Chris@42 506 T1T = VCONJ(VSUB(T1R, T1Q));
Chris@42 507 ST(&(Rm[WS(rs, 6)]), T1T, -ms, &(Rm[0]));
Chris@42 508 T1D = VCONJ(VSUB(T1B, T1n));
Chris@42 509 ST(&(Rm[0]), T1D, -ms, &(Rm[0]));
Chris@42 510 T1G = VADD(T1E, T1F);
Chris@42 511 ST(&(Rp[WS(rs, 9)]), T1G, ms, &(Rp[WS(rs, 1)]));
Chris@42 512 T1O = VADD(T1K, T1N);
Chris@42 513 ST(&(Rp[WS(rs, 3)]), T1O, ms, &(Rp[WS(rs, 1)]));
Chris@42 514 T1P = VCONJ(VSUB(T1N, T1K));
Chris@42 515 ST(&(Rm[WS(rs, 2)]), T1P, -ms, &(Rm[0]));
Chris@42 516 T1H = VCONJ(VSUB(T1F, T1E));
Chris@42 517 ST(&(Rm[WS(rs, 8)]), T1H, -ms, &(Rm[0]));
Chris@42 518 }
Chris@42 519 }
Chris@42 520 }
Chris@42 521 VLEAVE();
Chris@42 522 }
Chris@42 523
Chris@42 524 static const tw_instr twinstr[] = {
Chris@42 525 VTW(1, 1),
Chris@42 526 VTW(1, 2),
Chris@42 527 VTW(1, 3),
Chris@42 528 VTW(1, 4),
Chris@42 529 VTW(1, 5),
Chris@42 530 VTW(1, 6),
Chris@42 531 VTW(1, 7),
Chris@42 532 VTW(1, 8),
Chris@42 533 VTW(1, 9),
Chris@42 534 VTW(1, 10),
Chris@42 535 VTW(1, 11),
Chris@42 536 VTW(1, 12),
Chris@42 537 VTW(1, 13),
Chris@42 538 VTW(1, 14),
Chris@42 539 VTW(1, 15),
Chris@42 540 VTW(1, 16),
Chris@42 541 VTW(1, 17),
Chris@42 542 VTW(1, 18),
Chris@42 543 VTW(1, 19),
Chris@42 544 {TW_NEXT, VL, 0}
Chris@42 545 };
Chris@42 546
Chris@42 547 static const hc2c_desc desc = { 20, XSIMD_STRING("hc2cfdftv_20"), twinstr, &GENUS, {131, 65, 12, 0} };
Chris@42 548
Chris@42 549 void XSIMD(codelet_hc2cfdftv_20) (planner *p) {
Chris@42 550 X(khc2c_register) (p, hc2cfdftv_20, &desc, HC2C_VIA_DFT);
Chris@42 551 }
Chris@42 552 #endif /* HAVE_FMA */