Mercurial > hg > sv-dependency-builds

annotate src/fftw-3.3.3/dft/simd/common/q1bv_5.c @ 165:7e6e71a29886

Find changesets by keywords (author, files, the commit message), revision number or hash, or revset expression.
Update build for AUDIO_COMPONENT_FIX
author Chris Cannam <cannam@all-day-breakfast.com>
date Wed, 30 Oct 2019 12:40:34 +0000
parents 89f5e221ed7b
children
rev   line source
cannam@95 1 /*
cannam@95 2 * Copyright (c) 2003, 2007-11 Matteo Frigo
cannam@95 3 * Copyright (c) 2003, 2007-11 Massachusetts Institute of Technology
cannam@95 4 *
cannam@95 5 * This program is free software; you can redistribute it and/or modify
cannam@95 6 * it under the terms of the GNU General Public License as published by
cannam@95 7 * the Free Software Foundation; either version 2 of the License, or
cannam@95 8 * (at your option) any later version.
cannam@95 9 *
cannam@95 10 * This program is distributed in the hope that it will be useful,
cannam@95 11 * but WITHOUT ANY WARRANTY; without even the implied warranty of
cannam@95 12 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
cannam@95 13 * GNU General Public License for more details.
cannam@95 14 *
cannam@95 15 * You should have received a copy of the GNU General Public License
cannam@95 16 * along with this program; if not, write to the Free Software
cannam@95 17 * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
cannam@95 18 *
cannam@95 19 */
cannam@95 20
cannam@95 21 /* This file was automatically generated --- DO NOT EDIT */
cannam@95 22 /* Generated on Sun Nov 25 07:39:33 EST 2012 */
cannam@95 23
cannam@95 24 #include "codelet-dft.h"
cannam@95 25
cannam@95 26 #ifdef HAVE_FMA
cannam@95 27
cannam@95 28 /* Generated by: ../../../genfft/gen_twidsq_c.native -fma -reorder-insns -schedule-for-pipeline -simd -compact -variables 4 -pipeline-latency 8 -n 5 -dif -name q1bv_5 -include q1b.h -sign 1 */
cannam@95 29
cannam@95 30 /*
cannam@95 31 * This function contains 100 FP additions, 95 FP multiplications,
cannam@95 32 * (or, 55 additions, 50 multiplications, 45 fused multiply/add),
cannam@95 33 * 69 stack variables, 4 constants, and 50 memory accesses
cannam@95 34 */
cannam@95 35 #include "q1b.h"
cannam@95 36
cannam@95 37 static void q1bv_5(R *ri, R *ii, const R *W, stride rs, stride vs, INT mb, INT me, INT ms)
cannam@95 38 {
cannam@95 39 DVK(KP559016994, +0.559016994374947424102293417182819058860154590);
cannam@95 40 DVK(KP250000000, +0.250000000000000000000000000000000000000000000);
cannam@95 41 DVK(KP618033988, +0.618033988749894848204586834365638117720309180);
cannam@95 42 DVK(KP951056516, +0.951056516295153572116439333379382143405698634);
cannam@95 43 {
cannam@95 44 INT m;
cannam@95 45 R *x;
cannam@95 46 x = ii;
cannam@95 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(10, rs), MAKE_VOLATILE_STRIDE(10, vs)) {
cannam@95 48 V Te, T1w, Ty, TS, TW, Tb, T1t, Tv, T1g, T1c, TP, TV, T1f, T19, TY;
cannam@95 49 V TX;
cannam@95 50 {
cannam@95 51 V T1, T1j, Tl, Ti, Ta, T8, T1A, T1q, T1s, T9, TF, T1r, TZ, TR, TL;
cannam@95 52 V TC, Ts, Tu, TQ, TI, T15, T1b, T10, T11, Tt;
cannam@95 53 {
cannam@95 54 V T1n, T1o, T1k, T1l, T7, Td, T4, Tc;
cannam@95 55 {
cannam@95 56 V T5, T6, T2, T3;
cannam@95 57 T1 = LD(&(x[0]), ms, &(x[0]));
cannam@95 58 T5 = LD(&(x[WS(rs, 2)]), ms, &(x[0]));
cannam@95 59 T6 = LD(&(x[WS(rs, 3)]), ms, &(x[WS(rs, 1)]));
cannam@95 60 T2 = LD(&(x[WS(rs, 1)]), ms, &(x[WS(rs, 1)]));
cannam@95 61 T3 = LD(&(x[WS(rs, 4)]), ms, &(x[0]));
cannam@95 62 T1j = LD(&(x[WS(vs, 4)]), ms, &(x[WS(vs, 4)]));
cannam@95 63 T1n = LD(&(x[WS(vs, 4) + WS(rs, 2)]), ms, &(x[WS(vs, 4)]));
cannam@95 64 T1o = LD(&(x[WS(vs, 4) + WS(rs, 3)]), ms, &(x[WS(vs, 4) + WS(rs, 1)]));
cannam@95 65 T1k = LD(&(x[WS(vs, 4) + WS(rs, 1)]), ms, &(x[WS(vs, 4) + WS(rs, 1)]));
cannam@95 66 T1l = LD(&(x[WS(vs, 4) + WS(rs, 4)]), ms, &(x[WS(vs, 4)]));
cannam@95 67 T7 = VADD(T5, T6);
cannam@95 68 Td = VSUB(T5, T6);
cannam@95 69 T4 = VADD(T2, T3);
cannam@95 70 Tc = VSUB(T2, T3);
cannam@95 71 }
cannam@95 72 {
cannam@95 73 V Tm, Tn, Tr, Tx, T1v, T1p;
cannam@95 74 Tl = LD(&(x[WS(vs, 1)]), ms, &(x[WS(vs, 1)]));
cannam@95 75 T1v = VSUB(T1n, T1o);
cannam@95 76 T1p = VADD(T1n, T1o);
cannam@95 77 {
cannam@95 78 V T1u, T1m, Tp, Tq;
cannam@95 79 T1u = VSUB(T1k, T1l);
cannam@95 80 T1m = VADD(T1k, T1l);
cannam@95 81 Tp = LD(&(x[WS(vs, 1) + WS(rs, 2)]), ms, &(x[WS(vs, 1)]));
cannam@95 82 Ti = VMUL(LDK(KP951056516), VFNMS(LDK(KP618033988), Tc, Td));
cannam@95 83 Te = VMUL(LDK(KP951056516), VFMA(LDK(KP618033988), Td, Tc));
cannam@95 84 Ta = VSUB(T4, T7);
cannam@95 85 T8 = VADD(T4, T7);
cannam@95 86 Tq = LD(&(x[WS(vs, 1) + WS(rs, 3)]), ms, &(x[WS(vs, 1) + WS(rs, 1)]));
cannam@95 87 T1w = VMUL(LDK(KP951056516), VFMA(LDK(KP618033988), T1v, T1u));
cannam@95 88 T1A = VMUL(LDK(KP951056516), VFNMS(LDK(KP618033988), T1u, T1v));
cannam@95 89 T1q = VADD(T1m, T1p);
cannam@95 90 T1s = VSUB(T1m, T1p);
cannam@95 91 Tm = LD(&(x[WS(vs, 1) + WS(rs, 1)]), ms, &(x[WS(vs, 1) + WS(rs, 1)]));
cannam@95 92 T9 = VFNMS(LDK(KP250000000), T8, T1);
cannam@95 93 Tn = LD(&(x[WS(vs, 1) + WS(rs, 4)]), ms, &(x[WS(vs, 1)]));
cannam@95 94 Tr = VADD(Tp, Tq);
cannam@95 95 Tx = VSUB(Tp, Tq);
cannam@95 96 }
cannam@95 97 {
cannam@95 98 V TJ, TK, TG, Tw, To, TH, T13, T14;
cannam@95 99 TF = LD(&(x[WS(vs, 2)]), ms, &(x[WS(vs, 2)]));
cannam@95 100 T1r = VFNMS(LDK(KP250000000), T1q, T1j);
cannam@95 101 TJ = LD(&(x[WS(vs, 2) + WS(rs, 2)]), ms, &(x[WS(vs, 2)]));
cannam@95 102 TK = LD(&(x[WS(vs, 2) + WS(rs, 3)]), ms, &(x[WS(vs, 2) + WS(rs, 1)]));
cannam@95 103 TG = LD(&(x[WS(vs, 2) + WS(rs, 1)]), ms, &(x[WS(vs, 2) + WS(rs, 1)]));
cannam@95 104 Tw = VSUB(Tm, Tn);
cannam@95 105 To = VADD(Tm, Tn);
cannam@95 106 TH = LD(&(x[WS(vs, 2) + WS(rs, 4)]), ms, &(x[WS(vs, 2)]));
cannam@95 107 TZ = LD(&(x[WS(vs, 3)]), ms, &(x[WS(vs, 3)]));
cannam@95 108 T13 = LD(&(x[WS(vs, 3) + WS(rs, 2)]), ms, &(x[WS(vs, 3)]));
cannam@95 109 T14 = LD(&(x[WS(vs, 3) + WS(rs, 3)]), ms, &(x[WS(vs, 3) + WS(rs, 1)]));
cannam@95 110 TR = VSUB(TJ, TK);
cannam@95 111 TL = VADD(TJ, TK);
cannam@95 112 Ty = VMUL(LDK(KP951056516), VFMA(LDK(KP618033988), Tx, Tw));
cannam@95 113 TC = VMUL(LDK(KP951056516), VFNMS(LDK(KP618033988), Tw, Tx));
cannam@95 114 Ts = VADD(To, Tr);
cannam@95 115 Tu = VSUB(To, Tr);
cannam@95 116 TQ = VSUB(TG, TH);
cannam@95 117 TI = VADD(TG, TH);
cannam@95 118 T15 = VADD(T13, T14);
cannam@95 119 T1b = VSUB(T13, T14);
cannam@95 120 T10 = LD(&(x[WS(vs, 3) + WS(rs, 1)]), ms, &(x[WS(vs, 3) + WS(rs, 1)]));
cannam@95 121 T11 = LD(&(x[WS(vs, 3) + WS(rs, 4)]), ms, &(x[WS(vs, 3)]));
cannam@95 122 Tt = VFNMS(LDK(KP250000000), Ts, Tl);
cannam@95 123 }
cannam@95 124 }
cannam@95 125 }
cannam@95 126 {
cannam@95 127 V TO, T12, T1a, Th, T1z, TN, TM, T18, T17;
cannam@95 128 ST(&(x[0]), VADD(T1, T8), ms, &(x[0]));
cannam@95 129 TS = VMUL(LDK(KP951056516), VFMA(LDK(KP618033988), TR, TQ));
cannam@95 130 TW = VMUL(LDK(KP951056516), VFNMS(LDK(KP618033988), TQ, TR));
cannam@95 131 TM = VADD(TI, TL);
cannam@95 132 TO = VSUB(TI, TL);
cannam@95 133 ST(&(x[WS(rs, 4)]), VADD(T1j, T1q), ms, &(x[0]));
cannam@95 134 T12 = VADD(T10, T11);
cannam@95 135 T1a = VSUB(T10, T11);
cannam@95 136 ST(&(x[WS(rs, 1)]), VADD(Tl, Ts), ms, &(x[WS(rs, 1)]));
cannam@95 137 Th = VFNMS(LDK(KP559016994), Ta, T9);
cannam@95 138 Tb = VFMA(LDK(KP559016994), Ta, T9);
cannam@95 139 T1t = VFMA(LDK(KP559016994), T1s, T1r);
cannam@95 140 T1z = VFNMS(LDK(KP559016994), T1s, T1r);
cannam@95 141 ST(&(x[WS(rs, 2)]), VADD(TF, TM), ms, &(x[0]));
cannam@95 142 TN = VFNMS(LDK(KP250000000), TM, TF);
cannam@95 143 {
cannam@95 144 V T16, Tk, Tj, T1C, T1B, TD, TE, TB;
cannam@95 145 TB = VFNMS(LDK(KP559016994), Tu, Tt);
cannam@95 146 Tv = VFMA(LDK(KP559016994), Tu, Tt);
cannam@95 147 T1g = VMUL(LDK(KP951056516), VFNMS(LDK(KP618033988), T1a, T1b));
cannam@95 148 T1c = VMUL(LDK(KP951056516), VFMA(LDK(KP618033988), T1b, T1a));
cannam@95 149 T18 = VSUB(T12, T15);
cannam@95 150 T16 = VADD(T12, T15);
cannam@95 151 Tk = BYTW(&(W[TWVL * 4]), VFMAI(Ti, Th));
cannam@95 152 Tj = BYTW(&(W[TWVL * 2]), VFNMSI(Ti, Th));
cannam@95 153 T1C = BYTW(&(W[TWVL * 4]), VFMAI(T1A, T1z));
cannam@95 154 T1B = BYTW(&(W[TWVL * 2]), VFNMSI(T1A, T1z));
cannam@95 155 TD = BYTW(&(W[TWVL * 2]), VFNMSI(TC, TB));
cannam@95 156 TE = BYTW(&(W[TWVL * 4]), VFMAI(TC, TB));
cannam@95 157 ST(&(x[WS(rs, 3)]), VADD(TZ, T16), ms, &(x[WS(rs, 1)]));
cannam@95 158 T17 = VFNMS(LDK(KP250000000), T16, TZ);
cannam@95 159 ST(&(x[WS(vs, 3)]), Tk, ms, &(x[WS(vs, 3)]));
cannam@95 160 ST(&(x[WS(vs, 2)]), Tj, ms, &(x[WS(vs, 2)]));
cannam@95 161 ST(&(x[WS(vs, 3) + WS(rs, 4)]), T1C, ms, &(x[WS(vs, 3)]));
cannam@95 162 ST(&(x[WS(vs, 2) + WS(rs, 4)]), T1B, ms, &(x[WS(vs, 2)]));
cannam@95 163 ST(&(x[WS(vs, 2) + WS(rs, 1)]), TD, ms, &(x[WS(vs, 2) + WS(rs, 1)]));
cannam@95 164 ST(&(x[WS(vs, 3) + WS(rs, 1)]), TE, ms, &(x[WS(vs, 3) + WS(rs, 1)]));
cannam@95 165 }
cannam@95 166 TP = VFMA(LDK(KP559016994), TO, TN);
cannam@95 167 TV = VFNMS(LDK(KP559016994), TO, TN);
cannam@95 168 T1f = VFNMS(LDK(KP559016994), T18, T17);
cannam@95 169 T19 = VFMA(LDK(KP559016994), T18, T17);
cannam@95 170 }
cannam@95 171 }
cannam@95 172 TY = BYTW(&(W[TWVL * 4]), VFMAI(TW, TV));
cannam@95 173 TX = BYTW(&(W[TWVL * 2]), VFNMSI(TW, TV));
cannam@95 174 {
cannam@95 175 V T1i, T1h, TU, TT;
cannam@95 176 T1i = BYTW(&(W[TWVL * 4]), VFMAI(T1g, T1f));
cannam@95 177 T1h = BYTW(&(W[TWVL * 2]), VFNMSI(T1g, T1f));
cannam@95 178 TU = BYTW(&(W[TWVL * 6]), VFNMSI(TS, TP));
cannam@95 179 TT = BYTW(&(W[0]), VFMAI(TS, TP));
cannam@95 180 {
cannam@95 181 V Tg, Tf, TA, Tz;
cannam@95 182 Tg = BYTW(&(W[TWVL * 6]), VFNMSI(Te, Tb));
cannam@95 183 Tf = BYTW(&(W[0]), VFMAI(Te, Tb));
cannam@95 184 TA = BYTW(&(W[TWVL * 6]), VFNMSI(Ty, Tv));
cannam@95 185 Tz = BYTW(&(W[0]), VFMAI(Ty, Tv));
cannam@95 186 {
cannam@95 187 V T1e, T1d, T1y, T1x;
cannam@95 188 T1e = BYTW(&(W[TWVL * 6]), VFNMSI(T1c, T19));
cannam@95 189 T1d = BYTW(&(W[0]), VFMAI(T1c, T19));
cannam@95 190 T1y = BYTW(&(W[TWVL * 6]), VFNMSI(T1w, T1t));
cannam@95 191 T1x = BYTW(&(W[0]), VFMAI(T1w, T1t));
cannam@95 192 ST(&(x[WS(vs, 3) + WS(rs, 2)]), TY, ms, &(x[WS(vs, 3)]));
cannam@95 193 ST(&(x[WS(vs, 2) + WS(rs, 2)]), TX, ms, &(x[WS(vs, 2)]));
cannam@95 194 ST(&(x[WS(vs, 3) + WS(rs, 3)]), T1i, ms, &(x[WS(vs, 3) + WS(rs, 1)]));
cannam@95 195 ST(&(x[WS(vs, 2) + WS(rs, 3)]), T1h, ms, &(x[WS(vs, 2) + WS(rs, 1)]));
cannam@95 196 ST(&(x[WS(vs, 4) + WS(rs, 2)]), TU, ms, &(x[WS(vs, 4)]));
cannam@95 197 ST(&(x[WS(vs, 1) + WS(rs, 2)]), TT, ms, &(x[WS(vs, 1)]));
cannam@95 198 ST(&(x[WS(vs, 4)]), Tg, ms, &(x[WS(vs, 4)]));
cannam@95 199 ST(&(x[WS(vs, 1)]), Tf, ms, &(x[WS(vs, 1)]));
cannam@95 200 ST(&(x[WS(vs, 4) + WS(rs, 1)]), TA, ms, &(x[WS(vs, 4) + WS(rs, 1)]));
cannam@95 201 ST(&(x[WS(vs, 1) + WS(rs, 1)]), Tz, ms, &(x[WS(vs, 1) + WS(rs, 1)]));
cannam@95 202 ST(&(x[WS(vs, 4) + WS(rs, 3)]), T1e, ms, &(x[WS(vs, 4) + WS(rs, 1)]));
cannam@95 203 ST(&(x[WS(vs, 1) + WS(rs, 3)]), T1d, ms, &(x[WS(vs, 1) + WS(rs, 1)]));
cannam@95 204 ST(&(x[WS(vs, 4) + WS(rs, 4)]), T1y, ms, &(x[WS(vs, 4)]));
cannam@95 205 ST(&(x[WS(vs, 1) + WS(rs, 4)]), T1x, ms, &(x[WS(vs, 1)]));
cannam@95 206 }
cannam@95 207 }
cannam@95 208 }
cannam@95 209 }
cannam@95 210 }
cannam@95 211 VLEAVE();
cannam@95 212 }
cannam@95 213
cannam@95 214 static const tw_instr twinstr[] = {
cannam@95 215 VTW(0, 1),
cannam@95 216 VTW(0, 2),
cannam@95 217 VTW(0, 3),
cannam@95 218 VTW(0, 4),
cannam@95 219 {TW_NEXT, VL, 0}
cannam@95 220 };
cannam@95 221
cannam@95 222 static const ct_desc desc = { 5, XSIMD_STRING("q1bv_5"), twinstr, &GENUS, {55, 50, 45, 0}, 0, 0, 0 };
cannam@95 223
cannam@95 224 void XSIMD(codelet_q1bv_5) (planner *p) {
cannam@95 225 X(kdft_difsq_register) (p, q1bv_5, &desc);
cannam@95 226 }
cannam@95 227 #else /* HAVE_FMA */
cannam@95 228
cannam@95 229 /* Generated by: ../../../genfft/gen_twidsq_c.native -simd -compact -variables 4 -pipeline-latency 8 -n 5 -dif -name q1bv_5 -include q1b.h -sign 1 */
cannam@95 230
cannam@95 231 /*
cannam@95 232 * This function contains 100 FP additions, 70 FP multiplications,
cannam@95 233 * (or, 85 additions, 55 multiplications, 15 fused multiply/add),
cannam@95 234 * 44 stack variables, 4 constants, and 50 memory accesses
cannam@95 235 */
cannam@95 236 #include "q1b.h"
cannam@95 237
cannam@95 238 static void q1bv_5(R *ri, R *ii, const R *W, stride rs, stride vs, INT mb, INT me, INT ms)
cannam@95 239 {
cannam@95 240 DVK(KP250000000, +0.250000000000000000000000000000000000000000000);
cannam@95 241 DVK(KP559016994, +0.559016994374947424102293417182819058860154590);
cannam@95 242 DVK(KP587785252, +0.587785252292473129168705954639072768597652438);
cannam@95 243 DVK(KP951056516, +0.951056516295153572116439333379382143405698634);
cannam@95 244 {
cannam@95 245 INT m;
cannam@95 246 R *x;
cannam@95 247 x = ii;
cannam@95 248 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(10, rs), MAKE_VOLATILE_STRIDE(10, vs)) {
cannam@95 249 V Tb, T7, Th, Ta, Tc, Td, T1t, T1p, T1z, T1s, T1u, T1v, Tv, Tr, TB;
cannam@95 250 V Tu, Tw, Tx, TP, TL, TV, TO, TQ, TR, T19, T15, T1f, T18, T1a, T1b;
cannam@95 251 {
cannam@95 252 V T6, T9, T3, T8;
cannam@95 253 Tb = LD(&(x[0]), ms, &(x[0]));
cannam@95 254 {
cannam@95 255 V T4, T5, T1, T2;
cannam@95 256 T4 = LD(&(x[WS(rs, 2)]), ms, &(x[0]));
cannam@95 257 T5 = LD(&(x[WS(rs, 3)]), ms, &(x[WS(rs, 1)]));
cannam@95 258 T6 = VSUB(T4, T5);
cannam@95 259 T9 = VADD(T4, T5);
cannam@95 260 T1 = LD(&(x[WS(rs, 1)]), ms, &(x[WS(rs, 1)]));
cannam@95 261 T2 = LD(&(x[WS(rs, 4)]), ms, &(x[0]));
cannam@95 262 T3 = VSUB(T1, T2);
cannam@95 263 T8 = VADD(T1, T2);
cannam@95 264 }
cannam@95 265 T7 = VBYI(VFMA(LDK(KP951056516), T3, VMUL(LDK(KP587785252), T6)));
cannam@95 266 Th = VBYI(VFNMS(LDK(KP951056516), T6, VMUL(LDK(KP587785252), T3)));
cannam@95 267 Ta = VMUL(LDK(KP559016994), VSUB(T8, T9));
cannam@95 268 Tc = VADD(T8, T9);
cannam@95 269 Td = VFNMS(LDK(KP250000000), Tc, Tb);
cannam@95 270 }
cannam@95 271 {
cannam@95 272 V T1o, T1r, T1l, T1q;
cannam@95 273 T1t = LD(&(x[WS(vs, 4)]), ms, &(x[WS(vs, 4)]));
cannam@95 274 {
cannam@95 275 V T1m, T1n, T1j, T1k;
cannam@95 276 T1m = LD(&(x[WS(vs, 4) + WS(rs, 2)]), ms, &(x[WS(vs, 4)]));
cannam@95 277 T1n = LD(&(x[WS(vs, 4) + WS(rs, 3)]), ms, &(x[WS(vs, 4) + WS(rs, 1)]));
cannam@95 278 T1o = VSUB(T1m, T1n);
cannam@95 279 T1r = VADD(T1m, T1n);
cannam@95 280 T1j = LD(&(x[WS(vs, 4) + WS(rs, 1)]), ms, &(x[WS(vs, 4) + WS(rs, 1)]));
cannam@95 281 T1k = LD(&(x[WS(vs, 4) + WS(rs, 4)]), ms, &(x[WS(vs, 4)]));
cannam@95 282 T1l = VSUB(T1j, T1k);
cannam@95 283 T1q = VADD(T1j, T1k);
cannam@95 284 }
cannam@95 285 T1p = VBYI(VFMA(LDK(KP951056516), T1l, VMUL(LDK(KP587785252), T1o)));
cannam@95 286 T1z = VBYI(VFNMS(LDK(KP951056516), T1o, VMUL(LDK(KP587785252), T1l)));
cannam@95 287 T1s = VMUL(LDK(KP559016994), VSUB(T1q, T1r));
cannam@95 288 T1u = VADD(T1q, T1r);
cannam@95 289 T1v = VFNMS(LDK(KP250000000), T1u, T1t);
cannam@95 290 }
cannam@95 291 {
cannam@95 292 V Tq, Tt, Tn, Ts;
cannam@95 293 Tv = LD(&(x[WS(vs, 1)]), ms, &(x[WS(vs, 1)]));
cannam@95 294 {
cannam@95 295 V To, Tp, Tl, Tm;
cannam@95 296 To = LD(&(x[WS(vs, 1) + WS(rs, 2)]), ms, &(x[WS(vs, 1)]));
cannam@95 297 Tp = LD(&(x[WS(vs, 1) + WS(rs, 3)]), ms, &(x[WS(vs, 1) + WS(rs, 1)]));
cannam@95 298 Tq = VSUB(To, Tp);
cannam@95 299 Tt = VADD(To, Tp);
cannam@95 300 Tl = LD(&(x[WS(vs, 1) + WS(rs, 1)]), ms, &(x[WS(vs, 1) + WS(rs, 1)]));
cannam@95 301 Tm = LD(&(x[WS(vs, 1) + WS(rs, 4)]), ms, &(x[WS(vs, 1)]));
cannam@95 302 Tn = VSUB(Tl, Tm);
cannam@95 303 Ts = VADD(Tl, Tm);
cannam@95 304 }
cannam@95 305 Tr = VBYI(VFMA(LDK(KP951056516), Tn, VMUL(LDK(KP587785252), Tq)));
cannam@95 306 TB = VBYI(VFNMS(LDK(KP951056516), Tq, VMUL(LDK(KP587785252), Tn)));
cannam@95 307 Tu = VMUL(LDK(KP559016994), VSUB(Ts, Tt));
cannam@95 308 Tw = VADD(Ts, Tt);
cannam@95 309 Tx = VFNMS(LDK(KP250000000), Tw, Tv);
cannam@95 310 }
cannam@95 311 {
cannam@95 312 V TK, TN, TH, TM;
cannam@95 313 TP = LD(&(x[WS(vs, 2)]), ms, &(x[WS(vs, 2)]));
cannam@95 314 {
cannam@95 315 V TI, TJ, TF, TG;
cannam@95 316 TI = LD(&(x[WS(vs, 2) + WS(rs, 2)]), ms, &(x[WS(vs, 2)]));
cannam@95 317 TJ = LD(&(x[WS(vs, 2) + WS(rs, 3)]), ms, &(x[WS(vs, 2) + WS(rs, 1)]));
cannam@95 318 TK = VSUB(TI, TJ);
cannam@95 319 TN = VADD(TI, TJ);
cannam@95 320 TF = LD(&(x[WS(vs, 2) + WS(rs, 1)]), ms, &(x[WS(vs, 2) + WS(rs, 1)]));
cannam@95 321 TG = LD(&(x[WS(vs, 2) + WS(rs, 4)]), ms, &(x[WS(vs, 2)]));
cannam@95 322 TH = VSUB(TF, TG);
cannam@95 323 TM = VADD(TF, TG);
cannam@95 324 }
cannam@95 325 TL = VBYI(VFMA(LDK(KP951056516), TH, VMUL(LDK(KP587785252), TK)));
cannam@95 326 TV = VBYI(VFNMS(LDK(KP951056516), TK, VMUL(LDK(KP587785252), TH)));
cannam@95 327 TO = VMUL(LDK(KP559016994), VSUB(TM, TN));
cannam@95 328 TQ = VADD(TM, TN);
cannam@95 329 TR = VFNMS(LDK(KP250000000), TQ, TP);
cannam@95 330 }
cannam@95 331 {
cannam@95 332 V T14, T17, T11, T16;
cannam@95 333 T19 = LD(&(x[WS(vs, 3)]), ms, &(x[WS(vs, 3)]));
cannam@95 334 {
cannam@95 335 V T12, T13, TZ, T10;
cannam@95 336 T12 = LD(&(x[WS(vs, 3) + WS(rs, 2)]), ms, &(x[WS(vs, 3)]));
cannam@95 337 T13 = LD(&(x[WS(vs, 3) + WS(rs, 3)]), ms, &(x[WS(vs, 3) + WS(rs, 1)]));
cannam@95 338 T14 = VSUB(T12, T13);
cannam@95 339 T17 = VADD(T12, T13);
cannam@95 340 TZ = LD(&(x[WS(vs, 3) + WS(rs, 1)]), ms, &(x[WS(vs, 3) + WS(rs, 1)]));
cannam@95 341 T10 = LD(&(x[WS(vs, 3) + WS(rs, 4)]), ms, &(x[WS(vs, 3)]));
cannam@95 342 T11 = VSUB(TZ, T10);
cannam@95 343 T16 = VADD(TZ, T10);
cannam@95 344 }
cannam@95 345 T15 = VBYI(VFMA(LDK(KP951056516), T11, VMUL(LDK(KP587785252), T14)));
cannam@95 346 T1f = VBYI(VFNMS(LDK(KP951056516), T14, VMUL(LDK(KP587785252), T11)));
cannam@95 347 T18 = VMUL(LDK(KP559016994), VSUB(T16, T17));
cannam@95 348 T1a = VADD(T16, T17);
cannam@95 349 T1b = VFNMS(LDK(KP250000000), T1a, T19);
cannam@95 350 }
cannam@95 351 ST(&(x[0]), VADD(Tb, Tc), ms, &(x[0]));
cannam@95 352 ST(&(x[WS(rs, 4)]), VADD(T1t, T1u), ms, &(x[0]));
cannam@95 353 ST(&(x[WS(rs, 2)]), VADD(TP, TQ), ms, &(x[0]));
cannam@95 354 ST(&(x[WS(rs, 3)]), VADD(T19, T1a), ms, &(x[WS(rs, 1)]));
cannam@95 355 ST(&(x[WS(rs, 1)]), VADD(Tv, Tw), ms, &(x[WS(rs, 1)]));
cannam@95 356 {
cannam@95 357 V Tj, Tk, Ti, T1B, T1C, T1A;
cannam@95 358 Ti = VSUB(Td, Ta);
cannam@95 359 Tj = BYTW(&(W[TWVL * 2]), VADD(Th, Ti));
cannam@95 360 Tk = BYTW(&(W[TWVL * 4]), VSUB(Ti, Th));
cannam@95 361 ST(&(x[WS(vs, 2)]), Tj, ms, &(x[WS(vs, 2)]));
cannam@95 362 ST(&(x[WS(vs, 3)]), Tk, ms, &(x[WS(vs, 3)]));
cannam@95 363 T1A = VSUB(T1v, T1s);
cannam@95 364 T1B = BYTW(&(W[TWVL * 2]), VADD(T1z, T1A));
cannam@95 365 T1C = BYTW(&(W[TWVL * 4]), VSUB(T1A, T1z));
cannam@95 366 ST(&(x[WS(vs, 2) + WS(rs, 4)]), T1B, ms, &(x[WS(vs, 2)]));
cannam@95 367 ST(&(x[WS(vs, 3) + WS(rs, 4)]), T1C, ms, &(x[WS(vs, 3)]));
cannam@95 368 }
cannam@95 369 {
cannam@95 370 V T1h, T1i, T1g, TD, TE, TC;
cannam@95 371 T1g = VSUB(T1b, T18);
cannam@95 372 T1h = BYTW(&(W[TWVL * 2]), VADD(T1f, T1g));
cannam@95 373 T1i = BYTW(&(W[TWVL * 4]), VSUB(T1g, T1f));
cannam@95 374 ST(&(x[WS(vs, 2) + WS(rs, 3)]), T1h, ms, &(x[WS(vs, 2) + WS(rs, 1)]));
cannam@95 375 ST(&(x[WS(vs, 3) + WS(rs, 3)]), T1i, ms, &(x[WS(vs, 3) + WS(rs, 1)]));
cannam@95 376 TC = VSUB(Tx, Tu);
cannam@95 377 TD = BYTW(&(W[TWVL * 2]), VADD(TB, TC));
cannam@95 378 TE = BYTW(&(W[TWVL * 4]), VSUB(TC, TB));
cannam@95 379 ST(&(x[WS(vs, 2) + WS(rs, 1)]), TD, ms, &(x[WS(vs, 2) + WS(rs, 1)]));
cannam@95 380 ST(&(x[WS(vs, 3) + WS(rs, 1)]), TE, ms, &(x[WS(vs, 3) + WS(rs, 1)]));
cannam@95 381 }
cannam@95 382 {
cannam@95 383 V TX, TY, TW, TT, TU, TS;
cannam@95 384 TW = VSUB(TR, TO);
cannam@95 385 TX = BYTW(&(W[TWVL * 2]), VADD(TV, TW));
cannam@95 386 TY = BYTW(&(W[TWVL * 4]), VSUB(TW, TV));
cannam@95 387 ST(&(x[WS(vs, 2) + WS(rs, 2)]), TX, ms, &(x[WS(vs, 2)]));
cannam@95 388 ST(&(x[WS(vs, 3) + WS(rs, 2)]), TY, ms, &(x[WS(vs, 3)]));
cannam@95 389 TS = VADD(TO, TR);
cannam@95 390 TT = BYTW(&(W[0]), VADD(TL, TS));
cannam@95 391 TU = BYTW(&(W[TWVL * 6]), VSUB(TS, TL));
cannam@95 392 ST(&(x[WS(vs, 1) + WS(rs, 2)]), TT, ms, &(x[WS(vs, 1)]));
cannam@95 393 ST(&(x[WS(vs, 4) + WS(rs, 2)]), TU, ms, &(x[WS(vs, 4)]));
cannam@95 394 }
cannam@95 395 {
cannam@95 396 V Tf, Tg, Te, Tz, TA, Ty;
cannam@95 397 Te = VADD(Ta, Td);
cannam@95 398 Tf = BYTW(&(W[0]), VADD(T7, Te));
cannam@95 399 Tg = BYTW(&(W[TWVL * 6]), VSUB(Te, T7));
cannam@95 400 ST(&(x[WS(vs, 1)]), Tf, ms, &(x[WS(vs, 1)]));
cannam@95 401 ST(&(x[WS(vs, 4)]), Tg, ms, &(x[WS(vs, 4)]));
cannam@95 402 Ty = VADD(Tu, Tx);
cannam@95 403 Tz = BYTW(&(W[0]), VADD(Tr, Ty));
cannam@95 404 TA = BYTW(&(W[TWVL * 6]), VSUB(Ty, Tr));
cannam@95 405 ST(&(x[WS(vs, 1) + WS(rs, 1)]), Tz, ms, &(x[WS(vs, 1) + WS(rs, 1)]));
cannam@95 406 ST(&(x[WS(vs, 4) + WS(rs, 1)]), TA, ms, &(x[WS(vs, 4) + WS(rs, 1)]));
cannam@95 407 }
cannam@95 408 {
cannam@95 409 V T1d, T1e, T1c, T1x, T1y, T1w;
cannam@95 410 T1c = VADD(T18, T1b);
cannam@95 411 T1d = BYTW(&(W[0]), VADD(T15, T1c));
cannam@95 412 T1e = BYTW(&(W[TWVL * 6]), VSUB(T1c, T15));
cannam@95 413 ST(&(x[WS(vs, 1) + WS(rs, 3)]), T1d, ms, &(x[WS(vs, 1) + WS(rs, 1)]));
cannam@95 414 ST(&(x[WS(vs, 4) + WS(rs, 3)]), T1e, ms, &(x[WS(vs, 4) + WS(rs, 1)]));
cannam@95 415 T1w = VADD(T1s, T1v);
cannam@95 416 T1x = BYTW(&(W[0]), VADD(T1p, T1w));
cannam@95 417 T1y = BYTW(&(W[TWVL * 6]), VSUB(T1w, T1p));
cannam@95 418 ST(&(x[WS(vs, 1) + WS(rs, 4)]), T1x, ms, &(x[WS(vs, 1)]));
cannam@95 419 ST(&(x[WS(vs, 4) + WS(rs, 4)]), T1y, ms, &(x[WS(vs, 4)]));
cannam@95 420 }
cannam@95 421 }
cannam@95 422 }
cannam@95 423 VLEAVE();
cannam@95 424 }
cannam@95 425
cannam@95 426 static const tw_instr twinstr[] = {
cannam@95 427 VTW(0, 1),
cannam@95 428 VTW(0, 2),
cannam@95 429 VTW(0, 3),
cannam@95 430 VTW(0, 4),
cannam@95 431 {TW_NEXT, VL, 0}
cannam@95 432 };
cannam@95 433
cannam@95 434 static const ct_desc desc = { 5, XSIMD_STRING("q1bv_5"), twinstr, &GENUS, {85, 55, 15, 0}, 0, 0, 0 };
cannam@95 435
cannam@95 436 void XSIMD(codelet_q1bv_5) (planner *p) {
cannam@95 437 X(kdft_difsq_register) (p, q1bv_5, &desc);
cannam@95 438 }
cannam@95 439 #endif /* HAVE_FMA */