Mercurial > hg > sv-dependency-builds

annotate src/fftw-3.3.8/dft/simd/common/q1bv_5.c @ 84:08ae793730bd

Find changesets by keywords (author, files, the commit message), revision number or hash, or revset expression.
Add null config files
author Chris Cannam
date Mon, 02 Mar 2020 14:03:47 +0000
parents d0c2a83c1364
children
rev   line source
Chris@82 1 /*
Chris@82 2 * Copyright (c) 2003, 2007-14 Matteo Frigo
Chris@82 3 * Copyright (c) 2003, 2007-14 Massachusetts Institute of Technology
Chris@82 4 *
Chris@82 5 * This program is free software; you can redistribute it and/or modify
Chris@82 6 * it under the terms of the GNU General Public License as published by
Chris@82 7 * the Free Software Foundation; either version 2 of the License, or
Chris@82 8 * (at your option) any later version.
Chris@82 9 *
Chris@82 10 * This program is distributed in the hope that it will be useful,
Chris@82 11 * but WITHOUT ANY WARRANTY; without even the implied warranty of
Chris@82 12 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
Chris@82 13 * GNU General Public License for more details.
Chris@82 14 *
Chris@82 15 * You should have received a copy of the GNU General Public License
Chris@82 16 * along with this program; if not, write to the Free Software
Chris@82 17 * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
Chris@82 18 *
Chris@82 19 */
Chris@82 20
Chris@82 21 /* This file was automatically generated --- DO NOT EDIT */
Chris@82 22 /* Generated on Thu May 24 08:06:14 EDT 2018 */
Chris@82 23
Chris@82 24 #include "dft/codelet-dft.h"
Chris@82 25
Chris@82 26 #if defined(ARCH_PREFERS_FMA) || defined(ISA_EXTENSION_PREFERS_FMA)
Chris@82 27
Chris@82 28 /* Generated by: ../../../genfft/gen_twidsq_c.native -fma -simd -compact -variables 4 -pipeline-latency 8 -n 5 -dif -name q1bv_5 -include dft/simd/q1b.h -sign 1 */
Chris@82 29
Chris@82 30 /*
Chris@82 31 * This function contains 100 FP additions, 95 FP multiplications,
Chris@82 32 * (or, 55 additions, 50 multiplications, 45 fused multiply/add),
Chris@82 33 * 44 stack variables, 4 constants, and 50 memory accesses
Chris@82 34 */
Chris@82 35 #include "dft/simd/q1b.h"
Chris@82 36
Chris@82 37 static void q1bv_5(R *ri, R *ii, const R *W, stride rs, stride vs, INT mb, INT me, INT ms)
Chris@82 38 {
Chris@82 39 DVK(KP559016994, +0.559016994374947424102293417182819058860154590);
Chris@82 40 DVK(KP250000000, +0.250000000000000000000000000000000000000000000);
Chris@82 41 DVK(KP618033988, +0.618033988749894848204586834365638117720309180);
Chris@82 42 DVK(KP951056516, +0.951056516295153572116439333379382143405698634);
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) * 8)); m < me; m = m + VL, x = x + (VL * ms), W = W + (TWVL * 8), MAKE_VOLATILE_STRIDE(10, rs), MAKE_VOLATILE_STRIDE(10, vs)) {
Chris@82 48 V T1, Ta, Ti, Te, T8, T9, T1j, T1s, T1A, T1w, T1q, T1r, Tl, Tu, TC;
Chris@82 49 V Ty, Ts, Tt, TF, TO, TW, TS, TM, TN, TZ, T18, T1g, T1c, T16, T17;
Chris@82 50 {
Chris@82 51 V T7, Td, T4, Tc;
Chris@82 52 T1 = LD(&(x[0]), ms, &(x[0]));
Chris@82 53 {
Chris@82 54 V T5, T6, T2, T3;
Chris@82 55 T5 = LD(&(x[WS(rs, 2)]), ms, &(x[0]));
Chris@82 56 T6 = LD(&(x[WS(rs, 3)]), ms, &(x[WS(rs, 1)]));
Chris@82 57 T7 = VADD(T5, T6);
Chris@82 58 Td = VSUB(T5, T6);
Chris@82 59 T2 = LD(&(x[WS(rs, 1)]), ms, &(x[WS(rs, 1)]));
Chris@82 60 T3 = LD(&(x[WS(rs, 4)]), ms, &(x[0]));
Chris@82 61 T4 = VADD(T2, T3);
Chris@82 62 Tc = VSUB(T2, T3);
Chris@82 63 }
Chris@82 64 Ta = VSUB(T4, T7);
Chris@82 65 Ti = VMUL(LDK(KP951056516), VFNMS(LDK(KP618033988), Tc, Td));
Chris@82 66 Te = VMUL(LDK(KP951056516), VFMA(LDK(KP618033988), Td, Tc));
Chris@82 67 T8 = VADD(T4, T7);
Chris@82 68 T9 = VFNMS(LDK(KP250000000), T8, T1);
Chris@82 69 }
Chris@82 70 {
Chris@82 71 V T1p, T1v, T1m, T1u;
Chris@82 72 T1j = LD(&(x[WS(vs, 4)]), ms, &(x[WS(vs, 4)]));
Chris@82 73 {
Chris@82 74 V T1n, T1o, T1k, T1l;
Chris@82 75 T1n = LD(&(x[WS(vs, 4) + WS(rs, 2)]), ms, &(x[WS(vs, 4)]));
Chris@82 76 T1o = LD(&(x[WS(vs, 4) + WS(rs, 3)]), ms, &(x[WS(vs, 4) + WS(rs, 1)]));
Chris@82 77 T1p = VADD(T1n, T1o);
Chris@82 78 T1v = VSUB(T1n, T1o);
Chris@82 79 T1k = LD(&(x[WS(vs, 4) + WS(rs, 1)]), ms, &(x[WS(vs, 4) + WS(rs, 1)]));
Chris@82 80 T1l = LD(&(x[WS(vs, 4) + WS(rs, 4)]), ms, &(x[WS(vs, 4)]));
Chris@82 81 T1m = VADD(T1k, T1l);
Chris@82 82 T1u = VSUB(T1k, T1l);
Chris@82 83 }
Chris@82 84 T1s = VSUB(T1m, T1p);
Chris@82 85 T1A = VMUL(LDK(KP951056516), VFNMS(LDK(KP618033988), T1u, T1v));
Chris@82 86 T1w = VMUL(LDK(KP951056516), VFMA(LDK(KP618033988), T1v, T1u));
Chris@82 87 T1q = VADD(T1m, T1p);
Chris@82 88 T1r = VFNMS(LDK(KP250000000), T1q, T1j);
Chris@82 89 }
Chris@82 90 {
Chris@82 91 V Tr, Tx, To, Tw;
Chris@82 92 Tl = LD(&(x[WS(vs, 1)]), ms, &(x[WS(vs, 1)]));
Chris@82 93 {
Chris@82 94 V Tp, Tq, Tm, Tn;
Chris@82 95 Tp = LD(&(x[WS(vs, 1) + WS(rs, 2)]), ms, &(x[WS(vs, 1)]));
Chris@82 96 Tq = LD(&(x[WS(vs, 1) + WS(rs, 3)]), ms, &(x[WS(vs, 1) + WS(rs, 1)]));
Chris@82 97 Tr = VADD(Tp, Tq);
Chris@82 98 Tx = VSUB(Tp, Tq);
Chris@82 99 Tm = LD(&(x[WS(vs, 1) + WS(rs, 1)]), ms, &(x[WS(vs, 1) + WS(rs, 1)]));
Chris@82 100 Tn = LD(&(x[WS(vs, 1) + WS(rs, 4)]), ms, &(x[WS(vs, 1)]));
Chris@82 101 To = VADD(Tm, Tn);
Chris@82 102 Tw = VSUB(Tm, Tn);
Chris@82 103 }
Chris@82 104 Tu = VSUB(To, Tr);
Chris@82 105 TC = VMUL(LDK(KP951056516), VFNMS(LDK(KP618033988), Tw, Tx));
Chris@82 106 Ty = VMUL(LDK(KP951056516), VFMA(LDK(KP618033988), Tx, Tw));
Chris@82 107 Ts = VADD(To, Tr);
Chris@82 108 Tt = VFNMS(LDK(KP250000000), Ts, Tl);
Chris@82 109 }
Chris@82 110 {
Chris@82 111 V TL, TR, TI, TQ;
Chris@82 112 TF = LD(&(x[WS(vs, 2)]), ms, &(x[WS(vs, 2)]));
Chris@82 113 {
Chris@82 114 V TJ, TK, TG, TH;
Chris@82 115 TJ = LD(&(x[WS(vs, 2) + WS(rs, 2)]), ms, &(x[WS(vs, 2)]));
Chris@82 116 TK = LD(&(x[WS(vs, 2) + WS(rs, 3)]), ms, &(x[WS(vs, 2) + WS(rs, 1)]));
Chris@82 117 TL = VADD(TJ, TK);
Chris@82 118 TR = VSUB(TJ, TK);
Chris@82 119 TG = LD(&(x[WS(vs, 2) + WS(rs, 1)]), ms, &(x[WS(vs, 2) + WS(rs, 1)]));
Chris@82 120 TH = LD(&(x[WS(vs, 2) + WS(rs, 4)]), ms, &(x[WS(vs, 2)]));
Chris@82 121 TI = VADD(TG, TH);
Chris@82 122 TQ = VSUB(TG, TH);
Chris@82 123 }
Chris@82 124 TO = VSUB(TI, TL);
Chris@82 125 TW = VMUL(LDK(KP951056516), VFNMS(LDK(KP618033988), TQ, TR));
Chris@82 126 TS = VMUL(LDK(KP951056516), VFMA(LDK(KP618033988), TR, TQ));
Chris@82 127 TM = VADD(TI, TL);
Chris@82 128 TN = VFNMS(LDK(KP250000000), TM, TF);
Chris@82 129 }
Chris@82 130 {
Chris@82 131 V T15, T1b, T12, T1a;
Chris@82 132 TZ = LD(&(x[WS(vs, 3)]), ms, &(x[WS(vs, 3)]));
Chris@82 133 {
Chris@82 134 V T13, T14, T10, T11;
Chris@82 135 T13 = LD(&(x[WS(vs, 3) + WS(rs, 2)]), ms, &(x[WS(vs, 3)]));
Chris@82 136 T14 = LD(&(x[WS(vs, 3) + WS(rs, 3)]), ms, &(x[WS(vs, 3) + WS(rs, 1)]));
Chris@82 137 T15 = VADD(T13, T14);
Chris@82 138 T1b = VSUB(T13, T14);
Chris@82 139 T10 = LD(&(x[WS(vs, 3) + WS(rs, 1)]), ms, &(x[WS(vs, 3) + WS(rs, 1)]));
Chris@82 140 T11 = LD(&(x[WS(vs, 3) + WS(rs, 4)]), ms, &(x[WS(vs, 3)]));
Chris@82 141 T12 = VADD(T10, T11);
Chris@82 142 T1a = VSUB(T10, T11);
Chris@82 143 }
Chris@82 144 T18 = VSUB(T12, T15);
Chris@82 145 T1g = VMUL(LDK(KP951056516), VFNMS(LDK(KP618033988), T1a, T1b));
Chris@82 146 T1c = VMUL(LDK(KP951056516), VFMA(LDK(KP618033988), T1b, T1a));
Chris@82 147 T16 = VADD(T12, T15);
Chris@82 148 T17 = VFNMS(LDK(KP250000000), T16, TZ);
Chris@82 149 }
Chris@82 150 ST(&(x[0]), VADD(T1, T8), ms, &(x[0]));
Chris@82 151 ST(&(x[WS(rs, 4)]), VADD(T1j, T1q), ms, &(x[0]));
Chris@82 152 ST(&(x[WS(rs, 2)]), VADD(TF, TM), ms, &(x[0]));
Chris@82 153 ST(&(x[WS(rs, 3)]), VADD(TZ, T16), ms, &(x[WS(rs, 1)]));
Chris@82 154 ST(&(x[WS(rs, 1)]), VADD(Tl, Ts), ms, &(x[WS(rs, 1)]));
Chris@82 155 {
Chris@82 156 V Tj, Tk, Th, T1B, T1C, T1z;
Chris@82 157 Th = VFNMS(LDK(KP559016994), Ta, T9);
Chris@82 158 Tj = BYTW(&(W[TWVL * 2]), VFNMSI(Ti, Th));
Chris@82 159 Tk = BYTW(&(W[TWVL * 4]), VFMAI(Ti, Th));
Chris@82 160 ST(&(x[WS(vs, 2)]), Tj, ms, &(x[WS(vs, 2)]));
Chris@82 161 ST(&(x[WS(vs, 3)]), Tk, ms, &(x[WS(vs, 3)]));
Chris@82 162 T1z = VFNMS(LDK(KP559016994), T1s, T1r);
Chris@82 163 T1B = BYTW(&(W[TWVL * 2]), VFNMSI(T1A, T1z));
Chris@82 164 T1C = BYTW(&(W[TWVL * 4]), VFMAI(T1A, T1z));
Chris@82 165 ST(&(x[WS(vs, 2) + WS(rs, 4)]), T1B, ms, &(x[WS(vs, 2)]));
Chris@82 166 ST(&(x[WS(vs, 3) + WS(rs, 4)]), T1C, ms, &(x[WS(vs, 3)]));
Chris@82 167 }
Chris@82 168 {
Chris@82 169 V T1h, T1i, T1f, TD, TE, TB;
Chris@82 170 T1f = VFNMS(LDK(KP559016994), T18, T17);
Chris@82 171 T1h = BYTW(&(W[TWVL * 2]), VFNMSI(T1g, T1f));
Chris@82 172 T1i = BYTW(&(W[TWVL * 4]), VFMAI(T1g, T1f));
Chris@82 173 ST(&(x[WS(vs, 2) + WS(rs, 3)]), T1h, ms, &(x[WS(vs, 2) + WS(rs, 1)]));
Chris@82 174 ST(&(x[WS(vs, 3) + WS(rs, 3)]), T1i, ms, &(x[WS(vs, 3) + WS(rs, 1)]));
Chris@82 175 TB = VFNMS(LDK(KP559016994), Tu, Tt);
Chris@82 176 TD = BYTW(&(W[TWVL * 2]), VFNMSI(TC, TB));
Chris@82 177 TE = BYTW(&(W[TWVL * 4]), VFMAI(TC, TB));
Chris@82 178 ST(&(x[WS(vs, 2) + WS(rs, 1)]), TD, ms, &(x[WS(vs, 2) + WS(rs, 1)]));
Chris@82 179 ST(&(x[WS(vs, 3) + WS(rs, 1)]), TE, ms, &(x[WS(vs, 3) + WS(rs, 1)]));
Chris@82 180 }
Chris@82 181 {
Chris@82 182 V TX, TY, TV, TT, TU, TP;
Chris@82 183 TV = VFNMS(LDK(KP559016994), TO, TN);
Chris@82 184 TX = BYTW(&(W[TWVL * 2]), VFNMSI(TW, TV));
Chris@82 185 TY = BYTW(&(W[TWVL * 4]), VFMAI(TW, TV));
Chris@82 186 ST(&(x[WS(vs, 2) + WS(rs, 2)]), TX, ms, &(x[WS(vs, 2)]));
Chris@82 187 ST(&(x[WS(vs, 3) + WS(rs, 2)]), TY, ms, &(x[WS(vs, 3)]));
Chris@82 188 TP = VFMA(LDK(KP559016994), TO, TN);
Chris@82 189 TT = BYTW(&(W[0]), VFMAI(TS, TP));
Chris@82 190 TU = BYTW(&(W[TWVL * 6]), VFNMSI(TS, TP));
Chris@82 191 ST(&(x[WS(vs, 1) + WS(rs, 2)]), TT, ms, &(x[WS(vs, 1)]));
Chris@82 192 ST(&(x[WS(vs, 4) + WS(rs, 2)]), TU, ms, &(x[WS(vs, 4)]));
Chris@82 193 }
Chris@82 194 {
Chris@82 195 V Tf, Tg, Tb, Tz, TA, Tv;
Chris@82 196 Tb = VFMA(LDK(KP559016994), Ta, T9);
Chris@82 197 Tf = BYTW(&(W[0]), VFMAI(Te, Tb));
Chris@82 198 Tg = BYTW(&(W[TWVL * 6]), VFNMSI(Te, Tb));
Chris@82 199 ST(&(x[WS(vs, 1)]), Tf, ms, &(x[WS(vs, 1)]));
Chris@82 200 ST(&(x[WS(vs, 4)]), Tg, ms, &(x[WS(vs, 4)]));
Chris@82 201 Tv = VFMA(LDK(KP559016994), Tu, Tt);
Chris@82 202 Tz = BYTW(&(W[0]), VFMAI(Ty, Tv));
Chris@82 203 TA = BYTW(&(W[TWVL * 6]), VFNMSI(Ty, Tv));
Chris@82 204 ST(&(x[WS(vs, 1) + WS(rs, 1)]), Tz, ms, &(x[WS(vs, 1) + WS(rs, 1)]));
Chris@82 205 ST(&(x[WS(vs, 4) + WS(rs, 1)]), TA, ms, &(x[WS(vs, 4) + WS(rs, 1)]));
Chris@82 206 }
Chris@82 207 {
Chris@82 208 V T1d, T1e, T19, T1x, T1y, T1t;
Chris@82 209 T19 = VFMA(LDK(KP559016994), T18, T17);
Chris@82 210 T1d = BYTW(&(W[0]), VFMAI(T1c, T19));
Chris@82 211 T1e = BYTW(&(W[TWVL * 6]), VFNMSI(T1c, T19));
Chris@82 212 ST(&(x[WS(vs, 1) + WS(rs, 3)]), T1d, ms, &(x[WS(vs, 1) + WS(rs, 1)]));
Chris@82 213 ST(&(x[WS(vs, 4) + WS(rs, 3)]), T1e, ms, &(x[WS(vs, 4) + WS(rs, 1)]));
Chris@82 214 T1t = VFMA(LDK(KP559016994), T1s, T1r);
Chris@82 215 T1x = BYTW(&(W[0]), VFMAI(T1w, T1t));
Chris@82 216 T1y = BYTW(&(W[TWVL * 6]), VFNMSI(T1w, T1t));
Chris@82 217 ST(&(x[WS(vs, 1) + WS(rs, 4)]), T1x, ms, &(x[WS(vs, 1)]));
Chris@82 218 ST(&(x[WS(vs, 4) + WS(rs, 4)]), T1y, ms, &(x[WS(vs, 4)]));
Chris@82 219 }
Chris@82 220 }
Chris@82 221 }
Chris@82 222 VLEAVE();
Chris@82 223 }
Chris@82 224
Chris@82 225 static const tw_instr twinstr[] = {
Chris@82 226 VTW(0, 1),
Chris@82 227 VTW(0, 2),
Chris@82 228 VTW(0, 3),
Chris@82 229 VTW(0, 4),
Chris@82 230 {TW_NEXT, VL, 0}
Chris@82 231 };
Chris@82 232
Chris@82 233 static const ct_desc desc = { 5, XSIMD_STRING("q1bv_5"), twinstr, &GENUS, {55, 50, 45, 0}, 0, 0, 0 };
Chris@82 234
Chris@82 235 void XSIMD(codelet_q1bv_5) (planner *p) {
Chris@82 236 X(kdft_difsq_register) (p, q1bv_5, &desc);
Chris@82 237 }
Chris@82 238 #else
Chris@82 239
Chris@82 240 /* Generated by: ../../../genfft/gen_twidsq_c.native -simd -compact -variables 4 -pipeline-latency 8 -n 5 -dif -name q1bv_5 -include dft/simd/q1b.h -sign 1 */
Chris@82 241
Chris@82 242 /*
Chris@82 243 * This function contains 100 FP additions, 70 FP multiplications,
Chris@82 244 * (or, 85 additions, 55 multiplications, 15 fused multiply/add),
Chris@82 245 * 44 stack variables, 4 constants, and 50 memory accesses
Chris@82 246 */
Chris@82 247 #include "dft/simd/q1b.h"
Chris@82 248
Chris@82 249 static void q1bv_5(R *ri, R *ii, const R *W, stride rs, stride vs, INT mb, INT me, INT ms)
Chris@82 250 {
Chris@82 251 DVK(KP250000000, +0.250000000000000000000000000000000000000000000);
Chris@82 252 DVK(KP559016994, +0.559016994374947424102293417182819058860154590);
Chris@82 253 DVK(KP587785252, +0.587785252292473129168705954639072768597652438);
Chris@82 254 DVK(KP951056516, +0.951056516295153572116439333379382143405698634);
Chris@82 255 {
Chris@82 256 INT m;
Chris@82 257 R *x;
Chris@82 258 x = ii;
Chris@82 259 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)) {
Chris@82 260 V Tb, T7, Th, Ta, Tc, Td, T1t, T1p, T1z, T1s, T1u, T1v, Tv, Tr, TB;
Chris@82 261 V Tu, Tw, Tx, TP, TL, TV, TO, TQ, TR, T19, T15, T1f, T18, T1a, T1b;
Chris@82 262 {
Chris@82 263 V T6, T9, T3, T8;
Chris@82 264 Tb = LD(&(x[0]), ms, &(x[0]));
Chris@82 265 {
Chris@82 266 V T4, T5, T1, T2;
Chris@82 267 T4 = LD(&(x[WS(rs, 2)]), ms, &(x[0]));
Chris@82 268 T5 = LD(&(x[WS(rs, 3)]), ms, &(x[WS(rs, 1)]));
Chris@82 269 T6 = VSUB(T4, T5);
Chris@82 270 T9 = VADD(T4, T5);
Chris@82 271 T1 = LD(&(x[WS(rs, 1)]), ms, &(x[WS(rs, 1)]));
Chris@82 272 T2 = LD(&(x[WS(rs, 4)]), ms, &(x[0]));
Chris@82 273 T3 = VSUB(T1, T2);
Chris@82 274 T8 = VADD(T1, T2);
Chris@82 275 }
Chris@82 276 T7 = VBYI(VFMA(LDK(KP951056516), T3, VMUL(LDK(KP587785252), T6)));
Chris@82 277 Th = VBYI(VFNMS(LDK(KP951056516), T6, VMUL(LDK(KP587785252), T3)));
Chris@82 278 Ta = VMUL(LDK(KP559016994), VSUB(T8, T9));
Chris@82 279 Tc = VADD(T8, T9);
Chris@82 280 Td = VFNMS(LDK(KP250000000), Tc, Tb);
Chris@82 281 }
Chris@82 282 {
Chris@82 283 V T1o, T1r, T1l, T1q;
Chris@82 284 T1t = LD(&(x[WS(vs, 4)]), ms, &(x[WS(vs, 4)]));
Chris@82 285 {
Chris@82 286 V T1m, T1n, T1j, T1k;
Chris@82 287 T1m = LD(&(x[WS(vs, 4) + WS(rs, 2)]), ms, &(x[WS(vs, 4)]));
Chris@82 288 T1n = LD(&(x[WS(vs, 4) + WS(rs, 3)]), ms, &(x[WS(vs, 4) + WS(rs, 1)]));
Chris@82 289 T1o = VSUB(T1m, T1n);
Chris@82 290 T1r = VADD(T1m, T1n);
Chris@82 291 T1j = LD(&(x[WS(vs, 4) + WS(rs, 1)]), ms, &(x[WS(vs, 4) + WS(rs, 1)]));
Chris@82 292 T1k = LD(&(x[WS(vs, 4) + WS(rs, 4)]), ms, &(x[WS(vs, 4)]));
Chris@82 293 T1l = VSUB(T1j, T1k);
Chris@82 294 T1q = VADD(T1j, T1k);
Chris@82 295 }
Chris@82 296 T1p = VBYI(VFMA(LDK(KP951056516), T1l, VMUL(LDK(KP587785252), T1o)));
Chris@82 297 T1z = VBYI(VFNMS(LDK(KP951056516), T1o, VMUL(LDK(KP587785252), T1l)));
Chris@82 298 T1s = VMUL(LDK(KP559016994), VSUB(T1q, T1r));
Chris@82 299 T1u = VADD(T1q, T1r);
Chris@82 300 T1v = VFNMS(LDK(KP250000000), T1u, T1t);
Chris@82 301 }
Chris@82 302 {
Chris@82 303 V Tq, Tt, Tn, Ts;
Chris@82 304 Tv = LD(&(x[WS(vs, 1)]), ms, &(x[WS(vs, 1)]));
Chris@82 305 {
Chris@82 306 V To, Tp, Tl, Tm;
Chris@82 307 To = LD(&(x[WS(vs, 1) + WS(rs, 2)]), ms, &(x[WS(vs, 1)]));
Chris@82 308 Tp = LD(&(x[WS(vs, 1) + WS(rs, 3)]), ms, &(x[WS(vs, 1) + WS(rs, 1)]));
Chris@82 309 Tq = VSUB(To, Tp);
Chris@82 310 Tt = VADD(To, Tp);
Chris@82 311 Tl = LD(&(x[WS(vs, 1) + WS(rs, 1)]), ms, &(x[WS(vs, 1) + WS(rs, 1)]));
Chris@82 312 Tm = LD(&(x[WS(vs, 1) + WS(rs, 4)]), ms, &(x[WS(vs, 1)]));
Chris@82 313 Tn = VSUB(Tl, Tm);
Chris@82 314 Ts = VADD(Tl, Tm);
Chris@82 315 }
Chris@82 316 Tr = VBYI(VFMA(LDK(KP951056516), Tn, VMUL(LDK(KP587785252), Tq)));
Chris@82 317 TB = VBYI(VFNMS(LDK(KP951056516), Tq, VMUL(LDK(KP587785252), Tn)));
Chris@82 318 Tu = VMUL(LDK(KP559016994), VSUB(Ts, Tt));
Chris@82 319 Tw = VADD(Ts, Tt);
Chris@82 320 Tx = VFNMS(LDK(KP250000000), Tw, Tv);
Chris@82 321 }
Chris@82 322 {
Chris@82 323 V TK, TN, TH, TM;
Chris@82 324 TP = LD(&(x[WS(vs, 2)]), ms, &(x[WS(vs, 2)]));
Chris@82 325 {
Chris@82 326 V TI, TJ, TF, TG;
Chris@82 327 TI = LD(&(x[WS(vs, 2) + WS(rs, 2)]), ms, &(x[WS(vs, 2)]));
Chris@82 328 TJ = LD(&(x[WS(vs, 2) + WS(rs, 3)]), ms, &(x[WS(vs, 2) + WS(rs, 1)]));
Chris@82 329 TK = VSUB(TI, TJ);
Chris@82 330 TN = VADD(TI, TJ);
Chris@82 331 TF = LD(&(x[WS(vs, 2) + WS(rs, 1)]), ms, &(x[WS(vs, 2) + WS(rs, 1)]));
Chris@82 332 TG = LD(&(x[WS(vs, 2) + WS(rs, 4)]), ms, &(x[WS(vs, 2)]));
Chris@82 333 TH = VSUB(TF, TG);
Chris@82 334 TM = VADD(TF, TG);
Chris@82 335 }
Chris@82 336 TL = VBYI(VFMA(LDK(KP951056516), TH, VMUL(LDK(KP587785252), TK)));
Chris@82 337 TV = VBYI(VFNMS(LDK(KP951056516), TK, VMUL(LDK(KP587785252), TH)));
Chris@82 338 TO = VMUL(LDK(KP559016994), VSUB(TM, TN));
Chris@82 339 TQ = VADD(TM, TN);
Chris@82 340 TR = VFNMS(LDK(KP250000000), TQ, TP);
Chris@82 341 }
Chris@82 342 {
Chris@82 343 V T14, T17, T11, T16;
Chris@82 344 T19 = LD(&(x[WS(vs, 3)]), ms, &(x[WS(vs, 3)]));
Chris@82 345 {
Chris@82 346 V T12, T13, TZ, T10;
Chris@82 347 T12 = LD(&(x[WS(vs, 3) + WS(rs, 2)]), ms, &(x[WS(vs, 3)]));
Chris@82 348 T13 = LD(&(x[WS(vs, 3) + WS(rs, 3)]), ms, &(x[WS(vs, 3) + WS(rs, 1)]));
Chris@82 349 T14 = VSUB(T12, T13);
Chris@82 350 T17 = VADD(T12, T13);
Chris@82 351 TZ = LD(&(x[WS(vs, 3) + WS(rs, 1)]), ms, &(x[WS(vs, 3) + WS(rs, 1)]));
Chris@82 352 T10 = LD(&(x[WS(vs, 3) + WS(rs, 4)]), ms, &(x[WS(vs, 3)]));
Chris@82 353 T11 = VSUB(TZ, T10);
Chris@82 354 T16 = VADD(TZ, T10);
Chris@82 355 }
Chris@82 356 T15 = VBYI(VFMA(LDK(KP951056516), T11, VMUL(LDK(KP587785252), T14)));
Chris@82 357 T1f = VBYI(VFNMS(LDK(KP951056516), T14, VMUL(LDK(KP587785252), T11)));
Chris@82 358 T18 = VMUL(LDK(KP559016994), VSUB(T16, T17));
Chris@82 359 T1a = VADD(T16, T17);
Chris@82 360 T1b = VFNMS(LDK(KP250000000), T1a, T19);
Chris@82 361 }
Chris@82 362 ST(&(x[0]), VADD(Tb, Tc), ms, &(x[0]));
Chris@82 363 ST(&(x[WS(rs, 4)]), VADD(T1t, T1u), ms, &(x[0]));
Chris@82 364 ST(&(x[WS(rs, 2)]), VADD(TP, TQ), ms, &(x[0]));
Chris@82 365 ST(&(x[WS(rs, 3)]), VADD(T19, T1a), ms, &(x[WS(rs, 1)]));
Chris@82 366 ST(&(x[WS(rs, 1)]), VADD(Tv, Tw), ms, &(x[WS(rs, 1)]));
Chris@82 367 {
Chris@82 368 V Tj, Tk, Ti, T1B, T1C, T1A;
Chris@82 369 Ti = VSUB(Td, Ta);
Chris@82 370 Tj = BYTW(&(W[TWVL * 2]), VADD(Th, Ti));
Chris@82 371 Tk = BYTW(&(W[TWVL * 4]), VSUB(Ti, Th));
Chris@82 372 ST(&(x[WS(vs, 2)]), Tj, ms, &(x[WS(vs, 2)]));
Chris@82 373 ST(&(x[WS(vs, 3)]), Tk, ms, &(x[WS(vs, 3)]));
Chris@82 374 T1A = VSUB(T1v, T1s);
Chris@82 375 T1B = BYTW(&(W[TWVL * 2]), VADD(T1z, T1A));
Chris@82 376 T1C = BYTW(&(W[TWVL * 4]), VSUB(T1A, T1z));
Chris@82 377 ST(&(x[WS(vs, 2) + WS(rs, 4)]), T1B, ms, &(x[WS(vs, 2)]));
Chris@82 378 ST(&(x[WS(vs, 3) + WS(rs, 4)]), T1C, ms, &(x[WS(vs, 3)]));
Chris@82 379 }
Chris@82 380 {
Chris@82 381 V T1h, T1i, T1g, TD, TE, TC;
Chris@82 382 T1g = VSUB(T1b, T18);
Chris@82 383 T1h = BYTW(&(W[TWVL * 2]), VADD(T1f, T1g));
Chris@82 384 T1i = BYTW(&(W[TWVL * 4]), VSUB(T1g, T1f));
Chris@82 385 ST(&(x[WS(vs, 2) + WS(rs, 3)]), T1h, ms, &(x[WS(vs, 2) + WS(rs, 1)]));
Chris@82 386 ST(&(x[WS(vs, 3) + WS(rs, 3)]), T1i, ms, &(x[WS(vs, 3) + WS(rs, 1)]));
Chris@82 387 TC = VSUB(Tx, Tu);
Chris@82 388 TD = BYTW(&(W[TWVL * 2]), VADD(TB, TC));
Chris@82 389 TE = BYTW(&(W[TWVL * 4]), VSUB(TC, TB));
Chris@82 390 ST(&(x[WS(vs, 2) + WS(rs, 1)]), TD, ms, &(x[WS(vs, 2) + WS(rs, 1)]));
Chris@82 391 ST(&(x[WS(vs, 3) + WS(rs, 1)]), TE, ms, &(x[WS(vs, 3) + WS(rs, 1)]));
Chris@82 392 }
Chris@82 393 {
Chris@82 394 V TX, TY, TW, TT, TU, TS;
Chris@82 395 TW = VSUB(TR, TO);
Chris@82 396 TX = BYTW(&(W[TWVL * 2]), VADD(TV, TW));
Chris@82 397 TY = BYTW(&(W[TWVL * 4]), VSUB(TW, TV));
Chris@82 398 ST(&(x[WS(vs, 2) + WS(rs, 2)]), TX, ms, &(x[WS(vs, 2)]));
Chris@82 399 ST(&(x[WS(vs, 3) + WS(rs, 2)]), TY, ms, &(x[WS(vs, 3)]));
Chris@82 400 TS = VADD(TO, TR);
Chris@82 401 TT = BYTW(&(W[0]), VADD(TL, TS));
Chris@82 402 TU = BYTW(&(W[TWVL * 6]), VSUB(TS, TL));
Chris@82 403 ST(&(x[WS(vs, 1) + WS(rs, 2)]), TT, ms, &(x[WS(vs, 1)]));
Chris@82 404 ST(&(x[WS(vs, 4) + WS(rs, 2)]), TU, ms, &(x[WS(vs, 4)]));
Chris@82 405 }
Chris@82 406 {
Chris@82 407 V Tf, Tg, Te, Tz, TA, Ty;
Chris@82 408 Te = VADD(Ta, Td);
Chris@82 409 Tf = BYTW(&(W[0]), VADD(T7, Te));
Chris@82 410 Tg = BYTW(&(W[TWVL * 6]), VSUB(Te, T7));
Chris@82 411 ST(&(x[WS(vs, 1)]), Tf, ms, &(x[WS(vs, 1)]));
Chris@82 412 ST(&(x[WS(vs, 4)]), Tg, ms, &(x[WS(vs, 4)]));
Chris@82 413 Ty = VADD(Tu, Tx);
Chris@82 414 Tz = BYTW(&(W[0]), VADD(Tr, Ty));
Chris@82 415 TA = BYTW(&(W[TWVL * 6]), VSUB(Ty, Tr));
Chris@82 416 ST(&(x[WS(vs, 1) + WS(rs, 1)]), Tz, ms, &(x[WS(vs, 1) + WS(rs, 1)]));
Chris@82 417 ST(&(x[WS(vs, 4) + WS(rs, 1)]), TA, ms, &(x[WS(vs, 4) + WS(rs, 1)]));
Chris@82 418 }
Chris@82 419 {
Chris@82 420 V T1d, T1e, T1c, T1x, T1y, T1w;
Chris@82 421 T1c = VADD(T18, T1b);
Chris@82 422 T1d = BYTW(&(W[0]), VADD(T15, T1c));
Chris@82 423 T1e = BYTW(&(W[TWVL * 6]), VSUB(T1c, T15));
Chris@82 424 ST(&(x[WS(vs, 1) + WS(rs, 3)]), T1d, ms, &(x[WS(vs, 1) + WS(rs, 1)]));
Chris@82 425 ST(&(x[WS(vs, 4) + WS(rs, 3)]), T1e, ms, &(x[WS(vs, 4) + WS(rs, 1)]));
Chris@82 426 T1w = VADD(T1s, T1v);
Chris@82 427 T1x = BYTW(&(W[0]), VADD(T1p, T1w));
Chris@82 428 T1y = BYTW(&(W[TWVL * 6]), VSUB(T1w, T1p));
Chris@82 429 ST(&(x[WS(vs, 1) + WS(rs, 4)]), T1x, ms, &(x[WS(vs, 1)]));
Chris@82 430 ST(&(x[WS(vs, 4) + WS(rs, 4)]), T1y, ms, &(x[WS(vs, 4)]));
Chris@82 431 }
Chris@82 432 }
Chris@82 433 }
Chris@82 434 VLEAVE();
Chris@82 435 }
Chris@82 436
Chris@82 437 static const tw_instr twinstr[] = {
Chris@82 438 VTW(0, 1),
Chris@82 439 VTW(0, 2),
Chris@82 440 VTW(0, 3),
Chris@82 441 VTW(0, 4),
Chris@82 442 {TW_NEXT, VL, 0}
Chris@82 443 };
Chris@82 444
Chris@82 445 static const ct_desc desc = { 5, XSIMD_STRING("q1bv_5"), twinstr, &GENUS, {85, 55, 15, 0}, 0, 0, 0 };
Chris@82 446
Chris@82 447 void XSIMD(codelet_q1bv_5) (planner *p) {
Chris@82 448 X(kdft_difsq_register) (p, q1bv_5, &desc);
Chris@82 449 }
Chris@82 450 #endif