Mercurial > hg > sv-dependency-builds

annotate src/fftw-3.3.3/dft/simd/common/q1fv_8.c @ 95:89f5e221ed7b

Find changesets by keywords (author, files, the commit message), revision number or hash, or revset expression.
Add FFTW3
author Chris Cannam <cannam@all-day-breakfast.com>
date Wed, 20 Mar 2013 15:35:50 +0000
parents
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 07:39:31 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 8 -dif -name q1fv_8 -include q1f.h */
cannam@95 29
cannam@95 30 /*
cannam@95 31 * This function contains 264 FP additions, 192 FP multiplications,
cannam@95 32 * (or, 184 additions, 112 multiplications, 80 fused multiply/add),
cannam@95 33 * 117 stack variables, 1 constants, and 128 memory accesses
cannam@95 34 */
cannam@95 35 #include "q1f.h"
cannam@95 36
cannam@95 37 static void q1fv_8(R *ri, R *ii, const R *W, stride rs, stride vs, INT mb, INT me, INT ms)
cannam@95 38 {
cannam@95 39 DVK(KP707106781, +0.707106781186547524400844362104849039284835938);
cannam@95 40 {
cannam@95 41 INT m;
cannam@95 42 R *x;
cannam@95 43 x = ri;
cannam@95 44 for (m = mb, W = W + (mb * ((TWVL / VL) * 14)); m < me; m = m + VL, x = x + (VL * ms), W = W + (TWVL * 14), MAKE_VOLATILE_STRIDE(16, rs), MAKE_VOLATILE_STRIDE(16, vs)) {
cannam@95 45 V T42, T43, T1U, T1V, T2Y, T2Z, TT, TS;
cannam@95 46 {
cannam@95 47 V T3, Te, T1E, T1P, Tu, Tp, T25, T20, T2b, T2m, T3M, T2x, T2C, T3X, TA;
cannam@95 48 V TL, T48, T4d, T17, T11, TW, T1i, T2I, T1y, T1t, T2T, T3f, T3q, T34, T39;
cannam@95 49 V T3G, T3B, Ts, Tv, Tf, Ta, T23, T26, T1Q, T1L, T2A, T2D, T2n, T2i, T4b;
cannam@95 50 V T4e, T3Y, T3T, TZ, T12, TM, TH, T35, T2L, T3j, T1w, T1z, T1j, T1e, T36;
cannam@95 51 V T2O, T3C, T3i, T3k;
cannam@95 52 {
cannam@95 53 V T3d, T32, T3e, T3o, T3p, T33;
cannam@95 54 {
cannam@95 55 V T2v, T2w, T3V, T46, T3W;
cannam@95 56 {
cannam@95 57 V T1, T2, Tc, Td, T1C, T1D, T1N, T1O;
cannam@95 58 T1 = LD(&(x[0]), ms, &(x[0]));
cannam@95 59 T2 = LD(&(x[WS(rs, 4)]), ms, &(x[0]));
cannam@95 60 Tc = LD(&(x[WS(rs, 2)]), ms, &(x[0]));
cannam@95 61 Td = LD(&(x[WS(rs, 6)]), ms, &(x[0]));
cannam@95 62 T1C = LD(&(x[WS(vs, 3)]), ms, &(x[WS(vs, 3)]));
cannam@95 63 T1D = LD(&(x[WS(vs, 3) + WS(rs, 4)]), ms, &(x[WS(vs, 3)]));
cannam@95 64 T1N = LD(&(x[WS(vs, 3) + WS(rs, 2)]), ms, &(x[WS(vs, 3)]));
cannam@95 65 T1O = LD(&(x[WS(vs, 3) + WS(rs, 6)]), ms, &(x[WS(vs, 3)]));
cannam@95 66 {
cannam@95 67 V T29, T1Y, T1Z, T2a, T2k, T2l, Tn, To, T3K, T3L;
cannam@95 68 T29 = LD(&(x[WS(vs, 4)]), ms, &(x[WS(vs, 4)]));
cannam@95 69 T3 = VSUB(T1, T2);
cannam@95 70 Tn = VADD(T1, T2);
cannam@95 71 Te = VSUB(Tc, Td);
cannam@95 72 To = VADD(Tc, Td);
cannam@95 73 T1E = VSUB(T1C, T1D);
cannam@95 74 T1Y = VADD(T1C, T1D);
cannam@95 75 T1P = VSUB(T1N, T1O);
cannam@95 76 T1Z = VADD(T1N, T1O);
cannam@95 77 T2a = LD(&(x[WS(vs, 4) + WS(rs, 4)]), ms, &(x[WS(vs, 4)]));
cannam@95 78 T2k = LD(&(x[WS(vs, 4) + WS(rs, 2)]), ms, &(x[WS(vs, 4)]));
cannam@95 79 T2l = LD(&(x[WS(vs, 4) + WS(rs, 6)]), ms, &(x[WS(vs, 4)]));
cannam@95 80 Tu = VSUB(Tn, To);
cannam@95 81 Tp = VADD(Tn, To);
cannam@95 82 T3K = LD(&(x[WS(vs, 7)]), ms, &(x[WS(vs, 7)]));
cannam@95 83 T3L = LD(&(x[WS(vs, 7) + WS(rs, 4)]), ms, &(x[WS(vs, 7)]));
cannam@95 84 T25 = VSUB(T1Y, T1Z);
cannam@95 85 T20 = VADD(T1Y, T1Z);
cannam@95 86 T2v = VADD(T29, T2a);
cannam@95 87 T2b = VSUB(T29, T2a);
cannam@95 88 T2w = VADD(T2k, T2l);
cannam@95 89 T2m = VSUB(T2k, T2l);
cannam@95 90 T3V = LD(&(x[WS(vs, 7) + WS(rs, 2)]), ms, &(x[WS(vs, 7)]));
cannam@95 91 T46 = VADD(T3K, T3L);
cannam@95 92 T3M = VSUB(T3K, T3L);
cannam@95 93 T3W = LD(&(x[WS(vs, 7) + WS(rs, 6)]), ms, &(x[WS(vs, 7)]));
cannam@95 94 }
cannam@95 95 }
cannam@95 96 {
cannam@95 97 V T15, TU, T16, T1g, TV, T1h;
cannam@95 98 {
cannam@95 99 V Ty, Tz, TJ, TK, T47;
cannam@95 100 Ty = LD(&(x[WS(vs, 1)]), ms, &(x[WS(vs, 1)]));
cannam@95 101 Tz = LD(&(x[WS(vs, 1) + WS(rs, 4)]), ms, &(x[WS(vs, 1)]));
cannam@95 102 TJ = LD(&(x[WS(vs, 1) + WS(rs, 2)]), ms, &(x[WS(vs, 1)]));
cannam@95 103 T2x = VADD(T2v, T2w);
cannam@95 104 T2C = VSUB(T2v, T2w);
cannam@95 105 TK = LD(&(x[WS(vs, 1) + WS(rs, 6)]), ms, &(x[WS(vs, 1)]));
cannam@95 106 T47 = VADD(T3V, T3W);
cannam@95 107 T3X = VSUB(T3V, T3W);
cannam@95 108 T15 = LD(&(x[WS(vs, 2)]), ms, &(x[WS(vs, 2)]));
cannam@95 109 TA = VSUB(Ty, Tz);
cannam@95 110 TU = VADD(Ty, Tz);
cannam@95 111 T16 = LD(&(x[WS(vs, 2) + WS(rs, 4)]), ms, &(x[WS(vs, 2)]));
cannam@95 112 T1g = LD(&(x[WS(vs, 2) + WS(rs, 2)]), ms, &(x[WS(vs, 2)]));
cannam@95 113 TL = VSUB(TJ, TK);
cannam@95 114 TV = VADD(TJ, TK);
cannam@95 115 T48 = VADD(T46, T47);
cannam@95 116 T4d = VSUB(T46, T47);
cannam@95 117 T1h = LD(&(x[WS(vs, 2) + WS(rs, 6)]), ms, &(x[WS(vs, 2)]));
cannam@95 118 }
cannam@95 119 {
cannam@95 120 V T2G, T1r, T2H, T2R, T1s, T2S;
cannam@95 121 T2G = LD(&(x[WS(vs, 5)]), ms, &(x[WS(vs, 5)]));
cannam@95 122 T17 = VSUB(T15, T16);
cannam@95 123 T1r = VADD(T15, T16);
cannam@95 124 T2H = LD(&(x[WS(vs, 5) + WS(rs, 4)]), ms, &(x[WS(vs, 5)]));
cannam@95 125 T11 = VSUB(TU, TV);
cannam@95 126 TW = VADD(TU, TV);
cannam@95 127 T2R = LD(&(x[WS(vs, 5) + WS(rs, 2)]), ms, &(x[WS(vs, 5)]));
cannam@95 128 T1i = VSUB(T1g, T1h);
cannam@95 129 T1s = VADD(T1g, T1h);
cannam@95 130 T2S = LD(&(x[WS(vs, 5) + WS(rs, 6)]), ms, &(x[WS(vs, 5)]));
cannam@95 131 T3d = LD(&(x[WS(vs, 6)]), ms, &(x[WS(vs, 6)]));
cannam@95 132 T2I = VSUB(T2G, T2H);
cannam@95 133 T32 = VADD(T2G, T2H);
cannam@95 134 T3e = LD(&(x[WS(vs, 6) + WS(rs, 4)]), ms, &(x[WS(vs, 6)]));
cannam@95 135 T3o = LD(&(x[WS(vs, 6) + WS(rs, 2)]), ms, &(x[WS(vs, 6)]));
cannam@95 136 T3p = LD(&(x[WS(vs, 6) + WS(rs, 6)]), ms, &(x[WS(vs, 6)]));
cannam@95 137 T1y = VSUB(T1r, T1s);
cannam@95 138 T1t = VADD(T1r, T1s);
cannam@95 139 T33 = VADD(T2R, T2S);
cannam@95 140 T2T = VSUB(T2R, T2S);
cannam@95 141 }
cannam@95 142 }
cannam@95 143 }
cannam@95 144 {
cannam@95 145 V T2y, T2e, T3Q, T2z, T2h, T49, T3P, T3R;
cannam@95 146 {
cannam@95 147 V T6, Tq, T1I, Tr, T9, T21, T1H, T1J;
cannam@95 148 {
cannam@95 149 V T4, T3z, T3A, T5, T7, T8, T1F, T1G;
cannam@95 150 T4 = LD(&(x[WS(rs, 1)]), ms, &(x[WS(rs, 1)]));
cannam@95 151 T3f = VSUB(T3d, T3e);
cannam@95 152 T3z = VADD(T3d, T3e);
cannam@95 153 T3q = VSUB(T3o, T3p);
cannam@95 154 T3A = VADD(T3o, T3p);
cannam@95 155 T5 = LD(&(x[WS(rs, 5)]), ms, &(x[WS(rs, 1)]));
cannam@95 156 T7 = LD(&(x[WS(rs, 7)]), ms, &(x[WS(rs, 1)]));
cannam@95 157 T34 = VADD(T32, T33);
cannam@95 158 T39 = VSUB(T32, T33);
cannam@95 159 T8 = LD(&(x[WS(rs, 3)]), ms, &(x[WS(rs, 1)]));
cannam@95 160 T1F = LD(&(x[WS(vs, 3) + WS(rs, 1)]), ms, &(x[WS(vs, 3) + WS(rs, 1)]));
cannam@95 161 T1G = LD(&(x[WS(vs, 3) + WS(rs, 5)]), ms, &(x[WS(vs, 3) + WS(rs, 1)]));
cannam@95 162 T3G = VSUB(T3z, T3A);
cannam@95 163 T3B = VADD(T3z, T3A);
cannam@95 164 T6 = VSUB(T4, T5);
cannam@95 165 Tq = VADD(T4, T5);
cannam@95 166 T1I = LD(&(x[WS(vs, 3) + WS(rs, 7)]), ms, &(x[WS(vs, 3) + WS(rs, 1)]));
cannam@95 167 Tr = VADD(T7, T8);
cannam@95 168 T9 = VSUB(T7, T8);
cannam@95 169 T21 = VADD(T1F, T1G);
cannam@95 170 T1H = VSUB(T1F, T1G);
cannam@95 171 T1J = LD(&(x[WS(vs, 3) + WS(rs, 3)]), ms, &(x[WS(vs, 3) + WS(rs, 1)]));
cannam@95 172 }
cannam@95 173 {
cannam@95 174 V T2f, T22, T1K, T2g, T2c, T2d, T3N, T3O;
cannam@95 175 T2c = LD(&(x[WS(vs, 4) + WS(rs, 1)]), ms, &(x[WS(vs, 4) + WS(rs, 1)]));
cannam@95 176 T2d = LD(&(x[WS(vs, 4) + WS(rs, 5)]), ms, &(x[WS(vs, 4) + WS(rs, 1)]));
cannam@95 177 T2f = LD(&(x[WS(vs, 4) + WS(rs, 7)]), ms, &(x[WS(vs, 4) + WS(rs, 1)]));
cannam@95 178 Ts = VADD(Tq, Tr);
cannam@95 179 Tv = VSUB(Tr, Tq);
cannam@95 180 Tf = VSUB(T9, T6);
cannam@95 181 Ta = VADD(T6, T9);
cannam@95 182 T22 = VADD(T1I, T1J);
cannam@95 183 T1K = VSUB(T1I, T1J);
cannam@95 184 T2y = VADD(T2c, T2d);
cannam@95 185 T2e = VSUB(T2c, T2d);
cannam@95 186 T2g = LD(&(x[WS(vs, 4) + WS(rs, 3)]), ms, &(x[WS(vs, 4) + WS(rs, 1)]));
cannam@95 187 T3N = LD(&(x[WS(vs, 7) + WS(rs, 1)]), ms, &(x[WS(vs, 7) + WS(rs, 1)]));
cannam@95 188 T3O = LD(&(x[WS(vs, 7) + WS(rs, 5)]), ms, &(x[WS(vs, 7) + WS(rs, 1)]));
cannam@95 189 T3Q = LD(&(x[WS(vs, 7) + WS(rs, 7)]), ms, &(x[WS(vs, 7) + WS(rs, 1)]));
cannam@95 190 T23 = VADD(T21, T22);
cannam@95 191 T26 = VSUB(T22, T21);
cannam@95 192 T1Q = VSUB(T1K, T1H);
cannam@95 193 T1L = VADD(T1H, T1K);
cannam@95 194 T2z = VADD(T2f, T2g);
cannam@95 195 T2h = VSUB(T2f, T2g);
cannam@95 196 T49 = VADD(T3N, T3O);
cannam@95 197 T3P = VSUB(T3N, T3O);
cannam@95 198 T3R = LD(&(x[WS(vs, 7) + WS(rs, 3)]), ms, &(x[WS(vs, 7) + WS(rs, 1)]));
cannam@95 199 }
cannam@95 200 }
cannam@95 201 {
cannam@95 202 V TX, TD, T1b, TY, TG, T1u, T1a, T1c;
cannam@95 203 {
cannam@95 204 V TE, T4a, T3S, TF, TB, TC, T18, T19;
cannam@95 205 TB = LD(&(x[WS(vs, 1) + WS(rs, 1)]), ms, &(x[WS(vs, 1) + WS(rs, 1)]));
cannam@95 206 TC = LD(&(x[WS(vs, 1) + WS(rs, 5)]), ms, &(x[WS(vs, 1) + WS(rs, 1)]));
cannam@95 207 TE = LD(&(x[WS(vs, 1) + WS(rs, 7)]), ms, &(x[WS(vs, 1) + WS(rs, 1)]));
cannam@95 208 T2A = VADD(T2y, T2z);
cannam@95 209 T2D = VSUB(T2z, T2y);
cannam@95 210 T2n = VSUB(T2h, T2e);
cannam@95 211 T2i = VADD(T2e, T2h);
cannam@95 212 T4a = VADD(T3Q, T3R);
cannam@95 213 T3S = VSUB(T3Q, T3R);
cannam@95 214 TX = VADD(TB, TC);
cannam@95 215 TD = VSUB(TB, TC);
cannam@95 216 TF = LD(&(x[WS(vs, 1) + WS(rs, 3)]), ms, &(x[WS(vs, 1) + WS(rs, 1)]));
cannam@95 217 T18 = LD(&(x[WS(vs, 2) + WS(rs, 1)]), ms, &(x[WS(vs, 2) + WS(rs, 1)]));
cannam@95 218 T19 = LD(&(x[WS(vs, 2) + WS(rs, 5)]), ms, &(x[WS(vs, 2) + WS(rs, 1)]));
cannam@95 219 T1b = LD(&(x[WS(vs, 2) + WS(rs, 7)]), ms, &(x[WS(vs, 2) + WS(rs, 1)]));
cannam@95 220 T4b = VADD(T49, T4a);
cannam@95 221 T4e = VSUB(T4a, T49);
cannam@95 222 T3Y = VSUB(T3S, T3P);
cannam@95 223 T3T = VADD(T3P, T3S);
cannam@95 224 TY = VADD(TE, TF);
cannam@95 225 TG = VSUB(TE, TF);
cannam@95 226 T1u = VADD(T18, T19);
cannam@95 227 T1a = VSUB(T18, T19);
cannam@95 228 T1c = LD(&(x[WS(vs, 2) + WS(rs, 3)]), ms, &(x[WS(vs, 2) + WS(rs, 1)]));
cannam@95 229 }
cannam@95 230 {
cannam@95 231 V T2M, T1v, T1d, T2N, T2J, T2K, T3g, T3h;
cannam@95 232 T2J = LD(&(x[WS(vs, 5) + WS(rs, 1)]), ms, &(x[WS(vs, 5) + WS(rs, 1)]));
cannam@95 233 T2K = LD(&(x[WS(vs, 5) + WS(rs, 5)]), ms, &(x[WS(vs, 5) + WS(rs, 1)]));
cannam@95 234 T2M = LD(&(x[WS(vs, 5) + WS(rs, 7)]), ms, &(x[WS(vs, 5) + WS(rs, 1)]));
cannam@95 235 TZ = VADD(TX, TY);
cannam@95 236 T12 = VSUB(TY, TX);
cannam@95 237 TM = VSUB(TG, TD);
cannam@95 238 TH = VADD(TD, TG);
cannam@95 239 T1v = VADD(T1b, T1c);
cannam@95 240 T1d = VSUB(T1b, T1c);
cannam@95 241 T35 = VADD(T2J, T2K);
cannam@95 242 T2L = VSUB(T2J, T2K);
cannam@95 243 T2N = LD(&(x[WS(vs, 5) + WS(rs, 3)]), ms, &(x[WS(vs, 5) + WS(rs, 1)]));
cannam@95 244 T3g = LD(&(x[WS(vs, 6) + WS(rs, 1)]), ms, &(x[WS(vs, 6) + WS(rs, 1)]));
cannam@95 245 T3h = LD(&(x[WS(vs, 6) + WS(rs, 5)]), ms, &(x[WS(vs, 6) + WS(rs, 1)]));
cannam@95 246 T3j = LD(&(x[WS(vs, 6) + WS(rs, 7)]), ms, &(x[WS(vs, 6) + WS(rs, 1)]));
cannam@95 247 T1w = VADD(T1u, T1v);
cannam@95 248 T1z = VSUB(T1v, T1u);
cannam@95 249 T1j = VSUB(T1d, T1a);
cannam@95 250 T1e = VADD(T1a, T1d);
cannam@95 251 T36 = VADD(T2M, T2N);
cannam@95 252 T2O = VSUB(T2M, T2N);
cannam@95 253 T3C = VADD(T3g, T3h);
cannam@95 254 T3i = VSUB(T3g, T3h);
cannam@95 255 T3k = LD(&(x[WS(vs, 6) + WS(rs, 3)]), ms, &(x[WS(vs, 6) + WS(rs, 1)]));
cannam@95 256 }
cannam@95 257 }
cannam@95 258 }
cannam@95 259 }
cannam@95 260 {
cannam@95 261 V T3a, T2U, T2P, T3H, T3r, T3m, T13, T27, T3b, T4f;
cannam@95 262 {
cannam@95 263 V T37, T3E, T2B, T24;
cannam@95 264 {
cannam@95 265 V T3D, T3l, Tt, T4c;
cannam@95 266 ST(&(x[0]), VADD(Tp, Ts), ms, &(x[0]));
cannam@95 267 ST(&(x[WS(rs, 2)]), VADD(T1t, T1w), ms, &(x[0]));
cannam@95 268 ST(&(x[WS(rs, 7)]), VADD(T48, T4b), ms, &(x[WS(rs, 1)]));
cannam@95 269 T37 = VADD(T35, T36);
cannam@95 270 T3a = VSUB(T36, T35);
cannam@95 271 T2U = VSUB(T2O, T2L);
cannam@95 272 T2P = VADD(T2L, T2O);
cannam@95 273 T3D = VADD(T3j, T3k);
cannam@95 274 T3l = VSUB(T3j, T3k);
cannam@95 275 ST(&(x[WS(rs, 4)]), VADD(T2x, T2A), ms, &(x[0]));
cannam@95 276 ST(&(x[WS(rs, 3)]), VADD(T20, T23), ms, &(x[WS(rs, 1)]));
cannam@95 277 ST(&(x[WS(rs, 5)]), VADD(T34, T37), ms, &(x[WS(rs, 1)]));
cannam@95 278 ST(&(x[WS(rs, 1)]), VADD(TW, TZ), ms, &(x[WS(rs, 1)]));
cannam@95 279 Tt = BYTWJ(&(W[TWVL * 6]), VSUB(Tp, Ts));
cannam@95 280 T4c = BYTWJ(&(W[TWVL * 6]), VSUB(T48, T4b));
cannam@95 281 T3E = VADD(T3C, T3D);
cannam@95 282 T3H = VSUB(T3D, T3C);
cannam@95 283 T3r = VSUB(T3l, T3i);
cannam@95 284 T3m = VADD(T3i, T3l);
cannam@95 285 T2B = BYTWJ(&(W[TWVL * 6]), VSUB(T2x, T2A));
cannam@95 286 T24 = BYTWJ(&(W[TWVL * 6]), VSUB(T20, T23));
cannam@95 287 ST(&(x[WS(vs, 4)]), Tt, ms, &(x[WS(vs, 4)]));
cannam@95 288 ST(&(x[WS(vs, 4) + WS(rs, 7)]), T4c, ms, &(x[WS(vs, 4) + WS(rs, 1)]));
cannam@95 289 ST(&(x[WS(rs, 6)]), VADD(T3B, T3E), ms, &(x[0]));
cannam@95 290 }
cannam@95 291 {
cannam@95 292 V T38, T1A, Tw, T10, T1x, T3F, T2E, T3I;
cannam@95 293 T10 = BYTWJ(&(W[TWVL * 6]), VSUB(TW, TZ));
cannam@95 294 T1x = BYTWJ(&(W[TWVL * 6]), VSUB(T1t, T1w));
cannam@95 295 T3F = BYTWJ(&(W[TWVL * 6]), VSUB(T3B, T3E));
cannam@95 296 ST(&(x[WS(vs, 4) + WS(rs, 4)]), T2B, ms, &(x[WS(vs, 4)]));
cannam@95 297 ST(&(x[WS(vs, 4) + WS(rs, 3)]), T24, ms, &(x[WS(vs, 4) + WS(rs, 1)]));
cannam@95 298 T38 = BYTWJ(&(W[TWVL * 6]), VSUB(T34, T37));
cannam@95 299 T1A = BYTWJ(&(W[TWVL * 10]), VFNMSI(T1z, T1y));
cannam@95 300 Tw = BYTWJ(&(W[TWVL * 10]), VFNMSI(Tv, Tu));
cannam@95 301 ST(&(x[WS(vs, 4) + WS(rs, 1)]), T10, ms, &(x[WS(vs, 4) + WS(rs, 1)]));
cannam@95 302 ST(&(x[WS(vs, 4) + WS(rs, 2)]), T1x, ms, &(x[WS(vs, 4)]));
cannam@95 303 ST(&(x[WS(vs, 4) + WS(rs, 6)]), T3F, ms, &(x[WS(vs, 4)]));
cannam@95 304 T2E = BYTWJ(&(W[TWVL * 10]), VFNMSI(T2D, T2C));
cannam@95 305 T3I = BYTWJ(&(W[TWVL * 10]), VFNMSI(T3H, T3G));
cannam@95 306 ST(&(x[WS(vs, 4) + WS(rs, 5)]), T38, ms, &(x[WS(vs, 4) + WS(rs, 1)]));
cannam@95 307 ST(&(x[WS(vs, 6) + WS(rs, 2)]), T1A, ms, &(x[WS(vs, 6)]));
cannam@95 308 ST(&(x[WS(vs, 6)]), Tw, ms, &(x[WS(vs, 6)]));
cannam@95 309 T13 = BYTWJ(&(W[TWVL * 10]), VFNMSI(T12, T11));
cannam@95 310 T27 = BYTWJ(&(W[TWVL * 10]), VFNMSI(T26, T25));
cannam@95 311 T3b = BYTWJ(&(W[TWVL * 10]), VFNMSI(T3a, T39));
cannam@95 312 ST(&(x[WS(vs, 6) + WS(rs, 4)]), T2E, ms, &(x[WS(vs, 6)]));
cannam@95 313 ST(&(x[WS(vs, 6) + WS(rs, 6)]), T3I, ms, &(x[WS(vs, 6)]));
cannam@95 314 T4f = BYTWJ(&(W[TWVL * 10]), VFNMSI(T4e, T4d));
cannam@95 315 }
cannam@95 316 }
cannam@95 317 {
cannam@95 318 V Tj, Tk, T2r, T2j, Ti, Th, T2o, T2s, T1M, T1R, T41, T40;
cannam@95 319 {
cannam@95 320 V T3c, T4g, T3J, T2F, Tx, T1B;
cannam@95 321 Tx = BYTWJ(&(W[TWVL * 2]), VFMAI(Tv, Tu));
cannam@95 322 T1B = BYTWJ(&(W[TWVL * 2]), VFMAI(T1z, T1y));
cannam@95 323 ST(&(x[WS(vs, 6) + WS(rs, 1)]), T13, ms, &(x[WS(vs, 6) + WS(rs, 1)]));
cannam@95 324 ST(&(x[WS(vs, 6) + WS(rs, 3)]), T27, ms, &(x[WS(vs, 6) + WS(rs, 1)]));
cannam@95 325 ST(&(x[WS(vs, 6) + WS(rs, 5)]), T3b, ms, &(x[WS(vs, 6) + WS(rs, 1)]));
cannam@95 326 T3c = BYTWJ(&(W[TWVL * 2]), VFMAI(T3a, T39));
cannam@95 327 T4g = BYTWJ(&(W[TWVL * 2]), VFMAI(T4e, T4d));
cannam@95 328 ST(&(x[WS(vs, 6) + WS(rs, 7)]), T4f, ms, &(x[WS(vs, 6) + WS(rs, 1)]));
cannam@95 329 ST(&(x[WS(vs, 2)]), Tx, ms, &(x[WS(vs, 2)]));
cannam@95 330 ST(&(x[WS(vs, 2) + WS(rs, 2)]), T1B, ms, &(x[WS(vs, 2)]));
cannam@95 331 T3J = BYTWJ(&(W[TWVL * 2]), VFMAI(T3H, T3G));
cannam@95 332 T2F = BYTWJ(&(W[TWVL * 2]), VFMAI(T2D, T2C));
cannam@95 333 {
cannam@95 334 V T14, Tb, Tg, T28, T3U, T3Z;
cannam@95 335 T28 = BYTWJ(&(W[TWVL * 2]), VFMAI(T26, T25));
cannam@95 336 ST(&(x[WS(vs, 2) + WS(rs, 5)]), T3c, ms, &(x[WS(vs, 2) + WS(rs, 1)]));
cannam@95 337 ST(&(x[WS(vs, 2) + WS(rs, 7)]), T4g, ms, &(x[WS(vs, 2) + WS(rs, 1)]));
cannam@95 338 T14 = BYTWJ(&(W[TWVL * 2]), VFMAI(T12, T11));
cannam@95 339 Tj = VFNMS(LDK(KP707106781), Ta, T3);
cannam@95 340 Tb = VFMA(LDK(KP707106781), Ta, T3);
cannam@95 341 Tg = VFNMS(LDK(KP707106781), Tf, Te);
cannam@95 342 Tk = VFMA(LDK(KP707106781), Tf, Te);
cannam@95 343 ST(&(x[WS(vs, 2) + WS(rs, 6)]), T3J, ms, &(x[WS(vs, 2)]));
cannam@95 344 ST(&(x[WS(vs, 2) + WS(rs, 4)]), T2F, ms, &(x[WS(vs, 2)]));
cannam@95 345 ST(&(x[WS(vs, 2) + WS(rs, 3)]), T28, ms, &(x[WS(vs, 2) + WS(rs, 1)]));
cannam@95 346 T3U = VFMA(LDK(KP707106781), T3T, T3M);
cannam@95 347 T42 = VFNMS(LDK(KP707106781), T3T, T3M);
cannam@95 348 T43 = VFMA(LDK(KP707106781), T3Y, T3X);
cannam@95 349 T3Z = VFNMS(LDK(KP707106781), T3Y, T3X);
cannam@95 350 ST(&(x[WS(vs, 2) + WS(rs, 1)]), T14, ms, &(x[WS(vs, 2) + WS(rs, 1)]));
cannam@95 351 T2r = VFNMS(LDK(KP707106781), T2i, T2b);
cannam@95 352 T2j = VFMA(LDK(KP707106781), T2i, T2b);
cannam@95 353 Ti = BYTWJ(&(W[TWVL * 12]), VFMAI(Tg, Tb));
cannam@95 354 Th = BYTWJ(&(W[0]), VFNMSI(Tg, Tb));
cannam@95 355 T2o = VFNMS(LDK(KP707106781), T2n, T2m);
cannam@95 356 T2s = VFMA(LDK(KP707106781), T2n, T2m);
cannam@95 357 T1U = VFNMS(LDK(KP707106781), T1L, T1E);
cannam@95 358 T1M = VFMA(LDK(KP707106781), T1L, T1E);
cannam@95 359 T1R = VFNMS(LDK(KP707106781), T1Q, T1P);
cannam@95 360 T1V = VFMA(LDK(KP707106781), T1Q, T1P);
cannam@95 361 T41 = BYTWJ(&(W[TWVL * 12]), VFMAI(T3Z, T3U));
cannam@95 362 T40 = BYTWJ(&(W[0]), VFNMSI(T3Z, T3U));
cannam@95 363 }
cannam@95 364 }
cannam@95 365 {
cannam@95 366 V TQ, TR, T1n, T1o, T3v, T3w;
cannam@95 367 {
cannam@95 368 V T1f, T1k, T3n, TP, TO, T3s, T2Q, T2V;
cannam@95 369 {
cannam@95 370 V TI, T2q, T2p, T1T, T1S, TN;
cannam@95 371 TQ = VFNMS(LDK(KP707106781), TH, TA);
cannam@95 372 TI = VFMA(LDK(KP707106781), TH, TA);
cannam@95 373 ST(&(x[WS(vs, 7)]), Ti, ms, &(x[WS(vs, 7)]));
cannam@95 374 ST(&(x[WS(vs, 1)]), Th, ms, &(x[WS(vs, 1)]));
cannam@95 375 T2q = BYTWJ(&(W[TWVL * 12]), VFMAI(T2o, T2j));
cannam@95 376 T2p = BYTWJ(&(W[0]), VFNMSI(T2o, T2j));
cannam@95 377 T1T = BYTWJ(&(W[TWVL * 12]), VFMAI(T1R, T1M));
cannam@95 378 T1S = BYTWJ(&(W[0]), VFNMSI(T1R, T1M));
cannam@95 379 ST(&(x[WS(vs, 7) + WS(rs, 7)]), T41, ms, &(x[WS(vs, 7) + WS(rs, 1)]));
cannam@95 380 ST(&(x[WS(vs, 1) + WS(rs, 7)]), T40, ms, &(x[WS(vs, 1) + WS(rs, 1)]));
cannam@95 381 TN = VFNMS(LDK(KP707106781), TM, TL);
cannam@95 382 TR = VFMA(LDK(KP707106781), TM, TL);
cannam@95 383 T1n = VFNMS(LDK(KP707106781), T1e, T17);
cannam@95 384 T1f = VFMA(LDK(KP707106781), T1e, T17);
cannam@95 385 ST(&(x[WS(vs, 7) + WS(rs, 4)]), T2q, ms, &(x[WS(vs, 7)]));
cannam@95 386 ST(&(x[WS(vs, 1) + WS(rs, 4)]), T2p, ms, &(x[WS(vs, 1)]));
cannam@95 387 ST(&(x[WS(vs, 7) + WS(rs, 3)]), T1T, ms, &(x[WS(vs, 7) + WS(rs, 1)]));
cannam@95 388 ST(&(x[WS(vs, 1) + WS(rs, 3)]), T1S, ms, &(x[WS(vs, 1) + WS(rs, 1)]));
cannam@95 389 T1k = VFNMS(LDK(KP707106781), T1j, T1i);
cannam@95 390 T1o = VFMA(LDK(KP707106781), T1j, T1i);
cannam@95 391 T3v = VFNMS(LDK(KP707106781), T3m, T3f);
cannam@95 392 T3n = VFMA(LDK(KP707106781), T3m, T3f);
cannam@95 393 TP = BYTWJ(&(W[TWVL * 12]), VFMAI(TN, TI));
cannam@95 394 TO = BYTWJ(&(W[0]), VFNMSI(TN, TI));
cannam@95 395 T3s = VFNMS(LDK(KP707106781), T3r, T3q);
cannam@95 396 T3w = VFMA(LDK(KP707106781), T3r, T3q);
cannam@95 397 }
cannam@95 398 T2Y = VFNMS(LDK(KP707106781), T2P, T2I);
cannam@95 399 T2Q = VFMA(LDK(KP707106781), T2P, T2I);
cannam@95 400 T2V = VFNMS(LDK(KP707106781), T2U, T2T);
cannam@95 401 T2Z = VFMA(LDK(KP707106781), T2U, T2T);
cannam@95 402 {
cannam@95 403 V T3u, T3t, T2X, T2W, T1m, T1l;
cannam@95 404 T1m = BYTWJ(&(W[TWVL * 12]), VFMAI(T1k, T1f));
cannam@95 405 T1l = BYTWJ(&(W[0]), VFNMSI(T1k, T1f));
cannam@95 406 ST(&(x[WS(vs, 7) + WS(rs, 1)]), TP, ms, &(x[WS(vs, 7) + WS(rs, 1)]));
cannam@95 407 ST(&(x[WS(vs, 1) + WS(rs, 1)]), TO, ms, &(x[WS(vs, 1) + WS(rs, 1)]));
cannam@95 408 T3u = BYTWJ(&(W[TWVL * 12]), VFMAI(T3s, T3n));
cannam@95 409 T3t = BYTWJ(&(W[0]), VFNMSI(T3s, T3n));
cannam@95 410 T2X = BYTWJ(&(W[TWVL * 12]), VFMAI(T2V, T2Q));
cannam@95 411 T2W = BYTWJ(&(W[0]), VFNMSI(T2V, T2Q));
cannam@95 412 ST(&(x[WS(vs, 7) + WS(rs, 2)]), T1m, ms, &(x[WS(vs, 7)]));
cannam@95 413 ST(&(x[WS(vs, 1) + WS(rs, 2)]), T1l, ms, &(x[WS(vs, 1)]));
cannam@95 414 ST(&(x[WS(vs, 7) + WS(rs, 6)]), T3u, ms, &(x[WS(vs, 7)]));
cannam@95 415 ST(&(x[WS(vs, 1) + WS(rs, 6)]), T3t, ms, &(x[WS(vs, 1)]));
cannam@95 416 ST(&(x[WS(vs, 7) + WS(rs, 5)]), T2X, ms, &(x[WS(vs, 7) + WS(rs, 1)]));
cannam@95 417 ST(&(x[WS(vs, 1) + WS(rs, 5)]), T2W, ms, &(x[WS(vs, 1) + WS(rs, 1)]));
cannam@95 418 }
cannam@95 419 }
cannam@95 420 {
cannam@95 421 V T2u, T2t, T3y, T3x;
cannam@95 422 {
cannam@95 423 V T1q, T1p, Tm, Tl;
cannam@95 424 T1q = BYTWJ(&(W[TWVL * 4]), VFMAI(T1o, T1n));
cannam@95 425 T1p = BYTWJ(&(W[TWVL * 8]), VFNMSI(T1o, T1n));
cannam@95 426 Tm = BYTWJ(&(W[TWVL * 4]), VFMAI(Tk, Tj));
cannam@95 427 Tl = BYTWJ(&(W[TWVL * 8]), VFNMSI(Tk, Tj));
cannam@95 428 ST(&(x[WS(vs, 3) + WS(rs, 2)]), T1q, ms, &(x[WS(vs, 3)]));
cannam@95 429 ST(&(x[WS(vs, 5) + WS(rs, 2)]), T1p, ms, &(x[WS(vs, 5)]));
cannam@95 430 T2u = BYTWJ(&(W[TWVL * 4]), VFMAI(T2s, T2r));
cannam@95 431 T2t = BYTWJ(&(W[TWVL * 8]), VFNMSI(T2s, T2r));
cannam@95 432 T3y = BYTWJ(&(W[TWVL * 4]), VFMAI(T3w, T3v));
cannam@95 433 T3x = BYTWJ(&(W[TWVL * 8]), VFNMSI(T3w, T3v));
cannam@95 434 ST(&(x[WS(vs, 3)]), Tm, ms, &(x[WS(vs, 3)]));
cannam@95 435 ST(&(x[WS(vs, 5)]), Tl, ms, &(x[WS(vs, 5)]));
cannam@95 436 }
cannam@95 437 ST(&(x[WS(vs, 3) + WS(rs, 4)]), T2u, ms, &(x[WS(vs, 3)]));
cannam@95 438 ST(&(x[WS(vs, 5) + WS(rs, 4)]), T2t, ms, &(x[WS(vs, 5)]));
cannam@95 439 ST(&(x[WS(vs, 3) + WS(rs, 6)]), T3y, ms, &(x[WS(vs, 3)]));
cannam@95 440 ST(&(x[WS(vs, 5) + WS(rs, 6)]), T3x, ms, &(x[WS(vs, 5)]));
cannam@95 441 TT = BYTWJ(&(W[TWVL * 4]), VFMAI(TR, TQ));
cannam@95 442 TS = BYTWJ(&(W[TWVL * 8]), VFNMSI(TR, TQ));
cannam@95 443 }
cannam@95 444 }
cannam@95 445 }
cannam@95 446 }
cannam@95 447 }
cannam@95 448 {
cannam@95 449 V T31, T30, T45, T44, T1X, T1W;
cannam@95 450 T1X = BYTWJ(&(W[TWVL * 4]), VFMAI(T1V, T1U));
cannam@95 451 T1W = BYTWJ(&(W[TWVL * 8]), VFNMSI(T1V, T1U));
cannam@95 452 ST(&(x[WS(vs, 3) + WS(rs, 1)]), TT, ms, &(x[WS(vs, 3) + WS(rs, 1)]));
cannam@95 453 ST(&(x[WS(vs, 5) + WS(rs, 1)]), TS, ms, &(x[WS(vs, 5) + WS(rs, 1)]));
cannam@95 454 T31 = BYTWJ(&(W[TWVL * 4]), VFMAI(T2Z, T2Y));
cannam@95 455 T30 = BYTWJ(&(W[TWVL * 8]), VFNMSI(T2Z, T2Y));
cannam@95 456 T45 = BYTWJ(&(W[TWVL * 4]), VFMAI(T43, T42));
cannam@95 457 T44 = BYTWJ(&(W[TWVL * 8]), VFNMSI(T43, T42));
cannam@95 458 ST(&(x[WS(vs, 3) + WS(rs, 3)]), T1X, ms, &(x[WS(vs, 3) + WS(rs, 1)]));
cannam@95 459 ST(&(x[WS(vs, 5) + WS(rs, 3)]), T1W, ms, &(x[WS(vs, 5) + WS(rs, 1)]));
cannam@95 460 ST(&(x[WS(vs, 3) + WS(rs, 5)]), T31, ms, &(x[WS(vs, 3) + WS(rs, 1)]));
cannam@95 461 ST(&(x[WS(vs, 5) + WS(rs, 5)]), T30, ms, &(x[WS(vs, 5) + WS(rs, 1)]));
cannam@95 462 ST(&(x[WS(vs, 3) + WS(rs, 7)]), T45, ms, &(x[WS(vs, 3) + WS(rs, 1)]));
cannam@95 463 ST(&(x[WS(vs, 5) + WS(rs, 7)]), T44, ms, &(x[WS(vs, 5) + WS(rs, 1)]));
cannam@95 464 }
cannam@95 465 }
cannam@95 466 }
cannam@95 467 VLEAVE();
cannam@95 468 }
cannam@95 469
cannam@95 470 static const tw_instr twinstr[] = {
cannam@95 471 VTW(0, 1),
cannam@95 472 VTW(0, 2),
cannam@95 473 VTW(0, 3),
cannam@95 474 VTW(0, 4),
cannam@95 475 VTW(0, 5),
cannam@95 476 VTW(0, 6),
cannam@95 477 VTW(0, 7),
cannam@95 478 {TW_NEXT, VL, 0}
cannam@95 479 };
cannam@95 480
cannam@95 481 static const ct_desc desc = { 8, XSIMD_STRING("q1fv_8"), twinstr, &GENUS, {184, 112, 80, 0}, 0, 0, 0 };
cannam@95 482
cannam@95 483 void XSIMD(codelet_q1fv_8) (planner *p) {
cannam@95 484 X(kdft_difsq_register) (p, q1fv_8, &desc);
cannam@95 485 }
cannam@95 486 #else /* HAVE_FMA */
cannam@95 487
cannam@95 488 /* Generated by: ../../../genfft/gen_twidsq_c.native -simd -compact -variables 4 -pipeline-latency 8 -n 8 -dif -name q1fv_8 -include q1f.h */
cannam@95 489
cannam@95 490 /*
cannam@95 491 * This function contains 264 FP additions, 128 FP multiplications,
cannam@95 492 * (or, 264 additions, 128 multiplications, 0 fused multiply/add),
cannam@95 493 * 77 stack variables, 1 constants, and 128 memory accesses
cannam@95 494 */
cannam@95 495 #include "q1f.h"
cannam@95 496
cannam@95 497 static void q1fv_8(R *ri, R *ii, const R *W, stride rs, stride vs, INT mb, INT me, INT ms)
cannam@95 498 {
cannam@95 499 DVK(KP707106781, +0.707106781186547524400844362104849039284835938);
cannam@95 500 {
cannam@95 501 INT m;
cannam@95 502 R *x;
cannam@95 503 x = ri;
cannam@95 504 for (m = mb, W = W + (mb * ((TWVL / VL) * 14)); m < me; m = m + VL, x = x + (VL * ms), W = W + (TWVL * 14), MAKE_VOLATILE_STRIDE(16, rs), MAKE_VOLATILE_STRIDE(16, vs)) {
cannam@95 505 V T3, Tu, Tf, Tp, T1E, T25, T1Q, T20, T2b, T2C, T2n, T2x, T3M, T4d, T3Y;
cannam@95 506 V T48, TA, T11, TM, TW, T17, T1y, T1j, T1t, T2I, T39, T2U, T34, T3f, T3G;
cannam@95 507 V T3r, T3B, Ta, Tv, Tc, Ts, T1L, T26, T1N, T23, T2i, T2D, T2k, T2A, T3T;
cannam@95 508 V T4e, T3V, T4b, TH, T12, TJ, TZ, T1e, T1z, T1g, T1w, T2P, T3a, T2R, T37;
cannam@95 509 V T3m, T3H, T3o, T3E, T28, T14;
cannam@95 510 {
cannam@95 511 V T1, T2, Tn, Td, Te, To;
cannam@95 512 T1 = LD(&(x[0]), ms, &(x[0]));
cannam@95 513 T2 = LD(&(x[WS(rs, 4)]), ms, &(x[0]));
cannam@95 514 Tn = VADD(T1, T2);
cannam@95 515 Td = LD(&(x[WS(rs, 2)]), ms, &(x[0]));
cannam@95 516 Te = LD(&(x[WS(rs, 6)]), ms, &(x[0]));
cannam@95 517 To = VADD(Td, Te);
cannam@95 518 T3 = VSUB(T1, T2);
cannam@95 519 Tu = VSUB(Tn, To);
cannam@95 520 Tf = VSUB(Td, Te);
cannam@95 521 Tp = VADD(Tn, To);
cannam@95 522 }
cannam@95 523 {
cannam@95 524 V T1C, T1D, T1Y, T1O, T1P, T1Z;
cannam@95 525 T1C = LD(&(x[WS(vs, 3)]), ms, &(x[WS(vs, 3)]));
cannam@95 526 T1D = LD(&(x[WS(vs, 3) + WS(rs, 4)]), ms, &(x[WS(vs, 3)]));
cannam@95 527 T1Y = VADD(T1C, T1D);
cannam@95 528 T1O = LD(&(x[WS(vs, 3) + WS(rs, 2)]), ms, &(x[WS(vs, 3)]));
cannam@95 529 T1P = LD(&(x[WS(vs, 3) + WS(rs, 6)]), ms, &(x[WS(vs, 3)]));
cannam@95 530 T1Z = VADD(T1O, T1P);
cannam@95 531 T1E = VSUB(T1C, T1D);
cannam@95 532 T25 = VSUB(T1Y, T1Z);
cannam@95 533 T1Q = VSUB(T1O, T1P);
cannam@95 534 T20 = VADD(T1Y, T1Z);
cannam@95 535 }
cannam@95 536 {
cannam@95 537 V T29, T2a, T2v, T2l, T2m, T2w;
cannam@95 538 T29 = LD(&(x[WS(vs, 4)]), ms, &(x[WS(vs, 4)]));
cannam@95 539 T2a = LD(&(x[WS(vs, 4) + WS(rs, 4)]), ms, &(x[WS(vs, 4)]));
cannam@95 540 T2v = VADD(T29, T2a);
cannam@95 541 T2l = LD(&(x[WS(vs, 4) + WS(rs, 2)]), ms, &(x[WS(vs, 4)]));
cannam@95 542 T2m = LD(&(x[WS(vs, 4) + WS(rs, 6)]), ms, &(x[WS(vs, 4)]));
cannam@95 543 T2w = VADD(T2l, T2m);
cannam@95 544 T2b = VSUB(T29, T2a);
cannam@95 545 T2C = VSUB(T2v, T2w);
cannam@95 546 T2n = VSUB(T2l, T2m);
cannam@95 547 T2x = VADD(T2v, T2w);
cannam@95 548 }
cannam@95 549 {
cannam@95 550 V T3K, T3L, T46, T3W, T3X, T47;
cannam@95 551 T3K = LD(&(x[WS(vs, 7)]), ms, &(x[WS(vs, 7)]));
cannam@95 552 T3L = LD(&(x[WS(vs, 7) + WS(rs, 4)]), ms, &(x[WS(vs, 7)]));
cannam@95 553 T46 = VADD(T3K, T3L);
cannam@95 554 T3W = LD(&(x[WS(vs, 7) + WS(rs, 2)]), ms, &(x[WS(vs, 7)]));
cannam@95 555 T3X = LD(&(x[WS(vs, 7) + WS(rs, 6)]), ms, &(x[WS(vs, 7)]));
cannam@95 556 T47 = VADD(T3W, T3X);
cannam@95 557 T3M = VSUB(T3K, T3L);
cannam@95 558 T4d = VSUB(T46, T47);
cannam@95 559 T3Y = VSUB(T3W, T3X);
cannam@95 560 T48 = VADD(T46, T47);
cannam@95 561 }
cannam@95 562 {
cannam@95 563 V Ty, Tz, TU, TK, TL, TV;
cannam@95 564 Ty = LD(&(x[WS(vs, 1)]), ms, &(x[WS(vs, 1)]));
cannam@95 565 Tz = LD(&(x[WS(vs, 1) + WS(rs, 4)]), ms, &(x[WS(vs, 1)]));
cannam@95 566 TU = VADD(Ty, Tz);
cannam@95 567 TK = LD(&(x[WS(vs, 1) + WS(rs, 2)]), ms, &(x[WS(vs, 1)]));
cannam@95 568 TL = LD(&(x[WS(vs, 1) + WS(rs, 6)]), ms, &(x[WS(vs, 1)]));
cannam@95 569 TV = VADD(TK, TL);
cannam@95 570 TA = VSUB(Ty, Tz);
cannam@95 571 T11 = VSUB(TU, TV);
cannam@95 572 TM = VSUB(TK, TL);
cannam@95 573 TW = VADD(TU, TV);
cannam@95 574 }
cannam@95 575 {
cannam@95 576 V T15, T16, T1r, T1h, T1i, T1s;
cannam@95 577 T15 = LD(&(x[WS(vs, 2)]), ms, &(x[WS(vs, 2)]));
cannam@95 578 T16 = LD(&(x[WS(vs, 2) + WS(rs, 4)]), ms, &(x[WS(vs, 2)]));
cannam@95 579 T1r = VADD(T15, T16);
cannam@95 580 T1h = LD(&(x[WS(vs, 2) + WS(rs, 2)]), ms, &(x[WS(vs, 2)]));
cannam@95 581 T1i = LD(&(x[WS(vs, 2) + WS(rs, 6)]), ms, &(x[WS(vs, 2)]));
cannam@95 582 T1s = VADD(T1h, T1i);
cannam@95 583 T17 = VSUB(T15, T16);
cannam@95 584 T1y = VSUB(T1r, T1s);
cannam@95 585 T1j = VSUB(T1h, T1i);
cannam@95 586 T1t = VADD(T1r, T1s);
cannam@95 587 }
cannam@95 588 {
cannam@95 589 V T2G, T2H, T32, T2S, T2T, T33;
cannam@95 590 T2G = LD(&(x[WS(vs, 5)]), ms, &(x[WS(vs, 5)]));
cannam@95 591 T2H = LD(&(x[WS(vs, 5) + WS(rs, 4)]), ms, &(x[WS(vs, 5)]));
cannam@95 592 T32 = VADD(T2G, T2H);
cannam@95 593 T2S = LD(&(x[WS(vs, 5) + WS(rs, 2)]), ms, &(x[WS(vs, 5)]));
cannam@95 594 T2T = LD(&(x[WS(vs, 5) + WS(rs, 6)]), ms, &(x[WS(vs, 5)]));
cannam@95 595 T33 = VADD(T2S, T2T);
cannam@95 596 T2I = VSUB(T2G, T2H);
cannam@95 597 T39 = VSUB(T32, T33);
cannam@95 598 T2U = VSUB(T2S, T2T);
cannam@95 599 T34 = VADD(T32, T33);
cannam@95 600 }
cannam@95 601 {
cannam@95 602 V T3d, T3e, T3z, T3p, T3q, T3A;
cannam@95 603 T3d = LD(&(x[WS(vs, 6)]), ms, &(x[WS(vs, 6)]));
cannam@95 604 T3e = LD(&(x[WS(vs, 6) + WS(rs, 4)]), ms, &(x[WS(vs, 6)]));
cannam@95 605 T3z = VADD(T3d, T3e);
cannam@95 606 T3p = LD(&(x[WS(vs, 6) + WS(rs, 2)]), ms, &(x[WS(vs, 6)]));
cannam@95 607 T3q = LD(&(x[WS(vs, 6) + WS(rs, 6)]), ms, &(x[WS(vs, 6)]));
cannam@95 608 T3A = VADD(T3p, T3q);
cannam@95 609 T3f = VSUB(T3d, T3e);
cannam@95 610 T3G = VSUB(T3z, T3A);
cannam@95 611 T3r = VSUB(T3p, T3q);
cannam@95 612 T3B = VADD(T3z, T3A);
cannam@95 613 }
cannam@95 614 {
cannam@95 615 V T6, Tq, T9, Tr;
cannam@95 616 {
cannam@95 617 V T4, T5, T7, T8;
cannam@95 618 T4 = LD(&(x[WS(rs, 1)]), ms, &(x[WS(rs, 1)]));
cannam@95 619 T5 = LD(&(x[WS(rs, 5)]), ms, &(x[WS(rs, 1)]));
cannam@95 620 T6 = VSUB(T4, T5);
cannam@95 621 Tq = VADD(T4, T5);
cannam@95 622 T7 = LD(&(x[WS(rs, 7)]), ms, &(x[WS(rs, 1)]));
cannam@95 623 T8 = LD(&(x[WS(rs, 3)]), ms, &(x[WS(rs, 1)]));
cannam@95 624 T9 = VSUB(T7, T8);
cannam@95 625 Tr = VADD(T7, T8);
cannam@95 626 }
cannam@95 627 Ta = VMUL(LDK(KP707106781), VADD(T6, T9));
cannam@95 628 Tv = VBYI(VSUB(Tr, Tq));
cannam@95 629 Tc = VMUL(LDK(KP707106781), VSUB(T9, T6));
cannam@95 630 Ts = VADD(Tq, Tr);
cannam@95 631 }
cannam@95 632 {
cannam@95 633 V T1H, T21, T1K, T22;
cannam@95 634 {
cannam@95 635 V T1F, T1G, T1I, T1J;
cannam@95 636 T1F = LD(&(x[WS(vs, 3) + WS(rs, 1)]), ms, &(x[WS(vs, 3) + WS(rs, 1)]));
cannam@95 637 T1G = LD(&(x[WS(vs, 3) + WS(rs, 5)]), ms, &(x[WS(vs, 3) + WS(rs, 1)]));
cannam@95 638 T1H = VSUB(T1F, T1G);
cannam@95 639 T21 = VADD(T1F, T1G);
cannam@95 640 T1I = LD(&(x[WS(vs, 3) + WS(rs, 7)]), ms, &(x[WS(vs, 3) + WS(rs, 1)]));
cannam@95 641 T1J = LD(&(x[WS(vs, 3) + WS(rs, 3)]), ms, &(x[WS(vs, 3) + WS(rs, 1)]));
cannam@95 642 T1K = VSUB(T1I, T1J);
cannam@95 643 T22 = VADD(T1I, T1J);
cannam@95 644 }
cannam@95 645 T1L = VMUL(LDK(KP707106781), VADD(T1H, T1K));
cannam@95 646 T26 = VBYI(VSUB(T22, T21));
cannam@95 647 T1N = VMUL(LDK(KP707106781), VSUB(T1K, T1H));
cannam@95 648 T23 = VADD(T21, T22);
cannam@95 649 }
cannam@95 650 {
cannam@95 651 V T2e, T2y, T2h, T2z;
cannam@95 652 {
cannam@95 653 V T2c, T2d, T2f, T2g;
cannam@95 654 T2c = LD(&(x[WS(vs, 4) + WS(rs, 1)]), ms, &(x[WS(vs, 4) + WS(rs, 1)]));
cannam@95 655 T2d = LD(&(x[WS(vs, 4) + WS(rs, 5)]), ms, &(x[WS(vs, 4) + WS(rs, 1)]));
cannam@95 656 T2e = VSUB(T2c, T2d);
cannam@95 657 T2y = VADD(T2c, T2d);
cannam@95 658 T2f = LD(&(x[WS(vs, 4) + WS(rs, 7)]), ms, &(x[WS(vs, 4) + WS(rs, 1)]));
cannam@95 659 T2g = LD(&(x[WS(vs, 4) + WS(rs, 3)]), ms, &(x[WS(vs, 4) + WS(rs, 1)]));
cannam@95 660 T2h = VSUB(T2f, T2g);
cannam@95 661 T2z = VADD(T2f, T2g);
cannam@95 662 }
cannam@95 663 T2i = VMUL(LDK(KP707106781), VADD(T2e, T2h));
cannam@95 664 T2D = VBYI(VSUB(T2z, T2y));
cannam@95 665 T2k = VMUL(LDK(KP707106781), VSUB(T2h, T2e));
cannam@95 666 T2A = VADD(T2y, T2z);
cannam@95 667 }
cannam@95 668 {
cannam@95 669 V T3P, T49, T3S, T4a;
cannam@95 670 {
cannam@95 671 V T3N, T3O, T3Q, T3R;
cannam@95 672 T3N = LD(&(x[WS(vs, 7) + WS(rs, 1)]), ms, &(x[WS(vs, 7) + WS(rs, 1)]));
cannam@95 673 T3O = LD(&(x[WS(vs, 7) + WS(rs, 5)]), ms, &(x[WS(vs, 7) + WS(rs, 1)]));
cannam@95 674 T3P = VSUB(T3N, T3O);
cannam@95 675 T49 = VADD(T3N, T3O);
cannam@95 676 T3Q = LD(&(x[WS(vs, 7) + WS(rs, 7)]), ms, &(x[WS(vs, 7) + WS(rs, 1)]));
cannam@95 677 T3R = LD(&(x[WS(vs, 7) + WS(rs, 3)]), ms, &(x[WS(vs, 7) + WS(rs, 1)]));
cannam@95 678 T3S = VSUB(T3Q, T3R);
cannam@95 679 T4a = VADD(T3Q, T3R);
cannam@95 680 }
cannam@95 681 T3T = VMUL(LDK(KP707106781), VADD(T3P, T3S));
cannam@95 682 T4e = VBYI(VSUB(T4a, T49));
cannam@95 683 T3V = VMUL(LDK(KP707106781), VSUB(T3S, T3P));
cannam@95 684 T4b = VADD(T49, T4a);
cannam@95 685 }
cannam@95 686 {
cannam@95 687 V TD, TX, TG, TY;
cannam@95 688 {
cannam@95 689 V TB, TC, TE, TF;
cannam@95 690 TB = LD(&(x[WS(vs, 1) + WS(rs, 1)]), ms, &(x[WS(vs, 1) + WS(rs, 1)]));
cannam@95 691 TC = LD(&(x[WS(vs, 1) + WS(rs, 5)]), ms, &(x[WS(vs, 1) + WS(rs, 1)]));
cannam@95 692 TD = VSUB(TB, TC);
cannam@95 693 TX = VADD(TB, TC);
cannam@95 694 TE = LD(&(x[WS(vs, 1) + WS(rs, 7)]), ms, &(x[WS(vs, 1) + WS(rs, 1)]));
cannam@95 695 TF = LD(&(x[WS(vs, 1) + WS(rs, 3)]), ms, &(x[WS(vs, 1) + WS(rs, 1)]));
cannam@95 696 TG = VSUB(TE, TF);
cannam@95 697 TY = VADD(TE, TF);
cannam@95 698 }
cannam@95 699 TH = VMUL(LDK(KP707106781), VADD(TD, TG));
cannam@95 700 T12 = VBYI(VSUB(TY, TX));
cannam@95 701 TJ = VMUL(LDK(KP707106781), VSUB(TG, TD));
cannam@95 702 TZ = VADD(TX, TY);
cannam@95 703 }
cannam@95 704 {
cannam@95 705 V T1a, T1u, T1d, T1v;
cannam@95 706 {
cannam@95 707 V T18, T19, T1b, T1c;
cannam@95 708 T18 = LD(&(x[WS(vs, 2) + WS(rs, 1)]), ms, &(x[WS(vs, 2) + WS(rs, 1)]));
cannam@95 709 T19 = LD(&(x[WS(vs, 2) + WS(rs, 5)]), ms, &(x[WS(vs, 2) + WS(rs, 1)]));
cannam@95 710 T1a = VSUB(T18, T19);
cannam@95 711 T1u = VADD(T18, T19);
cannam@95 712 T1b = LD(&(x[WS(vs, 2) + WS(rs, 7)]), ms, &(x[WS(vs, 2) + WS(rs, 1)]));
cannam@95 713 T1c = LD(&(x[WS(vs, 2) + WS(rs, 3)]), ms, &(x[WS(vs, 2) + WS(rs, 1)]));
cannam@95 714 T1d = VSUB(T1b, T1c);
cannam@95 715 T1v = VADD(T1b, T1c);
cannam@95 716 }
cannam@95 717 T1e = VMUL(LDK(KP707106781), VADD(T1a, T1d));
cannam@95 718 T1z = VBYI(VSUB(T1v, T1u));
cannam@95 719 T1g = VMUL(LDK(KP707106781), VSUB(T1d, T1a));
cannam@95 720 T1w = VADD(T1u, T1v);
cannam@95 721 }
cannam@95 722 {
cannam@95 723 V T2L, T35, T2O, T36;
cannam@95 724 {
cannam@95 725 V T2J, T2K, T2M, T2N;
cannam@95 726 T2J = LD(&(x[WS(vs, 5) + WS(rs, 1)]), ms, &(x[WS(vs, 5) + WS(rs, 1)]));
cannam@95 727 T2K = LD(&(x[WS(vs, 5) + WS(rs, 5)]), ms, &(x[WS(vs, 5) + WS(rs, 1)]));
cannam@95 728 T2L = VSUB(T2J, T2K);
cannam@95 729 T35 = VADD(T2J, T2K);
cannam@95 730 T2M = LD(&(x[WS(vs, 5) + WS(rs, 7)]), ms, &(x[WS(vs, 5) + WS(rs, 1)]));
cannam@95 731 T2N = LD(&(x[WS(vs, 5) + WS(rs, 3)]), ms, &(x[WS(vs, 5) + WS(rs, 1)]));
cannam@95 732 T2O = VSUB(T2M, T2N);
cannam@95 733 T36 = VADD(T2M, T2N);
cannam@95 734 }
cannam@95 735 T2P = VMUL(LDK(KP707106781), VADD(T2L, T2O));
cannam@95 736 T3a = VBYI(VSUB(T36, T35));
cannam@95 737 T2R = VMUL(LDK(KP707106781), VSUB(T2O, T2L));
cannam@95 738 T37 = VADD(T35, T36);
cannam@95 739 }
cannam@95 740 {
cannam@95 741 V T3i, T3C, T3l, T3D;
cannam@95 742 {
cannam@95 743 V T3g, T3h, T3j, T3k;
cannam@95 744 T3g = LD(&(x[WS(vs, 6) + WS(rs, 1)]), ms, &(x[WS(vs, 6) + WS(rs, 1)]));
cannam@95 745 T3h = LD(&(x[WS(vs, 6) + WS(rs, 5)]), ms, &(x[WS(vs, 6) + WS(rs, 1)]));
cannam@95 746 T3i = VSUB(T3g, T3h);
cannam@95 747 T3C = VADD(T3g, T3h);
cannam@95 748 T3j = LD(&(x[WS(vs, 6) + WS(rs, 7)]), ms, &(x[WS(vs, 6) + WS(rs, 1)]));
cannam@95 749 T3k = LD(&(x[WS(vs, 6) + WS(rs, 3)]), ms, &(x[WS(vs, 6) + WS(rs, 1)]));
cannam@95 750 T3l = VSUB(T3j, T3k);
cannam@95 751 T3D = VADD(T3j, T3k);
cannam@95 752 }
cannam@95 753 T3m = VMUL(LDK(KP707106781), VADD(T3i, T3l));
cannam@95 754 T3H = VBYI(VSUB(T3D, T3C));
cannam@95 755 T3o = VMUL(LDK(KP707106781), VSUB(T3l, T3i));
cannam@95 756 T3E = VADD(T3C, T3D);
cannam@95 757 }
cannam@95 758 ST(&(x[0]), VADD(Tp, Ts), ms, &(x[0]));
cannam@95 759 ST(&(x[WS(rs, 2)]), VADD(T1t, T1w), ms, &(x[0]));
cannam@95 760 ST(&(x[WS(rs, 5)]), VADD(T34, T37), ms, &(x[WS(rs, 1)]));
cannam@95 761 ST(&(x[WS(rs, 7)]), VADD(T48, T4b), ms, &(x[WS(rs, 1)]));
cannam@95 762 ST(&(x[WS(rs, 6)]), VADD(T3B, T3E), ms, &(x[0]));
cannam@95 763 ST(&(x[WS(rs, 4)]), VADD(T2x, T2A), ms, &(x[0]));
cannam@95 764 {
cannam@95 765 V Tt, T4c, T2B, T24;
cannam@95 766 ST(&(x[WS(rs, 3)]), VADD(T20, T23), ms, &(x[WS(rs, 1)]));
cannam@95 767 ST(&(x[WS(rs, 1)]), VADD(TW, TZ), ms, &(x[WS(rs, 1)]));
cannam@95 768 Tt = BYTWJ(&(W[TWVL * 6]), VSUB(Tp, Ts));
cannam@95 769 ST(&(x[WS(vs, 4)]), Tt, ms, &(x[WS(vs, 4)]));
cannam@95 770 T4c = BYTWJ(&(W[TWVL * 6]), VSUB(T48, T4b));
cannam@95 771 ST(&(x[WS(vs, 4) + WS(rs, 7)]), T4c, ms, &(x[WS(vs, 4) + WS(rs, 1)]));
cannam@95 772 T2B = BYTWJ(&(W[TWVL * 6]), VSUB(T2x, T2A));
cannam@95 773 ST(&(x[WS(vs, 4) + WS(rs, 4)]), T2B, ms, &(x[WS(vs, 4)]));
cannam@95 774 T24 = BYTWJ(&(W[TWVL * 6]), VSUB(T20, T23));
cannam@95 775 ST(&(x[WS(vs, 4) + WS(rs, 3)]), T24, ms, &(x[WS(vs, 4) + WS(rs, 1)]));
cannam@95 776 }
cannam@95 777 {
cannam@95 778 V T10, T1x, T3F, T38, T1A, Tw;
cannam@95 779 T10 = BYTWJ(&(W[TWVL * 6]), VSUB(TW, TZ));
cannam@95 780 ST(&(x[WS(vs, 4) + WS(rs, 1)]), T10, ms, &(x[WS(vs, 4) + WS(rs, 1)]));
cannam@95 781 T1x = BYTWJ(&(W[TWVL * 6]), VSUB(T1t, T1w));
cannam@95 782 ST(&(x[WS(vs, 4) + WS(rs, 2)]), T1x, ms, &(x[WS(vs, 4)]));
cannam@95 783 T3F = BYTWJ(&(W[TWVL * 6]), VSUB(T3B, T3E));
cannam@95 784 ST(&(x[WS(vs, 4) + WS(rs, 6)]), T3F, ms, &(x[WS(vs, 4)]));
cannam@95 785 T38 = BYTWJ(&(W[TWVL * 6]), VSUB(T34, T37));
cannam@95 786 ST(&(x[WS(vs, 4) + WS(rs, 5)]), T38, ms, &(x[WS(vs, 4) + WS(rs, 1)]));
cannam@95 787 T1A = BYTWJ(&(W[TWVL * 10]), VSUB(T1y, T1z));
cannam@95 788 ST(&(x[WS(vs, 6) + WS(rs, 2)]), T1A, ms, &(x[WS(vs, 6)]));
cannam@95 789 Tw = BYTWJ(&(W[TWVL * 10]), VSUB(Tu, Tv));
cannam@95 790 ST(&(x[WS(vs, 6)]), Tw, ms, &(x[WS(vs, 6)]));
cannam@95 791 }
cannam@95 792 {
cannam@95 793 V T2E, T3I, T13, T27, T3b, T4f;
cannam@95 794 T2E = BYTWJ(&(W[TWVL * 10]), VSUB(T2C, T2D));
cannam@95 795 ST(&(x[WS(vs, 6) + WS(rs, 4)]), T2E, ms, &(x[WS(vs, 6)]));
cannam@95 796 T3I = BYTWJ(&(W[TWVL * 10]), VSUB(T3G, T3H));
cannam@95 797 ST(&(x[WS(vs, 6) + WS(rs, 6)]), T3I, ms, &(x[WS(vs, 6)]));
cannam@95 798 T13 = BYTWJ(&(W[TWVL * 10]), VSUB(T11, T12));
cannam@95 799 ST(&(x[WS(vs, 6) + WS(rs, 1)]), T13, ms, &(x[WS(vs, 6) + WS(rs, 1)]));
cannam@95 800 T27 = BYTWJ(&(W[TWVL * 10]), VSUB(T25, T26));
cannam@95 801 ST(&(x[WS(vs, 6) + WS(rs, 3)]), T27, ms, &(x[WS(vs, 6) + WS(rs, 1)]));
cannam@95 802 T3b = BYTWJ(&(W[TWVL * 10]), VSUB(T39, T3a));
cannam@95 803 ST(&(x[WS(vs, 6) + WS(rs, 5)]), T3b, ms, &(x[WS(vs, 6) + WS(rs, 1)]));
cannam@95 804 T4f = BYTWJ(&(W[TWVL * 10]), VSUB(T4d, T4e));
cannam@95 805 ST(&(x[WS(vs, 6) + WS(rs, 7)]), T4f, ms, &(x[WS(vs, 6) + WS(rs, 1)]));
cannam@95 806 }
cannam@95 807 {
cannam@95 808 V Tx, T1B, T3c, T4g, T3J, T2F;
cannam@95 809 Tx = BYTWJ(&(W[TWVL * 2]), VADD(Tu, Tv));
cannam@95 810 ST(&(x[WS(vs, 2)]), Tx, ms, &(x[WS(vs, 2)]));
cannam@95 811 T1B = BYTWJ(&(W[TWVL * 2]), VADD(T1y, T1z));
cannam@95 812 ST(&(x[WS(vs, 2) + WS(rs, 2)]), T1B, ms, &(x[WS(vs, 2)]));
cannam@95 813 T3c = BYTWJ(&(W[TWVL * 2]), VADD(T39, T3a));
cannam@95 814 ST(&(x[WS(vs, 2) + WS(rs, 5)]), T3c, ms, &(x[WS(vs, 2) + WS(rs, 1)]));
cannam@95 815 T4g = BYTWJ(&(W[TWVL * 2]), VADD(T4d, T4e));
cannam@95 816 ST(&(x[WS(vs, 2) + WS(rs, 7)]), T4g, ms, &(x[WS(vs, 2) + WS(rs, 1)]));
cannam@95 817 T3J = BYTWJ(&(W[TWVL * 2]), VADD(T3G, T3H));
cannam@95 818 ST(&(x[WS(vs, 2) + WS(rs, 6)]), T3J, ms, &(x[WS(vs, 2)]));
cannam@95 819 T2F = BYTWJ(&(W[TWVL * 2]), VADD(T2C, T2D));
cannam@95 820 ST(&(x[WS(vs, 2) + WS(rs, 4)]), T2F, ms, &(x[WS(vs, 2)]));
cannam@95 821 }
cannam@95 822 T28 = BYTWJ(&(W[TWVL * 2]), VADD(T25, T26));
cannam@95 823 ST(&(x[WS(vs, 2) + WS(rs, 3)]), T28, ms, &(x[WS(vs, 2) + WS(rs, 1)]));
cannam@95 824 T14 = BYTWJ(&(W[TWVL * 2]), VADD(T11, T12));
cannam@95 825 ST(&(x[WS(vs, 2) + WS(rs, 1)]), T14, ms, &(x[WS(vs, 2) + WS(rs, 1)]));
cannam@95 826 {
cannam@95 827 V Th, Ti, Tb, Tg;
cannam@95 828 Tb = VADD(T3, Ta);
cannam@95 829 Tg = VBYI(VSUB(Tc, Tf));
cannam@95 830 Th = BYTWJ(&(W[TWVL * 12]), VSUB(Tb, Tg));
cannam@95 831 Ti = BYTWJ(&(W[0]), VADD(Tb, Tg));
cannam@95 832 ST(&(x[WS(vs, 7)]), Th, ms, &(x[WS(vs, 7)]));
cannam@95 833 ST(&(x[WS(vs, 1)]), Ti, ms, &(x[WS(vs, 1)]));
cannam@95 834 }
cannam@95 835 {
cannam@95 836 V T40, T41, T3U, T3Z;
cannam@95 837 T3U = VADD(T3M, T3T);
cannam@95 838 T3Z = VBYI(VSUB(T3V, T3Y));
cannam@95 839 T40 = BYTWJ(&(W[TWVL * 12]), VSUB(T3U, T3Z));
cannam@95 840 T41 = BYTWJ(&(W[0]), VADD(T3U, T3Z));
cannam@95 841 ST(&(x[WS(vs, 7) + WS(rs, 7)]), T40, ms, &(x[WS(vs, 7) + WS(rs, 1)]));
cannam@95 842 ST(&(x[WS(vs, 1) + WS(rs, 7)]), T41, ms, &(x[WS(vs, 1) + WS(rs, 1)]));
cannam@95 843 }
cannam@95 844 {
cannam@95 845 V T2p, T2q, T2j, T2o;
cannam@95 846 T2j = VADD(T2b, T2i);
cannam@95 847 T2o = VBYI(VSUB(T2k, T2n));
cannam@95 848 T2p = BYTWJ(&(W[TWVL * 12]), VSUB(T2j, T2o));
cannam@95 849 T2q = BYTWJ(&(W[0]), VADD(T2j, T2o));
cannam@95 850 ST(&(x[WS(vs, 7) + WS(rs, 4)]), T2p, ms, &(x[WS(vs, 7)]));
cannam@95 851 ST(&(x[WS(vs, 1) + WS(rs, 4)]), T2q, ms, &(x[WS(vs, 1)]));
cannam@95 852 }
cannam@95 853 {
cannam@95 854 V T1S, T1T, T1M, T1R;
cannam@95 855 T1M = VADD(T1E, T1L);
cannam@95 856 T1R = VBYI(VSUB(T1N, T1Q));
cannam@95 857 T1S = BYTWJ(&(W[TWVL * 12]), VSUB(T1M, T1R));
cannam@95 858 T1T = BYTWJ(&(W[0]), VADD(T1M, T1R));
cannam@95 859 ST(&(x[WS(vs, 7) + WS(rs, 3)]), T1S, ms, &(x[WS(vs, 7) + WS(rs, 1)]));
cannam@95 860 ST(&(x[WS(vs, 1) + WS(rs, 3)]), T1T, ms, &(x[WS(vs, 1) + WS(rs, 1)]));
cannam@95 861 }
cannam@95 862 {
cannam@95 863 V TO, TP, TI, TN;
cannam@95 864 TI = VADD(TA, TH);
cannam@95 865 TN = VBYI(VSUB(TJ, TM));
cannam@95 866 TO = BYTWJ(&(W[TWVL * 12]), VSUB(TI, TN));
cannam@95 867 TP = BYTWJ(&(W[0]), VADD(TI, TN));
cannam@95 868 ST(&(x[WS(vs, 7) + WS(rs, 1)]), TO, ms, &(x[WS(vs, 7) + WS(rs, 1)]));
cannam@95 869 ST(&(x[WS(vs, 1) + WS(rs, 1)]), TP, ms, &(x[WS(vs, 1) + WS(rs, 1)]));
cannam@95 870 }
cannam@95 871 {
cannam@95 872 V T1l, T1m, T1f, T1k;
cannam@95 873 T1f = VADD(T17, T1e);
cannam@95 874 T1k = VBYI(VSUB(T1g, T1j));
cannam@95 875 T1l = BYTWJ(&(W[TWVL * 12]), VSUB(T1f, T1k));
cannam@95 876 T1m = BYTWJ(&(W[0]), VADD(T1f, T1k));
cannam@95 877 ST(&(x[WS(vs, 7) + WS(rs, 2)]), T1l, ms, &(x[WS(vs, 7)]));
cannam@95 878 ST(&(x[WS(vs, 1) + WS(rs, 2)]), T1m, ms, &(x[WS(vs, 1)]));
cannam@95 879 }
cannam@95 880 {
cannam@95 881 V T3t, T3u, T3n, T3s;
cannam@95 882 T3n = VADD(T3f, T3m);
cannam@95 883 T3s = VBYI(VSUB(T3o, T3r));
cannam@95 884 T3t = BYTWJ(&(W[TWVL * 12]), VSUB(T3n, T3s));
cannam@95 885 T3u = BYTWJ(&(W[0]), VADD(T3n, T3s));
cannam@95 886 ST(&(x[WS(vs, 7) + WS(rs, 6)]), T3t, ms, &(x[WS(vs, 7)]));
cannam@95 887 ST(&(x[WS(vs, 1) + WS(rs, 6)]), T3u, ms, &(x[WS(vs, 1)]));
cannam@95 888 }
cannam@95 889 {
cannam@95 890 V T2W, T2X, T2Q, T2V;
cannam@95 891 T2Q = VADD(T2I, T2P);
cannam@95 892 T2V = VBYI(VSUB(T2R, T2U));
cannam@95 893 T2W = BYTWJ(&(W[TWVL * 12]), VSUB(T2Q, T2V));
cannam@95 894 T2X = BYTWJ(&(W[0]), VADD(T2Q, T2V));
cannam@95 895 ST(&(x[WS(vs, 7) + WS(rs, 5)]), T2W, ms, &(x[WS(vs, 7) + WS(rs, 1)]));
cannam@95 896 ST(&(x[WS(vs, 1) + WS(rs, 5)]), T2X, ms, &(x[WS(vs, 1) + WS(rs, 1)]));
cannam@95 897 }
cannam@95 898 {
cannam@95 899 V T1p, T1q, T1n, T1o;
cannam@95 900 T1n = VSUB(T17, T1e);
cannam@95 901 T1o = VBYI(VADD(T1j, T1g));
cannam@95 902 T1p = BYTWJ(&(W[TWVL * 8]), VSUB(T1n, T1o));
cannam@95 903 T1q = BYTWJ(&(W[TWVL * 4]), VADD(T1n, T1o));
cannam@95 904 ST(&(x[WS(vs, 5) + WS(rs, 2)]), T1p, ms, &(x[WS(vs, 5)]));
cannam@95 905 ST(&(x[WS(vs, 3) + WS(rs, 2)]), T1q, ms, &(x[WS(vs, 3)]));
cannam@95 906 }
cannam@95 907 {
cannam@95 908 V Tl, Tm, Tj, Tk;
cannam@95 909 Tj = VSUB(T3, Ta);
cannam@95 910 Tk = VBYI(VADD(Tf, Tc));
cannam@95 911 Tl = BYTWJ(&(W[TWVL * 8]), VSUB(Tj, Tk));
cannam@95 912 Tm = BYTWJ(&(W[TWVL * 4]), VADD(Tj, Tk));
cannam@95 913 ST(&(x[WS(vs, 5)]), Tl, ms, &(x[WS(vs, 5)]));
cannam@95 914 ST(&(x[WS(vs, 3)]), Tm, ms, &(x[WS(vs, 3)]));
cannam@95 915 }
cannam@95 916 {
cannam@95 917 V T2t, T2u, T2r, T2s;
cannam@95 918 T2r = VSUB(T2b, T2i);
cannam@95 919 T2s = VBYI(VADD(T2n, T2k));
cannam@95 920 T2t = BYTWJ(&(W[TWVL * 8]), VSUB(T2r, T2s));
cannam@95 921 T2u = BYTWJ(&(W[TWVL * 4]), VADD(T2r, T2s));
cannam@95 922 ST(&(x[WS(vs, 5) + WS(rs, 4)]), T2t, ms, &(x[WS(vs, 5)]));
cannam@95 923 ST(&(x[WS(vs, 3) + WS(rs, 4)]), T2u, ms, &(x[WS(vs, 3)]));
cannam@95 924 }
cannam@95 925 {
cannam@95 926 V T3x, T3y, T3v, T3w;
cannam@95 927 T3v = VSUB(T3f, T3m);
cannam@95 928 T3w = VBYI(VADD(T3r, T3o));
cannam@95 929 T3x = BYTWJ(&(W[TWVL * 8]), VSUB(T3v, T3w));
cannam@95 930 T3y = BYTWJ(&(W[TWVL * 4]), VADD(T3v, T3w));
cannam@95 931 ST(&(x[WS(vs, 5) + WS(rs, 6)]), T3x, ms, &(x[WS(vs, 5)]));
cannam@95 932 ST(&(x[WS(vs, 3) + WS(rs, 6)]), T3y, ms, &(x[WS(vs, 3)]));
cannam@95 933 }
cannam@95 934 {
cannam@95 935 V TS, TT, TQ, TR;
cannam@95 936 TQ = VSUB(TA, TH);
cannam@95 937 TR = VBYI(VADD(TM, TJ));
cannam@95 938 TS = BYTWJ(&(W[TWVL * 8]), VSUB(TQ, TR));
cannam@95 939 TT = BYTWJ(&(W[TWVL * 4]), VADD(TQ, TR));
cannam@95 940 ST(&(x[WS(vs, 5) + WS(rs, 1)]), TS, ms, &(x[WS(vs, 5) + WS(rs, 1)]));
cannam@95 941 ST(&(x[WS(vs, 3) + WS(rs, 1)]), TT, ms, &(x[WS(vs, 3) + WS(rs, 1)]));
cannam@95 942 }
cannam@95 943 {
cannam@95 944 V T1W, T1X, T1U, T1V;
cannam@95 945 T1U = VSUB(T1E, T1L);
cannam@95 946 T1V = VBYI(VADD(T1Q, T1N));
cannam@95 947 T1W = BYTWJ(&(W[TWVL * 8]), VSUB(T1U, T1V));
cannam@95 948 T1X = BYTWJ(&(W[TWVL * 4]), VADD(T1U, T1V));
cannam@95 949 ST(&(x[WS(vs, 5) + WS(rs, 3)]), T1W, ms, &(x[WS(vs, 5) + WS(rs, 1)]));
cannam@95 950 ST(&(x[WS(vs, 3) + WS(rs, 3)]), T1X, ms, &(x[WS(vs, 3) + WS(rs, 1)]));
cannam@95 951 }
cannam@95 952 {
cannam@95 953 V T30, T31, T2Y, T2Z;
cannam@95 954 T2Y = VSUB(T2I, T2P);
cannam@95 955 T2Z = VBYI(VADD(T2U, T2R));
cannam@95 956 T30 = BYTWJ(&(W[TWVL * 8]), VSUB(T2Y, T2Z));
cannam@95 957 T31 = BYTWJ(&(W[TWVL * 4]), VADD(T2Y, T2Z));
cannam@95 958 ST(&(x[WS(vs, 5) + WS(rs, 5)]), T30, ms, &(x[WS(vs, 5) + WS(rs, 1)]));
cannam@95 959 ST(&(x[WS(vs, 3) + WS(rs, 5)]), T31, ms, &(x[WS(vs, 3) + WS(rs, 1)]));
cannam@95 960 }
cannam@95 961 {
cannam@95 962 V T44, T45, T42, T43;
cannam@95 963 T42 = VSUB(T3M, T3T);
cannam@95 964 T43 = VBYI(VADD(T3Y, T3V));
cannam@95 965 T44 = BYTWJ(&(W[TWVL * 8]), VSUB(T42, T43));
cannam@95 966 T45 = BYTWJ(&(W[TWVL * 4]), VADD(T42, T43));
cannam@95 967 ST(&(x[WS(vs, 5) + WS(rs, 7)]), T44, ms, &(x[WS(vs, 5) + WS(rs, 1)]));
cannam@95 968 ST(&(x[WS(vs, 3) + WS(rs, 7)]), T45, ms, &(x[WS(vs, 3) + WS(rs, 1)]));
cannam@95 969 }
cannam@95 970 }
cannam@95 971 }
cannam@95 972 VLEAVE();
cannam@95 973 }
cannam@95 974
cannam@95 975 static const tw_instr twinstr[] = {
cannam@95 976 VTW(0, 1),
cannam@95 977 VTW(0, 2),
cannam@95 978 VTW(0, 3),
cannam@95 979 VTW(0, 4),
cannam@95 980 VTW(0, 5),
cannam@95 981 VTW(0, 6),
cannam@95 982 VTW(0, 7),
cannam@95 983 {TW_NEXT, VL, 0}
cannam@95 984 };
cannam@95 985
cannam@95 986 static const ct_desc desc = { 8, XSIMD_STRING("q1fv_8"), twinstr, &GENUS, {264, 128, 0, 0}, 0, 0, 0 };
cannam@95 987
cannam@95 988 void XSIMD(codelet_q1fv_8) (planner *p) {
cannam@95 989 X(kdft_difsq_register) (p, q1fv_8, &desc);
cannam@95 990 }
cannam@95 991 #endif /* HAVE_FMA */