Mercurial > hg > sv-dependency-builds

annotate src/fftw-3.3.5/dft/simd/common/n1bv_14.c @ 157:570d27da3fb5

Find changesets by keywords (author, files, the commit message), revision number or hash, or revset expression.
Update exclusion list
author Chris Cannam <cannam@all-day-breakfast.com>
date Fri, 25 Jan 2019 13:49:22 +0000
parents 7867fa7e1b6b
children
rev   line source
cannam@127 1 /*
cannam@127 2 * Copyright (c) 2003, 2007-14 Matteo Frigo
cannam@127 3 * Copyright (c) 2003, 2007-14 Massachusetts Institute of Technology
cannam@127 4 *
cannam@127 5 * This program is free software; you can redistribute it and/or modify
cannam@127 6 * it under the terms of the GNU General Public License as published by
cannam@127 7 * the Free Software Foundation; either version 2 of the License, or
cannam@127 8 * (at your option) any later version.
cannam@127 9 *
cannam@127 10 * This program is distributed in the hope that it will be useful,
cannam@127 11 * but WITHOUT ANY WARRANTY; without even the implied warranty of
cannam@127 12 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
cannam@127 13 * GNU General Public License for more details.
cannam@127 14 *
cannam@127 15 * You should have received a copy of the GNU General Public License
cannam@127 16 * along with this program; if not, write to the Free Software
cannam@127 17 * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
cannam@127 18 *
cannam@127 19 */
cannam@127 20
cannam@127 21 /* This file was automatically generated --- DO NOT EDIT */
cannam@127 22 /* Generated on Sat Jul 30 16:39:15 EDT 2016 */
cannam@127 23
cannam@127 24 #include "codelet-dft.h"
cannam@127 25
cannam@127 26 #ifdef HAVE_FMA
cannam@127 27
cannam@127 28 /* Generated by: ../../../genfft/gen_notw_c.native -fma -reorder-insns -schedule-for-pipeline -simd -compact -variables 4 -pipeline-latency 8 -sign 1 -n 14 -name n1bv_14 -include n1b.h */
cannam@127 29
cannam@127 30 /*
cannam@127 31 * This function contains 74 FP additions, 48 FP multiplications,
cannam@127 32 * (or, 32 additions, 6 multiplications, 42 fused multiply/add),
cannam@127 33 * 63 stack variables, 6 constants, and 28 memory accesses
cannam@127 34 */
cannam@127 35 #include "n1b.h"
cannam@127 36
cannam@127 37 static void n1bv_14(const R *ri, const R *ii, R *ro, R *io, stride is, stride os, INT v, INT ivs, INT ovs)
cannam@127 38 {
cannam@127 39 DVK(KP900968867, +0.900968867902419126236102319507445051165919162);
cannam@127 40 DVK(KP801937735, +0.801937735804838252472204639014890102331838324);
cannam@127 41 DVK(KP974927912, +0.974927912181823607018131682993931217232785801);
cannam@127 42 DVK(KP692021471, +0.692021471630095869627814897002069140197260599);
cannam@127 43 DVK(KP554958132, +0.554958132087371191422194871006410481067288862);
cannam@127 44 DVK(KP356895867, +0.356895867892209443894399510021300583399127187);
cannam@127 45 {
cannam@127 46 INT i;
cannam@127 47 const R *xi;
cannam@127 48 R *xo;
cannam@127 49 xi = ii;
cannam@127 50 xo = io;
cannam@127 51 for (i = v; i > 0; i = i - VL, xi = xi + (VL * ivs), xo = xo + (VL * ovs), MAKE_VOLATILE_STRIDE(28, is), MAKE_VOLATILE_STRIDE(28, os)) {
cannam@127 52 V TH, T3, TP, Tn, Ta, Tu, TU, TK, TO, Tk, TM, Tg, TL, Td, T1;
cannam@127 53 V T2;
cannam@127 54 T1 = LD(&(xi[0]), ivs, &(xi[0]));
cannam@127 55 T2 = LD(&(xi[WS(is, 7)]), ivs, &(xi[WS(is, 1)]));
cannam@127 56 {
cannam@127 57 V Ti, TI, T6, TJ, T9, Tj, Te, Tf, Tb, Tc;
cannam@127 58 {
cannam@127 59 V T4, T5, T7, T8, Tl, Tm;
cannam@127 60 T4 = LD(&(xi[WS(is, 2)]), ivs, &(xi[0]));
cannam@127 61 T5 = LD(&(xi[WS(is, 9)]), ivs, &(xi[WS(is, 1)]));
cannam@127 62 T7 = LD(&(xi[WS(is, 12)]), ivs, &(xi[0]));
cannam@127 63 T8 = LD(&(xi[WS(is, 5)]), ivs, &(xi[WS(is, 1)]));
cannam@127 64 Tl = LD(&(xi[WS(is, 8)]), ivs, &(xi[0]));
cannam@127 65 Tm = LD(&(xi[WS(is, 1)]), ivs, &(xi[WS(is, 1)]));
cannam@127 66 Ti = LD(&(xi[WS(is, 6)]), ivs, &(xi[0]));
cannam@127 67 TH = VADD(T1, T2);
cannam@127 68 T3 = VSUB(T1, T2);
cannam@127 69 TI = VADD(T4, T5);
cannam@127 70 T6 = VSUB(T4, T5);
cannam@127 71 TJ = VADD(T7, T8);
cannam@127 72 T9 = VSUB(T7, T8);
cannam@127 73 TP = VADD(Tl, Tm);
cannam@127 74 Tn = VSUB(Tl, Tm);
cannam@127 75 Tj = LD(&(xi[WS(is, 13)]), ivs, &(xi[WS(is, 1)]));
cannam@127 76 Te = LD(&(xi[WS(is, 10)]), ivs, &(xi[0]));
cannam@127 77 Tf = LD(&(xi[WS(is, 3)]), ivs, &(xi[WS(is, 1)]));
cannam@127 78 Tb = LD(&(xi[WS(is, 4)]), ivs, &(xi[0]));
cannam@127 79 Tc = LD(&(xi[WS(is, 11)]), ivs, &(xi[WS(is, 1)]));
cannam@127 80 }
cannam@127 81 Ta = VADD(T6, T9);
cannam@127 82 Tu = VSUB(T6, T9);
cannam@127 83 TU = VSUB(TI, TJ);
cannam@127 84 TK = VADD(TI, TJ);
cannam@127 85 TO = VADD(Ti, Tj);
cannam@127 86 Tk = VSUB(Ti, Tj);
cannam@127 87 TM = VADD(Te, Tf);
cannam@127 88 Tg = VSUB(Te, Tf);
cannam@127 89 TL = VADD(Tb, Tc);
cannam@127 90 Td = VSUB(Tb, Tc);
cannam@127 91 }
cannam@127 92 {
cannam@127 93 V T13, TG, TY, T18, TB, Tw, TT, Tz, T11, T16, TE, Tr, TV, TQ;
cannam@127 94 TV = VSUB(TP, TO);
cannam@127 95 TQ = VADD(TO, TP);
cannam@127 96 {
cannam@127 97 V Ts, To, TW, TN;
cannam@127 98 Ts = VSUB(Tk, Tn);
cannam@127 99 To = VADD(Tk, Tn);
cannam@127 100 TW = VSUB(TM, TL);
cannam@127 101 TN = VADD(TL, TM);
cannam@127 102 {
cannam@127 103 V Tt, Th, TR, T12;
cannam@127 104 Tt = VSUB(Td, Tg);
cannam@127 105 Th = VADD(Td, Tg);
cannam@127 106 TR = VFNMS(LDK(KP356895867), TK, TQ);
cannam@127 107 T12 = VFNMS(LDK(KP554958132), TV, TU);
cannam@127 108 {
cannam@127 109 V Tx, TF, TZ, T14;
cannam@127 110 Tx = VFNMS(LDK(KP356895867), Ta, To);
cannam@127 111 TF = VFMA(LDK(KP554958132), Ts, Tu);
cannam@127 112 ST(&(xo[0]), VADD(TH, VADD(TK, VADD(TN, TQ))), ovs, &(xo[0]));
cannam@127 113 TZ = VFNMS(LDK(KP356895867), TN, TK);
cannam@127 114 T14 = VFNMS(LDK(KP356895867), TQ, TN);
cannam@127 115 {
cannam@127 116 V TX, T17, TC, Tp;
cannam@127 117 TX = VFMA(LDK(KP554958132), TW, TV);
cannam@127 118 T17 = VFMA(LDK(KP554958132), TU, TW);
cannam@127 119 ST(&(xo[WS(os, 7)]), VADD(T3, VADD(Ta, VADD(Th, To))), ovs, &(xo[WS(os, 1)]));
cannam@127 120 TC = VFNMS(LDK(KP356895867), Th, Ta);
cannam@127 121 Tp = VFNMS(LDK(KP356895867), To, Th);
cannam@127 122 {
cannam@127 123 V TA, Tv, TS, Ty;
cannam@127 124 TA = VFMA(LDK(KP554958132), Tt, Ts);
cannam@127 125 Tv = VFNMS(LDK(KP554958132), Tu, Tt);
cannam@127 126 TS = VFNMS(LDK(KP692021471), TR, TN);
cannam@127 127 T13 = VMUL(LDK(KP974927912), VFNMS(LDK(KP801937735), T12, TW));
cannam@127 128 Ty = VFNMS(LDK(KP692021471), Tx, Th);
cannam@127 129 TG = VMUL(LDK(KP974927912), VFMA(LDK(KP801937735), TF, Tt));
cannam@127 130 {
cannam@127 131 V T10, T15, TD, Tq;
cannam@127 132 T10 = VFNMS(LDK(KP692021471), TZ, TQ);
cannam@127 133 T15 = VFNMS(LDK(KP692021471), T14, TK);
cannam@127 134 TY = VMUL(LDK(KP974927912), VFMA(LDK(KP801937735), TX, TU));
cannam@127 135 T18 = VMUL(LDK(KP974927912), VFNMS(LDK(KP801937735), T17, TV));
cannam@127 136 TD = VFNMS(LDK(KP692021471), TC, To);
cannam@127 137 Tq = VFNMS(LDK(KP692021471), Tp, Ta);
cannam@127 138 TB = VMUL(LDK(KP974927912), VFNMS(LDK(KP801937735), TA, Tu));
cannam@127 139 Tw = VMUL(LDK(KP974927912), VFNMS(LDK(KP801937735), Tv, Ts));
cannam@127 140 TT = VFNMS(LDK(KP900968867), TS, TH);
cannam@127 141 Tz = VFNMS(LDK(KP900968867), Ty, T3);
cannam@127 142 T11 = VFNMS(LDK(KP900968867), T10, TH);
cannam@127 143 T16 = VFNMS(LDK(KP900968867), T15, TH);
cannam@127 144 TE = VFNMS(LDK(KP900968867), TD, T3);
cannam@127 145 Tr = VFNMS(LDK(KP900968867), Tq, T3);
cannam@127 146 }
cannam@127 147 }
cannam@127 148 }
cannam@127 149 }
cannam@127 150 }
cannam@127 151 }
cannam@127 152 ST(&(xo[WS(os, 2)]), VFMAI(TY, TT), ovs, &(xo[0]));
cannam@127 153 ST(&(xo[WS(os, 12)]), VFNMSI(TY, TT), ovs, &(xo[0]));
cannam@127 154 ST(&(xo[WS(os, 9)]), VFMAI(TB, Tz), ovs, &(xo[WS(os, 1)]));
cannam@127 155 ST(&(xo[WS(os, 5)]), VFNMSI(TB, Tz), ovs, &(xo[WS(os, 1)]));
cannam@127 156 ST(&(xo[WS(os, 6)]), VFMAI(T13, T11), ovs, &(xo[0]));
cannam@127 157 ST(&(xo[WS(os, 8)]), VFNMSI(T13, T11), ovs, &(xo[0]));
cannam@127 158 ST(&(xo[WS(os, 4)]), VFMAI(T18, T16), ovs, &(xo[0]));
cannam@127 159 ST(&(xo[WS(os, 10)]), VFNMSI(T18, T16), ovs, &(xo[0]));
cannam@127 160 ST(&(xo[WS(os, 13)]), VFNMSI(TG, TE), ovs, &(xo[WS(os, 1)]));
cannam@127 161 ST(&(xo[WS(os, 1)]), VFMAI(TG, TE), ovs, &(xo[WS(os, 1)]));
cannam@127 162 ST(&(xo[WS(os, 11)]), VFNMSI(Tw, Tr), ovs, &(xo[WS(os, 1)]));
cannam@127 163 ST(&(xo[WS(os, 3)]), VFMAI(Tw, Tr), ovs, &(xo[WS(os, 1)]));
cannam@127 164 }
cannam@127 165 }
cannam@127 166 }
cannam@127 167 VLEAVE();
cannam@127 168 }
cannam@127 169
cannam@127 170 static const kdft_desc desc = { 14, XSIMD_STRING("n1bv_14"), {32, 6, 42, 0}, &GENUS, 0, 0, 0, 0 };
cannam@127 171
cannam@127 172 void XSIMD(codelet_n1bv_14) (planner *p) {
cannam@127 173 X(kdft_register) (p, n1bv_14, &desc);
cannam@127 174 }
cannam@127 175
cannam@127 176 #else /* HAVE_FMA */
cannam@127 177
cannam@127 178 /* Generated by: ../../../genfft/gen_notw_c.native -simd -compact -variables 4 -pipeline-latency 8 -sign 1 -n 14 -name n1bv_14 -include n1b.h */
cannam@127 179
cannam@127 180 /*
cannam@127 181 * This function contains 74 FP additions, 36 FP multiplications,
cannam@127 182 * (or, 50 additions, 12 multiplications, 24 fused multiply/add),
cannam@127 183 * 33 stack variables, 6 constants, and 28 memory accesses
cannam@127 184 */
cannam@127 185 #include "n1b.h"
cannam@127 186
cannam@127 187 static void n1bv_14(const R *ri, const R *ii, R *ro, R *io, stride is, stride os, INT v, INT ivs, INT ovs)
cannam@127 188 {
cannam@127 189 DVK(KP900968867, +0.900968867902419126236102319507445051165919162);
cannam@127 190 DVK(KP222520933, +0.222520933956314404288902564496794759466355569);
cannam@127 191 DVK(KP623489801, +0.623489801858733530525004884004239810632274731);
cannam@127 192 DVK(KP781831482, +0.781831482468029808708444526674057750232334519);
cannam@127 193 DVK(KP974927912, +0.974927912181823607018131682993931217232785801);
cannam@127 194 DVK(KP433883739, +0.433883739117558120475768332848358754609990728);
cannam@127 195 {
cannam@127 196 INT i;
cannam@127 197 const R *xi;
cannam@127 198 R *xo;
cannam@127 199 xi = ii;
cannam@127 200 xo = io;
cannam@127 201 for (i = v; i > 0; i = i - VL, xi = xi + (VL * ivs), xo = xo + (VL * ovs), MAKE_VOLATILE_STRIDE(28, is), MAKE_VOLATILE_STRIDE(28, os)) {
cannam@127 202 V Tp, Ty, Tl, TL, Tq, TE, T7, TJ, Ts, TB, Te, TK, Tr, TH, Tn;
cannam@127 203 V To;
cannam@127 204 Tn = LD(&(xi[0]), ivs, &(xi[0]));
cannam@127 205 To = LD(&(xi[WS(is, 7)]), ivs, &(xi[WS(is, 1)]));
cannam@127 206 Tp = VSUB(Tn, To);
cannam@127 207 Ty = VADD(Tn, To);
cannam@127 208 {
cannam@127 209 V Th, TC, Tk, TD;
cannam@127 210 {
cannam@127 211 V Tf, Tg, Ti, Tj;
cannam@127 212 Tf = LD(&(xi[WS(is, 4)]), ivs, &(xi[0]));
cannam@127 213 Tg = LD(&(xi[WS(is, 11)]), ivs, &(xi[WS(is, 1)]));
cannam@127 214 Th = VSUB(Tf, Tg);
cannam@127 215 TC = VADD(Tf, Tg);
cannam@127 216 Ti = LD(&(xi[WS(is, 10)]), ivs, &(xi[0]));
cannam@127 217 Tj = LD(&(xi[WS(is, 3)]), ivs, &(xi[WS(is, 1)]));
cannam@127 218 Tk = VSUB(Ti, Tj);
cannam@127 219 TD = VADD(Ti, Tj);
cannam@127 220 }
cannam@127 221 Tl = VSUB(Th, Tk);
cannam@127 222 TL = VSUB(TD, TC);
cannam@127 223 Tq = VADD(Th, Tk);
cannam@127 224 TE = VADD(TC, TD);
cannam@127 225 }
cannam@127 226 {
cannam@127 227 V T3, Tz, T6, TA;
cannam@127 228 {
cannam@127 229 V T1, T2, T4, T5;
cannam@127 230 T1 = LD(&(xi[WS(is, 2)]), ivs, &(xi[0]));
cannam@127 231 T2 = LD(&(xi[WS(is, 9)]), ivs, &(xi[WS(is, 1)]));
cannam@127 232 T3 = VSUB(T1, T2);
cannam@127 233 Tz = VADD(T1, T2);
cannam@127 234 T4 = LD(&(xi[WS(is, 12)]), ivs, &(xi[0]));
cannam@127 235 T5 = LD(&(xi[WS(is, 5)]), ivs, &(xi[WS(is, 1)]));
cannam@127 236 T6 = VSUB(T4, T5);
cannam@127 237 TA = VADD(T4, T5);
cannam@127 238 }
cannam@127 239 T7 = VSUB(T3, T6);
cannam@127 240 TJ = VSUB(Tz, TA);
cannam@127 241 Ts = VADD(T3, T6);
cannam@127 242 TB = VADD(Tz, TA);
cannam@127 243 }
cannam@127 244 {
cannam@127 245 V Ta, TF, Td, TG;
cannam@127 246 {
cannam@127 247 V T8, T9, Tb, Tc;
cannam@127 248 T8 = LD(&(xi[WS(is, 6)]), ivs, &(xi[0]));
cannam@127 249 T9 = LD(&(xi[WS(is, 13)]), ivs, &(xi[WS(is, 1)]));
cannam@127 250 Ta = VSUB(T8, T9);
cannam@127 251 TF = VADD(T8, T9);
cannam@127 252 Tb = LD(&(xi[WS(is, 8)]), ivs, &(xi[0]));
cannam@127 253 Tc = LD(&(xi[WS(is, 1)]), ivs, &(xi[WS(is, 1)]));
cannam@127 254 Td = VSUB(Tb, Tc);
cannam@127 255 TG = VADD(Tb, Tc);
cannam@127 256 }
cannam@127 257 Te = VSUB(Ta, Td);
cannam@127 258 TK = VSUB(TG, TF);
cannam@127 259 Tr = VADD(Ta, Td);
cannam@127 260 TH = VADD(TF, TG);
cannam@127 261 }
cannam@127 262 ST(&(xo[WS(os, 7)]), VADD(Tp, VADD(Ts, VADD(Tq, Tr))), ovs, &(xo[WS(os, 1)]));
cannam@127 263 ST(&(xo[0]), VADD(Ty, VADD(TB, VADD(TE, TH))), ovs, &(xo[0]));
cannam@127 264 {
cannam@127 265 V Tm, Tt, TQ, TP;
cannam@127 266 Tm = VBYI(VFMA(LDK(KP433883739), T7, VFNMS(LDK(KP781831482), Tl, VMUL(LDK(KP974927912), Te))));
cannam@127 267 Tt = VFMA(LDK(KP623489801), Tq, VFNMS(LDK(KP222520933), Tr, VFNMS(LDK(KP900968867), Ts, Tp)));
cannam@127 268 ST(&(xo[WS(os, 3)]), VADD(Tm, Tt), ovs, &(xo[WS(os, 1)]));
cannam@127 269 ST(&(xo[WS(os, 11)]), VSUB(Tt, Tm), ovs, &(xo[WS(os, 1)]));
cannam@127 270 TQ = VBYI(VFMA(LDK(KP974927912), TJ, VFMA(LDK(KP433883739), TL, VMUL(LDK(KP781831482), TK))));
cannam@127 271 TP = VFMA(LDK(KP623489801), TH, VFNMS(LDK(KP900968867), TE, VFNMS(LDK(KP222520933), TB, Ty)));
cannam@127 272 ST(&(xo[WS(os, 12)]), VSUB(TP, TQ), ovs, &(xo[0]));
cannam@127 273 ST(&(xo[WS(os, 2)]), VADD(TP, TQ), ovs, &(xo[0]));
cannam@127 274 }
cannam@127 275 {
cannam@127 276 V Tu, Tv, TM, TI;
cannam@127 277 Tu = VBYI(VFMA(LDK(KP781831482), T7, VFMA(LDK(KP974927912), Tl, VMUL(LDK(KP433883739), Te))));
cannam@127 278 Tv = VFMA(LDK(KP623489801), Ts, VFNMS(LDK(KP900968867), Tr, VFNMS(LDK(KP222520933), Tq, Tp)));
cannam@127 279 ST(&(xo[WS(os, 1)]), VADD(Tu, Tv), ovs, &(xo[WS(os, 1)]));
cannam@127 280 ST(&(xo[WS(os, 13)]), VSUB(Tv, Tu), ovs, &(xo[WS(os, 1)]));
cannam@127 281 TM = VBYI(VFNMS(LDK(KP433883739), TK, VFNMS(LDK(KP974927912), TL, VMUL(LDK(KP781831482), TJ))));
cannam@127 282 TI = VFMA(LDK(KP623489801), TB, VFNMS(LDK(KP900968867), TH, VFNMS(LDK(KP222520933), TE, Ty)));
cannam@127 283 ST(&(xo[WS(os, 6)]), VSUB(TI, TM), ovs, &(xo[0]));
cannam@127 284 ST(&(xo[WS(os, 8)]), VADD(TI, TM), ovs, &(xo[0]));
cannam@127 285 }
cannam@127 286 {
cannam@127 287 V TO, TN, Tx, Tw;
cannam@127 288 TO = VBYI(VFMA(LDK(KP433883739), TJ, VFNMS(LDK(KP974927912), TK, VMUL(LDK(KP781831482), TL))));
cannam@127 289 TN = VFMA(LDK(KP623489801), TE, VFNMS(LDK(KP222520933), TH, VFNMS(LDK(KP900968867), TB, Ty)));
cannam@127 290 ST(&(xo[WS(os, 4)]), VSUB(TN, TO), ovs, &(xo[0]));
cannam@127 291 ST(&(xo[WS(os, 10)]), VADD(TN, TO), ovs, &(xo[0]));
cannam@127 292 Tx = VBYI(VFNMS(LDK(KP781831482), Te, VFNMS(LDK(KP433883739), Tl, VMUL(LDK(KP974927912), T7))));
cannam@127 293 Tw = VFMA(LDK(KP623489801), Tr, VFNMS(LDK(KP900968867), Tq, VFNMS(LDK(KP222520933), Ts, Tp)));
cannam@127 294 ST(&(xo[WS(os, 5)]), VSUB(Tw, Tx), ovs, &(xo[WS(os, 1)]));
cannam@127 295 ST(&(xo[WS(os, 9)]), VADD(Tx, Tw), ovs, &(xo[WS(os, 1)]));
cannam@127 296 }
cannam@127 297 }
cannam@127 298 }
cannam@127 299 VLEAVE();
cannam@127 300 }
cannam@127 301
cannam@127 302 static const kdft_desc desc = { 14, XSIMD_STRING("n1bv_14"), {50, 12, 24, 0}, &GENUS, 0, 0, 0, 0 };
cannam@127 303
cannam@127 304 void XSIMD(codelet_n1bv_14) (planner *p) {
cannam@127 305 X(kdft_register) (p, n1bv_14, &desc);
cannam@127 306 }
cannam@127 307
cannam@127 308 #endif /* HAVE_FMA */