Mercurial > hg > sv-dependency-builds

annotate src/fftw-3.3.8/dft/simd/common/n1fv_16.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 08:04:52 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_notw_c.native -fma -simd -compact -variables 4 -pipeline-latency 8 -n 16 -name n1fv_16 -include dft/simd/n1f.h */
Chris@82 29
Chris@82 30 /*
Chris@82 31 * This function contains 72 FP additions, 34 FP multiplications,
Chris@82 32 * (or, 38 additions, 0 multiplications, 34 fused multiply/add),
Chris@82 33 * 30 stack variables, 3 constants, and 32 memory accesses
Chris@82 34 */
Chris@82 35 #include "dft/simd/n1f.h"
Chris@82 36
Chris@82 37 static void n1fv_16(const R *ri, const R *ii, R *ro, R *io, stride is, stride os, INT v, INT ivs, INT ovs)
Chris@82 38 {
Chris@82 39 DVK(KP923879532, +0.923879532511286756128183189396788286822416626);
Chris@82 40 DVK(KP707106781, +0.707106781186547524400844362104849039284835938);
Chris@82 41 DVK(KP414213562, +0.414213562373095048801688724209698078569671875);
Chris@82 42 {
Chris@82 43 INT i;
Chris@82 44 const R *xi;
Chris@82 45 R *xo;
Chris@82 46 xi = ri;
Chris@82 47 xo = ro;
Chris@82 48 for (i = v; i > 0; i = i - VL, xi = xi + (VL * ivs), xo = xo + (VL * ovs), MAKE_VOLATILE_STRIDE(32, is), MAKE_VOLATILE_STRIDE(32, os)) {
Chris@82 49 V T7, TU, Tz, TH, Tu, TV, TA, TK, Te, TX, TC, TO, Tl, TY, TD;
Chris@82 50 V TR;
Chris@82 51 {
Chris@82 52 V T1, T2, T3, T4, T5, T6;
Chris@82 53 T1 = LD(&(xi[0]), ivs, &(xi[0]));
Chris@82 54 T2 = LD(&(xi[WS(is, 8)]), ivs, &(xi[0]));
Chris@82 55 T3 = VADD(T1, T2);
Chris@82 56 T4 = LD(&(xi[WS(is, 4)]), ivs, &(xi[0]));
Chris@82 57 T5 = LD(&(xi[WS(is, 12)]), ivs, &(xi[0]));
Chris@82 58 T6 = VADD(T4, T5);
Chris@82 59 T7 = VSUB(T3, T6);
Chris@82 60 TU = VSUB(T4, T5);
Chris@82 61 Tz = VADD(T3, T6);
Chris@82 62 TH = VSUB(T1, T2);
Chris@82 63 }
Chris@82 64 {
Chris@82 65 V Tq, TJ, Tt, TI;
Chris@82 66 {
Chris@82 67 V To, Tp, Tr, Ts;
Chris@82 68 To = LD(&(xi[WS(is, 14)]), ivs, &(xi[0]));
Chris@82 69 Tp = LD(&(xi[WS(is, 6)]), ivs, &(xi[0]));
Chris@82 70 Tq = VADD(To, Tp);
Chris@82 71 TJ = VSUB(To, Tp);
Chris@82 72 Tr = LD(&(xi[WS(is, 2)]), ivs, &(xi[0]));
Chris@82 73 Ts = LD(&(xi[WS(is, 10)]), ivs, &(xi[0]));
Chris@82 74 Tt = VADD(Tr, Ts);
Chris@82 75 TI = VSUB(Tr, Ts);
Chris@82 76 }
Chris@82 77 Tu = VSUB(Tq, Tt);
Chris@82 78 TV = VSUB(TJ, TI);
Chris@82 79 TA = VADD(Tt, Tq);
Chris@82 80 TK = VADD(TI, TJ);
Chris@82 81 }
Chris@82 82 {
Chris@82 83 V Ta, TM, Td, TN;
Chris@82 84 {
Chris@82 85 V T8, T9, Tb, Tc;
Chris@82 86 T8 = LD(&(xi[WS(is, 1)]), ivs, &(xi[WS(is, 1)]));
Chris@82 87 T9 = LD(&(xi[WS(is, 9)]), ivs, &(xi[WS(is, 1)]));
Chris@82 88 Ta = VADD(T8, T9);
Chris@82 89 TM = VSUB(T8, T9);
Chris@82 90 Tb = LD(&(xi[WS(is, 5)]), ivs, &(xi[WS(is, 1)]));
Chris@82 91 Tc = LD(&(xi[WS(is, 13)]), ivs, &(xi[WS(is, 1)]));
Chris@82 92 Td = VADD(Tb, Tc);
Chris@82 93 TN = VSUB(Tb, Tc);
Chris@82 94 }
Chris@82 95 Te = VSUB(Ta, Td);
Chris@82 96 TX = VFMA(LDK(KP414213562), TM, TN);
Chris@82 97 TC = VADD(Ta, Td);
Chris@82 98 TO = VFNMS(LDK(KP414213562), TN, TM);
Chris@82 99 }
Chris@82 100 {
Chris@82 101 V Th, TP, Tk, TQ;
Chris@82 102 {
Chris@82 103 V Tf, Tg, Ti, Tj;
Chris@82 104 Tf = LD(&(xi[WS(is, 15)]), ivs, &(xi[WS(is, 1)]));
Chris@82 105 Tg = LD(&(xi[WS(is, 7)]), ivs, &(xi[WS(is, 1)]));
Chris@82 106 Th = VADD(Tf, Tg);
Chris@82 107 TP = VSUB(Tf, Tg);
Chris@82 108 Ti = LD(&(xi[WS(is, 3)]), ivs, &(xi[WS(is, 1)]));
Chris@82 109 Tj = LD(&(xi[WS(is, 11)]), ivs, &(xi[WS(is, 1)]));
Chris@82 110 Tk = VADD(Ti, Tj);
Chris@82 111 TQ = VSUB(Tj, Ti);
Chris@82 112 }
Chris@82 113 Tl = VSUB(Th, Tk);
Chris@82 114 TY = VFMA(LDK(KP414213562), TP, TQ);
Chris@82 115 TD = VADD(Th, Tk);
Chris@82 116 TR = VFNMS(LDK(KP414213562), TQ, TP);
Chris@82 117 }
Chris@82 118 {
Chris@82 119 V TB, TE, TF, TG;
Chris@82 120 TB = VADD(Tz, TA);
Chris@82 121 TE = VADD(TC, TD);
Chris@82 122 ST(&(xo[WS(os, 8)]), VSUB(TB, TE), ovs, &(xo[0]));
Chris@82 123 ST(&(xo[0]), VADD(TB, TE), ovs, &(xo[0]));
Chris@82 124 TF = VSUB(Tz, TA);
Chris@82 125 TG = VSUB(TD, TC);
Chris@82 126 ST(&(xo[WS(os, 12)]), VFNMSI(TG, TF), ovs, &(xo[0]));
Chris@82 127 ST(&(xo[WS(os, 4)]), VFMAI(TG, TF), ovs, &(xo[0]));
Chris@82 128 }
Chris@82 129 {
Chris@82 130 V Tn, Tx, Tw, Ty, Tm, Tv;
Chris@82 131 Tm = VADD(Te, Tl);
Chris@82 132 Tn = VFNMS(LDK(KP707106781), Tm, T7);
Chris@82 133 Tx = VFMA(LDK(KP707106781), Tm, T7);
Chris@82 134 Tv = VSUB(Tl, Te);
Chris@82 135 Tw = VFNMS(LDK(KP707106781), Tv, Tu);
Chris@82 136 Ty = VFMA(LDK(KP707106781), Tv, Tu);
Chris@82 137 ST(&(xo[WS(os, 6)]), VFNMSI(Tw, Tn), ovs, &(xo[0]));
Chris@82 138 ST(&(xo[WS(os, 2)]), VFMAI(Ty, Tx), ovs, &(xo[0]));
Chris@82 139 ST(&(xo[WS(os, 10)]), VFMAI(Tw, Tn), ovs, &(xo[0]));
Chris@82 140 ST(&(xo[WS(os, 14)]), VFNMSI(Ty, Tx), ovs, &(xo[0]));
Chris@82 141 }
Chris@82 142 {
Chris@82 143 V TT, T11, T10, T12;
Chris@82 144 {
Chris@82 145 V TL, TS, TW, TZ;
Chris@82 146 TL = VFMA(LDK(KP707106781), TK, TH);
Chris@82 147 TS = VADD(TO, TR);
Chris@82 148 TT = VFNMS(LDK(KP923879532), TS, TL);
Chris@82 149 T11 = VFMA(LDK(KP923879532), TS, TL);
Chris@82 150 TW = VFNMS(LDK(KP707106781), TV, TU);
Chris@82 151 TZ = VSUB(TX, TY);
Chris@82 152 T10 = VFNMS(LDK(KP923879532), TZ, TW);
Chris@82 153 T12 = VFMA(LDK(KP923879532), TZ, TW);
Chris@82 154 }
Chris@82 155 ST(&(xo[WS(os, 9)]), VFNMSI(T10, TT), ovs, &(xo[WS(os, 1)]));
Chris@82 156 ST(&(xo[WS(os, 15)]), VFMAI(T12, T11), ovs, &(xo[WS(os, 1)]));
Chris@82 157 ST(&(xo[WS(os, 7)]), VFMAI(T10, TT), ovs, &(xo[WS(os, 1)]));
Chris@82 158 ST(&(xo[WS(os, 1)]), VFNMSI(T12, T11), ovs, &(xo[WS(os, 1)]));
Chris@82 159 }
Chris@82 160 {
Chris@82 161 V T15, T19, T18, T1a;
Chris@82 162 {
Chris@82 163 V T13, T14, T16, T17;
Chris@82 164 T13 = VFNMS(LDK(KP707106781), TK, TH);
Chris@82 165 T14 = VADD(TX, TY);
Chris@82 166 T15 = VFNMS(LDK(KP923879532), T14, T13);
Chris@82 167 T19 = VFMA(LDK(KP923879532), T14, T13);
Chris@82 168 T16 = VFMA(LDK(KP707106781), TV, TU);
Chris@82 169 T17 = VSUB(TR, TO);
Chris@82 170 T18 = VFNMS(LDK(KP923879532), T17, T16);
Chris@82 171 T1a = VFMA(LDK(KP923879532), T17, T16);
Chris@82 172 }
Chris@82 173 ST(&(xo[WS(os, 5)]), VFNMSI(T18, T15), ovs, &(xo[WS(os, 1)]));
Chris@82 174 ST(&(xo[WS(os, 13)]), VFNMSI(T1a, T19), ovs, &(xo[WS(os, 1)]));
Chris@82 175 ST(&(xo[WS(os, 11)]), VFMAI(T18, T15), ovs, &(xo[WS(os, 1)]));
Chris@82 176 ST(&(xo[WS(os, 3)]), VFMAI(T1a, T19), ovs, &(xo[WS(os, 1)]));
Chris@82 177 }
Chris@82 178 }
Chris@82 179 }
Chris@82 180 VLEAVE();
Chris@82 181 }
Chris@82 182
Chris@82 183 static const kdft_desc desc = { 16, XSIMD_STRING("n1fv_16"), {38, 0, 34, 0}, &GENUS, 0, 0, 0, 0 };
Chris@82 184
Chris@82 185 void XSIMD(codelet_n1fv_16) (planner *p) {
Chris@82 186 X(kdft_register) (p, n1fv_16, &desc);
Chris@82 187 }
Chris@82 188
Chris@82 189 #else
Chris@82 190
Chris@82 191 /* Generated by: ../../../genfft/gen_notw_c.native -simd -compact -variables 4 -pipeline-latency 8 -n 16 -name n1fv_16 -include dft/simd/n1f.h */
Chris@82 192
Chris@82 193 /*
Chris@82 194 * This function contains 72 FP additions, 12 FP multiplications,
Chris@82 195 * (or, 68 additions, 8 multiplications, 4 fused multiply/add),
Chris@82 196 * 30 stack variables, 3 constants, and 32 memory accesses
Chris@82 197 */
Chris@82 198 #include "dft/simd/n1f.h"
Chris@82 199
Chris@82 200 static void n1fv_16(const R *ri, const R *ii, R *ro, R *io, stride is, stride os, INT v, INT ivs, INT ovs)
Chris@82 201 {
Chris@82 202 DVK(KP923879532, +0.923879532511286756128183189396788286822416626);
Chris@82 203 DVK(KP382683432, +0.382683432365089771728459984030398866761344562);
Chris@82 204 DVK(KP707106781, +0.707106781186547524400844362104849039284835938);
Chris@82 205 {
Chris@82 206 INT i;
Chris@82 207 const R *xi;
Chris@82 208 R *xo;
Chris@82 209 xi = ri;
Chris@82 210 xo = ro;
Chris@82 211 for (i = v; i > 0; i = i - VL, xi = xi + (VL * ivs), xo = xo + (VL * ovs), MAKE_VOLATILE_STRIDE(32, is), MAKE_VOLATILE_STRIDE(32, os)) {
Chris@82 212 V Tp, T13, Tu, TN, Tm, T14, Tv, TY, T7, T17, Ty, TT, Te, T16, Tx;
Chris@82 213 V TQ;
Chris@82 214 {
Chris@82 215 V Tn, To, TM, Ts, Tt, TL;
Chris@82 216 Tn = LD(&(xi[WS(is, 4)]), ivs, &(xi[0]));
Chris@82 217 To = LD(&(xi[WS(is, 12)]), ivs, &(xi[0]));
Chris@82 218 TM = VADD(Tn, To);
Chris@82 219 Ts = LD(&(xi[0]), ivs, &(xi[0]));
Chris@82 220 Tt = LD(&(xi[WS(is, 8)]), ivs, &(xi[0]));
Chris@82 221 TL = VADD(Ts, Tt);
Chris@82 222 Tp = VSUB(Tn, To);
Chris@82 223 T13 = VADD(TL, TM);
Chris@82 224 Tu = VSUB(Ts, Tt);
Chris@82 225 TN = VSUB(TL, TM);
Chris@82 226 }
Chris@82 227 {
Chris@82 228 V Ti, TW, Tl, TX;
Chris@82 229 {
Chris@82 230 V Tg, Th, Tj, Tk;
Chris@82 231 Tg = LD(&(xi[WS(is, 14)]), ivs, &(xi[0]));
Chris@82 232 Th = LD(&(xi[WS(is, 6)]), ivs, &(xi[0]));
Chris@82 233 Ti = VSUB(Tg, Th);
Chris@82 234 TW = VADD(Tg, Th);
Chris@82 235 Tj = LD(&(xi[WS(is, 2)]), ivs, &(xi[0]));
Chris@82 236 Tk = LD(&(xi[WS(is, 10)]), ivs, &(xi[0]));
Chris@82 237 Tl = VSUB(Tj, Tk);
Chris@82 238 TX = VADD(Tj, Tk);
Chris@82 239 }
Chris@82 240 Tm = VMUL(LDK(KP707106781), VSUB(Ti, Tl));
Chris@82 241 T14 = VADD(TX, TW);
Chris@82 242 Tv = VMUL(LDK(KP707106781), VADD(Tl, Ti));
Chris@82 243 TY = VSUB(TW, TX);
Chris@82 244 }
Chris@82 245 {
Chris@82 246 V T3, TR, T6, TS;
Chris@82 247 {
Chris@82 248 V T1, T2, T4, T5;
Chris@82 249 T1 = LD(&(xi[WS(is, 15)]), ivs, &(xi[WS(is, 1)]));
Chris@82 250 T2 = LD(&(xi[WS(is, 7)]), ivs, &(xi[WS(is, 1)]));
Chris@82 251 T3 = VSUB(T1, T2);
Chris@82 252 TR = VADD(T1, T2);
Chris@82 253 T4 = LD(&(xi[WS(is, 3)]), ivs, &(xi[WS(is, 1)]));
Chris@82 254 T5 = LD(&(xi[WS(is, 11)]), ivs, &(xi[WS(is, 1)]));
Chris@82 255 T6 = VSUB(T4, T5);
Chris@82 256 TS = VADD(T4, T5);
Chris@82 257 }
Chris@82 258 T7 = VFNMS(LDK(KP923879532), T6, VMUL(LDK(KP382683432), T3));
Chris@82 259 T17 = VADD(TR, TS);
Chris@82 260 Ty = VFMA(LDK(KP923879532), T3, VMUL(LDK(KP382683432), T6));
Chris@82 261 TT = VSUB(TR, TS);
Chris@82 262 }
Chris@82 263 {
Chris@82 264 V Ta, TO, Td, TP;
Chris@82 265 {
Chris@82 266 V T8, T9, Tb, Tc;
Chris@82 267 T8 = LD(&(xi[WS(is, 1)]), ivs, &(xi[WS(is, 1)]));
Chris@82 268 T9 = LD(&(xi[WS(is, 9)]), ivs, &(xi[WS(is, 1)]));
Chris@82 269 Ta = VSUB(T8, T9);
Chris@82 270 TO = VADD(T8, T9);
Chris@82 271 Tb = LD(&(xi[WS(is, 5)]), ivs, &(xi[WS(is, 1)]));
Chris@82 272 Tc = LD(&(xi[WS(is, 13)]), ivs, &(xi[WS(is, 1)]));
Chris@82 273 Td = VSUB(Tb, Tc);
Chris@82 274 TP = VADD(Tb, Tc);
Chris@82 275 }
Chris@82 276 Te = VFMA(LDK(KP382683432), Ta, VMUL(LDK(KP923879532), Td));
Chris@82 277 T16 = VADD(TO, TP);
Chris@82 278 Tx = VFNMS(LDK(KP382683432), Td, VMUL(LDK(KP923879532), Ta));
Chris@82 279 TQ = VSUB(TO, TP);
Chris@82 280 }
Chris@82 281 {
Chris@82 282 V T15, T18, T19, T1a;
Chris@82 283 T15 = VADD(T13, T14);
Chris@82 284 T18 = VADD(T16, T17);
Chris@82 285 ST(&(xo[WS(os, 8)]), VSUB(T15, T18), ovs, &(xo[0]));
Chris@82 286 ST(&(xo[0]), VADD(T15, T18), ovs, &(xo[0]));
Chris@82 287 T19 = VSUB(T13, T14);
Chris@82 288 T1a = VBYI(VSUB(T17, T16));
Chris@82 289 ST(&(xo[WS(os, 12)]), VSUB(T19, T1a), ovs, &(xo[0]));
Chris@82 290 ST(&(xo[WS(os, 4)]), VADD(T19, T1a), ovs, &(xo[0]));
Chris@82 291 }
Chris@82 292 {
Chris@82 293 V TV, T11, T10, T12, TU, TZ;
Chris@82 294 TU = VMUL(LDK(KP707106781), VADD(TQ, TT));
Chris@82 295 TV = VADD(TN, TU);
Chris@82 296 T11 = VSUB(TN, TU);
Chris@82 297 TZ = VMUL(LDK(KP707106781), VSUB(TT, TQ));
Chris@82 298 T10 = VBYI(VADD(TY, TZ));
Chris@82 299 T12 = VBYI(VSUB(TZ, TY));
Chris@82 300 ST(&(xo[WS(os, 14)]), VSUB(TV, T10), ovs, &(xo[0]));
Chris@82 301 ST(&(xo[WS(os, 6)]), VADD(T11, T12), ovs, &(xo[0]));
Chris@82 302 ST(&(xo[WS(os, 2)]), VADD(TV, T10), ovs, &(xo[0]));
Chris@82 303 ST(&(xo[WS(os, 10)]), VSUB(T11, T12), ovs, &(xo[0]));
Chris@82 304 }
Chris@82 305 {
Chris@82 306 V Tr, TB, TA, TC;
Chris@82 307 {
Chris@82 308 V Tf, Tq, Tw, Tz;
Chris@82 309 Tf = VSUB(T7, Te);
Chris@82 310 Tq = VSUB(Tm, Tp);
Chris@82 311 Tr = VBYI(VSUB(Tf, Tq));
Chris@82 312 TB = VBYI(VADD(Tq, Tf));
Chris@82 313 Tw = VADD(Tu, Tv);
Chris@82 314 Tz = VADD(Tx, Ty);
Chris@82 315 TA = VSUB(Tw, Tz);
Chris@82 316 TC = VADD(Tw, Tz);
Chris@82 317 }
Chris@82 318 ST(&(xo[WS(os, 7)]), VADD(Tr, TA), ovs, &(xo[WS(os, 1)]));
Chris@82 319 ST(&(xo[WS(os, 15)]), VSUB(TC, TB), ovs, &(xo[WS(os, 1)]));
Chris@82 320 ST(&(xo[WS(os, 9)]), VSUB(TA, Tr), ovs, &(xo[WS(os, 1)]));
Chris@82 321 ST(&(xo[WS(os, 1)]), VADD(TB, TC), ovs, &(xo[WS(os, 1)]));
Chris@82 322 }
Chris@82 323 {
Chris@82 324 V TF, TJ, TI, TK;
Chris@82 325 {
Chris@82 326 V TD, TE, TG, TH;
Chris@82 327 TD = VSUB(Tu, Tv);
Chris@82 328 TE = VADD(Te, T7);
Chris@82 329 TF = VADD(TD, TE);
Chris@82 330 TJ = VSUB(TD, TE);
Chris@82 331 TG = VADD(Tp, Tm);
Chris@82 332 TH = VSUB(Ty, Tx);
Chris@82 333 TI = VBYI(VADD(TG, TH));
Chris@82 334 TK = VBYI(VSUB(TH, TG));
Chris@82 335 }
Chris@82 336 ST(&(xo[WS(os, 13)]), VSUB(TF, TI), ovs, &(xo[WS(os, 1)]));
Chris@82 337 ST(&(xo[WS(os, 5)]), VADD(TJ, TK), ovs, &(xo[WS(os, 1)]));
Chris@82 338 ST(&(xo[WS(os, 3)]), VADD(TF, TI), ovs, &(xo[WS(os, 1)]));
Chris@82 339 ST(&(xo[WS(os, 11)]), VSUB(TJ, TK), ovs, &(xo[WS(os, 1)]));
Chris@82 340 }
Chris@82 341 }
Chris@82 342 }
Chris@82 343 VLEAVE();
Chris@82 344 }
Chris@82 345
Chris@82 346 static const kdft_desc desc = { 16, XSIMD_STRING("n1fv_16"), {68, 8, 4, 0}, &GENUS, 0, 0, 0, 0 };
Chris@82 347
Chris@82 348 void XSIMD(codelet_n1fv_16) (planner *p) {
Chris@82 349 X(kdft_register) (p, n1fv_16, &desc);
Chris@82 350 }
Chris@82 351
Chris@82 352 #endif