Mercurial > hg > sv-dependency-builds

annotate src/fftw-3.3.8/dft/simd/common/n2sv_8.c @ 82:d0c2a83c1364

Find changesets by keywords (author, files, the commit message), revision number or hash, or revset expression.
Add FFTW 3.3.8 source, and a Linux build
author Chris Cannam
date Tue, 19 Nov 2019 14:52:55 +0000
parents
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 08:05:19 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.native -fma -simd -compact -variables 4 -pipeline-latency 8 -n 8 -name n2sv_8 -with-ostride 1 -include dft/simd/n2s.h -store-multiple 4 */
Chris@82 29
Chris@82 30 /*
Chris@82 31 * This function contains 52 FP additions, 8 FP multiplications,
Chris@82 32 * (or, 44 additions, 0 multiplications, 8 fused multiply/add),
Chris@82 33 * 34 stack variables, 1 constants, and 36 memory accesses
Chris@82 34 */
Chris@82 35 #include "dft/simd/n2s.h"
Chris@82 36
Chris@82 37 static void n2sv_8(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(KP707106781, +0.707106781186547524400844362104849039284835938);
Chris@82 40 {
Chris@82 41 INT i;
Chris@82 42 for (i = v; i > 0; i = i - (2 * VL), ri = ri + ((2 * VL) * ivs), ii = ii + ((2 * VL) * ivs), ro = ro + ((2 * VL) * ovs), io = io + ((2 * VL) * ovs), MAKE_VOLATILE_STRIDE(32, is), MAKE_VOLATILE_STRIDE(32, os)) {
Chris@82 43 V T3, Tn, Ti, TC, T6, TB, Tl, To, Td, TN, Tz, TH, Ta, TM, Tu;
Chris@82 44 V TG;
Chris@82 45 {
Chris@82 46 V T1, T2, Tj, Tk;
Chris@82 47 T1 = LD(&(ri[0]), ivs, &(ri[0]));
Chris@82 48 T2 = LD(&(ri[WS(is, 4)]), ivs, &(ri[0]));
Chris@82 49 T3 = VADD(T1, T2);
Chris@82 50 Tn = VSUB(T1, T2);
Chris@82 51 {
Chris@82 52 V Tg, Th, T4, T5;
Chris@82 53 Tg = LD(&(ii[0]), ivs, &(ii[0]));
Chris@82 54 Th = LD(&(ii[WS(is, 4)]), ivs, &(ii[0]));
Chris@82 55 Ti = VADD(Tg, Th);
Chris@82 56 TC = VSUB(Tg, Th);
Chris@82 57 T4 = LD(&(ri[WS(is, 2)]), ivs, &(ri[0]));
Chris@82 58 T5 = LD(&(ri[WS(is, 6)]), ivs, &(ri[0]));
Chris@82 59 T6 = VADD(T4, T5);
Chris@82 60 TB = VSUB(T4, T5);
Chris@82 61 }
Chris@82 62 Tj = LD(&(ii[WS(is, 2)]), ivs, &(ii[0]));
Chris@82 63 Tk = LD(&(ii[WS(is, 6)]), ivs, &(ii[0]));
Chris@82 64 Tl = VADD(Tj, Tk);
Chris@82 65 To = VSUB(Tj, Tk);
Chris@82 66 {
Chris@82 67 V Tb, Tc, Tv, Tw, Tx, Ty;
Chris@82 68 Tb = LD(&(ri[WS(is, 7)]), ivs, &(ri[WS(is, 1)]));
Chris@82 69 Tc = LD(&(ri[WS(is, 3)]), ivs, &(ri[WS(is, 1)]));
Chris@82 70 Tv = VSUB(Tb, Tc);
Chris@82 71 Tw = LD(&(ii[WS(is, 7)]), ivs, &(ii[WS(is, 1)]));
Chris@82 72 Tx = LD(&(ii[WS(is, 3)]), ivs, &(ii[WS(is, 1)]));
Chris@82 73 Ty = VSUB(Tw, Tx);
Chris@82 74 Td = VADD(Tb, Tc);
Chris@82 75 TN = VADD(Tw, Tx);
Chris@82 76 Tz = VSUB(Tv, Ty);
Chris@82 77 TH = VADD(Tv, Ty);
Chris@82 78 }
Chris@82 79 {
Chris@82 80 V T8, T9, Tq, Tr, Ts, Tt;
Chris@82 81 T8 = LD(&(ri[WS(is, 1)]), ivs, &(ri[WS(is, 1)]));
Chris@82 82 T9 = LD(&(ri[WS(is, 5)]), ivs, &(ri[WS(is, 1)]));
Chris@82 83 Tq = VSUB(T8, T9);
Chris@82 84 Tr = LD(&(ii[WS(is, 1)]), ivs, &(ii[WS(is, 1)]));
Chris@82 85 Ts = LD(&(ii[WS(is, 5)]), ivs, &(ii[WS(is, 1)]));
Chris@82 86 Tt = VSUB(Tr, Ts);
Chris@82 87 Ta = VADD(T8, T9);
Chris@82 88 TM = VADD(Tr, Ts);
Chris@82 89 Tu = VADD(Tq, Tt);
Chris@82 90 TG = VSUB(Tt, Tq);
Chris@82 91 }
Chris@82 92 }
Chris@82 93 {
Chris@82 94 V TR, TS, TT, TU, TV, TW, TX, TY;
Chris@82 95 {
Chris@82 96 V T7, Te, TP, TQ;
Chris@82 97 T7 = VADD(T3, T6);
Chris@82 98 Te = VADD(Ta, Td);
Chris@82 99 TR = VSUB(T7, Te);
Chris@82 100 STM4(&(ro[4]), TR, ovs, &(ro[0]));
Chris@82 101 TS = VADD(T7, Te);
Chris@82 102 STM4(&(ro[0]), TS, ovs, &(ro[0]));
Chris@82 103 TP = VADD(Ti, Tl);
Chris@82 104 TQ = VADD(TM, TN);
Chris@82 105 TT = VSUB(TP, TQ);
Chris@82 106 STM4(&(io[4]), TT, ovs, &(io[0]));
Chris@82 107 TU = VADD(TP, TQ);
Chris@82 108 STM4(&(io[0]), TU, ovs, &(io[0]));
Chris@82 109 }
Chris@82 110 {
Chris@82 111 V Tf, Tm, TL, TO;
Chris@82 112 Tf = VSUB(Td, Ta);
Chris@82 113 Tm = VSUB(Ti, Tl);
Chris@82 114 TV = VADD(Tf, Tm);
Chris@82 115 STM4(&(io[2]), TV, ovs, &(io[0]));
Chris@82 116 TW = VSUB(Tm, Tf);
Chris@82 117 STM4(&(io[6]), TW, ovs, &(io[0]));
Chris@82 118 TL = VSUB(T3, T6);
Chris@82 119 TO = VSUB(TM, TN);
Chris@82 120 TX = VSUB(TL, TO);
Chris@82 121 STM4(&(ro[6]), TX, ovs, &(ro[0]));
Chris@82 122 TY = VADD(TL, TO);
Chris@82 123 STM4(&(ro[2]), TY, ovs, &(ro[0]));
Chris@82 124 }
Chris@82 125 {
Chris@82 126 V TZ, T10, T11, T12;
Chris@82 127 {
Chris@82 128 V Tp, TA, TJ, TK;
Chris@82 129 Tp = VADD(Tn, To);
Chris@82 130 TA = VADD(Tu, Tz);
Chris@82 131 TZ = VFNMS(LDK(KP707106781), TA, Tp);
Chris@82 132 STM4(&(ro[5]), TZ, ovs, &(ro[1]));
Chris@82 133 T10 = VFMA(LDK(KP707106781), TA, Tp);
Chris@82 134 STM4(&(ro[1]), T10, ovs, &(ro[1]));
Chris@82 135 TJ = VSUB(TC, TB);
Chris@82 136 TK = VADD(TG, TH);
Chris@82 137 T11 = VFNMS(LDK(KP707106781), TK, TJ);
Chris@82 138 STM4(&(io[5]), T11, ovs, &(io[1]));
Chris@82 139 T12 = VFMA(LDK(KP707106781), TK, TJ);
Chris@82 140 STM4(&(io[1]), T12, ovs, &(io[1]));
Chris@82 141 }
Chris@82 142 {
Chris@82 143 V TD, TE, T13, T14;
Chris@82 144 TD = VADD(TB, TC);
Chris@82 145 TE = VSUB(Tz, Tu);
Chris@82 146 T13 = VFNMS(LDK(KP707106781), TE, TD);
Chris@82 147 STM4(&(io[7]), T13, ovs, &(io[1]));
Chris@82 148 STN4(&(io[4]), TT, T11, TW, T13, ovs);
Chris@82 149 T14 = VFMA(LDK(KP707106781), TE, TD);
Chris@82 150 STM4(&(io[3]), T14, ovs, &(io[1]));
Chris@82 151 STN4(&(io[0]), TU, T12, TV, T14, ovs);
Chris@82 152 }
Chris@82 153 {
Chris@82 154 V TF, TI, T15, T16;
Chris@82 155 TF = VSUB(Tn, To);
Chris@82 156 TI = VSUB(TG, TH);
Chris@82 157 T15 = VFNMS(LDK(KP707106781), TI, TF);
Chris@82 158 STM4(&(ro[7]), T15, ovs, &(ro[1]));
Chris@82 159 STN4(&(ro[4]), TR, TZ, TX, T15, ovs);
Chris@82 160 T16 = VFMA(LDK(KP707106781), TI, TF);
Chris@82 161 STM4(&(ro[3]), T16, ovs, &(ro[1]));
Chris@82 162 STN4(&(ro[0]), TS, T10, TY, T16, ovs);
Chris@82 163 }
Chris@82 164 }
Chris@82 165 }
Chris@82 166 }
Chris@82 167 }
Chris@82 168 VLEAVE();
Chris@82 169 }
Chris@82 170
Chris@82 171 static const kdft_desc desc = { 8, XSIMD_STRING("n2sv_8"), {44, 0, 8, 0}, &GENUS, 0, 1, 0, 0 };
Chris@82 172
Chris@82 173 void XSIMD(codelet_n2sv_8) (planner *p) {
Chris@82 174 X(kdft_register) (p, n2sv_8, &desc);
Chris@82 175 }
Chris@82 176
Chris@82 177 #else
Chris@82 178
Chris@82 179 /* Generated by: ../../../genfft/gen_notw.native -simd -compact -variables 4 -pipeline-latency 8 -n 8 -name n2sv_8 -with-ostride 1 -include dft/simd/n2s.h -store-multiple 4 */
Chris@82 180
Chris@82 181 /*
Chris@82 182 * This function contains 52 FP additions, 4 FP multiplications,
Chris@82 183 * (or, 52 additions, 4 multiplications, 0 fused multiply/add),
Chris@82 184 * 34 stack variables, 1 constants, and 36 memory accesses
Chris@82 185 */
Chris@82 186 #include "dft/simd/n2s.h"
Chris@82 187
Chris@82 188 static void n2sv_8(const R *ri, const R *ii, R *ro, R *io, stride is, stride os, INT v, INT ivs, INT ovs)
Chris@82 189 {
Chris@82 190 DVK(KP707106781, +0.707106781186547524400844362104849039284835938);
Chris@82 191 {
Chris@82 192 INT i;
Chris@82 193 for (i = v; i > 0; i = i - (2 * VL), ri = ri + ((2 * VL) * ivs), ii = ii + ((2 * VL) * ivs), ro = ro + ((2 * VL) * ovs), io = io + ((2 * VL) * ovs), MAKE_VOLATILE_STRIDE(32, is), MAKE_VOLATILE_STRIDE(32, os)) {
Chris@82 194 V T3, Tn, Ti, TC, T6, TB, Tl, To, Td, TN, Tz, TH, Ta, TM, Tu;
Chris@82 195 V TG;
Chris@82 196 {
Chris@82 197 V T1, T2, Tj, Tk;
Chris@82 198 T1 = LD(&(ri[0]), ivs, &(ri[0]));
Chris@82 199 T2 = LD(&(ri[WS(is, 4)]), ivs, &(ri[0]));
Chris@82 200 T3 = VADD(T1, T2);
Chris@82 201 Tn = VSUB(T1, T2);
Chris@82 202 {
Chris@82 203 V Tg, Th, T4, T5;
Chris@82 204 Tg = LD(&(ii[0]), ivs, &(ii[0]));
Chris@82 205 Th = LD(&(ii[WS(is, 4)]), ivs, &(ii[0]));
Chris@82 206 Ti = VADD(Tg, Th);
Chris@82 207 TC = VSUB(Tg, Th);
Chris@82 208 T4 = LD(&(ri[WS(is, 2)]), ivs, &(ri[0]));
Chris@82 209 T5 = LD(&(ri[WS(is, 6)]), ivs, &(ri[0]));
Chris@82 210 T6 = VADD(T4, T5);
Chris@82 211 TB = VSUB(T4, T5);
Chris@82 212 }
Chris@82 213 Tj = LD(&(ii[WS(is, 2)]), ivs, &(ii[0]));
Chris@82 214 Tk = LD(&(ii[WS(is, 6)]), ivs, &(ii[0]));
Chris@82 215 Tl = VADD(Tj, Tk);
Chris@82 216 To = VSUB(Tj, Tk);
Chris@82 217 {
Chris@82 218 V Tb, Tc, Tv, Tw, Tx, Ty;
Chris@82 219 Tb = LD(&(ri[WS(is, 7)]), ivs, &(ri[WS(is, 1)]));
Chris@82 220 Tc = LD(&(ri[WS(is, 3)]), ivs, &(ri[WS(is, 1)]));
Chris@82 221 Tv = VSUB(Tb, Tc);
Chris@82 222 Tw = LD(&(ii[WS(is, 7)]), ivs, &(ii[WS(is, 1)]));
Chris@82 223 Tx = LD(&(ii[WS(is, 3)]), ivs, &(ii[WS(is, 1)]));
Chris@82 224 Ty = VSUB(Tw, Tx);
Chris@82 225 Td = VADD(Tb, Tc);
Chris@82 226 TN = VADD(Tw, Tx);
Chris@82 227 Tz = VSUB(Tv, Ty);
Chris@82 228 TH = VADD(Tv, Ty);
Chris@82 229 }
Chris@82 230 {
Chris@82 231 V T8, T9, Tq, Tr, Ts, Tt;
Chris@82 232 T8 = LD(&(ri[WS(is, 1)]), ivs, &(ri[WS(is, 1)]));
Chris@82 233 T9 = LD(&(ri[WS(is, 5)]), ivs, &(ri[WS(is, 1)]));
Chris@82 234 Tq = VSUB(T8, T9);
Chris@82 235 Tr = LD(&(ii[WS(is, 1)]), ivs, &(ii[WS(is, 1)]));
Chris@82 236 Ts = LD(&(ii[WS(is, 5)]), ivs, &(ii[WS(is, 1)]));
Chris@82 237 Tt = VSUB(Tr, Ts);
Chris@82 238 Ta = VADD(T8, T9);
Chris@82 239 TM = VADD(Tr, Ts);
Chris@82 240 Tu = VADD(Tq, Tt);
Chris@82 241 TG = VSUB(Tt, Tq);
Chris@82 242 }
Chris@82 243 }
Chris@82 244 {
Chris@82 245 V TR, TS, TT, TU, TV, TW, TX, TY;
Chris@82 246 {
Chris@82 247 V T7, Te, TP, TQ;
Chris@82 248 T7 = VADD(T3, T6);
Chris@82 249 Te = VADD(Ta, Td);
Chris@82 250 TR = VSUB(T7, Te);
Chris@82 251 STM4(&(ro[4]), TR, ovs, &(ro[0]));
Chris@82 252 TS = VADD(T7, Te);
Chris@82 253 STM4(&(ro[0]), TS, ovs, &(ro[0]));
Chris@82 254 TP = VADD(Ti, Tl);
Chris@82 255 TQ = VADD(TM, TN);
Chris@82 256 TT = VSUB(TP, TQ);
Chris@82 257 STM4(&(io[4]), TT, ovs, &(io[0]));
Chris@82 258 TU = VADD(TP, TQ);
Chris@82 259 STM4(&(io[0]), TU, ovs, &(io[0]));
Chris@82 260 }
Chris@82 261 {
Chris@82 262 V Tf, Tm, TL, TO;
Chris@82 263 Tf = VSUB(Td, Ta);
Chris@82 264 Tm = VSUB(Ti, Tl);
Chris@82 265 TV = VADD(Tf, Tm);
Chris@82 266 STM4(&(io[2]), TV, ovs, &(io[0]));
Chris@82 267 TW = VSUB(Tm, Tf);
Chris@82 268 STM4(&(io[6]), TW, ovs, &(io[0]));
Chris@82 269 TL = VSUB(T3, T6);
Chris@82 270 TO = VSUB(TM, TN);
Chris@82 271 TX = VSUB(TL, TO);
Chris@82 272 STM4(&(ro[6]), TX, ovs, &(ro[0]));
Chris@82 273 TY = VADD(TL, TO);
Chris@82 274 STM4(&(ro[2]), TY, ovs, &(ro[0]));
Chris@82 275 }
Chris@82 276 {
Chris@82 277 V TZ, T10, T11, T12;
Chris@82 278 {
Chris@82 279 V Tp, TA, TJ, TK;
Chris@82 280 Tp = VADD(Tn, To);
Chris@82 281 TA = VMUL(LDK(KP707106781), VADD(Tu, Tz));
Chris@82 282 TZ = VSUB(Tp, TA);
Chris@82 283 STM4(&(ro[5]), TZ, ovs, &(ro[1]));
Chris@82 284 T10 = VADD(Tp, TA);
Chris@82 285 STM4(&(ro[1]), T10, ovs, &(ro[1]));
Chris@82 286 TJ = VSUB(TC, TB);
Chris@82 287 TK = VMUL(LDK(KP707106781), VADD(TG, TH));
Chris@82 288 T11 = VSUB(TJ, TK);
Chris@82 289 STM4(&(io[5]), T11, ovs, &(io[1]));
Chris@82 290 T12 = VADD(TJ, TK);
Chris@82 291 STM4(&(io[1]), T12, ovs, &(io[1]));
Chris@82 292 }
Chris@82 293 {
Chris@82 294 V TD, TE, T13, T14;
Chris@82 295 TD = VADD(TB, TC);
Chris@82 296 TE = VMUL(LDK(KP707106781), VSUB(Tz, Tu));
Chris@82 297 T13 = VSUB(TD, TE);
Chris@82 298 STM4(&(io[7]), T13, ovs, &(io[1]));
Chris@82 299 STN4(&(io[4]), TT, T11, TW, T13, ovs);
Chris@82 300 T14 = VADD(TD, TE);
Chris@82 301 STM4(&(io[3]), T14, ovs, &(io[1]));
Chris@82 302 STN4(&(io[0]), TU, T12, TV, T14, ovs);
Chris@82 303 }
Chris@82 304 {
Chris@82 305 V TF, TI, T15, T16;
Chris@82 306 TF = VSUB(Tn, To);
Chris@82 307 TI = VMUL(LDK(KP707106781), VSUB(TG, TH));
Chris@82 308 T15 = VSUB(TF, TI);
Chris@82 309 STM4(&(ro[7]), T15, ovs, &(ro[1]));
Chris@82 310 STN4(&(ro[4]), TR, TZ, TX, T15, ovs);
Chris@82 311 T16 = VADD(TF, TI);
Chris@82 312 STM4(&(ro[3]), T16, ovs, &(ro[1]));
Chris@82 313 STN4(&(ro[0]), TS, T10, TY, T16, ovs);
Chris@82 314 }
Chris@82 315 }
Chris@82 316 }
Chris@82 317 }
Chris@82 318 }
Chris@82 319 VLEAVE();
Chris@82 320 }
Chris@82 321
Chris@82 322 static const kdft_desc desc = { 8, XSIMD_STRING("n2sv_8"), {52, 4, 0, 0}, &GENUS, 0, 1, 0, 0 };
Chris@82 323
Chris@82 324 void XSIMD(codelet_n2sv_8) (planner *p) {
Chris@82 325 X(kdft_register) (p, n2sv_8, &desc);
Chris@82 326 }
Chris@82 327
Chris@82 328 #endif