Mercurial > hg > sv-dependency-builds

annotate src/fftw-3.3.5/dft/simd/common/t1fv_12.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 2cd0e3b3e1fd
children
rev   line source
Chris@42 1 /*
Chris@42 2 * Copyright (c) 2003, 2007-14 Matteo Frigo
Chris@42 3 * Copyright (c) 2003, 2007-14 Massachusetts Institute of Technology
Chris@42 4 *
Chris@42 5 * This program is free software; you can redistribute it and/or modify
Chris@42 6 * it under the terms of the GNU General Public License as published by
Chris@42 7 * the Free Software Foundation; either version 2 of the License, or
Chris@42 8 * (at your option) any later version.
Chris@42 9 *
Chris@42 10 * This program is distributed in the hope that it will be useful,
Chris@42 11 * but WITHOUT ANY WARRANTY; without even the implied warranty of
Chris@42 12 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
Chris@42 13 * GNU General Public License for more details.
Chris@42 14 *
Chris@42 15 * You should have received a copy of the GNU General Public License
Chris@42 16 * along with this program; if not, write to the Free Software
Chris@42 17 * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
Chris@42 18 *
Chris@42 19 */
Chris@42 20
Chris@42 21 /* This file was automatically generated --- DO NOT EDIT */
Chris@42 22 /* Generated on Sat Jul 30 16:42:01 EDT 2016 */
Chris@42 23
Chris@42 24 #include "codelet-dft.h"
Chris@42 25
Chris@42 26 #ifdef HAVE_FMA
Chris@42 27
Chris@42 28 /* Generated by: ../../../genfft/gen_twiddle_c.native -fma -reorder-insns -schedule-for-pipeline -simd -compact -variables 4 -pipeline-latency 8 -n 12 -name t1fv_12 -include t1f.h */
Chris@42 29
Chris@42 30 /*
Chris@42 31 * This function contains 59 FP additions, 42 FP multiplications,
Chris@42 32 * (or, 41 additions, 24 multiplications, 18 fused multiply/add),
Chris@42 33 * 41 stack variables, 2 constants, and 24 memory accesses
Chris@42 34 */
Chris@42 35 #include "t1f.h"
Chris@42 36
Chris@42 37 static void t1fv_12(R *ri, R *ii, const R *W, stride rs, INT mb, INT me, INT ms)
Chris@42 38 {
Chris@42 39 DVK(KP866025403, +0.866025403784438646763723170752936183471402627);
Chris@42 40 DVK(KP500000000, +0.500000000000000000000000000000000000000000000);
Chris@42 41 {
Chris@42 42 INT m;
Chris@42 43 R *x;
Chris@42 44 x = ri;
Chris@42 45 for (m = mb, W = W + (mb * ((TWVL / VL) * 22)); m < me; m = m + VL, x = x + (VL * ms), W = W + (TWVL * 22), MAKE_VOLATILE_STRIDE(12, rs)) {
Chris@42 46 V Tq, Ti, T7, TQ, Tu, TA, TU, Tk, TR, Tf, TE, TM;
Chris@42 47 {
Chris@42 48 V T9, TC, Tj, TD, Te;
Chris@42 49 {
Chris@42 50 V T1, T4, T2, Tm, Tx, To;
Chris@42 51 T1 = LD(&(x[0]), ms, &(x[0]));
Chris@42 52 T4 = LD(&(x[WS(rs, 8)]), ms, &(x[0]));
Chris@42 53 T2 = LD(&(x[WS(rs, 4)]), ms, &(x[0]));
Chris@42 54 Tm = LD(&(x[WS(rs, 1)]), ms, &(x[WS(rs, 1)]));
Chris@42 55 Tx = LD(&(x[WS(rs, 9)]), ms, &(x[WS(rs, 1)]));
Chris@42 56 To = LD(&(x[WS(rs, 5)]), ms, &(x[WS(rs, 1)]));
Chris@42 57 {
Chris@42 58 V T5, T3, Tn, Ty, Tp, Td, Tb, T8, Tc, Ta;
Chris@42 59 T8 = LD(&(x[WS(rs, 6)]), ms, &(x[0]));
Chris@42 60 Tc = LD(&(x[WS(rs, 2)]), ms, &(x[0]));
Chris@42 61 Ta = LD(&(x[WS(rs, 10)]), ms, &(x[0]));
Chris@42 62 T5 = BYTWJ(&(W[TWVL * 14]), T4);
Chris@42 63 T3 = BYTWJ(&(W[TWVL * 6]), T2);
Chris@42 64 Tn = BYTWJ(&(W[0]), Tm);
Chris@42 65 Ty = BYTWJ(&(W[TWVL * 16]), Tx);
Chris@42 66 Tp = BYTWJ(&(W[TWVL * 8]), To);
Chris@42 67 T9 = BYTWJ(&(W[TWVL * 10]), T8);
Chris@42 68 Td = BYTWJ(&(W[TWVL * 2]), Tc);
Chris@42 69 Tb = BYTWJ(&(W[TWVL * 18]), Ta);
Chris@42 70 {
Chris@42 71 V Th, T6, Tt, Tz;
Chris@42 72 Th = LD(&(x[WS(rs, 11)]), ms, &(x[WS(rs, 1)]));
Chris@42 73 TC = VSUB(T5, T3);
Chris@42 74 T6 = VADD(T3, T5);
Chris@42 75 Tt = LD(&(x[WS(rs, 3)]), ms, &(x[WS(rs, 1)]));
Chris@42 76 Tz = VADD(Tn, Tp);
Chris@42 77 Tq = VSUB(Tn, Tp);
Chris@42 78 Tj = LD(&(x[WS(rs, 7)]), ms, &(x[WS(rs, 1)]));
Chris@42 79 TD = VSUB(Td, Tb);
Chris@42 80 Te = VADD(Tb, Td);
Chris@42 81 Ti = BYTWJ(&(W[TWVL * 20]), Th);
Chris@42 82 T7 = VFNMS(LDK(KP500000000), T6, T1);
Chris@42 83 TQ = VADD(T1, T6);
Chris@42 84 Tu = BYTWJ(&(W[TWVL * 4]), Tt);
Chris@42 85 TA = VFNMS(LDK(KP500000000), Tz, Ty);
Chris@42 86 TU = VADD(Ty, Tz);
Chris@42 87 }
Chris@42 88 }
Chris@42 89 }
Chris@42 90 Tk = BYTWJ(&(W[TWVL * 12]), Tj);
Chris@42 91 TR = VADD(T9, Te);
Chris@42 92 Tf = VFNMS(LDK(KP500000000), Te, T9);
Chris@42 93 TE = VSUB(TC, TD);
Chris@42 94 TM = VADD(TC, TD);
Chris@42 95 }
Chris@42 96 {
Chris@42 97 V Tv, Tl, TI, Tg, TW, TS;
Chris@42 98 Tv = VADD(Tk, Ti);
Chris@42 99 Tl = VSUB(Ti, Tk);
Chris@42 100 TI = VADD(T7, Tf);
Chris@42 101 Tg = VSUB(T7, Tf);
Chris@42 102 TW = VADD(TQ, TR);
Chris@42 103 TS = VSUB(TQ, TR);
Chris@42 104 {
Chris@42 105 V TT, Tw, TL, Tr;
Chris@42 106 TT = VADD(Tu, Tv);
Chris@42 107 Tw = VFNMS(LDK(KP500000000), Tv, Tu);
Chris@42 108 TL = VSUB(Tl, Tq);
Chris@42 109 Tr = VADD(Tl, Tq);
Chris@42 110 {
Chris@42 111 V TP, TN, TG, Ts, TO, TK, TH, TF;
Chris@42 112 {
Chris@42 113 V TX, TV, TJ, TB;
Chris@42 114 TX = VADD(TT, TU);
Chris@42 115 TV = VSUB(TT, TU);
Chris@42 116 TJ = VADD(Tw, TA);
Chris@42 117 TB = VSUB(Tw, TA);
Chris@42 118 TP = VMUL(LDK(KP866025403), VADD(TM, TL));
Chris@42 119 TN = VMUL(LDK(KP866025403), VSUB(TL, TM));
Chris@42 120 TG = VFNMS(LDK(KP866025403), Tr, Tg);
Chris@42 121 Ts = VFMA(LDK(KP866025403), Tr, Tg);
Chris@42 122 ST(&(x[WS(rs, 6)]), VSUB(TW, TX), ms, &(x[0]));
Chris@42 123 ST(&(x[0]), VADD(TW, TX), ms, &(x[0]));
Chris@42 124 ST(&(x[WS(rs, 3)]), VFMAI(TV, TS), ms, &(x[WS(rs, 1)]));
Chris@42 125 ST(&(x[WS(rs, 9)]), VFNMSI(TV, TS), ms, &(x[WS(rs, 1)]));
Chris@42 126 TO = VADD(TI, TJ);
Chris@42 127 TK = VSUB(TI, TJ);
Chris@42 128 TH = VFMA(LDK(KP866025403), TE, TB);
Chris@42 129 TF = VFNMS(LDK(KP866025403), TE, TB);
Chris@42 130 }
Chris@42 131 ST(&(x[WS(rs, 4)]), VFMAI(TP, TO), ms, &(x[0]));
Chris@42 132 ST(&(x[WS(rs, 8)]), VFNMSI(TP, TO), ms, &(x[0]));
Chris@42 133 ST(&(x[WS(rs, 10)]), VFNMSI(TN, TK), ms, &(x[0]));
Chris@42 134 ST(&(x[WS(rs, 2)]), VFMAI(TN, TK), ms, &(x[0]));
Chris@42 135 ST(&(x[WS(rs, 5)]), VFNMSI(TH, TG), ms, &(x[WS(rs, 1)]));
Chris@42 136 ST(&(x[WS(rs, 7)]), VFMAI(TH, TG), ms, &(x[WS(rs, 1)]));
Chris@42 137 ST(&(x[WS(rs, 11)]), VFMAI(TF, Ts), ms, &(x[WS(rs, 1)]));
Chris@42 138 ST(&(x[WS(rs, 1)]), VFNMSI(TF, Ts), ms, &(x[WS(rs, 1)]));
Chris@42 139 }
Chris@42 140 }
Chris@42 141 }
Chris@42 142 }
Chris@42 143 }
Chris@42 144 VLEAVE();
Chris@42 145 }
Chris@42 146
Chris@42 147 static const tw_instr twinstr[] = {
Chris@42 148 VTW(0, 1),
Chris@42 149 VTW(0, 2),
Chris@42 150 VTW(0, 3),
Chris@42 151 VTW(0, 4),
Chris@42 152 VTW(0, 5),
Chris@42 153 VTW(0, 6),
Chris@42 154 VTW(0, 7),
Chris@42 155 VTW(0, 8),
Chris@42 156 VTW(0, 9),
Chris@42 157 VTW(0, 10),
Chris@42 158 VTW(0, 11),
Chris@42 159 {TW_NEXT, VL, 0}
Chris@42 160 };
Chris@42 161
Chris@42 162 static const ct_desc desc = { 12, XSIMD_STRING("t1fv_12"), twinstr, &GENUS, {41, 24, 18, 0}, 0, 0, 0 };
Chris@42 163
Chris@42 164 void XSIMD(codelet_t1fv_12) (planner *p) {
Chris@42 165 X(kdft_dit_register) (p, t1fv_12, &desc);
Chris@42 166 }
Chris@42 167 #else /* HAVE_FMA */
Chris@42 168
Chris@42 169 /* Generated by: ../../../genfft/gen_twiddle_c.native -simd -compact -variables 4 -pipeline-latency 8 -n 12 -name t1fv_12 -include t1f.h */
Chris@42 170
Chris@42 171 /*
Chris@42 172 * This function contains 59 FP additions, 30 FP multiplications,
Chris@42 173 * (or, 55 additions, 26 multiplications, 4 fused multiply/add),
Chris@42 174 * 28 stack variables, 2 constants, and 24 memory accesses
Chris@42 175 */
Chris@42 176 #include "t1f.h"
Chris@42 177
Chris@42 178 static void t1fv_12(R *ri, R *ii, const R *W, stride rs, INT mb, INT me, INT ms)
Chris@42 179 {
Chris@42 180 DVK(KP866025403, +0.866025403784438646763723170752936183471402627);
Chris@42 181 DVK(KP500000000, +0.500000000000000000000000000000000000000000000);
Chris@42 182 {
Chris@42 183 INT m;
Chris@42 184 R *x;
Chris@42 185 x = ri;
Chris@42 186 for (m = mb, W = W + (mb * ((TWVL / VL) * 22)); m < me; m = m + VL, x = x + (VL * ms), W = W + (TWVL * 22), MAKE_VOLATILE_STRIDE(12, rs)) {
Chris@42 187 V T1, TH, T6, TA, Tq, TE, Tv, TL, T9, TI, Te, TB, Ti, TD, Tn;
Chris@42 188 V TK;
Chris@42 189 {
Chris@42 190 V T5, T3, T4, T2;
Chris@42 191 T1 = LD(&(x[0]), ms, &(x[0]));
Chris@42 192 T4 = LD(&(x[WS(rs, 8)]), ms, &(x[0]));
Chris@42 193 T5 = BYTWJ(&(W[TWVL * 14]), T4);
Chris@42 194 T2 = LD(&(x[WS(rs, 4)]), ms, &(x[0]));
Chris@42 195 T3 = BYTWJ(&(W[TWVL * 6]), T2);
Chris@42 196 TH = VSUB(T5, T3);
Chris@42 197 T6 = VADD(T3, T5);
Chris@42 198 TA = VFNMS(LDK(KP500000000), T6, T1);
Chris@42 199 }
Chris@42 200 {
Chris@42 201 V Tu, Ts, Tp, Tt, Tr;
Chris@42 202 Tp = LD(&(x[WS(rs, 9)]), ms, &(x[WS(rs, 1)]));
Chris@42 203 Tq = BYTWJ(&(W[TWVL * 16]), Tp);
Chris@42 204 Tt = LD(&(x[WS(rs, 5)]), ms, &(x[WS(rs, 1)]));
Chris@42 205 Tu = BYTWJ(&(W[TWVL * 8]), Tt);
Chris@42 206 Tr = LD(&(x[WS(rs, 1)]), ms, &(x[WS(rs, 1)]));
Chris@42 207 Ts = BYTWJ(&(W[0]), Tr);
Chris@42 208 TE = VSUB(Tu, Ts);
Chris@42 209 Tv = VADD(Ts, Tu);
Chris@42 210 TL = VFNMS(LDK(KP500000000), Tv, Tq);
Chris@42 211 }
Chris@42 212 {
Chris@42 213 V Td, Tb, T8, Tc, Ta;
Chris@42 214 T8 = LD(&(x[WS(rs, 6)]), ms, &(x[0]));
Chris@42 215 T9 = BYTWJ(&(W[TWVL * 10]), T8);
Chris@42 216 Tc = LD(&(x[WS(rs, 2)]), ms, &(x[0]));
Chris@42 217 Td = BYTWJ(&(W[TWVL * 2]), Tc);
Chris@42 218 Ta = LD(&(x[WS(rs, 10)]), ms, &(x[0]));
Chris@42 219 Tb = BYTWJ(&(W[TWVL * 18]), Ta);
Chris@42 220 TI = VSUB(Td, Tb);
Chris@42 221 Te = VADD(Tb, Td);
Chris@42 222 TB = VFNMS(LDK(KP500000000), Te, T9);
Chris@42 223 }
Chris@42 224 {
Chris@42 225 V Tm, Tk, Th, Tl, Tj;
Chris@42 226 Th = LD(&(x[WS(rs, 3)]), ms, &(x[WS(rs, 1)]));
Chris@42 227 Ti = BYTWJ(&(W[TWVL * 4]), Th);
Chris@42 228 Tl = LD(&(x[WS(rs, 11)]), ms, &(x[WS(rs, 1)]));
Chris@42 229 Tm = BYTWJ(&(W[TWVL * 20]), Tl);
Chris@42 230 Tj = LD(&(x[WS(rs, 7)]), ms, &(x[WS(rs, 1)]));
Chris@42 231 Tk = BYTWJ(&(W[TWVL * 12]), Tj);
Chris@42 232 TD = VSUB(Tm, Tk);
Chris@42 233 Tn = VADD(Tk, Tm);
Chris@42 234 TK = VFNMS(LDK(KP500000000), Tn, Ti);
Chris@42 235 }
Chris@42 236 {
Chris@42 237 V Tg, Ty, Tx, Tz;
Chris@42 238 {
Chris@42 239 V T7, Tf, To, Tw;
Chris@42 240 T7 = VADD(T1, T6);
Chris@42 241 Tf = VADD(T9, Te);
Chris@42 242 Tg = VSUB(T7, Tf);
Chris@42 243 Ty = VADD(T7, Tf);
Chris@42 244 To = VADD(Ti, Tn);
Chris@42 245 Tw = VADD(Tq, Tv);
Chris@42 246 Tx = VBYI(VSUB(To, Tw));
Chris@42 247 Tz = VADD(To, Tw);
Chris@42 248 }
Chris@42 249 ST(&(x[WS(rs, 9)]), VSUB(Tg, Tx), ms, &(x[WS(rs, 1)]));
Chris@42 250 ST(&(x[0]), VADD(Ty, Tz), ms, &(x[0]));
Chris@42 251 ST(&(x[WS(rs, 3)]), VADD(Tg, Tx), ms, &(x[WS(rs, 1)]));
Chris@42 252 ST(&(x[WS(rs, 6)]), VSUB(Ty, Tz), ms, &(x[0]));
Chris@42 253 }
Chris@42 254 {
Chris@42 255 V TS, TW, TV, TX;
Chris@42 256 {
Chris@42 257 V TQ, TR, TT, TU;
Chris@42 258 TQ = VADD(TA, TB);
Chris@42 259 TR = VADD(TK, TL);
Chris@42 260 TS = VSUB(TQ, TR);
Chris@42 261 TW = VADD(TQ, TR);
Chris@42 262 TT = VADD(TD, TE);
Chris@42 263 TU = VADD(TH, TI);
Chris@42 264 TV = VBYI(VMUL(LDK(KP866025403), VSUB(TT, TU)));
Chris@42 265 TX = VBYI(VMUL(LDK(KP866025403), VADD(TU, TT)));
Chris@42 266 }
Chris@42 267 ST(&(x[WS(rs, 10)]), VSUB(TS, TV), ms, &(x[0]));
Chris@42 268 ST(&(x[WS(rs, 4)]), VADD(TW, TX), ms, &(x[0]));
Chris@42 269 ST(&(x[WS(rs, 2)]), VADD(TS, TV), ms, &(x[0]));
Chris@42 270 ST(&(x[WS(rs, 8)]), VSUB(TW, TX), ms, &(x[0]));
Chris@42 271 }
Chris@42 272 {
Chris@42 273 V TG, TP, TN, TO;
Chris@42 274 {
Chris@42 275 V TC, TF, TJ, TM;
Chris@42 276 TC = VSUB(TA, TB);
Chris@42 277 TF = VMUL(LDK(KP866025403), VSUB(TD, TE));
Chris@42 278 TG = VSUB(TC, TF);
Chris@42 279 TP = VADD(TC, TF);
Chris@42 280 TJ = VMUL(LDK(KP866025403), VSUB(TH, TI));
Chris@42 281 TM = VSUB(TK, TL);
Chris@42 282 TN = VBYI(VADD(TJ, TM));
Chris@42 283 TO = VBYI(VSUB(TJ, TM));
Chris@42 284 }
Chris@42 285 ST(&(x[WS(rs, 5)]), VSUB(TG, TN), ms, &(x[WS(rs, 1)]));
Chris@42 286 ST(&(x[WS(rs, 11)]), VSUB(TP, TO), ms, &(x[WS(rs, 1)]));
Chris@42 287 ST(&(x[WS(rs, 7)]), VADD(TN, TG), ms, &(x[WS(rs, 1)]));
Chris@42 288 ST(&(x[WS(rs, 1)]), VADD(TO, TP), ms, &(x[WS(rs, 1)]));
Chris@42 289 }
Chris@42 290 }
Chris@42 291 }
Chris@42 292 VLEAVE();
Chris@42 293 }
Chris@42 294
Chris@42 295 static const tw_instr twinstr[] = {
Chris@42 296 VTW(0, 1),
Chris@42 297 VTW(0, 2),
Chris@42 298 VTW(0, 3),
Chris@42 299 VTW(0, 4),
Chris@42 300 VTW(0, 5),
Chris@42 301 VTW(0, 6),
Chris@42 302 VTW(0, 7),
Chris@42 303 VTW(0, 8),
Chris@42 304 VTW(0, 9),
Chris@42 305 VTW(0, 10),
Chris@42 306 VTW(0, 11),
Chris@42 307 {TW_NEXT, VL, 0}
Chris@42 308 };
Chris@42 309
Chris@42 310 static const ct_desc desc = { 12, XSIMD_STRING("t1fv_12"), twinstr, &GENUS, {55, 26, 4, 0}, 0, 0, 0 };
Chris@42 311
Chris@42 312 void XSIMD(codelet_t1fv_12) (planner *p) {
Chris@42 313 X(kdft_dit_register) (p, t1fv_12, &desc);
Chris@42 314 }
Chris@42 315 #endif /* HAVE_FMA */