Mercurial > hg > sv-dependency-builds

annotate src/fftw-3.3.5/dft/simd/common/t1fuv_10.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 16:41:53 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 10 -name t1fuv_10 -include t1fu.h */
Chris@42 29
Chris@42 30 /*
Chris@42 31 * This function contains 51 FP additions, 40 FP multiplications,
Chris@42 32 * (or, 33 additions, 22 multiplications, 18 fused multiply/add),
Chris@42 33 * 43 stack variables, 4 constants, and 20 memory accesses
Chris@42 34 */
Chris@42 35 #include "t1fu.h"
Chris@42 36
Chris@42 37 static void t1fuv_10(R *ri, R *ii, const R *W, stride rs, INT mb, INT me, INT ms)
Chris@42 38 {
Chris@42 39 DVK(KP559016994, +0.559016994374947424102293417182819058860154590);
Chris@42 40 DVK(KP250000000, +0.250000000000000000000000000000000000000000000);
Chris@42 41 DVK(KP618033988, +0.618033988749894848204586834365638117720309180);
Chris@42 42 DVK(KP951056516, +0.951056516295153572116439333379382143405698634);
Chris@42 43 {
Chris@42 44 INT m;
Chris@42 45 R *x;
Chris@42 46 x = ri;
Chris@42 47 for (m = mb, W = W + (mb * ((TWVL / VL) * 18)); m < me; m = m + VL, x = x + (VL * ms), W = W + (TWVL * 18), MAKE_VOLATILE_STRIDE(10, rs)) {
Chris@42 48 V Td, TA, T4, Ta, Tk, TE, Tp, TF, TB, T9, T1, T2, Tb;
Chris@42 49 T1 = LD(&(x[0]), ms, &(x[0]));
Chris@42 50 T2 = LD(&(x[WS(rs, 5)]), ms, &(x[WS(rs, 1)]));
Chris@42 51 {
Chris@42 52 V Tg, Tn, Ti, Tl;
Chris@42 53 Tg = LD(&(x[WS(rs, 4)]), ms, &(x[0]));
Chris@42 54 Tn = LD(&(x[WS(rs, 1)]), ms, &(x[WS(rs, 1)]));
Chris@42 55 Ti = LD(&(x[WS(rs, 9)]), ms, &(x[WS(rs, 1)]));
Chris@42 56 Tl = LD(&(x[WS(rs, 6)]), ms, &(x[0]));
Chris@42 57 {
Chris@42 58 V T6, T8, T5, Tc;
Chris@42 59 T5 = LD(&(x[WS(rs, 2)]), ms, &(x[0]));
Chris@42 60 Tc = LD(&(x[WS(rs, 3)]), ms, &(x[WS(rs, 1)]));
Chris@42 61 {
Chris@42 62 V T3, Th, To, Tj, Tm, T7;
Chris@42 63 T7 = LD(&(x[WS(rs, 7)]), ms, &(x[WS(rs, 1)]));
Chris@42 64 T3 = BYTWJ(&(W[TWVL * 8]), T2);
Chris@42 65 Th = BYTWJ(&(W[TWVL * 6]), Tg);
Chris@42 66 To = BYTWJ(&(W[0]), Tn);
Chris@42 67 Tj = BYTWJ(&(W[TWVL * 16]), Ti);
Chris@42 68 Tm = BYTWJ(&(W[TWVL * 10]), Tl);
Chris@42 69 T6 = BYTWJ(&(W[TWVL * 2]), T5);
Chris@42 70 Td = BYTWJ(&(W[TWVL * 4]), Tc);
Chris@42 71 T8 = BYTWJ(&(W[TWVL * 12]), T7);
Chris@42 72 TA = VADD(T1, T3);
Chris@42 73 T4 = VSUB(T1, T3);
Chris@42 74 Ta = LD(&(x[WS(rs, 8)]), ms, &(x[0]));
Chris@42 75 Tk = VSUB(Th, Tj);
Chris@42 76 TE = VADD(Th, Tj);
Chris@42 77 Tp = VSUB(Tm, To);
Chris@42 78 TF = VADD(Tm, To);
Chris@42 79 }
Chris@42 80 TB = VADD(T6, T8);
Chris@42 81 T9 = VSUB(T6, T8);
Chris@42 82 }
Chris@42 83 }
Chris@42 84 Tb = BYTWJ(&(W[TWVL * 14]), Ta);
Chris@42 85 {
Chris@42 86 V TL, TG, Tw, Tq, TC, Te;
Chris@42 87 TL = VSUB(TE, TF);
Chris@42 88 TG = VADD(TE, TF);
Chris@42 89 Tw = VSUB(Tk, Tp);
Chris@42 90 Tq = VADD(Tk, Tp);
Chris@42 91 TC = VADD(Tb, Td);
Chris@42 92 Te = VSUB(Tb, Td);
Chris@42 93 {
Chris@42 94 V TM, TD, Tv, Tf;
Chris@42 95 TM = VSUB(TB, TC);
Chris@42 96 TD = VADD(TB, TC);
Chris@42 97 Tv = VSUB(T9, Te);
Chris@42 98 Tf = VADD(T9, Te);
Chris@42 99 {
Chris@42 100 V TP, TN, TH, TJ, Tz, Tx, Tr, Tt, TI, Ts;
Chris@42 101 TP = VMUL(LDK(KP951056516), VFMA(LDK(KP618033988), TL, TM));
Chris@42 102 TN = VMUL(LDK(KP951056516), VFNMS(LDK(KP618033988), TM, TL));
Chris@42 103 TH = VADD(TD, TG);
Chris@42 104 TJ = VSUB(TD, TG);
Chris@42 105 Tz = VMUL(LDK(KP951056516), VFNMS(LDK(KP618033988), Tv, Tw));
Chris@42 106 Tx = VMUL(LDK(KP951056516), VFMA(LDK(KP618033988), Tw, Tv));
Chris@42 107 Tr = VADD(Tf, Tq);
Chris@42 108 Tt = VSUB(Tf, Tq);
Chris@42 109 ST(&(x[0]), VADD(TA, TH), ms, &(x[0]));
Chris@42 110 TI = VFNMS(LDK(KP250000000), TH, TA);
Chris@42 111 ST(&(x[WS(rs, 5)]), VADD(T4, Tr), ms, &(x[WS(rs, 1)]));
Chris@42 112 Ts = VFNMS(LDK(KP250000000), Tr, T4);
Chris@42 113 {
Chris@42 114 V TK, TO, Tu, Ty;
Chris@42 115 TK = VFNMS(LDK(KP559016994), TJ, TI);
Chris@42 116 TO = VFMA(LDK(KP559016994), TJ, TI);
Chris@42 117 Tu = VFMA(LDK(KP559016994), Tt, Ts);
Chris@42 118 Ty = VFNMS(LDK(KP559016994), Tt, Ts);
Chris@42 119 ST(&(x[WS(rs, 8)]), VFNMSI(TN, TK), ms, &(x[0]));
Chris@42 120 ST(&(x[WS(rs, 2)]), VFMAI(TN, TK), ms, &(x[0]));
Chris@42 121 ST(&(x[WS(rs, 6)]), VFNMSI(TP, TO), ms, &(x[0]));
Chris@42 122 ST(&(x[WS(rs, 4)]), VFMAI(TP, TO), ms, &(x[0]));
Chris@42 123 ST(&(x[WS(rs, 9)]), VFMAI(Tx, Tu), ms, &(x[WS(rs, 1)]));
Chris@42 124 ST(&(x[WS(rs, 1)]), VFNMSI(Tx, Tu), ms, &(x[WS(rs, 1)]));
Chris@42 125 ST(&(x[WS(rs, 7)]), VFMAI(Tz, Ty), ms, &(x[WS(rs, 1)]));
Chris@42 126 ST(&(x[WS(rs, 3)]), VFNMSI(Tz, Ty), ms, &(x[WS(rs, 1)]));
Chris@42 127 }
Chris@42 128 }
Chris@42 129 }
Chris@42 130 }
Chris@42 131 }
Chris@42 132 }
Chris@42 133 VLEAVE();
Chris@42 134 }
Chris@42 135
Chris@42 136 static const tw_instr twinstr[] = {
Chris@42 137 VTW(0, 1),
Chris@42 138 VTW(0, 2),
Chris@42 139 VTW(0, 3),
Chris@42 140 VTW(0, 4),
Chris@42 141 VTW(0, 5),
Chris@42 142 VTW(0, 6),
Chris@42 143 VTW(0, 7),
Chris@42 144 VTW(0, 8),
Chris@42 145 VTW(0, 9),
Chris@42 146 {TW_NEXT, VL, 0}
Chris@42 147 };
Chris@42 148
Chris@42 149 static const ct_desc desc = { 10, XSIMD_STRING("t1fuv_10"), twinstr, &GENUS, {33, 22, 18, 0}, 0, 0, 0 };
Chris@42 150
Chris@42 151 void XSIMD(codelet_t1fuv_10) (planner *p) {
Chris@42 152 X(kdft_dit_register) (p, t1fuv_10, &desc);
Chris@42 153 }
Chris@42 154 #else /* HAVE_FMA */
Chris@42 155
Chris@42 156 /* Generated by: ../../../genfft/gen_twiddle_c.native -simd -compact -variables 4 -pipeline-latency 8 -n 10 -name t1fuv_10 -include t1fu.h */
Chris@42 157
Chris@42 158 /*
Chris@42 159 * This function contains 51 FP additions, 30 FP multiplications,
Chris@42 160 * (or, 45 additions, 24 multiplications, 6 fused multiply/add),
Chris@42 161 * 32 stack variables, 4 constants, and 20 memory accesses
Chris@42 162 */
Chris@42 163 #include "t1fu.h"
Chris@42 164
Chris@42 165 static void t1fuv_10(R *ri, R *ii, const R *W, stride rs, INT mb, INT me, INT ms)
Chris@42 166 {
Chris@42 167 DVK(KP587785252, +0.587785252292473129168705954639072768597652438);
Chris@42 168 DVK(KP951056516, +0.951056516295153572116439333379382143405698634);
Chris@42 169 DVK(KP250000000, +0.250000000000000000000000000000000000000000000);
Chris@42 170 DVK(KP559016994, +0.559016994374947424102293417182819058860154590);
Chris@42 171 {
Chris@42 172 INT m;
Chris@42 173 R *x;
Chris@42 174 x = ri;
Chris@42 175 for (m = mb, W = W + (mb * ((TWVL / VL) * 18)); m < me; m = m + VL, x = x + (VL * ms), W = W + (TWVL * 18), MAKE_VOLATILE_STRIDE(10, rs)) {
Chris@42 176 V Tr, TH, Tg, Tl, Tm, TA, TB, TJ, T5, Ta, Tb, TD, TE, TI, To;
Chris@42 177 V Tq, Tp;
Chris@42 178 To = LD(&(x[0]), ms, &(x[0]));
Chris@42 179 Tp = LD(&(x[WS(rs, 5)]), ms, &(x[WS(rs, 1)]));
Chris@42 180 Tq = BYTWJ(&(W[TWVL * 8]), Tp);
Chris@42 181 Tr = VSUB(To, Tq);
Chris@42 182 TH = VADD(To, Tq);
Chris@42 183 {
Chris@42 184 V Td, Tk, Tf, Ti;
Chris@42 185 {
Chris@42 186 V Tc, Tj, Te, Th;
Chris@42 187 Tc = LD(&(x[WS(rs, 4)]), ms, &(x[0]));
Chris@42 188 Td = BYTWJ(&(W[TWVL * 6]), Tc);
Chris@42 189 Tj = LD(&(x[WS(rs, 1)]), ms, &(x[WS(rs, 1)]));
Chris@42 190 Tk = BYTWJ(&(W[0]), Tj);
Chris@42 191 Te = LD(&(x[WS(rs, 9)]), ms, &(x[WS(rs, 1)]));
Chris@42 192 Tf = BYTWJ(&(W[TWVL * 16]), Te);
Chris@42 193 Th = LD(&(x[WS(rs, 6)]), ms, &(x[0]));
Chris@42 194 Ti = BYTWJ(&(W[TWVL * 10]), Th);
Chris@42 195 }
Chris@42 196 Tg = VSUB(Td, Tf);
Chris@42 197 Tl = VSUB(Ti, Tk);
Chris@42 198 Tm = VADD(Tg, Tl);
Chris@42 199 TA = VADD(Td, Tf);
Chris@42 200 TB = VADD(Ti, Tk);
Chris@42 201 TJ = VADD(TA, TB);
Chris@42 202 }
Chris@42 203 {
Chris@42 204 V T2, T9, T4, T7;
Chris@42 205 {
Chris@42 206 V T1, T8, T3, T6;
Chris@42 207 T1 = LD(&(x[WS(rs, 2)]), ms, &(x[0]));
Chris@42 208 T2 = BYTWJ(&(W[TWVL * 2]), T1);
Chris@42 209 T8 = LD(&(x[WS(rs, 3)]), ms, &(x[WS(rs, 1)]));
Chris@42 210 T9 = BYTWJ(&(W[TWVL * 4]), T8);
Chris@42 211 T3 = LD(&(x[WS(rs, 7)]), ms, &(x[WS(rs, 1)]));
Chris@42 212 T4 = BYTWJ(&(W[TWVL * 12]), T3);
Chris@42 213 T6 = LD(&(x[WS(rs, 8)]), ms, &(x[0]));
Chris@42 214 T7 = BYTWJ(&(W[TWVL * 14]), T6);
Chris@42 215 }
Chris@42 216 T5 = VSUB(T2, T4);
Chris@42 217 Ta = VSUB(T7, T9);
Chris@42 218 Tb = VADD(T5, Ta);
Chris@42 219 TD = VADD(T2, T4);
Chris@42 220 TE = VADD(T7, T9);
Chris@42 221 TI = VADD(TD, TE);
Chris@42 222 }
Chris@42 223 {
Chris@42 224 V Tn, Ts, Tt, Tx, Tz, Tv, Tw, Ty, Tu;
Chris@42 225 Tn = VMUL(LDK(KP559016994), VSUB(Tb, Tm));
Chris@42 226 Ts = VADD(Tb, Tm);
Chris@42 227 Tt = VFNMS(LDK(KP250000000), Ts, Tr);
Chris@42 228 Tv = VSUB(T5, Ta);
Chris@42 229 Tw = VSUB(Tg, Tl);
Chris@42 230 Tx = VBYI(VFMA(LDK(KP951056516), Tv, VMUL(LDK(KP587785252), Tw)));
Chris@42 231 Tz = VBYI(VFNMS(LDK(KP587785252), Tv, VMUL(LDK(KP951056516), Tw)));
Chris@42 232 ST(&(x[WS(rs, 5)]), VADD(Tr, Ts), ms, &(x[WS(rs, 1)]));
Chris@42 233 Ty = VSUB(Tt, Tn);
Chris@42 234 ST(&(x[WS(rs, 3)]), VSUB(Ty, Tz), ms, &(x[WS(rs, 1)]));
Chris@42 235 ST(&(x[WS(rs, 7)]), VADD(Tz, Ty), ms, &(x[WS(rs, 1)]));
Chris@42 236 Tu = VADD(Tn, Tt);
Chris@42 237 ST(&(x[WS(rs, 1)]), VSUB(Tu, Tx), ms, &(x[WS(rs, 1)]));
Chris@42 238 ST(&(x[WS(rs, 9)]), VADD(Tx, Tu), ms, &(x[WS(rs, 1)]));
Chris@42 239 }
Chris@42 240 {
Chris@42 241 V TM, TK, TL, TG, TO, TC, TF, TP, TN;
Chris@42 242 TM = VMUL(LDK(KP559016994), VSUB(TI, TJ));
Chris@42 243 TK = VADD(TI, TJ);
Chris@42 244 TL = VFNMS(LDK(KP250000000), TK, TH);
Chris@42 245 TC = VSUB(TA, TB);
Chris@42 246 TF = VSUB(TD, TE);
Chris@42 247 TG = VBYI(VFNMS(LDK(KP587785252), TF, VMUL(LDK(KP951056516), TC)));
Chris@42 248 TO = VBYI(VFMA(LDK(KP951056516), TF, VMUL(LDK(KP587785252), TC)));
Chris@42 249 ST(&(x[0]), VADD(TH, TK), ms, &(x[0]));
Chris@42 250 TP = VADD(TM, TL);
Chris@42 251 ST(&(x[WS(rs, 4)]), VADD(TO, TP), ms, &(x[0]));
Chris@42 252 ST(&(x[WS(rs, 6)]), VSUB(TP, TO), ms, &(x[0]));
Chris@42 253 TN = VSUB(TL, TM);
Chris@42 254 ST(&(x[WS(rs, 2)]), VADD(TG, TN), ms, &(x[0]));
Chris@42 255 ST(&(x[WS(rs, 8)]), VSUB(TN, TG), ms, &(x[0]));
Chris@42 256 }
Chris@42 257 }
Chris@42 258 }
Chris@42 259 VLEAVE();
Chris@42 260 }
Chris@42 261
Chris@42 262 static const tw_instr twinstr[] = {
Chris@42 263 VTW(0, 1),
Chris@42 264 VTW(0, 2),
Chris@42 265 VTW(0, 3),
Chris@42 266 VTW(0, 4),
Chris@42 267 VTW(0, 5),
Chris@42 268 VTW(0, 6),
Chris@42 269 VTW(0, 7),
Chris@42 270 VTW(0, 8),
Chris@42 271 VTW(0, 9),
Chris@42 272 {TW_NEXT, VL, 0}
Chris@42 273 };
Chris@42 274
Chris@42 275 static const ct_desc desc = { 10, XSIMD_STRING("t1fuv_10"), twinstr, &GENUS, {45, 24, 6, 0}, 0, 0, 0 };
Chris@42 276
Chris@42 277 void XSIMD(codelet_t1fuv_10) (planner *p) {
Chris@42 278 X(kdft_dit_register) (p, t1fuv_10, &desc);
Chris@42 279 }
Chris@42 280 #endif /* HAVE_FMA */