Mercurial > hg > sv-dependency-builds

annotate src/fftw-3.3.5/dft/simd/common/t1buv_7.c @ 79:91c729825bca pa_catalina

Find changesets by keywords (author, files, the commit message), revision number or hash, or revset expression.
Update build for AUDIO_COMPONENT_FIX
author Chris Cannam
date Wed, 30 Oct 2019 12:40:34 +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 16:44:08 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 7 -name t1buv_7 -include t1bu.h -sign 1 */
Chris@42 29
Chris@42 30 /*
Chris@42 31 * This function contains 36 FP additions, 36 FP multiplications,
Chris@42 32 * (or, 15 additions, 15 multiplications, 21 fused multiply/add),
Chris@42 33 * 42 stack variables, 6 constants, and 14 memory accesses
Chris@42 34 */
Chris@42 35 #include "t1bu.h"
Chris@42 36
Chris@42 37 static void t1buv_7(R *ri, R *ii, const R *W, stride rs, INT mb, INT me, INT ms)
Chris@42 38 {
Chris@42 39 DVK(KP900968867, +0.900968867902419126236102319507445051165919162);
Chris@42 40 DVK(KP801937735, +0.801937735804838252472204639014890102331838324);
Chris@42 41 DVK(KP974927912, +0.974927912181823607018131682993931217232785801);
Chris@42 42 DVK(KP692021471, +0.692021471630095869627814897002069140197260599);
Chris@42 43 DVK(KP554958132, +0.554958132087371191422194871006410481067288862);
Chris@42 44 DVK(KP356895867, +0.356895867892209443894399510021300583399127187);
Chris@42 45 {
Chris@42 46 INT m;
Chris@42 47 R *x;
Chris@42 48 x = ii;
Chris@42 49 for (m = mb, W = W + (mb * ((TWVL / VL) * 12)); m < me; m = m + VL, x = x + (VL * ms), W = W + (TWVL * 12), MAKE_VOLATILE_STRIDE(7, rs)) {
Chris@42 50 V T1, T2, T4, Te, Tc, T9, T7;
Chris@42 51 T1 = LD(&(x[0]), ms, &(x[0]));
Chris@42 52 T2 = LD(&(x[WS(rs, 1)]), ms, &(x[WS(rs, 1)]));
Chris@42 53 T4 = LD(&(x[WS(rs, 6)]), ms, &(x[0]));
Chris@42 54 Te = LD(&(x[WS(rs, 4)]), ms, &(x[0]));
Chris@42 55 Tc = LD(&(x[WS(rs, 3)]), ms, &(x[WS(rs, 1)]));
Chris@42 56 T9 = LD(&(x[WS(rs, 5)]), ms, &(x[WS(rs, 1)]));
Chris@42 57 T7 = LD(&(x[WS(rs, 2)]), ms, &(x[0]));
Chris@42 58 {
Chris@42 59 V T3, T5, Tf, Td, Ta, T8;
Chris@42 60 T3 = BYTW(&(W[0]), T2);
Chris@42 61 T5 = BYTW(&(W[TWVL * 10]), T4);
Chris@42 62 Tf = BYTW(&(W[TWVL * 6]), Te);
Chris@42 63 Td = BYTW(&(W[TWVL * 4]), Tc);
Chris@42 64 Ta = BYTW(&(W[TWVL * 8]), T9);
Chris@42 65 T8 = BYTW(&(W[TWVL * 2]), T7);
Chris@42 66 {
Chris@42 67 V T6, Tm, Tg, Tk, Tb, Tl;
Chris@42 68 T6 = VADD(T3, T5);
Chris@42 69 Tm = VSUB(T3, T5);
Chris@42 70 Tg = VADD(Td, Tf);
Chris@42 71 Tk = VSUB(Td, Tf);
Chris@42 72 Tb = VADD(T8, Ta);
Chris@42 73 Tl = VSUB(T8, Ta);
Chris@42 74 {
Chris@42 75 V Tp, Tx, Tu, Th, Ts, Tn, Tq, Ty;
Chris@42 76 Tp = VFNMS(LDK(KP356895867), T6, Tg);
Chris@42 77 Tx = VFMA(LDK(KP554958132), Tk, Tm);
Chris@42 78 ST(&(x[0]), VADD(T1, VADD(T6, VADD(Tb, Tg))), ms, &(x[0]));
Chris@42 79 Tu = VFNMS(LDK(KP356895867), Tb, T6);
Chris@42 80 Th = VFNMS(LDK(KP356895867), Tg, Tb);
Chris@42 81 Ts = VFMA(LDK(KP554958132), Tl, Tk);
Chris@42 82 Tn = VFNMS(LDK(KP554958132), Tm, Tl);
Chris@42 83 Tq = VFNMS(LDK(KP692021471), Tp, Tb);
Chris@42 84 Ty = VMUL(LDK(KP974927912), VFMA(LDK(KP801937735), Tx, Tl));
Chris@42 85 {
Chris@42 86 V Tv, Ti, Tt, To, Tr, Tw, Tj;
Chris@42 87 Tv = VFNMS(LDK(KP692021471), Tu, Tg);
Chris@42 88 Ti = VFNMS(LDK(KP692021471), Th, T6);
Chris@42 89 Tt = VMUL(LDK(KP974927912), VFNMS(LDK(KP801937735), Ts, Tm));
Chris@42 90 To = VMUL(LDK(KP974927912), VFNMS(LDK(KP801937735), Tn, Tk));
Chris@42 91 Tr = VFNMS(LDK(KP900968867), Tq, T1);
Chris@42 92 Tw = VFNMS(LDK(KP900968867), Tv, T1);
Chris@42 93 Tj = VFNMS(LDK(KP900968867), Ti, T1);
Chris@42 94 ST(&(x[WS(rs, 5)]), VFNMSI(Tt, Tr), ms, &(x[WS(rs, 1)]));
Chris@42 95 ST(&(x[WS(rs, 2)]), VFMAI(Tt, Tr), ms, &(x[0]));
Chris@42 96 ST(&(x[WS(rs, 6)]), VFNMSI(Ty, Tw), ms, &(x[0]));
Chris@42 97 ST(&(x[WS(rs, 1)]), VFMAI(Ty, Tw), ms, &(x[WS(rs, 1)]));
Chris@42 98 ST(&(x[WS(rs, 4)]), VFNMSI(To, Tj), ms, &(x[0]));
Chris@42 99 ST(&(x[WS(rs, 3)]), VFMAI(To, Tj), ms, &(x[WS(rs, 1)]));
Chris@42 100 }
Chris@42 101 }
Chris@42 102 }
Chris@42 103 }
Chris@42 104 }
Chris@42 105 }
Chris@42 106 VLEAVE();
Chris@42 107 }
Chris@42 108
Chris@42 109 static const tw_instr twinstr[] = {
Chris@42 110 VTW(0, 1),
Chris@42 111 VTW(0, 2),
Chris@42 112 VTW(0, 3),
Chris@42 113 VTW(0, 4),
Chris@42 114 VTW(0, 5),
Chris@42 115 VTW(0, 6),
Chris@42 116 {TW_NEXT, VL, 0}
Chris@42 117 };
Chris@42 118
Chris@42 119 static const ct_desc desc = { 7, XSIMD_STRING("t1buv_7"), twinstr, &GENUS, {15, 15, 21, 0}, 0, 0, 0 };
Chris@42 120
Chris@42 121 void XSIMD(codelet_t1buv_7) (planner *p) {
Chris@42 122 X(kdft_dit_register) (p, t1buv_7, &desc);
Chris@42 123 }
Chris@42 124 #else /* HAVE_FMA */
Chris@42 125
Chris@42 126 /* Generated by: ../../../genfft/gen_twiddle_c.native -simd -compact -variables 4 -pipeline-latency 8 -n 7 -name t1buv_7 -include t1bu.h -sign 1 */
Chris@42 127
Chris@42 128 /*
Chris@42 129 * This function contains 36 FP additions, 30 FP multiplications,
Chris@42 130 * (or, 24 additions, 18 multiplications, 12 fused multiply/add),
Chris@42 131 * 21 stack variables, 6 constants, and 14 memory accesses
Chris@42 132 */
Chris@42 133 #include "t1bu.h"
Chris@42 134
Chris@42 135 static void t1buv_7(R *ri, R *ii, const R *W, stride rs, INT mb, INT me, INT ms)
Chris@42 136 {
Chris@42 137 DVK(KP222520933, +0.222520933956314404288902564496794759466355569);
Chris@42 138 DVK(KP900968867, +0.900968867902419126236102319507445051165919162);
Chris@42 139 DVK(KP623489801, +0.623489801858733530525004884004239810632274731);
Chris@42 140 DVK(KP433883739, +0.433883739117558120475768332848358754609990728);
Chris@42 141 DVK(KP781831482, +0.781831482468029808708444526674057750232334519);
Chris@42 142 DVK(KP974927912, +0.974927912181823607018131682993931217232785801);
Chris@42 143 {
Chris@42 144 INT m;
Chris@42 145 R *x;
Chris@42 146 x = ii;
Chris@42 147 for (m = mb, W = W + (mb * ((TWVL / VL) * 12)); m < me; m = m + VL, x = x + (VL * ms), W = W + (TWVL * 12), MAKE_VOLATILE_STRIDE(7, rs)) {
Chris@42 148 V Th, Tf, Ti, T5, Tk, Ta, Tj, To, Tp;
Chris@42 149 Th = LD(&(x[0]), ms, &(x[0]));
Chris@42 150 {
Chris@42 151 V Tc, Te, Tb, Td;
Chris@42 152 Tb = LD(&(x[WS(rs, 2)]), ms, &(x[0]));
Chris@42 153 Tc = BYTW(&(W[TWVL * 2]), Tb);
Chris@42 154 Td = LD(&(x[WS(rs, 5)]), ms, &(x[WS(rs, 1)]));
Chris@42 155 Te = BYTW(&(W[TWVL * 8]), Td);
Chris@42 156 Tf = VSUB(Tc, Te);
Chris@42 157 Ti = VADD(Tc, Te);
Chris@42 158 }
Chris@42 159 {
Chris@42 160 V T2, T4, T1, T3;
Chris@42 161 T1 = LD(&(x[WS(rs, 1)]), ms, &(x[WS(rs, 1)]));
Chris@42 162 T2 = BYTW(&(W[0]), T1);
Chris@42 163 T3 = LD(&(x[WS(rs, 6)]), ms, &(x[0]));
Chris@42 164 T4 = BYTW(&(W[TWVL * 10]), T3);
Chris@42 165 T5 = VSUB(T2, T4);
Chris@42 166 Tk = VADD(T2, T4);
Chris@42 167 }
Chris@42 168 {
Chris@42 169 V T7, T9, T6, T8;
Chris@42 170 T6 = LD(&(x[WS(rs, 3)]), ms, &(x[WS(rs, 1)]));
Chris@42 171 T7 = BYTW(&(W[TWVL * 4]), T6);
Chris@42 172 T8 = LD(&(x[WS(rs, 4)]), ms, &(x[0]));
Chris@42 173 T9 = BYTW(&(W[TWVL * 6]), T8);
Chris@42 174 Ta = VSUB(T7, T9);
Chris@42 175 Tj = VADD(T7, T9);
Chris@42 176 }
Chris@42 177 ST(&(x[0]), VADD(Th, VADD(Tk, VADD(Ti, Tj))), ms, &(x[0]));
Chris@42 178 To = VBYI(VFNMS(LDK(KP781831482), Ta, VFNMS(LDK(KP433883739), Tf, VMUL(LDK(KP974927912), T5))));
Chris@42 179 Tp = VFMA(LDK(KP623489801), Tj, VFNMS(LDK(KP900968867), Ti, VFNMS(LDK(KP222520933), Tk, Th)));
Chris@42 180 ST(&(x[WS(rs, 2)]), VADD(To, Tp), ms, &(x[0]));
Chris@42 181 ST(&(x[WS(rs, 5)]), VSUB(Tp, To), ms, &(x[WS(rs, 1)]));
Chris@42 182 {
Chris@42 183 V Tg, Tl, Tm, Tn;
Chris@42 184 Tg = VBYI(VFMA(LDK(KP433883739), T5, VFNMS(LDK(KP781831482), Tf, VMUL(LDK(KP974927912), Ta))));
Chris@42 185 Tl = VFMA(LDK(KP623489801), Ti, VFNMS(LDK(KP222520933), Tj, VFNMS(LDK(KP900968867), Tk, Th)));
Chris@42 186 ST(&(x[WS(rs, 3)]), VADD(Tg, Tl), ms, &(x[WS(rs, 1)]));
Chris@42 187 ST(&(x[WS(rs, 4)]), VSUB(Tl, Tg), ms, &(x[0]));
Chris@42 188 Tm = VBYI(VFMA(LDK(KP781831482), T5, VFMA(LDK(KP974927912), Tf, VMUL(LDK(KP433883739), Ta))));
Chris@42 189 Tn = VFMA(LDK(KP623489801), Tk, VFNMS(LDK(KP900968867), Tj, VFNMS(LDK(KP222520933), Ti, Th)));
Chris@42 190 ST(&(x[WS(rs, 1)]), VADD(Tm, Tn), ms, &(x[WS(rs, 1)]));
Chris@42 191 ST(&(x[WS(rs, 6)]), VSUB(Tn, Tm), ms, &(x[0]));
Chris@42 192 }
Chris@42 193 }
Chris@42 194 }
Chris@42 195 VLEAVE();
Chris@42 196 }
Chris@42 197
Chris@42 198 static const tw_instr twinstr[] = {
Chris@42 199 VTW(0, 1),
Chris@42 200 VTW(0, 2),
Chris@42 201 VTW(0, 3),
Chris@42 202 VTW(0, 4),
Chris@42 203 VTW(0, 5),
Chris@42 204 VTW(0, 6),
Chris@42 205 {TW_NEXT, VL, 0}
Chris@42 206 };
Chris@42 207
Chris@42 208 static const ct_desc desc = { 7, XSIMD_STRING("t1buv_7"), twinstr, &GENUS, {24, 18, 12, 0}, 0, 0, 0 };
Chris@42 209
Chris@42 210 void XSIMD(codelet_t1buv_7) (planner *p) {
Chris@42 211 X(kdft_dit_register) (p, t1buv_7, &desc);
Chris@42 212 }
Chris@42 213 #endif /* HAVE_FMA */