Mercurial > hg > sv-dependency-builds

annotate src/fftw-3.3.5/dft/simd/common/t1fv_6.c @ 83:ae30d91d2ffe

Find changesets by keywords (author, files, the commit message), revision number or hash, or revset expression.
Replace these with versions built using an older toolset (so as to avoid ABI compatibilities when linking on Ubuntu 14.04 for packaging purposes)
author Chris Cannam
date Fri, 07 Feb 2020 11:51:13 +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 16:41:55 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 6 -name t1fv_6 -include t1f.h */
Chris@42 29
Chris@42 30 /*
Chris@42 31 * This function contains 23 FP additions, 18 FP multiplications,
Chris@42 32 * (or, 17 additions, 12 multiplications, 6 fused multiply/add),
Chris@42 33 * 27 stack variables, 2 constants, and 12 memory accesses
Chris@42 34 */
Chris@42 35 #include "t1f.h"
Chris@42 36
Chris@42 37 static void t1fv_6(R *ri, R *ii, const R *W, stride rs, INT mb, INT me, INT ms)
Chris@42 38 {
Chris@42 39 DVK(KP500000000, +0.500000000000000000000000000000000000000000000);
Chris@42 40 DVK(KP866025403, +0.866025403784438646763723170752936183471402627);
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) * 10)); m < me; m = m + VL, x = x + (VL * ms), W = W + (TWVL * 10), MAKE_VOLATILE_STRIDE(6, rs)) {
Chris@42 46 V T1, T2, Ta, Tc, T5, T7;
Chris@42 47 T1 = LD(&(x[0]), ms, &(x[0]));
Chris@42 48 T2 = LD(&(x[WS(rs, 3)]), ms, &(x[WS(rs, 1)]));
Chris@42 49 Ta = LD(&(x[WS(rs, 4)]), ms, &(x[0]));
Chris@42 50 Tc = LD(&(x[WS(rs, 1)]), ms, &(x[WS(rs, 1)]));
Chris@42 51 T5 = LD(&(x[WS(rs, 2)]), ms, &(x[0]));
Chris@42 52 T7 = LD(&(x[WS(rs, 5)]), ms, &(x[WS(rs, 1)]));
Chris@42 53 {
Chris@42 54 V T3, Tb, Td, T6, T8;
Chris@42 55 T3 = BYTWJ(&(W[TWVL * 4]), T2);
Chris@42 56 Tb = BYTWJ(&(W[TWVL * 6]), Ta);
Chris@42 57 Td = BYTWJ(&(W[0]), Tc);
Chris@42 58 T6 = BYTWJ(&(W[TWVL * 2]), T5);
Chris@42 59 T8 = BYTWJ(&(W[TWVL * 8]), T7);
Chris@42 60 {
Chris@42 61 V Ti, T4, Tk, Te, Tj, T9;
Chris@42 62 Ti = VADD(T1, T3);
Chris@42 63 T4 = VSUB(T1, T3);
Chris@42 64 Tk = VADD(Tb, Td);
Chris@42 65 Te = VSUB(Tb, Td);
Chris@42 66 Tj = VADD(T6, T8);
Chris@42 67 T9 = VSUB(T6, T8);
Chris@42 68 {
Chris@42 69 V Tl, Tn, Tf, Th, Tm, Tg;
Chris@42 70 Tl = VADD(Tj, Tk);
Chris@42 71 Tn = VMUL(LDK(KP866025403), VSUB(Tk, Tj));
Chris@42 72 Tf = VADD(T9, Te);
Chris@42 73 Th = VMUL(LDK(KP866025403), VSUB(Te, T9));
Chris@42 74 ST(&(x[0]), VADD(Ti, Tl), ms, &(x[0]));
Chris@42 75 Tm = VFNMS(LDK(KP500000000), Tl, Ti);
Chris@42 76 ST(&(x[WS(rs, 3)]), VADD(T4, Tf), ms, &(x[WS(rs, 1)]));
Chris@42 77 Tg = VFNMS(LDK(KP500000000), Tf, T4);
Chris@42 78 ST(&(x[WS(rs, 2)]), VFNMSI(Tn, Tm), ms, &(x[0]));
Chris@42 79 ST(&(x[WS(rs, 4)]), VFMAI(Tn, Tm), ms, &(x[0]));
Chris@42 80 ST(&(x[WS(rs, 5)]), VFNMSI(Th, Tg), ms, &(x[WS(rs, 1)]));
Chris@42 81 ST(&(x[WS(rs, 1)]), VFMAI(Th, Tg), ms, &(x[WS(rs, 1)]));
Chris@42 82 }
Chris@42 83 }
Chris@42 84 }
Chris@42 85 }
Chris@42 86 }
Chris@42 87 VLEAVE();
Chris@42 88 }
Chris@42 89
Chris@42 90 static const tw_instr twinstr[] = {
Chris@42 91 VTW(0, 1),
Chris@42 92 VTW(0, 2),
Chris@42 93 VTW(0, 3),
Chris@42 94 VTW(0, 4),
Chris@42 95 VTW(0, 5),
Chris@42 96 {TW_NEXT, VL, 0}
Chris@42 97 };
Chris@42 98
Chris@42 99 static const ct_desc desc = { 6, XSIMD_STRING("t1fv_6"), twinstr, &GENUS, {17, 12, 6, 0}, 0, 0, 0 };
Chris@42 100
Chris@42 101 void XSIMD(codelet_t1fv_6) (planner *p) {
Chris@42 102 X(kdft_dit_register) (p, t1fv_6, &desc);
Chris@42 103 }
Chris@42 104 #else /* HAVE_FMA */
Chris@42 105
Chris@42 106 /* Generated by: ../../../genfft/gen_twiddle_c.native -simd -compact -variables 4 -pipeline-latency 8 -n 6 -name t1fv_6 -include t1f.h */
Chris@42 107
Chris@42 108 /*
Chris@42 109 * This function contains 23 FP additions, 14 FP multiplications,
Chris@42 110 * (or, 21 additions, 12 multiplications, 2 fused multiply/add),
Chris@42 111 * 19 stack variables, 2 constants, and 12 memory accesses
Chris@42 112 */
Chris@42 113 #include "t1f.h"
Chris@42 114
Chris@42 115 static void t1fv_6(R *ri, R *ii, const R *W, stride rs, INT mb, INT me, INT ms)
Chris@42 116 {
Chris@42 117 DVK(KP500000000, +0.500000000000000000000000000000000000000000000);
Chris@42 118 DVK(KP866025403, +0.866025403784438646763723170752936183471402627);
Chris@42 119 {
Chris@42 120 INT m;
Chris@42 121 R *x;
Chris@42 122 x = ri;
Chris@42 123 for (m = mb, W = W + (mb * ((TWVL / VL) * 10)); m < me; m = m + VL, x = x + (VL * ms), W = W + (TWVL * 10), MAKE_VOLATILE_STRIDE(6, rs)) {
Chris@42 124 V T4, Ti, Te, Tk, T9, Tj, T1, T3, T2;
Chris@42 125 T1 = LD(&(x[0]), ms, &(x[0]));
Chris@42 126 T2 = LD(&(x[WS(rs, 3)]), ms, &(x[WS(rs, 1)]));
Chris@42 127 T3 = BYTWJ(&(W[TWVL * 4]), T2);
Chris@42 128 T4 = VSUB(T1, T3);
Chris@42 129 Ti = VADD(T1, T3);
Chris@42 130 {
Chris@42 131 V Tb, Td, Ta, Tc;
Chris@42 132 Ta = LD(&(x[WS(rs, 4)]), ms, &(x[0]));
Chris@42 133 Tb = BYTWJ(&(W[TWVL * 6]), Ta);
Chris@42 134 Tc = LD(&(x[WS(rs, 1)]), ms, &(x[WS(rs, 1)]));
Chris@42 135 Td = BYTWJ(&(W[0]), Tc);
Chris@42 136 Te = VSUB(Tb, Td);
Chris@42 137 Tk = VADD(Tb, Td);
Chris@42 138 }
Chris@42 139 {
Chris@42 140 V T6, T8, T5, T7;
Chris@42 141 T5 = LD(&(x[WS(rs, 2)]), ms, &(x[0]));
Chris@42 142 T6 = BYTWJ(&(W[TWVL * 2]), T5);
Chris@42 143 T7 = LD(&(x[WS(rs, 5)]), ms, &(x[WS(rs, 1)]));
Chris@42 144 T8 = BYTWJ(&(W[TWVL * 8]), T7);
Chris@42 145 T9 = VSUB(T6, T8);
Chris@42 146 Tj = VADD(T6, T8);
Chris@42 147 }
Chris@42 148 {
Chris@42 149 V Th, Tf, Tg, Tn, Tl, Tm;
Chris@42 150 Th = VBYI(VMUL(LDK(KP866025403), VSUB(Te, T9)));
Chris@42 151 Tf = VADD(T9, Te);
Chris@42 152 Tg = VFNMS(LDK(KP500000000), Tf, T4);
Chris@42 153 ST(&(x[WS(rs, 3)]), VADD(T4, Tf), ms, &(x[WS(rs, 1)]));
Chris@42 154 ST(&(x[WS(rs, 1)]), VADD(Tg, Th), ms, &(x[WS(rs, 1)]));
Chris@42 155 ST(&(x[WS(rs, 5)]), VSUB(Tg, Th), ms, &(x[WS(rs, 1)]));
Chris@42 156 Tn = VBYI(VMUL(LDK(KP866025403), VSUB(Tk, Tj)));
Chris@42 157 Tl = VADD(Tj, Tk);
Chris@42 158 Tm = VFNMS(LDK(KP500000000), Tl, Ti);
Chris@42 159 ST(&(x[0]), VADD(Ti, Tl), ms, &(x[0]));
Chris@42 160 ST(&(x[WS(rs, 4)]), VADD(Tm, Tn), ms, &(x[0]));
Chris@42 161 ST(&(x[WS(rs, 2)]), VSUB(Tm, Tn), ms, &(x[0]));
Chris@42 162 }
Chris@42 163 }
Chris@42 164 }
Chris@42 165 VLEAVE();
Chris@42 166 }
Chris@42 167
Chris@42 168 static const tw_instr twinstr[] = {
Chris@42 169 VTW(0, 1),
Chris@42 170 VTW(0, 2),
Chris@42 171 VTW(0, 3),
Chris@42 172 VTW(0, 4),
Chris@42 173 VTW(0, 5),
Chris@42 174 {TW_NEXT, VL, 0}
Chris@42 175 };
Chris@42 176
Chris@42 177 static const ct_desc desc = { 6, XSIMD_STRING("t1fv_6"), twinstr, &GENUS, {21, 12, 2, 0}, 0, 0, 0 };
Chris@42 178
Chris@42 179 void XSIMD(codelet_t1fv_6) (planner *p) {
Chris@42 180 X(kdft_dit_register) (p, t1fv_6, &desc);
Chris@42 181 }
Chris@42 182 #endif /* HAVE_FMA */