Mercurial > hg > sv-dependency-builds

annotate src/fftw-3.3.5/dft/scalar/codelets/t1_3.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 16:36:05 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.native -fma -reorder-insns -schedule-for-pipeline -compact -variables 4 -pipeline-latency 4 -n 3 -name t1_3 -include t.h */
Chris@42 29
Chris@42 30 /*
Chris@42 31 * This function contains 16 FP additions, 14 FP multiplications,
Chris@42 32 * (or, 6 additions, 4 multiplications, 10 fused multiply/add),
Chris@42 33 * 21 stack variables, 2 constants, and 12 memory accesses
Chris@42 34 */
Chris@42 35 #include "t.h"
Chris@42 36
Chris@42 37 static void t1_3(R *ri, R *ii, const R *W, stride rs, INT mb, INT me, INT ms)
Chris@42 38 {
Chris@42 39 DK(KP866025403, +0.866025403784438646763723170752936183471402627);
Chris@42 40 DK(KP500000000, +0.500000000000000000000000000000000000000000000);
Chris@42 41 {
Chris@42 42 INT m;
Chris@42 43 for (m = mb, W = W + (mb * 4); m < me; m = m + 1, ri = ri + ms, ii = ii + ms, W = W + 4, MAKE_VOLATILE_STRIDE(6, rs)) {
Chris@42 44 E T1, Tm, T9, Tc, Tb, Th, T7, Ti, Ta, Tj, Td;
Chris@42 45 T1 = ri[0];
Chris@42 46 Tm = ii[0];
Chris@42 47 {
Chris@42 48 E T3, T6, T2, T5, Tg, T4, T8;
Chris@42 49 T3 = ri[WS(rs, 1)];
Chris@42 50 T6 = ii[WS(rs, 1)];
Chris@42 51 T2 = W[0];
Chris@42 52 T5 = W[1];
Chris@42 53 T9 = ri[WS(rs, 2)];
Chris@42 54 Tc = ii[WS(rs, 2)];
Chris@42 55 Tg = T2 * T6;
Chris@42 56 T4 = T2 * T3;
Chris@42 57 T8 = W[2];
Chris@42 58 Tb = W[3];
Chris@42 59 Th = FNMS(T5, T3, Tg);
Chris@42 60 T7 = FMA(T5, T6, T4);
Chris@42 61 Ti = T8 * Tc;
Chris@42 62 Ta = T8 * T9;
Chris@42 63 }
Chris@42 64 Tj = FNMS(Tb, T9, Ti);
Chris@42 65 Td = FMA(Tb, Tc, Ta);
Chris@42 66 {
Chris@42 67 E Tk, Te, To, Tn, Tl, Tf;
Chris@42 68 Tk = Th - Tj;
Chris@42 69 Tl = Th + Tj;
Chris@42 70 Te = T7 + Td;
Chris@42 71 To = Td - T7;
Chris@42 72 ii[0] = Tl + Tm;
Chris@42 73 Tn = FNMS(KP500000000, Tl, Tm);
Chris@42 74 ri[0] = T1 + Te;
Chris@42 75 Tf = FNMS(KP500000000, Te, T1);
Chris@42 76 ii[WS(rs, 1)] = FMA(KP866025403, To, Tn);
Chris@42 77 ii[WS(rs, 2)] = FNMS(KP866025403, To, Tn);
Chris@42 78 ri[WS(rs, 2)] = FNMS(KP866025403, Tk, Tf);
Chris@42 79 ri[WS(rs, 1)] = FMA(KP866025403, Tk, Tf);
Chris@42 80 }
Chris@42 81 }
Chris@42 82 }
Chris@42 83 }
Chris@42 84
Chris@42 85 static const tw_instr twinstr[] = {
Chris@42 86 {TW_FULL, 0, 3},
Chris@42 87 {TW_NEXT, 1, 0}
Chris@42 88 };
Chris@42 89
Chris@42 90 static const ct_desc desc = { 3, "t1_3", twinstr, &GENUS, {6, 4, 10, 0}, 0, 0, 0 };
Chris@42 91
Chris@42 92 void X(codelet_t1_3) (planner *p) {
Chris@42 93 X(kdft_dit_register) (p, t1_3, &desc);
Chris@42 94 }
Chris@42 95 #else /* HAVE_FMA */
Chris@42 96
Chris@42 97 /* Generated by: ../../../genfft/gen_twiddle.native -compact -variables 4 -pipeline-latency 4 -n 3 -name t1_3 -include t.h */
Chris@42 98
Chris@42 99 /*
Chris@42 100 * This function contains 16 FP additions, 12 FP multiplications,
Chris@42 101 * (or, 10 additions, 6 multiplications, 6 fused multiply/add),
Chris@42 102 * 15 stack variables, 2 constants, and 12 memory accesses
Chris@42 103 */
Chris@42 104 #include "t.h"
Chris@42 105
Chris@42 106 static void t1_3(R *ri, R *ii, const R *W, stride rs, INT mb, INT me, INT ms)
Chris@42 107 {
Chris@42 108 DK(KP866025403, +0.866025403784438646763723170752936183471402627);
Chris@42 109 DK(KP500000000, +0.500000000000000000000000000000000000000000000);
Chris@42 110 {
Chris@42 111 INT m;
Chris@42 112 for (m = mb, W = W + (mb * 4); m < me; m = m + 1, ri = ri + ms, ii = ii + ms, W = W + 4, MAKE_VOLATILE_STRIDE(6, rs)) {
Chris@42 113 E T1, Ti, T6, Te, Tb, Tf, Tc, Th;
Chris@42 114 T1 = ri[0];
Chris@42 115 Ti = ii[0];
Chris@42 116 {
Chris@42 117 E T3, T5, T2, T4;
Chris@42 118 T3 = ri[WS(rs, 1)];
Chris@42 119 T5 = ii[WS(rs, 1)];
Chris@42 120 T2 = W[0];
Chris@42 121 T4 = W[1];
Chris@42 122 T6 = FMA(T2, T3, T4 * T5);
Chris@42 123 Te = FNMS(T4, T3, T2 * T5);
Chris@42 124 }
Chris@42 125 {
Chris@42 126 E T8, Ta, T7, T9;
Chris@42 127 T8 = ri[WS(rs, 2)];
Chris@42 128 Ta = ii[WS(rs, 2)];
Chris@42 129 T7 = W[2];
Chris@42 130 T9 = W[3];
Chris@42 131 Tb = FMA(T7, T8, T9 * Ta);
Chris@42 132 Tf = FNMS(T9, T8, T7 * Ta);
Chris@42 133 }
Chris@42 134 Tc = T6 + Tb;
Chris@42 135 Th = Te + Tf;
Chris@42 136 ri[0] = T1 + Tc;
Chris@42 137 ii[0] = Th + Ti;
Chris@42 138 {
Chris@42 139 E Td, Tg, Tj, Tk;
Chris@42 140 Td = FNMS(KP500000000, Tc, T1);
Chris@42 141 Tg = KP866025403 * (Te - Tf);
Chris@42 142 ri[WS(rs, 2)] = Td - Tg;
Chris@42 143 ri[WS(rs, 1)] = Td + Tg;
Chris@42 144 Tj = KP866025403 * (Tb - T6);
Chris@42 145 Tk = FNMS(KP500000000, Th, Ti);
Chris@42 146 ii[WS(rs, 1)] = Tj + Tk;
Chris@42 147 ii[WS(rs, 2)] = Tk - Tj;
Chris@42 148 }
Chris@42 149 }
Chris@42 150 }
Chris@42 151 }
Chris@42 152
Chris@42 153 static const tw_instr twinstr[] = {
Chris@42 154 {TW_FULL, 0, 3},
Chris@42 155 {TW_NEXT, 1, 0}
Chris@42 156 };
Chris@42 157
Chris@42 158 static const ct_desc desc = { 3, "t1_3", twinstr, &GENUS, {10, 6, 6, 0}, 0, 0, 0 };
Chris@42 159
Chris@42 160 void X(codelet_t1_3) (planner *p) {
Chris@42 161 X(kdft_dit_register) (p, t1_3, &desc);
Chris@42 162 }
Chris@42 163 #endif /* HAVE_FMA */