Mercurial > hg > sv-dependency-builds

annotate src/fftw-3.3.8/dft/scalar/codelets/q1_2.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 d0c2a83c1364
children
rev   line source
Chris@82 1 /*
Chris@82 2 * Copyright (c) 2003, 2007-14 Matteo Frigo
Chris@82 3 * Copyright (c) 2003, 2007-14 Massachusetts Institute of Technology
Chris@82 4 *
Chris@82 5 * This program is free software; you can redistribute it and/or modify
Chris@82 6 * it under the terms of the GNU General Public License as published by
Chris@82 7 * the Free Software Foundation; either version 2 of the License, or
Chris@82 8 * (at your option) any later version.
Chris@82 9 *
Chris@82 10 * This program is distributed in the hope that it will be useful,
Chris@82 11 * but WITHOUT ANY WARRANTY; without even the implied warranty of
Chris@82 12 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
Chris@82 13 * GNU General Public License for more details.
Chris@82 14 *
Chris@82 15 * You should have received a copy of the GNU General Public License
Chris@82 16 * along with this program; if not, write to the Free Software
Chris@82 17 * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
Chris@82 18 *
Chris@82 19 */
Chris@82 20
Chris@82 21 /* This file was automatically generated --- DO NOT EDIT */
Chris@82 22 /* Generated on Thu May 24 08:04:29 EDT 2018 */
Chris@82 23
Chris@82 24 #include "dft/codelet-dft.h"
Chris@82 25
Chris@82 26 #if defined(ARCH_PREFERS_FMA) || defined(ISA_EXTENSION_PREFERS_FMA)
Chris@82 27
Chris@82 28 /* Generated by: ../../../genfft/gen_twidsq.native -fma -compact -variables 4 -pipeline-latency 4 -reload-twiddle -dif -n 2 -name q1_2 -include dft/scalar/q.h */
Chris@82 29
Chris@82 30 /*
Chris@82 31 * This function contains 12 FP additions, 8 FP multiplications,
Chris@82 32 * (or, 8 additions, 4 multiplications, 4 fused multiply/add),
Chris@82 33 * 17 stack variables, 0 constants, and 16 memory accesses
Chris@82 34 */
Chris@82 35 #include "dft/scalar/q.h"
Chris@82 36
Chris@82 37 static void q1_2(R *rio, R *iio, const R *W, stride rs, stride vs, INT mb, INT me, INT ms)
Chris@82 38 {
Chris@82 39 {
Chris@82 40 INT m;
Chris@82 41 for (m = mb, W = W + (mb * 2); m < me; m = m + 1, rio = rio + ms, iio = iio + ms, W = W + 2, MAKE_VOLATILE_STRIDE(4, rs), MAKE_VOLATILE_STRIDE(0, vs)) {
Chris@82 42 E T1, T2, T4, T7, T8, T9, Tb, Tc, Te, Th, Ti, Tj;
Chris@82 43 T1 = rio[0];
Chris@82 44 T2 = rio[WS(rs, 1)];
Chris@82 45 T4 = T1 - T2;
Chris@82 46 T7 = iio[0];
Chris@82 47 T8 = iio[WS(rs, 1)];
Chris@82 48 T9 = T7 - T8;
Chris@82 49 Tb = rio[WS(vs, 1)];
Chris@82 50 Tc = rio[WS(vs, 1) + WS(rs, 1)];
Chris@82 51 Te = Tb - Tc;
Chris@82 52 Th = iio[WS(vs, 1)];
Chris@82 53 Ti = iio[WS(vs, 1) + WS(rs, 1)];
Chris@82 54 Tj = Th - Ti;
Chris@82 55 rio[0] = T1 + T2;
Chris@82 56 iio[0] = T7 + T8;
Chris@82 57 rio[WS(rs, 1)] = Tb + Tc;
Chris@82 58 iio[WS(rs, 1)] = Th + Ti;
Chris@82 59 {
Chris@82 60 E Tf, Tk, Td, Tg;
Chris@82 61 Td = W[0];
Chris@82 62 Tf = Td * Te;
Chris@82 63 Tk = Td * Tj;
Chris@82 64 Tg = W[1];
Chris@82 65 rio[WS(vs, 1) + WS(rs, 1)] = FMA(Tg, Tj, Tf);
Chris@82 66 iio[WS(vs, 1) + WS(rs, 1)] = FNMS(Tg, Te, Tk);
Chris@82 67 }
Chris@82 68 {
Chris@82 69 E T5, Ta, T3, T6;
Chris@82 70 T3 = W[0];
Chris@82 71 T5 = T3 * T4;
Chris@82 72 Ta = T3 * T9;
Chris@82 73 T6 = W[1];
Chris@82 74 rio[WS(vs, 1)] = FMA(T6, T9, T5);
Chris@82 75 iio[WS(vs, 1)] = FNMS(T6, T4, Ta);
Chris@82 76 }
Chris@82 77 }
Chris@82 78 }
Chris@82 79 }
Chris@82 80
Chris@82 81 static const tw_instr twinstr[] = {
Chris@82 82 {TW_FULL, 0, 2},
Chris@82 83 {TW_NEXT, 1, 0}
Chris@82 84 };
Chris@82 85
Chris@82 86 static const ct_desc desc = { 2, "q1_2", twinstr, &GENUS, {8, 4, 4, 0}, 0, 0, 0 };
Chris@82 87
Chris@82 88 void X(codelet_q1_2) (planner *p) {
Chris@82 89 X(kdft_difsq_register) (p, q1_2, &desc);
Chris@82 90 }
Chris@82 91 #else
Chris@82 92
Chris@82 93 /* Generated by: ../../../genfft/gen_twidsq.native -compact -variables 4 -pipeline-latency 4 -reload-twiddle -dif -n 2 -name q1_2 -include dft/scalar/q.h */
Chris@82 94
Chris@82 95 /*
Chris@82 96 * This function contains 12 FP additions, 8 FP multiplications,
Chris@82 97 * (or, 8 additions, 4 multiplications, 4 fused multiply/add),
Chris@82 98 * 17 stack variables, 0 constants, and 16 memory accesses
Chris@82 99 */
Chris@82 100 #include "dft/scalar/q.h"
Chris@82 101
Chris@82 102 static void q1_2(R *rio, R *iio, const R *W, stride rs, stride vs, INT mb, INT me, INT ms)
Chris@82 103 {
Chris@82 104 {
Chris@82 105 INT m;
Chris@82 106 for (m = mb, W = W + (mb * 2); m < me; m = m + 1, rio = rio + ms, iio = iio + ms, W = W + 2, MAKE_VOLATILE_STRIDE(4, rs), MAKE_VOLATILE_STRIDE(0, vs)) {
Chris@82 107 E T1, T2, T4, T6, T7, T8, T9, Ta, Tc, Te, Tf, Tg;
Chris@82 108 T1 = rio[0];
Chris@82 109 T2 = rio[WS(rs, 1)];
Chris@82 110 T4 = T1 - T2;
Chris@82 111 T6 = iio[0];
Chris@82 112 T7 = iio[WS(rs, 1)];
Chris@82 113 T8 = T6 - T7;
Chris@82 114 T9 = rio[WS(vs, 1)];
Chris@82 115 Ta = rio[WS(vs, 1) + WS(rs, 1)];
Chris@82 116 Tc = T9 - Ta;
Chris@82 117 Te = iio[WS(vs, 1)];
Chris@82 118 Tf = iio[WS(vs, 1) + WS(rs, 1)];
Chris@82 119 Tg = Te - Tf;
Chris@82 120 rio[0] = T1 + T2;
Chris@82 121 iio[0] = T6 + T7;
Chris@82 122 rio[WS(rs, 1)] = T9 + Ta;
Chris@82 123 iio[WS(rs, 1)] = Te + Tf;
Chris@82 124 {
Chris@82 125 E Tb, Td, T3, T5;
Chris@82 126 Tb = W[0];
Chris@82 127 Td = W[1];
Chris@82 128 rio[WS(vs, 1) + WS(rs, 1)] = FMA(Tb, Tc, Td * Tg);
Chris@82 129 iio[WS(vs, 1) + WS(rs, 1)] = FNMS(Td, Tc, Tb * Tg);
Chris@82 130 T3 = W[0];
Chris@82 131 T5 = W[1];
Chris@82 132 rio[WS(vs, 1)] = FMA(T3, T4, T5 * T8);
Chris@82 133 iio[WS(vs, 1)] = FNMS(T5, T4, T3 * T8);
Chris@82 134 }
Chris@82 135 }
Chris@82 136 }
Chris@82 137 }
Chris@82 138
Chris@82 139 static const tw_instr twinstr[] = {
Chris@82 140 {TW_FULL, 0, 2},
Chris@82 141 {TW_NEXT, 1, 0}
Chris@82 142 };
Chris@82 143
Chris@82 144 static const ct_desc desc = { 2, "q1_2", twinstr, &GENUS, {8, 4, 4, 0}, 0, 0, 0 };
Chris@82 145
Chris@82 146 void X(codelet_q1_2) (planner *p) {
Chris@82 147 X(kdft_difsq_register) (p, q1_2, &desc);
Chris@82 148 }
Chris@82 149 #endif