Mercurial > hg > sv-dependency-builds

annotate src/fftw-3.3.8/dft/scalar/codelets/n1_6.c @ 82:d0c2a83c1364

Find changesets by keywords (author, files, the commit message), revision number or hash, or revset expression.
Add FFTW 3.3.8 source, and a Linux build
author Chris Cannam
date Tue, 19 Nov 2019 14:52:55 +0000
parents
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 08:04:10 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_notw.native -fma -compact -variables 4 -pipeline-latency 4 -n 6 -name n1_6 -include dft/scalar/n.h */
Chris@82 29
Chris@82 30 /*
Chris@82 31 * This function contains 36 FP additions, 12 FP multiplications,
Chris@82 32 * (or, 24 additions, 0 multiplications, 12 fused multiply/add),
Chris@82 33 * 23 stack variables, 2 constants, and 24 memory accesses
Chris@82 34 */
Chris@82 35 #include "dft/scalar/n.h"
Chris@82 36
Chris@82 37 static void n1_6(const R *ri, const R *ii, R *ro, R *io, stride is, stride os, INT v, INT ivs, INT ovs)
Chris@82 38 {
Chris@82 39 DK(KP866025403, +0.866025403784438646763723170752936183471402627);
Chris@82 40 DK(KP500000000, +0.500000000000000000000000000000000000000000000);
Chris@82 41 {
Chris@82 42 INT i;
Chris@82 43 for (i = v; i > 0; i = i - 1, ri = ri + ivs, ii = ii + ivs, ro = ro + ovs, io = io + ovs, MAKE_VOLATILE_STRIDE(24, is), MAKE_VOLATILE_STRIDE(24, os)) {
Chris@82 44 E T3, Tb, Tp, Tx, T6, Tc, T9, Td, Ta, Te, Ti, Tu, Tl, Tv, Tq;
Chris@82 45 E Ty;
Chris@82 46 {
Chris@82 47 E T1, T2, Tn, To;
Chris@82 48 T1 = ri[0];
Chris@82 49 T2 = ri[WS(is, 3)];
Chris@82 50 T3 = T1 - T2;
Chris@82 51 Tb = T1 + T2;
Chris@82 52 Tn = ii[0];
Chris@82 53 To = ii[WS(is, 3)];
Chris@82 54 Tp = Tn - To;
Chris@82 55 Tx = Tn + To;
Chris@82 56 }
Chris@82 57 {
Chris@82 58 E T4, T5, T7, T8;
Chris@82 59 T4 = ri[WS(is, 2)];
Chris@82 60 T5 = ri[WS(is, 5)];
Chris@82 61 T6 = T4 - T5;
Chris@82 62 Tc = T4 + T5;
Chris@82 63 T7 = ri[WS(is, 4)];
Chris@82 64 T8 = ri[WS(is, 1)];
Chris@82 65 T9 = T7 - T8;
Chris@82 66 Td = T7 + T8;
Chris@82 67 }
Chris@82 68 Ta = T6 + T9;
Chris@82 69 Te = Tc + Td;
Chris@82 70 {
Chris@82 71 E Tg, Th, Tj, Tk;
Chris@82 72 Tg = ii[WS(is, 2)];
Chris@82 73 Th = ii[WS(is, 5)];
Chris@82 74 Ti = Tg - Th;
Chris@82 75 Tu = Tg + Th;
Chris@82 76 Tj = ii[WS(is, 4)];
Chris@82 77 Tk = ii[WS(is, 1)];
Chris@82 78 Tl = Tj - Tk;
Chris@82 79 Tv = Tj + Tk;
Chris@82 80 }
Chris@82 81 Tq = Ti + Tl;
Chris@82 82 Ty = Tu + Tv;
Chris@82 83 ro[WS(os, 3)] = T3 + Ta;
Chris@82 84 io[WS(os, 3)] = Tp + Tq;
Chris@82 85 ro[0] = Tb + Te;
Chris@82 86 io[0] = Tx + Ty;
Chris@82 87 {
Chris@82 88 E Tf, Tm, Tr, Ts;
Chris@82 89 Tf = FNMS(KP500000000, Ta, T3);
Chris@82 90 Tm = Ti - Tl;
Chris@82 91 ro[WS(os, 5)] = FNMS(KP866025403, Tm, Tf);
Chris@82 92 ro[WS(os, 1)] = FMA(KP866025403, Tm, Tf);
Chris@82 93 Tr = FNMS(KP500000000, Tq, Tp);
Chris@82 94 Ts = T9 - T6;
Chris@82 95 io[WS(os, 1)] = FMA(KP866025403, Ts, Tr);
Chris@82 96 io[WS(os, 5)] = FNMS(KP866025403, Ts, Tr);
Chris@82 97 }
Chris@82 98 {
Chris@82 99 E Tt, Tw, Tz, TA;
Chris@82 100 Tt = FNMS(KP500000000, Te, Tb);
Chris@82 101 Tw = Tu - Tv;
Chris@82 102 ro[WS(os, 2)] = FNMS(KP866025403, Tw, Tt);
Chris@82 103 ro[WS(os, 4)] = FMA(KP866025403, Tw, Tt);
Chris@82 104 Tz = FNMS(KP500000000, Ty, Tx);
Chris@82 105 TA = Td - Tc;
Chris@82 106 io[WS(os, 2)] = FNMS(KP866025403, TA, Tz);
Chris@82 107 io[WS(os, 4)] = FMA(KP866025403, TA, Tz);
Chris@82 108 }
Chris@82 109 }
Chris@82 110 }
Chris@82 111 }
Chris@82 112
Chris@82 113 static const kdft_desc desc = { 6, "n1_6", {24, 0, 12, 0}, &GENUS, 0, 0, 0, 0 };
Chris@82 114
Chris@82 115 void X(codelet_n1_6) (planner *p) {
Chris@82 116 X(kdft_register) (p, n1_6, &desc);
Chris@82 117 }
Chris@82 118
Chris@82 119 #else
Chris@82 120
Chris@82 121 /* Generated by: ../../../genfft/gen_notw.native -compact -variables 4 -pipeline-latency 4 -n 6 -name n1_6 -include dft/scalar/n.h */
Chris@82 122
Chris@82 123 /*
Chris@82 124 * This function contains 36 FP additions, 8 FP multiplications,
Chris@82 125 * (or, 32 additions, 4 multiplications, 4 fused multiply/add),
Chris@82 126 * 23 stack variables, 2 constants, and 24 memory accesses
Chris@82 127 */
Chris@82 128 #include "dft/scalar/n.h"
Chris@82 129
Chris@82 130 static void n1_6(const R *ri, const R *ii, R *ro, R *io, stride is, stride os, INT v, INT ivs, INT ovs)
Chris@82 131 {
Chris@82 132 DK(KP866025403, +0.866025403784438646763723170752936183471402627);
Chris@82 133 DK(KP500000000, +0.500000000000000000000000000000000000000000000);
Chris@82 134 {
Chris@82 135 INT i;
Chris@82 136 for (i = v; i > 0; i = i - 1, ri = ri + ivs, ii = ii + ivs, ro = ro + ovs, io = io + ovs, MAKE_VOLATILE_STRIDE(24, is), MAKE_VOLATILE_STRIDE(24, os)) {
Chris@82 137 E T3, Tb, Tq, Tx, T6, Tc, T9, Td, Ta, Te, Ti, Tu, Tl, Tv, Tr;
Chris@82 138 E Ty;
Chris@82 139 {
Chris@82 140 E T1, T2, To, Tp;
Chris@82 141 T1 = ri[0];
Chris@82 142 T2 = ri[WS(is, 3)];
Chris@82 143 T3 = T1 - T2;
Chris@82 144 Tb = T1 + T2;
Chris@82 145 To = ii[0];
Chris@82 146 Tp = ii[WS(is, 3)];
Chris@82 147 Tq = To - Tp;
Chris@82 148 Tx = To + Tp;
Chris@82 149 }
Chris@82 150 {
Chris@82 151 E T4, T5, T7, T8;
Chris@82 152 T4 = ri[WS(is, 2)];
Chris@82 153 T5 = ri[WS(is, 5)];
Chris@82 154 T6 = T4 - T5;
Chris@82 155 Tc = T4 + T5;
Chris@82 156 T7 = ri[WS(is, 4)];
Chris@82 157 T8 = ri[WS(is, 1)];
Chris@82 158 T9 = T7 - T8;
Chris@82 159 Td = T7 + T8;
Chris@82 160 }
Chris@82 161 Ta = T6 + T9;
Chris@82 162 Te = Tc + Td;
Chris@82 163 {
Chris@82 164 E Tg, Th, Tj, Tk;
Chris@82 165 Tg = ii[WS(is, 2)];
Chris@82 166 Th = ii[WS(is, 5)];
Chris@82 167 Ti = Tg - Th;
Chris@82 168 Tu = Tg + Th;
Chris@82 169 Tj = ii[WS(is, 4)];
Chris@82 170 Tk = ii[WS(is, 1)];
Chris@82 171 Tl = Tj - Tk;
Chris@82 172 Tv = Tj + Tk;
Chris@82 173 }
Chris@82 174 Tr = Ti + Tl;
Chris@82 175 Ty = Tu + Tv;
Chris@82 176 ro[WS(os, 3)] = T3 + Ta;
Chris@82 177 io[WS(os, 3)] = Tq + Tr;
Chris@82 178 ro[0] = Tb + Te;
Chris@82 179 io[0] = Tx + Ty;
Chris@82 180 {
Chris@82 181 E Tf, Tm, Tn, Ts;
Chris@82 182 Tf = FNMS(KP500000000, Ta, T3);
Chris@82 183 Tm = KP866025403 * (Ti - Tl);
Chris@82 184 ro[WS(os, 5)] = Tf - Tm;
Chris@82 185 ro[WS(os, 1)] = Tf + Tm;
Chris@82 186 Tn = KP866025403 * (T9 - T6);
Chris@82 187 Ts = FNMS(KP500000000, Tr, Tq);
Chris@82 188 io[WS(os, 1)] = Tn + Ts;
Chris@82 189 io[WS(os, 5)] = Ts - Tn;
Chris@82 190 }
Chris@82 191 {
Chris@82 192 E Tt, Tw, Tz, TA;
Chris@82 193 Tt = FNMS(KP500000000, Te, Tb);
Chris@82 194 Tw = KP866025403 * (Tu - Tv);
Chris@82 195 ro[WS(os, 2)] = Tt - Tw;
Chris@82 196 ro[WS(os, 4)] = Tt + Tw;
Chris@82 197 Tz = FNMS(KP500000000, Ty, Tx);
Chris@82 198 TA = KP866025403 * (Td - Tc);
Chris@82 199 io[WS(os, 2)] = Tz - TA;
Chris@82 200 io[WS(os, 4)] = TA + Tz;
Chris@82 201 }
Chris@82 202 }
Chris@82 203 }
Chris@82 204 }
Chris@82 205
Chris@82 206 static const kdft_desc desc = { 6, "n1_6", {32, 4, 4, 0}, &GENUS, 0, 0, 0, 0 };
Chris@82 207
Chris@82 208 void X(codelet_n1_6) (planner *p) {
Chris@82 209 X(kdft_register) (p, n1_6, &desc);
Chris@82 210 }
Chris@82 211
Chris@82 212 #endif