Mercurial > hg > sv-dependency-builds

annotate src/fftw-3.3.8/dft/simd/common/n1bv_8.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 08:04:55 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_c.native -fma -simd -compact -variables 4 -pipeline-latency 8 -sign 1 -n 8 -name n1bv_8 -include dft/simd/n1b.h */
Chris@82 29
Chris@82 30 /*
Chris@82 31 * This function contains 26 FP additions, 10 FP multiplications,
Chris@82 32 * (or, 16 additions, 0 multiplications, 10 fused multiply/add),
Chris@82 33 * 22 stack variables, 1 constants, and 16 memory accesses
Chris@82 34 */
Chris@82 35 #include "dft/simd/n1b.h"
Chris@82 36
Chris@82 37 static void n1bv_8(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 DVK(KP707106781, +0.707106781186547524400844362104849039284835938);
Chris@82 40 {
Chris@82 41 INT i;
Chris@82 42 const R *xi;
Chris@82 43 R *xo;
Chris@82 44 xi = ii;
Chris@82 45 xo = io;
Chris@82 46 for (i = v; i > 0; i = i - VL, xi = xi + (VL * ivs), xo = xo + (VL * ovs), MAKE_VOLATILE_STRIDE(16, is), MAKE_VOLATILE_STRIDE(16, os)) {
Chris@82 47 V T3, Tj, Te, Tk, Ta, Tn, Tf, Tm;
Chris@82 48 {
Chris@82 49 V T1, T2, Tc, Td;
Chris@82 50 T1 = LD(&(xi[0]), ivs, &(xi[0]));
Chris@82 51 T2 = LD(&(xi[WS(is, 4)]), ivs, &(xi[0]));
Chris@82 52 T3 = VSUB(T1, T2);
Chris@82 53 Tj = VADD(T1, T2);
Chris@82 54 Tc = LD(&(xi[WS(is, 2)]), ivs, &(xi[0]));
Chris@82 55 Td = LD(&(xi[WS(is, 6)]), ivs, &(xi[0]));
Chris@82 56 Te = VSUB(Tc, Td);
Chris@82 57 Tk = VADD(Tc, Td);
Chris@82 58 {
Chris@82 59 V T4, T5, T6, T7, T8, T9;
Chris@82 60 T4 = LD(&(xi[WS(is, 1)]), ivs, &(xi[WS(is, 1)]));
Chris@82 61 T5 = LD(&(xi[WS(is, 5)]), ivs, &(xi[WS(is, 1)]));
Chris@82 62 T6 = VSUB(T4, T5);
Chris@82 63 T7 = LD(&(xi[WS(is, 7)]), ivs, &(xi[WS(is, 1)]));
Chris@82 64 T8 = LD(&(xi[WS(is, 3)]), ivs, &(xi[WS(is, 1)]));
Chris@82 65 T9 = VSUB(T7, T8);
Chris@82 66 Ta = VADD(T6, T9);
Chris@82 67 Tn = VADD(T7, T8);
Chris@82 68 Tf = VSUB(T6, T9);
Chris@82 69 Tm = VADD(T4, T5);
Chris@82 70 }
Chris@82 71 }
Chris@82 72 {
Chris@82 73 V Tb, Tg, Tp, Tq;
Chris@82 74 Tb = VFNMS(LDK(KP707106781), Ta, T3);
Chris@82 75 Tg = VFNMS(LDK(KP707106781), Tf, Te);
Chris@82 76 ST(&(xo[WS(os, 3)]), VFNMSI(Tg, Tb), ovs, &(xo[WS(os, 1)]));
Chris@82 77 ST(&(xo[WS(os, 5)]), VFMAI(Tg, Tb), ovs, &(xo[WS(os, 1)]));
Chris@82 78 Tp = VADD(Tj, Tk);
Chris@82 79 Tq = VADD(Tm, Tn);
Chris@82 80 ST(&(xo[WS(os, 4)]), VSUB(Tp, Tq), ovs, &(xo[0]));
Chris@82 81 ST(&(xo[0]), VADD(Tp, Tq), ovs, &(xo[0]));
Chris@82 82 }
Chris@82 83 {
Chris@82 84 V Th, Ti, Tl, To;
Chris@82 85 Th = VFMA(LDK(KP707106781), Ta, T3);
Chris@82 86 Ti = VFMA(LDK(KP707106781), Tf, Te);
Chris@82 87 ST(&(xo[WS(os, 1)]), VFMAI(Ti, Th), ovs, &(xo[WS(os, 1)]));
Chris@82 88 ST(&(xo[WS(os, 7)]), VFNMSI(Ti, Th), ovs, &(xo[WS(os, 1)]));
Chris@82 89 Tl = VSUB(Tj, Tk);
Chris@82 90 To = VSUB(Tm, Tn);
Chris@82 91 ST(&(xo[WS(os, 6)]), VFNMSI(To, Tl), ovs, &(xo[0]));
Chris@82 92 ST(&(xo[WS(os, 2)]), VFMAI(To, Tl), ovs, &(xo[0]));
Chris@82 93 }
Chris@82 94 }
Chris@82 95 }
Chris@82 96 VLEAVE();
Chris@82 97 }
Chris@82 98
Chris@82 99 static const kdft_desc desc = { 8, XSIMD_STRING("n1bv_8"), {16, 0, 10, 0}, &GENUS, 0, 0, 0, 0 };
Chris@82 100
Chris@82 101 void XSIMD(codelet_n1bv_8) (planner *p) {
Chris@82 102 X(kdft_register) (p, n1bv_8, &desc);
Chris@82 103 }
Chris@82 104
Chris@82 105 #else
Chris@82 106
Chris@82 107 /* Generated by: ../../../genfft/gen_notw_c.native -simd -compact -variables 4 -pipeline-latency 8 -sign 1 -n 8 -name n1bv_8 -include dft/simd/n1b.h */
Chris@82 108
Chris@82 109 /*
Chris@82 110 * This function contains 26 FP additions, 2 FP multiplications,
Chris@82 111 * (or, 26 additions, 2 multiplications, 0 fused multiply/add),
Chris@82 112 * 22 stack variables, 1 constants, and 16 memory accesses
Chris@82 113 */
Chris@82 114 #include "dft/simd/n1b.h"
Chris@82 115
Chris@82 116 static void n1bv_8(const R *ri, const R *ii, R *ro, R *io, stride is, stride os, INT v, INT ivs, INT ovs)
Chris@82 117 {
Chris@82 118 DVK(KP707106781, +0.707106781186547524400844362104849039284835938);
Chris@82 119 {
Chris@82 120 INT i;
Chris@82 121 const R *xi;
Chris@82 122 R *xo;
Chris@82 123 xi = ii;
Chris@82 124 xo = io;
Chris@82 125 for (i = v; i > 0; i = i - VL, xi = xi + (VL * ivs), xo = xo + (VL * ovs), MAKE_VOLATILE_STRIDE(16, is), MAKE_VOLATILE_STRIDE(16, os)) {
Chris@82 126 V Ta, Tk, Te, Tj, T7, Tn, Tf, Tm;
Chris@82 127 {
Chris@82 128 V T8, T9, Tc, Td;
Chris@82 129 T8 = LD(&(xi[WS(is, 2)]), ivs, &(xi[0]));
Chris@82 130 T9 = LD(&(xi[WS(is, 6)]), ivs, &(xi[0]));
Chris@82 131 Ta = VSUB(T8, T9);
Chris@82 132 Tk = VADD(T8, T9);
Chris@82 133 Tc = LD(&(xi[0]), ivs, &(xi[0]));
Chris@82 134 Td = LD(&(xi[WS(is, 4)]), ivs, &(xi[0]));
Chris@82 135 Te = VSUB(Tc, Td);
Chris@82 136 Tj = VADD(Tc, Td);
Chris@82 137 {
Chris@82 138 V T1, T2, T3, T4, T5, T6;
Chris@82 139 T1 = LD(&(xi[WS(is, 1)]), ivs, &(xi[WS(is, 1)]));
Chris@82 140 T2 = LD(&(xi[WS(is, 5)]), ivs, &(xi[WS(is, 1)]));
Chris@82 141 T3 = VSUB(T1, T2);
Chris@82 142 T4 = LD(&(xi[WS(is, 7)]), ivs, &(xi[WS(is, 1)]));
Chris@82 143 T5 = LD(&(xi[WS(is, 3)]), ivs, &(xi[WS(is, 1)]));
Chris@82 144 T6 = VSUB(T4, T5);
Chris@82 145 T7 = VMUL(LDK(KP707106781), VSUB(T3, T6));
Chris@82 146 Tn = VADD(T4, T5);
Chris@82 147 Tf = VMUL(LDK(KP707106781), VADD(T3, T6));
Chris@82 148 Tm = VADD(T1, T2);
Chris@82 149 }
Chris@82 150 }
Chris@82 151 {
Chris@82 152 V Tb, Tg, Tp, Tq;
Chris@82 153 Tb = VBYI(VSUB(T7, Ta));
Chris@82 154 Tg = VSUB(Te, Tf);
Chris@82 155 ST(&(xo[WS(os, 3)]), VADD(Tb, Tg), ovs, &(xo[WS(os, 1)]));
Chris@82 156 ST(&(xo[WS(os, 5)]), VSUB(Tg, Tb), ovs, &(xo[WS(os, 1)]));
Chris@82 157 Tp = VADD(Tj, Tk);
Chris@82 158 Tq = VADD(Tm, Tn);
Chris@82 159 ST(&(xo[WS(os, 4)]), VSUB(Tp, Tq), ovs, &(xo[0]));
Chris@82 160 ST(&(xo[0]), VADD(Tp, Tq), ovs, &(xo[0]));
Chris@82 161 }
Chris@82 162 {
Chris@82 163 V Th, Ti, Tl, To;
Chris@82 164 Th = VBYI(VADD(Ta, T7));
Chris@82 165 Ti = VADD(Te, Tf);
Chris@82 166 ST(&(xo[WS(os, 1)]), VADD(Th, Ti), ovs, &(xo[WS(os, 1)]));
Chris@82 167 ST(&(xo[WS(os, 7)]), VSUB(Ti, Th), ovs, &(xo[WS(os, 1)]));
Chris@82 168 Tl = VSUB(Tj, Tk);
Chris@82 169 To = VBYI(VSUB(Tm, Tn));
Chris@82 170 ST(&(xo[WS(os, 6)]), VSUB(Tl, To), ovs, &(xo[0]));
Chris@82 171 ST(&(xo[WS(os, 2)]), VADD(Tl, To), ovs, &(xo[0]));
Chris@82 172 }
Chris@82 173 }
Chris@82 174 }
Chris@82 175 VLEAVE();
Chris@82 176 }
Chris@82 177
Chris@82 178 static const kdft_desc desc = { 8, XSIMD_STRING("n1bv_8"), {26, 2, 0, 0}, &GENUS, 0, 0, 0, 0 };
Chris@82 179
Chris@82 180 void XSIMD(codelet_n1bv_8) (planner *p) {
Chris@82 181 X(kdft_register) (p, n1bv_8, &desc);
Chris@82 182 }
Chris@82 183
Chris@82 184 #endif