Mercurial > hg > sv-dependency-builds

annotate src/fftw-3.3.5/dft/simd/common/n2bv_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 16:40:33 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_notw_c.native -fma -reorder-insns -schedule-for-pipeline -simd -compact -variables 4 -pipeline-latency 8 -sign 1 -n 6 -name n2bv_6 -with-ostride 2 -include n2b.h -store-multiple 2 */
Chris@42 29
Chris@42 30 /*
Chris@42 31 * This function contains 18 FP additions, 8 FP multiplications,
Chris@42 32 * (or, 12 additions, 2 multiplications, 6 fused multiply/add),
Chris@42 33 * 29 stack variables, 2 constants, and 15 memory accesses
Chris@42 34 */
Chris@42 35 #include "n2b.h"
Chris@42 36
Chris@42 37 static void n2bv_6(const R *ri, const R *ii, R *ro, R *io, stride is, stride os, INT v, INT ivs, INT ovs)
Chris@42 38 {
Chris@42 39 DVK(KP500000000, +0.500000000000000000000000000000000000000000000);
Chris@42 40 DVK(KP866025403, +0.866025403784438646763723170752936183471402627);
Chris@42 41 {
Chris@42 42 INT i;
Chris@42 43 const R *xi;
Chris@42 44 R *xo;
Chris@42 45 xi = ii;
Chris@42 46 xo = io;
Chris@42 47 for (i = v; i > 0; i = i - VL, xi = xi + (VL * ivs), xo = xo + (VL * ovs), MAKE_VOLATILE_STRIDE(12, is), MAKE_VOLATILE_STRIDE(12, os)) {
Chris@42 48 V T1, T2, T4, T5, T7, T8;
Chris@42 49 T1 = LD(&(xi[0]), ivs, &(xi[0]));
Chris@42 50 T2 = LD(&(xi[WS(is, 3)]), ivs, &(xi[WS(is, 1)]));
Chris@42 51 T4 = LD(&(xi[WS(is, 2)]), ivs, &(xi[0]));
Chris@42 52 T5 = LD(&(xi[WS(is, 5)]), ivs, &(xi[WS(is, 1)]));
Chris@42 53 T7 = LD(&(xi[WS(is, 4)]), ivs, &(xi[0]));
Chris@42 54 T8 = LD(&(xi[WS(is, 1)]), ivs, &(xi[WS(is, 1)]));
Chris@42 55 {
Chris@42 56 V T3, Td, T6, Te, T9, Tf;
Chris@42 57 T3 = VSUB(T1, T2);
Chris@42 58 Td = VADD(T1, T2);
Chris@42 59 T6 = VSUB(T4, T5);
Chris@42 60 Te = VADD(T4, T5);
Chris@42 61 T9 = VSUB(T7, T8);
Chris@42 62 Tf = VADD(T7, T8);
Chris@42 63 {
Chris@42 64 V Tg, Ti, Ta, Tc;
Chris@42 65 Tg = VADD(Te, Tf);
Chris@42 66 Ti = VMUL(LDK(KP866025403), VSUB(Te, Tf));
Chris@42 67 Ta = VADD(T6, T9);
Chris@42 68 Tc = VMUL(LDK(KP866025403), VSUB(T6, T9));
Chris@42 69 {
Chris@42 70 V Th, Tj, Tb, Tk;
Chris@42 71 Th = VFNMS(LDK(KP500000000), Tg, Td);
Chris@42 72 Tj = VADD(Td, Tg);
Chris@42 73 STM2(&(xo[0]), Tj, ovs, &(xo[0]));
Chris@42 74 Tb = VFNMS(LDK(KP500000000), Ta, T3);
Chris@42 75 Tk = VADD(T3, Ta);
Chris@42 76 STM2(&(xo[6]), Tk, ovs, &(xo[2]));
Chris@42 77 {
Chris@42 78 V Tl, Tm, Tn, To;
Chris@42 79 Tl = VFMAI(Ti, Th);
Chris@42 80 STM2(&(xo[8]), Tl, ovs, &(xo[0]));
Chris@42 81 Tm = VFNMSI(Ti, Th);
Chris@42 82 STM2(&(xo[4]), Tm, ovs, &(xo[0]));
Chris@42 83 STN2(&(xo[4]), Tm, Tk, ovs);
Chris@42 84 Tn = VFNMSI(Tc, Tb);
Chris@42 85 STM2(&(xo[10]), Tn, ovs, &(xo[2]));
Chris@42 86 STN2(&(xo[8]), Tl, Tn, ovs);
Chris@42 87 To = VFMAI(Tc, Tb);
Chris@42 88 STM2(&(xo[2]), To, ovs, &(xo[2]));
Chris@42 89 STN2(&(xo[0]), Tj, To, ovs);
Chris@42 90 }
Chris@42 91 }
Chris@42 92 }
Chris@42 93 }
Chris@42 94 }
Chris@42 95 }
Chris@42 96 VLEAVE();
Chris@42 97 }
Chris@42 98
Chris@42 99 static const kdft_desc desc = { 6, XSIMD_STRING("n2bv_6"), {12, 2, 6, 0}, &GENUS, 0, 2, 0, 0 };
Chris@42 100
Chris@42 101 void XSIMD(codelet_n2bv_6) (planner *p) {
Chris@42 102 X(kdft_register) (p, n2bv_6, &desc);
Chris@42 103 }
Chris@42 104
Chris@42 105 #else /* HAVE_FMA */
Chris@42 106
Chris@42 107 /* Generated by: ../../../genfft/gen_notw_c.native -simd -compact -variables 4 -pipeline-latency 8 -sign 1 -n 6 -name n2bv_6 -with-ostride 2 -include n2b.h -store-multiple 2 */
Chris@42 108
Chris@42 109 /*
Chris@42 110 * This function contains 18 FP additions, 4 FP multiplications,
Chris@42 111 * (or, 16 additions, 2 multiplications, 2 fused multiply/add),
Chris@42 112 * 25 stack variables, 2 constants, and 15 memory accesses
Chris@42 113 */
Chris@42 114 #include "n2b.h"
Chris@42 115
Chris@42 116 static void n2bv_6(const R *ri, const R *ii, R *ro, R *io, stride is, stride os, INT v, INT ivs, INT ovs)
Chris@42 117 {
Chris@42 118 DVK(KP500000000, +0.500000000000000000000000000000000000000000000);
Chris@42 119 DVK(KP866025403, +0.866025403784438646763723170752936183471402627);
Chris@42 120 {
Chris@42 121 INT i;
Chris@42 122 const R *xi;
Chris@42 123 R *xo;
Chris@42 124 xi = ii;
Chris@42 125 xo = io;
Chris@42 126 for (i = v; i > 0; i = i - VL, xi = xi + (VL * ivs), xo = xo + (VL * ovs), MAKE_VOLATILE_STRIDE(12, is), MAKE_VOLATILE_STRIDE(12, os)) {
Chris@42 127 V Ta, Td, T3, Te, T6, Tf, Tb, Tg, T8, T9, Tj, Tk;
Chris@42 128 T8 = LD(&(xi[0]), ivs, &(xi[0]));
Chris@42 129 T9 = LD(&(xi[WS(is, 3)]), ivs, &(xi[WS(is, 1)]));
Chris@42 130 Ta = VSUB(T8, T9);
Chris@42 131 Td = VADD(T8, T9);
Chris@42 132 {
Chris@42 133 V T1, T2, T4, T5;
Chris@42 134 T1 = LD(&(xi[WS(is, 2)]), ivs, &(xi[0]));
Chris@42 135 T2 = LD(&(xi[WS(is, 5)]), ivs, &(xi[WS(is, 1)]));
Chris@42 136 T3 = VSUB(T1, T2);
Chris@42 137 Te = VADD(T1, T2);
Chris@42 138 T4 = LD(&(xi[WS(is, 4)]), ivs, &(xi[0]));
Chris@42 139 T5 = LD(&(xi[WS(is, 1)]), ivs, &(xi[WS(is, 1)]));
Chris@42 140 T6 = VSUB(T4, T5);
Chris@42 141 Tf = VADD(T4, T5);
Chris@42 142 }
Chris@42 143 Tb = VADD(T3, T6);
Chris@42 144 Tg = VADD(Te, Tf);
Chris@42 145 Tj = VADD(Ta, Tb);
Chris@42 146 STM2(&(xo[6]), Tj, ovs, &(xo[2]));
Chris@42 147 Tk = VADD(Td, Tg);
Chris@42 148 STM2(&(xo[0]), Tk, ovs, &(xo[0]));
Chris@42 149 {
Chris@42 150 V Tm, T7, Tc, Tl;
Chris@42 151 T7 = VBYI(VMUL(LDK(KP866025403), VSUB(T3, T6)));
Chris@42 152 Tc = VFNMS(LDK(KP500000000), Tb, Ta);
Chris@42 153 Tl = VADD(T7, Tc);
Chris@42 154 STM2(&(xo[2]), Tl, ovs, &(xo[2]));
Chris@42 155 STN2(&(xo[0]), Tk, Tl, ovs);
Chris@42 156 Tm = VSUB(Tc, T7);
Chris@42 157 STM2(&(xo[10]), Tm, ovs, &(xo[2]));
Chris@42 158 {
Chris@42 159 V Th, Ti, Tn, To;
Chris@42 160 Th = VFNMS(LDK(KP500000000), Tg, Td);
Chris@42 161 Ti = VBYI(VMUL(LDK(KP866025403), VSUB(Te, Tf)));
Chris@42 162 Tn = VSUB(Th, Ti);
Chris@42 163 STM2(&(xo[4]), Tn, ovs, &(xo[0]));
Chris@42 164 STN2(&(xo[4]), Tn, Tj, ovs);
Chris@42 165 To = VADD(Ti, Th);
Chris@42 166 STM2(&(xo[8]), To, ovs, &(xo[0]));
Chris@42 167 STN2(&(xo[8]), To, Tm, ovs);
Chris@42 168 }
Chris@42 169 }
Chris@42 170 }
Chris@42 171 }
Chris@42 172 VLEAVE();
Chris@42 173 }
Chris@42 174
Chris@42 175 static const kdft_desc desc = { 6, XSIMD_STRING("n2bv_6"), {16, 2, 2, 0}, &GENUS, 0, 2, 0, 0 };
Chris@42 176
Chris@42 177 void XSIMD(codelet_n2bv_6) (planner *p) {
Chris@42 178 X(kdft_register) (p, n2bv_6, &desc);
Chris@42 179 }
Chris@42 180
Chris@42 181 #endif /* HAVE_FMA */