Mercurial > hg > sv-dependency-builds

annotate src/fftw-3.3.5/dft/simd/common/n2fv_8.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 16:40:07 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 -n 8 -name n2fv_8 -with-ostride 2 -include n2f.h -store-multiple 2 */
Chris@42 29
Chris@42 30 /*
Chris@42 31 * This function contains 26 FP additions, 10 FP multiplications,
Chris@42 32 * (or, 16 additions, 0 multiplications, 10 fused multiply/add),
Chris@42 33 * 38 stack variables, 1 constants, and 20 memory accesses
Chris@42 34 */
Chris@42 35 #include "n2f.h"
Chris@42 36
Chris@42 37 static void n2fv_8(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(KP707106781, +0.707106781186547524400844362104849039284835938);
Chris@42 40 {
Chris@42 41 INT i;
Chris@42 42 const R *xi;
Chris@42 43 R *xo;
Chris@42 44 xi = ri;
Chris@42 45 xo = ro;
Chris@42 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@42 47 V T1, T2, Tc, Td, T4, T5, T7, T8;
Chris@42 48 T1 = LD(&(xi[0]), ivs, &(xi[0]));
Chris@42 49 T2 = LD(&(xi[WS(is, 4)]), ivs, &(xi[0]));
Chris@42 50 Tc = LD(&(xi[WS(is, 2)]), ivs, &(xi[0]));
Chris@42 51 Td = LD(&(xi[WS(is, 6)]), ivs, &(xi[0]));
Chris@42 52 T4 = LD(&(xi[WS(is, 1)]), ivs, &(xi[WS(is, 1)]));
Chris@42 53 T5 = LD(&(xi[WS(is, 5)]), ivs, &(xi[WS(is, 1)]));
Chris@42 54 T7 = LD(&(xi[WS(is, 7)]), ivs, &(xi[WS(is, 1)]));
Chris@42 55 T8 = LD(&(xi[WS(is, 3)]), ivs, &(xi[WS(is, 1)]));
Chris@42 56 {
Chris@42 57 V T3, Tj, Te, Tk, T6, Tm, T9, Tn, Tp, Tl;
Chris@42 58 T3 = VSUB(T1, T2);
Chris@42 59 Tj = VADD(T1, T2);
Chris@42 60 Te = VSUB(Tc, Td);
Chris@42 61 Tk = VADD(Tc, Td);
Chris@42 62 T6 = VSUB(T4, T5);
Chris@42 63 Tm = VADD(T4, T5);
Chris@42 64 T9 = VSUB(T7, T8);
Chris@42 65 Tn = VADD(T7, T8);
Chris@42 66 Tp = VSUB(Tj, Tk);
Chris@42 67 Tl = VADD(Tj, Tk);
Chris@42 68 {
Chris@42 69 V Tq, To, Ta, Tf;
Chris@42 70 Tq = VSUB(Tn, Tm);
Chris@42 71 To = VADD(Tm, Tn);
Chris@42 72 Ta = VADD(T6, T9);
Chris@42 73 Tf = VSUB(T9, T6);
Chris@42 74 {
Chris@42 75 V Tr, Ts, Tt, Tu, Tg, Ti, Tb, Th;
Chris@42 76 Tr = VADD(Tl, To);
Chris@42 77 STM2(&(xo[0]), Tr, ovs, &(xo[0]));
Chris@42 78 Ts = VSUB(Tl, To);
Chris@42 79 STM2(&(xo[8]), Ts, ovs, &(xo[0]));
Chris@42 80 Tt = VFMAI(Tq, Tp);
Chris@42 81 STM2(&(xo[4]), Tt, ovs, &(xo[0]));
Chris@42 82 Tu = VFNMSI(Tq, Tp);
Chris@42 83 STM2(&(xo[12]), Tu, ovs, &(xo[0]));
Chris@42 84 Tg = VFNMS(LDK(KP707106781), Tf, Te);
Chris@42 85 Ti = VFMA(LDK(KP707106781), Tf, Te);
Chris@42 86 Tb = VFMA(LDK(KP707106781), Ta, T3);
Chris@42 87 Th = VFNMS(LDK(KP707106781), Ta, T3);
Chris@42 88 {
Chris@42 89 V Tv, Tw, Tx, Ty;
Chris@42 90 Tv = VFMAI(Ti, Th);
Chris@42 91 STM2(&(xo[6]), Tv, ovs, &(xo[2]));
Chris@42 92 STN2(&(xo[4]), Tt, Tv, ovs);
Chris@42 93 Tw = VFNMSI(Ti, Th);
Chris@42 94 STM2(&(xo[10]), Tw, ovs, &(xo[2]));
Chris@42 95 STN2(&(xo[8]), Ts, Tw, ovs);
Chris@42 96 Tx = VFMAI(Tg, Tb);
Chris@42 97 STM2(&(xo[14]), Tx, ovs, &(xo[2]));
Chris@42 98 STN2(&(xo[12]), Tu, Tx, ovs);
Chris@42 99 Ty = VFNMSI(Tg, Tb);
Chris@42 100 STM2(&(xo[2]), Ty, ovs, &(xo[2]));
Chris@42 101 STN2(&(xo[0]), Tr, Ty, ovs);
Chris@42 102 }
Chris@42 103 }
Chris@42 104 }
Chris@42 105 }
Chris@42 106 }
Chris@42 107 }
Chris@42 108 VLEAVE();
Chris@42 109 }
Chris@42 110
Chris@42 111 static const kdft_desc desc = { 8, XSIMD_STRING("n2fv_8"), {16, 0, 10, 0}, &GENUS, 0, 2, 0, 0 };
Chris@42 112
Chris@42 113 void XSIMD(codelet_n2fv_8) (planner *p) {
Chris@42 114 X(kdft_register) (p, n2fv_8, &desc);
Chris@42 115 }
Chris@42 116
Chris@42 117 #else /* HAVE_FMA */
Chris@42 118
Chris@42 119 /* Generated by: ../../../genfft/gen_notw_c.native -simd -compact -variables 4 -pipeline-latency 8 -n 8 -name n2fv_8 -with-ostride 2 -include n2f.h -store-multiple 2 */
Chris@42 120
Chris@42 121 /*
Chris@42 122 * This function contains 26 FP additions, 2 FP multiplications,
Chris@42 123 * (or, 26 additions, 2 multiplications, 0 fused multiply/add),
Chris@42 124 * 24 stack variables, 1 constants, and 20 memory accesses
Chris@42 125 */
Chris@42 126 #include "n2f.h"
Chris@42 127
Chris@42 128 static void n2fv_8(const R *ri, const R *ii, R *ro, R *io, stride is, stride os, INT v, INT ivs, INT ovs)
Chris@42 129 {
Chris@42 130 DVK(KP707106781, +0.707106781186547524400844362104849039284835938);
Chris@42 131 {
Chris@42 132 INT i;
Chris@42 133 const R *xi;
Chris@42 134 R *xo;
Chris@42 135 xi = ri;
Chris@42 136 xo = ro;
Chris@42 137 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@42 138 V T3, Tj, Tf, Tk, Ta, Tn, Tc, Tm, Ts, Tu;
Chris@42 139 {
Chris@42 140 V T1, T2, Td, Te;
Chris@42 141 T1 = LD(&(xi[0]), ivs, &(xi[0]));
Chris@42 142 T2 = LD(&(xi[WS(is, 4)]), ivs, &(xi[0]));
Chris@42 143 T3 = VSUB(T1, T2);
Chris@42 144 Tj = VADD(T1, T2);
Chris@42 145 Td = LD(&(xi[WS(is, 2)]), ivs, &(xi[0]));
Chris@42 146 Te = LD(&(xi[WS(is, 6)]), ivs, &(xi[0]));
Chris@42 147 Tf = VSUB(Td, Te);
Chris@42 148 Tk = VADD(Td, Te);
Chris@42 149 {
Chris@42 150 V T4, T5, T6, T7, T8, T9;
Chris@42 151 T4 = LD(&(xi[WS(is, 1)]), ivs, &(xi[WS(is, 1)]));
Chris@42 152 T5 = LD(&(xi[WS(is, 5)]), ivs, &(xi[WS(is, 1)]));
Chris@42 153 T6 = VSUB(T4, T5);
Chris@42 154 T7 = LD(&(xi[WS(is, 7)]), ivs, &(xi[WS(is, 1)]));
Chris@42 155 T8 = LD(&(xi[WS(is, 3)]), ivs, &(xi[WS(is, 1)]));
Chris@42 156 T9 = VSUB(T7, T8);
Chris@42 157 Ta = VMUL(LDK(KP707106781), VADD(T6, T9));
Chris@42 158 Tn = VADD(T7, T8);
Chris@42 159 Tc = VMUL(LDK(KP707106781), VSUB(T9, T6));
Chris@42 160 Tm = VADD(T4, T5);
Chris@42 161 }
Chris@42 162 }
Chris@42 163 {
Chris@42 164 V Tr, Tb, Tg, Tp, Tq, Tt;
Chris@42 165 Tb = VADD(T3, Ta);
Chris@42 166 Tg = VBYI(VSUB(Tc, Tf));
Chris@42 167 Tr = VSUB(Tb, Tg);
Chris@42 168 STM2(&(xo[14]), Tr, ovs, &(xo[2]));
Chris@42 169 Ts = VADD(Tb, Tg);
Chris@42 170 STM2(&(xo[2]), Ts, ovs, &(xo[2]));
Chris@42 171 Tp = VSUB(Tj, Tk);
Chris@42 172 Tq = VBYI(VSUB(Tn, Tm));
Chris@42 173 Tt = VSUB(Tp, Tq);
Chris@42 174 STM2(&(xo[12]), Tt, ovs, &(xo[0]));
Chris@42 175 STN2(&(xo[12]), Tt, Tr, ovs);
Chris@42 176 Tu = VADD(Tp, Tq);
Chris@42 177 STM2(&(xo[4]), Tu, ovs, &(xo[0]));
Chris@42 178 }
Chris@42 179 {
Chris@42 180 V Tv, Th, Ti, Tw;
Chris@42 181 Th = VSUB(T3, Ta);
Chris@42 182 Ti = VBYI(VADD(Tf, Tc));
Chris@42 183 Tv = VSUB(Th, Ti);
Chris@42 184 STM2(&(xo[10]), Tv, ovs, &(xo[2]));
Chris@42 185 Tw = VADD(Th, Ti);
Chris@42 186 STM2(&(xo[6]), Tw, ovs, &(xo[2]));
Chris@42 187 STN2(&(xo[4]), Tu, Tw, ovs);
Chris@42 188 {
Chris@42 189 V Tl, To, Tx, Ty;
Chris@42 190 Tl = VADD(Tj, Tk);
Chris@42 191 To = VADD(Tm, Tn);
Chris@42 192 Tx = VSUB(Tl, To);
Chris@42 193 STM2(&(xo[8]), Tx, ovs, &(xo[0]));
Chris@42 194 STN2(&(xo[8]), Tx, Tv, ovs);
Chris@42 195 Ty = VADD(Tl, To);
Chris@42 196 STM2(&(xo[0]), Ty, ovs, &(xo[0]));
Chris@42 197 STN2(&(xo[0]), Ty, Ts, ovs);
Chris@42 198 }
Chris@42 199 }
Chris@42 200 }
Chris@42 201 }
Chris@42 202 VLEAVE();
Chris@42 203 }
Chris@42 204
Chris@42 205 static const kdft_desc desc = { 8, XSIMD_STRING("n2fv_8"), {26, 2, 0, 0}, &GENUS, 0, 2, 0, 0 };
Chris@42 206
Chris@42 207 void XSIMD(codelet_n2fv_8) (planner *p) {
Chris@42 208 X(kdft_register) (p, n2fv_8, &desc);
Chris@42 209 }
Chris@42 210
Chris@42 211 #endif /* HAVE_FMA */