Mercurial > hg > sv-dependency-builds

annotate src/fftw-3.3.8/dft/simd/common/t1fuv_6.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 08:05:25 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_twiddle_c.native -fma -simd -compact -variables 4 -pipeline-latency 8 -n 6 -name t1fuv_6 -include dft/simd/t1fu.h */
Chris@82 29
Chris@82 30 /*
Chris@82 31 * This function contains 23 FP additions, 18 FP multiplications,
Chris@82 32 * (or, 17 additions, 12 multiplications, 6 fused multiply/add),
Chris@82 33 * 19 stack variables, 2 constants, and 12 memory accesses
Chris@82 34 */
Chris@82 35 #include "dft/simd/t1fu.h"
Chris@82 36
Chris@82 37 static void t1fuv_6(R *ri, R *ii, const R *W, stride rs, INT mb, INT me, INT ms)
Chris@82 38 {
Chris@82 39 DVK(KP500000000, +0.500000000000000000000000000000000000000000000);
Chris@82 40 DVK(KP866025403, +0.866025403784438646763723170752936183471402627);
Chris@82 41 {
Chris@82 42 INT m;
Chris@82 43 R *x;
Chris@82 44 x = ri;
Chris@82 45 for (m = mb, W = W + (mb * ((TWVL / VL) * 10)); m < me; m = m + VL, x = x + (VL * ms), W = W + (TWVL * 10), MAKE_VOLATILE_STRIDE(6, rs)) {
Chris@82 46 V T4, Ti, Te, Tk, T9, Tj, T1, T3, T2;
Chris@82 47 T1 = LD(&(x[0]), ms, &(x[0]));
Chris@82 48 T2 = LD(&(x[WS(rs, 3)]), ms, &(x[WS(rs, 1)]));
Chris@82 49 T3 = BYTWJ(&(W[TWVL * 4]), T2);
Chris@82 50 T4 = VSUB(T1, T3);
Chris@82 51 Ti = VADD(T1, T3);
Chris@82 52 {
Chris@82 53 V Tb, Td, Ta, Tc;
Chris@82 54 Ta = LD(&(x[WS(rs, 4)]), ms, &(x[0]));
Chris@82 55 Tb = BYTWJ(&(W[TWVL * 6]), Ta);
Chris@82 56 Tc = LD(&(x[WS(rs, 1)]), ms, &(x[WS(rs, 1)]));
Chris@82 57 Td = BYTWJ(&(W[0]), Tc);
Chris@82 58 Te = VSUB(Tb, Td);
Chris@82 59 Tk = VADD(Tb, Td);
Chris@82 60 }
Chris@82 61 {
Chris@82 62 V T6, T8, T5, T7;
Chris@82 63 T5 = LD(&(x[WS(rs, 2)]), ms, &(x[0]));
Chris@82 64 T6 = BYTWJ(&(W[TWVL * 2]), T5);
Chris@82 65 T7 = LD(&(x[WS(rs, 5)]), ms, &(x[WS(rs, 1)]));
Chris@82 66 T8 = BYTWJ(&(W[TWVL * 8]), T7);
Chris@82 67 T9 = VSUB(T6, T8);
Chris@82 68 Tj = VADD(T6, T8);
Chris@82 69 }
Chris@82 70 {
Chris@82 71 V Th, Tf, Tg, Tn, Tl, Tm;
Chris@82 72 Th = VMUL(LDK(KP866025403), VSUB(Te, T9));
Chris@82 73 Tf = VADD(T9, Te);
Chris@82 74 Tg = VFNMS(LDK(KP500000000), Tf, T4);
Chris@82 75 ST(&(x[WS(rs, 3)]), VADD(T4, Tf), ms, &(x[WS(rs, 1)]));
Chris@82 76 ST(&(x[WS(rs, 1)]), VFMAI(Th, Tg), ms, &(x[WS(rs, 1)]));
Chris@82 77 ST(&(x[WS(rs, 5)]), VFNMSI(Th, Tg), ms, &(x[WS(rs, 1)]));
Chris@82 78 Tn = VMUL(LDK(KP866025403), VSUB(Tk, Tj));
Chris@82 79 Tl = VADD(Tj, Tk);
Chris@82 80 Tm = VFNMS(LDK(KP500000000), Tl, Ti);
Chris@82 81 ST(&(x[0]), VADD(Ti, Tl), ms, &(x[0]));
Chris@82 82 ST(&(x[WS(rs, 4)]), VFMAI(Tn, Tm), ms, &(x[0]));
Chris@82 83 ST(&(x[WS(rs, 2)]), VFNMSI(Tn, Tm), ms, &(x[0]));
Chris@82 84 }
Chris@82 85 }
Chris@82 86 }
Chris@82 87 VLEAVE();
Chris@82 88 }
Chris@82 89
Chris@82 90 static const tw_instr twinstr[] = {
Chris@82 91 VTW(0, 1),
Chris@82 92 VTW(0, 2),
Chris@82 93 VTW(0, 3),
Chris@82 94 VTW(0, 4),
Chris@82 95 VTW(0, 5),
Chris@82 96 {TW_NEXT, VL, 0}
Chris@82 97 };
Chris@82 98
Chris@82 99 static const ct_desc desc = { 6, XSIMD_STRING("t1fuv_6"), twinstr, &GENUS, {17, 12, 6, 0}, 0, 0, 0 };
Chris@82 100
Chris@82 101 void XSIMD(codelet_t1fuv_6) (planner *p) {
Chris@82 102 X(kdft_dit_register) (p, t1fuv_6, &desc);
Chris@82 103 }
Chris@82 104 #else
Chris@82 105
Chris@82 106 /* Generated by: ../../../genfft/gen_twiddle_c.native -simd -compact -variables 4 -pipeline-latency 8 -n 6 -name t1fuv_6 -include dft/simd/t1fu.h */
Chris@82 107
Chris@82 108 /*
Chris@82 109 * This function contains 23 FP additions, 14 FP multiplications,
Chris@82 110 * (or, 21 additions, 12 multiplications, 2 fused multiply/add),
Chris@82 111 * 19 stack variables, 2 constants, and 12 memory accesses
Chris@82 112 */
Chris@82 113 #include "dft/simd/t1fu.h"
Chris@82 114
Chris@82 115 static void t1fuv_6(R *ri, R *ii, const R *W, stride rs, INT mb, INT me, INT ms)
Chris@82 116 {
Chris@82 117 DVK(KP500000000, +0.500000000000000000000000000000000000000000000);
Chris@82 118 DVK(KP866025403, +0.866025403784438646763723170752936183471402627);
Chris@82 119 {
Chris@82 120 INT m;
Chris@82 121 R *x;
Chris@82 122 x = ri;
Chris@82 123 for (m = mb, W = W + (mb * ((TWVL / VL) * 10)); m < me; m = m + VL, x = x + (VL * ms), W = W + (TWVL * 10), MAKE_VOLATILE_STRIDE(6, rs)) {
Chris@82 124 V T4, Ti, Te, Tk, T9, Tj, T1, T3, T2;
Chris@82 125 T1 = LD(&(x[0]), ms, &(x[0]));
Chris@82 126 T2 = LD(&(x[WS(rs, 3)]), ms, &(x[WS(rs, 1)]));
Chris@82 127 T3 = BYTWJ(&(W[TWVL * 4]), T2);
Chris@82 128 T4 = VSUB(T1, T3);
Chris@82 129 Ti = VADD(T1, T3);
Chris@82 130 {
Chris@82 131 V Tb, Td, Ta, Tc;
Chris@82 132 Ta = LD(&(x[WS(rs, 4)]), ms, &(x[0]));
Chris@82 133 Tb = BYTWJ(&(W[TWVL * 6]), Ta);
Chris@82 134 Tc = LD(&(x[WS(rs, 1)]), ms, &(x[WS(rs, 1)]));
Chris@82 135 Td = BYTWJ(&(W[0]), Tc);
Chris@82 136 Te = VSUB(Tb, Td);
Chris@82 137 Tk = VADD(Tb, Td);
Chris@82 138 }
Chris@82 139 {
Chris@82 140 V T6, T8, T5, T7;
Chris@82 141 T5 = LD(&(x[WS(rs, 2)]), ms, &(x[0]));
Chris@82 142 T6 = BYTWJ(&(W[TWVL * 2]), T5);
Chris@82 143 T7 = LD(&(x[WS(rs, 5)]), ms, &(x[WS(rs, 1)]));
Chris@82 144 T8 = BYTWJ(&(W[TWVL * 8]), T7);
Chris@82 145 T9 = VSUB(T6, T8);
Chris@82 146 Tj = VADD(T6, T8);
Chris@82 147 }
Chris@82 148 {
Chris@82 149 V Th, Tf, Tg, Tn, Tl, Tm;
Chris@82 150 Th = VBYI(VMUL(LDK(KP866025403), VSUB(Te, T9)));
Chris@82 151 Tf = VADD(T9, Te);
Chris@82 152 Tg = VFNMS(LDK(KP500000000), Tf, T4);
Chris@82 153 ST(&(x[WS(rs, 3)]), VADD(T4, Tf), ms, &(x[WS(rs, 1)]));
Chris@82 154 ST(&(x[WS(rs, 1)]), VADD(Tg, Th), ms, &(x[WS(rs, 1)]));
Chris@82 155 ST(&(x[WS(rs, 5)]), VSUB(Tg, Th), ms, &(x[WS(rs, 1)]));
Chris@82 156 Tn = VBYI(VMUL(LDK(KP866025403), VSUB(Tk, Tj)));
Chris@82 157 Tl = VADD(Tj, Tk);
Chris@82 158 Tm = VFNMS(LDK(KP500000000), Tl, Ti);
Chris@82 159 ST(&(x[0]), VADD(Ti, Tl), ms, &(x[0]));
Chris@82 160 ST(&(x[WS(rs, 4)]), VADD(Tm, Tn), ms, &(x[0]));
Chris@82 161 ST(&(x[WS(rs, 2)]), VSUB(Tm, Tn), ms, &(x[0]));
Chris@82 162 }
Chris@82 163 }
Chris@82 164 }
Chris@82 165 VLEAVE();
Chris@82 166 }
Chris@82 167
Chris@82 168 static const tw_instr twinstr[] = {
Chris@82 169 VTW(0, 1),
Chris@82 170 VTW(0, 2),
Chris@82 171 VTW(0, 3),
Chris@82 172 VTW(0, 4),
Chris@82 173 VTW(0, 5),
Chris@82 174 {TW_NEXT, VL, 0}
Chris@82 175 };
Chris@82 176
Chris@82 177 static const ct_desc desc = { 6, XSIMD_STRING("t1fuv_6"), twinstr, &GENUS, {21, 12, 2, 0}, 0, 0, 0 };
Chris@82 178
Chris@82 179 void XSIMD(codelet_t1fuv_6) (planner *p) {
Chris@82 180 X(kdft_dit_register) (p, t1fuv_6, &desc);
Chris@82 181 }
Chris@82 182 #endif