Mercurial > hg > sv-dependency-builds

annotate src/fftw-3.3.8/dft/simd/common/t2sv_4.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 08:06:11 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.native -fma -simd -compact -variables 4 -pipeline-latency 8 -twiddle-log3 -precompute-twiddles -n 4 -name t2sv_4 -include dft/simd/ts.h */
Chris@82 29
Chris@82 30 /*
Chris@82 31 * This function contains 24 FP additions, 16 FP multiplications,
Chris@82 32 * (or, 16 additions, 8 multiplications, 8 fused multiply/add),
Chris@82 33 * 21 stack variables, 0 constants, and 16 memory accesses
Chris@82 34 */
Chris@82 35 #include "dft/simd/ts.h"
Chris@82 36
Chris@82 37 static void t2sv_4(R *ri, R *ii, const R *W, stride rs, INT mb, INT me, INT ms)
Chris@82 38 {
Chris@82 39 {
Chris@82 40 INT m;
Chris@82 41 for (m = mb, W = W + (mb * 4); m < me; m = m + (2 * VL), ri = ri + ((2 * VL) * ms), ii = ii + ((2 * VL) * ms), W = W + ((2 * VL) * 4), MAKE_VOLATILE_STRIDE(8, rs)) {
Chris@82 42 V T2, T6, T3, T5, T7, Tb, T4, Ta;
Chris@82 43 T2 = LDW(&(W[0]));
Chris@82 44 T6 = LDW(&(W[TWVL * 3]));
Chris@82 45 T3 = LDW(&(W[TWVL * 2]));
Chris@82 46 T4 = VMUL(T2, T3);
Chris@82 47 Ta = VMUL(T2, T6);
Chris@82 48 T5 = LDW(&(W[TWVL * 1]));
Chris@82 49 T7 = VFMA(T5, T6, T4);
Chris@82 50 Tb = VFNMS(T5, T3, Ta);
Chris@82 51 {
Chris@82 52 V T1, Tx, Td, Tw, Ti, Tq, Tm, Ts;
Chris@82 53 T1 = LD(&(ri[0]), ms, &(ri[0]));
Chris@82 54 Tx = LD(&(ii[0]), ms, &(ii[0]));
Chris@82 55 {
Chris@82 56 V T8, T9, Tc, Tv;
Chris@82 57 T8 = LD(&(ri[WS(rs, 2)]), ms, &(ri[0]));
Chris@82 58 T9 = VMUL(T7, T8);
Chris@82 59 Tc = LD(&(ii[WS(rs, 2)]), ms, &(ii[0]));
Chris@82 60 Tv = VMUL(T7, Tc);
Chris@82 61 Td = VFMA(Tb, Tc, T9);
Chris@82 62 Tw = VFNMS(Tb, T8, Tv);
Chris@82 63 }
Chris@82 64 {
Chris@82 65 V Tf, Tg, Th, Tp;
Chris@82 66 Tf = LD(&(ri[WS(rs, 1)]), ms, &(ri[WS(rs, 1)]));
Chris@82 67 Tg = VMUL(T2, Tf);
Chris@82 68 Th = LD(&(ii[WS(rs, 1)]), ms, &(ii[WS(rs, 1)]));
Chris@82 69 Tp = VMUL(T2, Th);
Chris@82 70 Ti = VFMA(T5, Th, Tg);
Chris@82 71 Tq = VFNMS(T5, Tf, Tp);
Chris@82 72 }
Chris@82 73 {
Chris@82 74 V Tj, Tk, Tl, Tr;
Chris@82 75 Tj = LD(&(ri[WS(rs, 3)]), ms, &(ri[WS(rs, 1)]));
Chris@82 76 Tk = VMUL(T3, Tj);
Chris@82 77 Tl = LD(&(ii[WS(rs, 3)]), ms, &(ii[WS(rs, 1)]));
Chris@82 78 Tr = VMUL(T3, Tl);
Chris@82 79 Tm = VFMA(T6, Tl, Tk);
Chris@82 80 Ts = VFNMS(T6, Tj, Tr);
Chris@82 81 }
Chris@82 82 {
Chris@82 83 V Te, Tn, Tu, Ty;
Chris@82 84 Te = VADD(T1, Td);
Chris@82 85 Tn = VADD(Ti, Tm);
Chris@82 86 ST(&(ri[WS(rs, 2)]), VSUB(Te, Tn), ms, &(ri[0]));
Chris@82 87 ST(&(ri[0]), VADD(Te, Tn), ms, &(ri[0]));
Chris@82 88 Tu = VADD(Tq, Ts);
Chris@82 89 Ty = VADD(Tw, Tx);
Chris@82 90 ST(&(ii[0]), VADD(Tu, Ty), ms, &(ii[0]));
Chris@82 91 ST(&(ii[WS(rs, 2)]), VSUB(Ty, Tu), ms, &(ii[0]));
Chris@82 92 }
Chris@82 93 {
Chris@82 94 V To, Tt, Tz, TA;
Chris@82 95 To = VSUB(T1, Td);
Chris@82 96 Tt = VSUB(Tq, Ts);
Chris@82 97 ST(&(ri[WS(rs, 3)]), VSUB(To, Tt), ms, &(ri[WS(rs, 1)]));
Chris@82 98 ST(&(ri[WS(rs, 1)]), VADD(To, Tt), ms, &(ri[WS(rs, 1)]));
Chris@82 99 Tz = VSUB(Tx, Tw);
Chris@82 100 TA = VSUB(Ti, Tm);
Chris@82 101 ST(&(ii[WS(rs, 1)]), VSUB(Tz, TA), ms, &(ii[WS(rs, 1)]));
Chris@82 102 ST(&(ii[WS(rs, 3)]), VADD(TA, Tz), ms, &(ii[WS(rs, 1)]));
Chris@82 103 }
Chris@82 104 }
Chris@82 105 }
Chris@82 106 }
Chris@82 107 VLEAVE();
Chris@82 108 }
Chris@82 109
Chris@82 110 static const tw_instr twinstr[] = {
Chris@82 111 VTW(0, 1),
Chris@82 112 VTW(0, 3),
Chris@82 113 {TW_NEXT, (2 * VL), 0}
Chris@82 114 };
Chris@82 115
Chris@82 116 static const ct_desc desc = { 4, XSIMD_STRING("t2sv_4"), twinstr, &GENUS, {16, 8, 8, 0}, 0, 0, 0 };
Chris@82 117
Chris@82 118 void XSIMD(codelet_t2sv_4) (planner *p) {
Chris@82 119 X(kdft_dit_register) (p, t2sv_4, &desc);
Chris@82 120 }
Chris@82 121 #else
Chris@82 122
Chris@82 123 /* Generated by: ../../../genfft/gen_twiddle.native -simd -compact -variables 4 -pipeline-latency 8 -twiddle-log3 -precompute-twiddles -n 4 -name t2sv_4 -include dft/simd/ts.h */
Chris@82 124
Chris@82 125 /*
Chris@82 126 * This function contains 24 FP additions, 16 FP multiplications,
Chris@82 127 * (or, 16 additions, 8 multiplications, 8 fused multiply/add),
Chris@82 128 * 21 stack variables, 0 constants, and 16 memory accesses
Chris@82 129 */
Chris@82 130 #include "dft/simd/ts.h"
Chris@82 131
Chris@82 132 static void t2sv_4(R *ri, R *ii, const R *W, stride rs, INT mb, INT me, INT ms)
Chris@82 133 {
Chris@82 134 {
Chris@82 135 INT m;
Chris@82 136 for (m = mb, W = W + (mb * 4); m < me; m = m + (2 * VL), ri = ri + ((2 * VL) * ms), ii = ii + ((2 * VL) * ms), W = W + ((2 * VL) * 4), MAKE_VOLATILE_STRIDE(8, rs)) {
Chris@82 137 V T2, T4, T3, T5, T6, T8;
Chris@82 138 T2 = LDW(&(W[0]));
Chris@82 139 T4 = LDW(&(W[TWVL * 1]));
Chris@82 140 T3 = LDW(&(W[TWVL * 2]));
Chris@82 141 T5 = LDW(&(W[TWVL * 3]));
Chris@82 142 T6 = VFMA(T2, T3, VMUL(T4, T5));
Chris@82 143 T8 = VFNMS(T4, T3, VMUL(T2, T5));
Chris@82 144 {
Chris@82 145 V T1, Tp, Ta, To, Te, Tk, Th, Tl, T7, T9;
Chris@82 146 T1 = LD(&(ri[0]), ms, &(ri[0]));
Chris@82 147 Tp = LD(&(ii[0]), ms, &(ii[0]));
Chris@82 148 T7 = LD(&(ri[WS(rs, 2)]), ms, &(ri[0]));
Chris@82 149 T9 = LD(&(ii[WS(rs, 2)]), ms, &(ii[0]));
Chris@82 150 Ta = VFMA(T6, T7, VMUL(T8, T9));
Chris@82 151 To = VFNMS(T8, T7, VMUL(T6, T9));
Chris@82 152 {
Chris@82 153 V Tc, Td, Tf, Tg;
Chris@82 154 Tc = LD(&(ri[WS(rs, 1)]), ms, &(ri[WS(rs, 1)]));
Chris@82 155 Td = LD(&(ii[WS(rs, 1)]), ms, &(ii[WS(rs, 1)]));
Chris@82 156 Te = VFMA(T2, Tc, VMUL(T4, Td));
Chris@82 157 Tk = VFNMS(T4, Tc, VMUL(T2, Td));
Chris@82 158 Tf = LD(&(ri[WS(rs, 3)]), ms, &(ri[WS(rs, 1)]));
Chris@82 159 Tg = LD(&(ii[WS(rs, 3)]), ms, &(ii[WS(rs, 1)]));
Chris@82 160 Th = VFMA(T3, Tf, VMUL(T5, Tg));
Chris@82 161 Tl = VFNMS(T5, Tf, VMUL(T3, Tg));
Chris@82 162 }
Chris@82 163 {
Chris@82 164 V Tb, Ti, Tn, Tq;
Chris@82 165 Tb = VADD(T1, Ta);
Chris@82 166 Ti = VADD(Te, Th);
Chris@82 167 ST(&(ri[WS(rs, 2)]), VSUB(Tb, Ti), ms, &(ri[0]));
Chris@82 168 ST(&(ri[0]), VADD(Tb, Ti), ms, &(ri[0]));
Chris@82 169 Tn = VADD(Tk, Tl);
Chris@82 170 Tq = VADD(To, Tp);
Chris@82 171 ST(&(ii[0]), VADD(Tn, Tq), ms, &(ii[0]));
Chris@82 172 ST(&(ii[WS(rs, 2)]), VSUB(Tq, Tn), ms, &(ii[0]));
Chris@82 173 }
Chris@82 174 {
Chris@82 175 V Tj, Tm, Tr, Ts;
Chris@82 176 Tj = VSUB(T1, Ta);
Chris@82 177 Tm = VSUB(Tk, Tl);
Chris@82 178 ST(&(ri[WS(rs, 3)]), VSUB(Tj, Tm), ms, &(ri[WS(rs, 1)]));
Chris@82 179 ST(&(ri[WS(rs, 1)]), VADD(Tj, Tm), ms, &(ri[WS(rs, 1)]));
Chris@82 180 Tr = VSUB(Tp, To);
Chris@82 181 Ts = VSUB(Te, Th);
Chris@82 182 ST(&(ii[WS(rs, 1)]), VSUB(Tr, Ts), ms, &(ii[WS(rs, 1)]));
Chris@82 183 ST(&(ii[WS(rs, 3)]), VADD(Ts, Tr), ms, &(ii[WS(rs, 1)]));
Chris@82 184 }
Chris@82 185 }
Chris@82 186 }
Chris@82 187 }
Chris@82 188 VLEAVE();
Chris@82 189 }
Chris@82 190
Chris@82 191 static const tw_instr twinstr[] = {
Chris@82 192 VTW(0, 1),
Chris@82 193 VTW(0, 3),
Chris@82 194 {TW_NEXT, (2 * VL), 0}
Chris@82 195 };
Chris@82 196
Chris@82 197 static const ct_desc desc = { 4, XSIMD_STRING("t2sv_4"), twinstr, &GENUS, {16, 8, 8, 0}, 0, 0, 0 };
Chris@82 198
Chris@82 199 void XSIMD(codelet_t2sv_4) (planner *p) {
Chris@82 200 X(kdft_dit_register) (p, t2sv_4, &desc);
Chris@82 201 }
Chris@82 202 #endif