Mercurial > hg > sv-dependency-builds

annotate src/fftw-3.3.8/dft/simd/common/t1buv_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 08:05:56 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 8 -name t1buv_8 -include dft/simd/t1bu.h -sign 1 */
Chris@82 29
Chris@82 30 /*
Chris@82 31 * This function contains 33 FP additions, 24 FP multiplications,
Chris@82 32 * (or, 23 additions, 14 multiplications, 10 fused multiply/add),
Chris@82 33 * 24 stack variables, 1 constants, and 16 memory accesses
Chris@82 34 */
Chris@82 35 #include "dft/simd/t1bu.h"
Chris@82 36
Chris@82 37 static void t1buv_8(R *ri, R *ii, const R *W, stride rs, INT mb, INT me, INT ms)
Chris@82 38 {
Chris@82 39 DVK(KP707106781, +0.707106781186547524400844362104849039284835938);
Chris@82 40 {
Chris@82 41 INT m;
Chris@82 42 R *x;
Chris@82 43 x = ii;
Chris@82 44 for (m = mb, W = W + (mb * ((TWVL / VL) * 14)); m < me; m = m + VL, x = x + (VL * ms), W = W + (TWVL * 14), MAKE_VOLATILE_STRIDE(8, rs)) {
Chris@82 45 V T4, Tq, Tl, Tr, T9, Tt, Te, Tu, T1, T3, T2;
Chris@82 46 T1 = LD(&(x[0]), ms, &(x[0]));
Chris@82 47 T2 = LD(&(x[WS(rs, 4)]), ms, &(x[0]));
Chris@82 48 T3 = BYTW(&(W[TWVL * 6]), T2);
Chris@82 49 T4 = VSUB(T1, T3);
Chris@82 50 Tq = VADD(T1, T3);
Chris@82 51 {
Chris@82 52 V Ti, Tk, Th, Tj;
Chris@82 53 Th = LD(&(x[WS(rs, 2)]), ms, &(x[0]));
Chris@82 54 Ti = BYTW(&(W[TWVL * 2]), Th);
Chris@82 55 Tj = LD(&(x[WS(rs, 6)]), ms, &(x[0]));
Chris@82 56 Tk = BYTW(&(W[TWVL * 10]), Tj);
Chris@82 57 Tl = VSUB(Ti, Tk);
Chris@82 58 Tr = VADD(Ti, Tk);
Chris@82 59 }
Chris@82 60 {
Chris@82 61 V T6, T8, T5, T7;
Chris@82 62 T5 = LD(&(x[WS(rs, 1)]), ms, &(x[WS(rs, 1)]));
Chris@82 63 T6 = BYTW(&(W[0]), T5);
Chris@82 64 T7 = LD(&(x[WS(rs, 5)]), ms, &(x[WS(rs, 1)]));
Chris@82 65 T8 = BYTW(&(W[TWVL * 8]), T7);
Chris@82 66 T9 = VSUB(T6, T8);
Chris@82 67 Tt = VADD(T6, T8);
Chris@82 68 }
Chris@82 69 {
Chris@82 70 V Tb, Td, Ta, Tc;
Chris@82 71 Ta = LD(&(x[WS(rs, 7)]), ms, &(x[WS(rs, 1)]));
Chris@82 72 Tb = BYTW(&(W[TWVL * 12]), Ta);
Chris@82 73 Tc = LD(&(x[WS(rs, 3)]), ms, &(x[WS(rs, 1)]));
Chris@82 74 Td = BYTW(&(W[TWVL * 4]), Tc);
Chris@82 75 Te = VSUB(Tb, Td);
Chris@82 76 Tu = VADD(Tb, Td);
Chris@82 77 }
Chris@82 78 {
Chris@82 79 V Ts, Tv, Tw, Tx;
Chris@82 80 Ts = VSUB(Tq, Tr);
Chris@82 81 Tv = VSUB(Tt, Tu);
Chris@82 82 ST(&(x[WS(rs, 6)]), VFNMSI(Tv, Ts), ms, &(x[0]));
Chris@82 83 ST(&(x[WS(rs, 2)]), VFMAI(Tv, Ts), ms, &(x[0]));
Chris@82 84 Tw = VADD(Tq, Tr);
Chris@82 85 Tx = VADD(Tt, Tu);
Chris@82 86 ST(&(x[WS(rs, 4)]), VSUB(Tw, Tx), ms, &(x[0]));
Chris@82 87 ST(&(x[0]), VADD(Tw, Tx), ms, &(x[0]));
Chris@82 88 {
Chris@82 89 V Tg, To, Tn, Tp, Tf, Tm;
Chris@82 90 Tf = VADD(T9, Te);
Chris@82 91 Tg = VFNMS(LDK(KP707106781), Tf, T4);
Chris@82 92 To = VFMA(LDK(KP707106781), Tf, T4);
Chris@82 93 Tm = VSUB(T9, Te);
Chris@82 94 Tn = VFNMS(LDK(KP707106781), Tm, Tl);
Chris@82 95 Tp = VFMA(LDK(KP707106781), Tm, Tl);
Chris@82 96 ST(&(x[WS(rs, 3)]), VFNMSI(Tn, Tg), ms, &(x[WS(rs, 1)]));
Chris@82 97 ST(&(x[WS(rs, 7)]), VFNMSI(Tp, To), ms, &(x[WS(rs, 1)]));
Chris@82 98 ST(&(x[WS(rs, 5)]), VFMAI(Tn, Tg), ms, &(x[WS(rs, 1)]));
Chris@82 99 ST(&(x[WS(rs, 1)]), VFMAI(Tp, To), ms, &(x[WS(rs, 1)]));
Chris@82 100 }
Chris@82 101 }
Chris@82 102 }
Chris@82 103 }
Chris@82 104 VLEAVE();
Chris@82 105 }
Chris@82 106
Chris@82 107 static const tw_instr twinstr[] = {
Chris@82 108 VTW(0, 1),
Chris@82 109 VTW(0, 2),
Chris@82 110 VTW(0, 3),
Chris@82 111 VTW(0, 4),
Chris@82 112 VTW(0, 5),
Chris@82 113 VTW(0, 6),
Chris@82 114 VTW(0, 7),
Chris@82 115 {TW_NEXT, VL, 0}
Chris@82 116 };
Chris@82 117
Chris@82 118 static const ct_desc desc = { 8, XSIMD_STRING("t1buv_8"), twinstr, &GENUS, {23, 14, 10, 0}, 0, 0, 0 };
Chris@82 119
Chris@82 120 void XSIMD(codelet_t1buv_8) (planner *p) {
Chris@82 121 X(kdft_dit_register) (p, t1buv_8, &desc);
Chris@82 122 }
Chris@82 123 #else
Chris@82 124
Chris@82 125 /* Generated by: ../../../genfft/gen_twiddle_c.native -simd -compact -variables 4 -pipeline-latency 8 -n 8 -name t1buv_8 -include dft/simd/t1bu.h -sign 1 */
Chris@82 126
Chris@82 127 /*
Chris@82 128 * This function contains 33 FP additions, 16 FP multiplications,
Chris@82 129 * (or, 33 additions, 16 multiplications, 0 fused multiply/add),
Chris@82 130 * 24 stack variables, 1 constants, and 16 memory accesses
Chris@82 131 */
Chris@82 132 #include "dft/simd/t1bu.h"
Chris@82 133
Chris@82 134 static void t1buv_8(R *ri, R *ii, const R *W, stride rs, INT mb, INT me, INT ms)
Chris@82 135 {
Chris@82 136 DVK(KP707106781, +0.707106781186547524400844362104849039284835938);
Chris@82 137 {
Chris@82 138 INT m;
Chris@82 139 R *x;
Chris@82 140 x = ii;
Chris@82 141 for (m = mb, W = W + (mb * ((TWVL / VL) * 14)); m < me; m = m + VL, x = x + (VL * ms), W = W + (TWVL * 14), MAKE_VOLATILE_STRIDE(8, rs)) {
Chris@82 142 V Tl, Tq, Tg, Tr, T5, Tt, Ta, Tu, Ti, Tk, Tj;
Chris@82 143 Ti = LD(&(x[0]), ms, &(x[0]));
Chris@82 144 Tj = LD(&(x[WS(rs, 4)]), ms, &(x[0]));
Chris@82 145 Tk = BYTW(&(W[TWVL * 6]), Tj);
Chris@82 146 Tl = VSUB(Ti, Tk);
Chris@82 147 Tq = VADD(Ti, Tk);
Chris@82 148 {
Chris@82 149 V Td, Tf, Tc, Te;
Chris@82 150 Tc = LD(&(x[WS(rs, 2)]), ms, &(x[0]));
Chris@82 151 Td = BYTW(&(W[TWVL * 2]), Tc);
Chris@82 152 Te = LD(&(x[WS(rs, 6)]), ms, &(x[0]));
Chris@82 153 Tf = BYTW(&(W[TWVL * 10]), Te);
Chris@82 154 Tg = VSUB(Td, Tf);
Chris@82 155 Tr = VADD(Td, Tf);
Chris@82 156 }
Chris@82 157 {
Chris@82 158 V T2, T4, T1, T3;
Chris@82 159 T1 = LD(&(x[WS(rs, 1)]), ms, &(x[WS(rs, 1)]));
Chris@82 160 T2 = BYTW(&(W[0]), T1);
Chris@82 161 T3 = LD(&(x[WS(rs, 5)]), ms, &(x[WS(rs, 1)]));
Chris@82 162 T4 = BYTW(&(W[TWVL * 8]), T3);
Chris@82 163 T5 = VSUB(T2, T4);
Chris@82 164 Tt = VADD(T2, T4);
Chris@82 165 }
Chris@82 166 {
Chris@82 167 V T7, T9, T6, T8;
Chris@82 168 T6 = LD(&(x[WS(rs, 7)]), ms, &(x[WS(rs, 1)]));
Chris@82 169 T7 = BYTW(&(W[TWVL * 12]), T6);
Chris@82 170 T8 = LD(&(x[WS(rs, 3)]), ms, &(x[WS(rs, 1)]));
Chris@82 171 T9 = BYTW(&(W[TWVL * 4]), T8);
Chris@82 172 Ta = VSUB(T7, T9);
Chris@82 173 Tu = VADD(T7, T9);
Chris@82 174 }
Chris@82 175 {
Chris@82 176 V Ts, Tv, Tw, Tx;
Chris@82 177 Ts = VSUB(Tq, Tr);
Chris@82 178 Tv = VBYI(VSUB(Tt, Tu));
Chris@82 179 ST(&(x[WS(rs, 6)]), VSUB(Ts, Tv), ms, &(x[0]));
Chris@82 180 ST(&(x[WS(rs, 2)]), VADD(Ts, Tv), ms, &(x[0]));
Chris@82 181 Tw = VADD(Tq, Tr);
Chris@82 182 Tx = VADD(Tt, Tu);
Chris@82 183 ST(&(x[WS(rs, 4)]), VSUB(Tw, Tx), ms, &(x[0]));
Chris@82 184 ST(&(x[0]), VADD(Tw, Tx), ms, &(x[0]));
Chris@82 185 {
Chris@82 186 V Th, To, Tn, Tp, Tb, Tm;
Chris@82 187 Tb = VMUL(LDK(KP707106781), VSUB(T5, Ta));
Chris@82 188 Th = VBYI(VSUB(Tb, Tg));
Chris@82 189 To = VBYI(VADD(Tg, Tb));
Chris@82 190 Tm = VMUL(LDK(KP707106781), VADD(T5, Ta));
Chris@82 191 Tn = VSUB(Tl, Tm);
Chris@82 192 Tp = VADD(Tl, Tm);
Chris@82 193 ST(&(x[WS(rs, 3)]), VADD(Th, Tn), ms, &(x[WS(rs, 1)]));
Chris@82 194 ST(&(x[WS(rs, 7)]), VSUB(Tp, To), ms, &(x[WS(rs, 1)]));
Chris@82 195 ST(&(x[WS(rs, 5)]), VSUB(Tn, Th), ms, &(x[WS(rs, 1)]));
Chris@82 196 ST(&(x[WS(rs, 1)]), VADD(To, Tp), ms, &(x[WS(rs, 1)]));
Chris@82 197 }
Chris@82 198 }
Chris@82 199 }
Chris@82 200 }
Chris@82 201 VLEAVE();
Chris@82 202 }
Chris@82 203
Chris@82 204 static const tw_instr twinstr[] = {
Chris@82 205 VTW(0, 1),
Chris@82 206 VTW(0, 2),
Chris@82 207 VTW(0, 3),
Chris@82 208 VTW(0, 4),
Chris@82 209 VTW(0, 5),
Chris@82 210 VTW(0, 6),
Chris@82 211 VTW(0, 7),
Chris@82 212 {TW_NEXT, VL, 0}
Chris@82 213 };
Chris@82 214
Chris@82 215 static const ct_desc desc = { 8, XSIMD_STRING("t1buv_8"), twinstr, &GENUS, {33, 16, 0, 0}, 0, 0, 0 };
Chris@82 216
Chris@82 217 void XSIMD(codelet_t1buv_8) (planner *p) {
Chris@82 218 X(kdft_dit_register) (p, t1buv_8, &desc);
Chris@82 219 }
Chris@82 220 #endif