Mercurial > hg > sv-dependency-builds

annotate src/fftw-3.3.8/dft/simd/common/q1bv_4.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 08:06:14 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_twidsq_c.native -fma -simd -compact -variables 4 -pipeline-latency 8 -n 4 -dif -name q1bv_4 -include dft/simd/q1b.h -sign 1 */
Chris@82 29
Chris@82 30 /*
Chris@82 31 * This function contains 44 FP additions, 32 FP multiplications,
Chris@82 32 * (or, 36 additions, 24 multiplications, 8 fused multiply/add),
Chris@82 33 * 22 stack variables, 0 constants, and 32 memory accesses
Chris@82 34 */
Chris@82 35 #include "dft/simd/q1b.h"
Chris@82 36
Chris@82 37 static void q1bv_4(R *ri, R *ii, const R *W, stride rs, stride vs, INT mb, INT me, INT ms)
Chris@82 38 {
Chris@82 39 {
Chris@82 40 INT m;
Chris@82 41 R *x;
Chris@82 42 x = ii;
Chris@82 43 for (m = mb, W = W + (mb * ((TWVL / VL) * 6)); m < me; m = m + VL, x = x + (VL * ms), W = W + (TWVL * 6), MAKE_VOLATILE_STRIDE(8, rs), MAKE_VOLATILE_STRIDE(8, vs)) {
Chris@82 44 V T3, T9, TA, TG, TD, TH, T6, Ta, Te, Tk, Tp, Tv, Ts, Tw, Th;
Chris@82 45 V Tl;
Chris@82 46 {
Chris@82 47 V T1, T2, Ty, Tz;
Chris@82 48 T1 = LD(&(x[0]), ms, &(x[0]));
Chris@82 49 T2 = LD(&(x[WS(rs, 2)]), ms, &(x[0]));
Chris@82 50 T3 = VSUB(T1, T2);
Chris@82 51 T9 = VADD(T1, T2);
Chris@82 52 Ty = LD(&(x[WS(vs, 3)]), ms, &(x[WS(vs, 3)]));
Chris@82 53 Tz = LD(&(x[WS(vs, 3) + WS(rs, 2)]), ms, &(x[WS(vs, 3)]));
Chris@82 54 TA = VSUB(Ty, Tz);
Chris@82 55 TG = VADD(Ty, Tz);
Chris@82 56 }
Chris@82 57 {
Chris@82 58 V TB, TC, T4, T5;
Chris@82 59 TB = LD(&(x[WS(vs, 3) + WS(rs, 1)]), ms, &(x[WS(vs, 3) + WS(rs, 1)]));
Chris@82 60 TC = LD(&(x[WS(vs, 3) + WS(rs, 3)]), ms, &(x[WS(vs, 3) + WS(rs, 1)]));
Chris@82 61 TD = VSUB(TB, TC);
Chris@82 62 TH = VADD(TB, TC);
Chris@82 63 T4 = LD(&(x[WS(rs, 1)]), ms, &(x[WS(rs, 1)]));
Chris@82 64 T5 = LD(&(x[WS(rs, 3)]), ms, &(x[WS(rs, 1)]));
Chris@82 65 T6 = VSUB(T4, T5);
Chris@82 66 Ta = VADD(T4, T5);
Chris@82 67 }
Chris@82 68 {
Chris@82 69 V Tc, Td, Tn, To;
Chris@82 70 Tc = LD(&(x[WS(vs, 1)]), ms, &(x[WS(vs, 1)]));
Chris@82 71 Td = LD(&(x[WS(vs, 1) + WS(rs, 2)]), ms, &(x[WS(vs, 1)]));
Chris@82 72 Te = VSUB(Tc, Td);
Chris@82 73 Tk = VADD(Tc, Td);
Chris@82 74 Tn = LD(&(x[WS(vs, 2)]), ms, &(x[WS(vs, 2)]));
Chris@82 75 To = LD(&(x[WS(vs, 2) + WS(rs, 2)]), ms, &(x[WS(vs, 2)]));
Chris@82 76 Tp = VSUB(Tn, To);
Chris@82 77 Tv = VADD(Tn, To);
Chris@82 78 }
Chris@82 79 {
Chris@82 80 V Tq, Tr, Tf, Tg;
Chris@82 81 Tq = LD(&(x[WS(vs, 2) + WS(rs, 1)]), ms, &(x[WS(vs, 2) + WS(rs, 1)]));
Chris@82 82 Tr = LD(&(x[WS(vs, 2) + WS(rs, 3)]), ms, &(x[WS(vs, 2) + WS(rs, 1)]));
Chris@82 83 Ts = VSUB(Tq, Tr);
Chris@82 84 Tw = VADD(Tq, Tr);
Chris@82 85 Tf = LD(&(x[WS(vs, 1) + WS(rs, 1)]), ms, &(x[WS(vs, 1) + WS(rs, 1)]));
Chris@82 86 Tg = LD(&(x[WS(vs, 1) + WS(rs, 3)]), ms, &(x[WS(vs, 1) + WS(rs, 1)]));
Chris@82 87 Th = VSUB(Tf, Tg);
Chris@82 88 Tl = VADD(Tf, Tg);
Chris@82 89 }
Chris@82 90 ST(&(x[0]), VADD(T9, Ta), ms, &(x[0]));
Chris@82 91 ST(&(x[WS(rs, 1)]), VADD(Tk, Tl), ms, &(x[WS(rs, 1)]));
Chris@82 92 ST(&(x[WS(rs, 2)]), VADD(Tv, Tw), ms, &(x[0]));
Chris@82 93 ST(&(x[WS(rs, 3)]), VADD(TG, TH), ms, &(x[WS(rs, 1)]));
Chris@82 94 {
Chris@82 95 V T7, Ti, Tt, TE;
Chris@82 96 T7 = BYTW(&(W[TWVL * 4]), VFNMSI(T6, T3));
Chris@82 97 ST(&(x[WS(vs, 3)]), T7, ms, &(x[WS(vs, 3)]));
Chris@82 98 Ti = BYTW(&(W[TWVL * 4]), VFNMSI(Th, Te));
Chris@82 99 ST(&(x[WS(vs, 3) + WS(rs, 1)]), Ti, ms, &(x[WS(vs, 3) + WS(rs, 1)]));
Chris@82 100 Tt = BYTW(&(W[TWVL * 4]), VFNMSI(Ts, Tp));
Chris@82 101 ST(&(x[WS(vs, 3) + WS(rs, 2)]), Tt, ms, &(x[WS(vs, 3)]));
Chris@82 102 TE = BYTW(&(W[TWVL * 4]), VFNMSI(TD, TA));
Chris@82 103 ST(&(x[WS(vs, 3) + WS(rs, 3)]), TE, ms, &(x[WS(vs, 3) + WS(rs, 1)]));
Chris@82 104 }
Chris@82 105 {
Chris@82 106 V T8, Tj, Tu, TF;
Chris@82 107 T8 = BYTW(&(W[0]), VFMAI(T6, T3));
Chris@82 108 ST(&(x[WS(vs, 1)]), T8, ms, &(x[WS(vs, 1)]));
Chris@82 109 Tj = BYTW(&(W[0]), VFMAI(Th, Te));
Chris@82 110 ST(&(x[WS(vs, 1) + WS(rs, 1)]), Tj, ms, &(x[WS(vs, 1) + WS(rs, 1)]));
Chris@82 111 Tu = BYTW(&(W[0]), VFMAI(Ts, Tp));
Chris@82 112 ST(&(x[WS(vs, 1) + WS(rs, 2)]), Tu, ms, &(x[WS(vs, 1)]));
Chris@82 113 TF = BYTW(&(W[0]), VFMAI(TD, TA));
Chris@82 114 ST(&(x[WS(vs, 1) + WS(rs, 3)]), TF, ms, &(x[WS(vs, 1) + WS(rs, 1)]));
Chris@82 115 }
Chris@82 116 {
Chris@82 117 V Tb, Tm, Tx, TI;
Chris@82 118 Tb = BYTW(&(W[TWVL * 2]), VSUB(T9, Ta));
Chris@82 119 ST(&(x[WS(vs, 2)]), Tb, ms, &(x[WS(vs, 2)]));
Chris@82 120 Tm = BYTW(&(W[TWVL * 2]), VSUB(Tk, Tl));
Chris@82 121 ST(&(x[WS(vs, 2) + WS(rs, 1)]), Tm, ms, &(x[WS(vs, 2) + WS(rs, 1)]));
Chris@82 122 Tx = BYTW(&(W[TWVL * 2]), VSUB(Tv, Tw));
Chris@82 123 ST(&(x[WS(vs, 2) + WS(rs, 2)]), Tx, ms, &(x[WS(vs, 2)]));
Chris@82 124 TI = BYTW(&(W[TWVL * 2]), VSUB(TG, TH));
Chris@82 125 ST(&(x[WS(vs, 2) + WS(rs, 3)]), TI, ms, &(x[WS(vs, 2) + WS(rs, 1)]));
Chris@82 126 }
Chris@82 127 }
Chris@82 128 }
Chris@82 129 VLEAVE();
Chris@82 130 }
Chris@82 131
Chris@82 132 static const tw_instr twinstr[] = {
Chris@82 133 VTW(0, 1),
Chris@82 134 VTW(0, 2),
Chris@82 135 VTW(0, 3),
Chris@82 136 {TW_NEXT, VL, 0}
Chris@82 137 };
Chris@82 138
Chris@82 139 static const ct_desc desc = { 4, XSIMD_STRING("q1bv_4"), twinstr, &GENUS, {36, 24, 8, 0}, 0, 0, 0 };
Chris@82 140
Chris@82 141 void XSIMD(codelet_q1bv_4) (planner *p) {
Chris@82 142 X(kdft_difsq_register) (p, q1bv_4, &desc);
Chris@82 143 }
Chris@82 144 #else
Chris@82 145
Chris@82 146 /* Generated by: ../../../genfft/gen_twidsq_c.native -simd -compact -variables 4 -pipeline-latency 8 -n 4 -dif -name q1bv_4 -include dft/simd/q1b.h -sign 1 */
Chris@82 147
Chris@82 148 /*
Chris@82 149 * This function contains 44 FP additions, 24 FP multiplications,
Chris@82 150 * (or, 44 additions, 24 multiplications, 0 fused multiply/add),
Chris@82 151 * 22 stack variables, 0 constants, and 32 memory accesses
Chris@82 152 */
Chris@82 153 #include "dft/simd/q1b.h"
Chris@82 154
Chris@82 155 static void q1bv_4(R *ri, R *ii, const R *W, stride rs, stride vs, INT mb, INT me, INT ms)
Chris@82 156 {
Chris@82 157 {
Chris@82 158 INT m;
Chris@82 159 R *x;
Chris@82 160 x = ii;
Chris@82 161 for (m = mb, W = W + (mb * ((TWVL / VL) * 6)); m < me; m = m + VL, x = x + (VL * ms), W = W + (TWVL * 6), MAKE_VOLATILE_STRIDE(8, rs), MAKE_VOLATILE_STRIDE(8, vs)) {
Chris@82 162 V T3, T9, TA, TG, TD, TH, T6, Ta, Te, Tk, Tp, Tv, Ts, Tw, Th;
Chris@82 163 V Tl;
Chris@82 164 {
Chris@82 165 V T1, T2, Ty, Tz;
Chris@82 166 T1 = LD(&(x[0]), ms, &(x[0]));
Chris@82 167 T2 = LD(&(x[WS(rs, 2)]), ms, &(x[0]));
Chris@82 168 T3 = VSUB(T1, T2);
Chris@82 169 T9 = VADD(T1, T2);
Chris@82 170 Ty = LD(&(x[WS(vs, 3)]), ms, &(x[WS(vs, 3)]));
Chris@82 171 Tz = LD(&(x[WS(vs, 3) + WS(rs, 2)]), ms, &(x[WS(vs, 3)]));
Chris@82 172 TA = VSUB(Ty, Tz);
Chris@82 173 TG = VADD(Ty, Tz);
Chris@82 174 }
Chris@82 175 {
Chris@82 176 V TB, TC, T4, T5;
Chris@82 177 TB = LD(&(x[WS(vs, 3) + WS(rs, 1)]), ms, &(x[WS(vs, 3) + WS(rs, 1)]));
Chris@82 178 TC = LD(&(x[WS(vs, 3) + WS(rs, 3)]), ms, &(x[WS(vs, 3) + WS(rs, 1)]));
Chris@82 179 TD = VBYI(VSUB(TB, TC));
Chris@82 180 TH = VADD(TB, TC);
Chris@82 181 T4 = LD(&(x[WS(rs, 1)]), ms, &(x[WS(rs, 1)]));
Chris@82 182 T5 = LD(&(x[WS(rs, 3)]), ms, &(x[WS(rs, 1)]));
Chris@82 183 T6 = VBYI(VSUB(T4, T5));
Chris@82 184 Ta = VADD(T4, T5);
Chris@82 185 }
Chris@82 186 {
Chris@82 187 V Tc, Td, Tn, To;
Chris@82 188 Tc = LD(&(x[WS(vs, 1)]), ms, &(x[WS(vs, 1)]));
Chris@82 189 Td = LD(&(x[WS(vs, 1) + WS(rs, 2)]), ms, &(x[WS(vs, 1)]));
Chris@82 190 Te = VSUB(Tc, Td);
Chris@82 191 Tk = VADD(Tc, Td);
Chris@82 192 Tn = LD(&(x[WS(vs, 2)]), ms, &(x[WS(vs, 2)]));
Chris@82 193 To = LD(&(x[WS(vs, 2) + WS(rs, 2)]), ms, &(x[WS(vs, 2)]));
Chris@82 194 Tp = VSUB(Tn, To);
Chris@82 195 Tv = VADD(Tn, To);
Chris@82 196 }
Chris@82 197 {
Chris@82 198 V Tq, Tr, Tf, Tg;
Chris@82 199 Tq = LD(&(x[WS(vs, 2) + WS(rs, 1)]), ms, &(x[WS(vs, 2) + WS(rs, 1)]));
Chris@82 200 Tr = LD(&(x[WS(vs, 2) + WS(rs, 3)]), ms, &(x[WS(vs, 2) + WS(rs, 1)]));
Chris@82 201 Ts = VBYI(VSUB(Tq, Tr));
Chris@82 202 Tw = VADD(Tq, Tr);
Chris@82 203 Tf = LD(&(x[WS(vs, 1) + WS(rs, 1)]), ms, &(x[WS(vs, 1) + WS(rs, 1)]));
Chris@82 204 Tg = LD(&(x[WS(vs, 1) + WS(rs, 3)]), ms, &(x[WS(vs, 1) + WS(rs, 1)]));
Chris@82 205 Th = VBYI(VSUB(Tf, Tg));
Chris@82 206 Tl = VADD(Tf, Tg);
Chris@82 207 }
Chris@82 208 ST(&(x[0]), VADD(T9, Ta), ms, &(x[0]));
Chris@82 209 ST(&(x[WS(rs, 1)]), VADD(Tk, Tl), ms, &(x[WS(rs, 1)]));
Chris@82 210 ST(&(x[WS(rs, 2)]), VADD(Tv, Tw), ms, &(x[0]));
Chris@82 211 ST(&(x[WS(rs, 3)]), VADD(TG, TH), ms, &(x[WS(rs, 1)]));
Chris@82 212 {
Chris@82 213 V T7, Ti, Tt, TE;
Chris@82 214 T7 = BYTW(&(W[TWVL * 4]), VSUB(T3, T6));
Chris@82 215 ST(&(x[WS(vs, 3)]), T7, ms, &(x[WS(vs, 3)]));
Chris@82 216 Ti = BYTW(&(W[TWVL * 4]), VSUB(Te, Th));
Chris@82 217 ST(&(x[WS(vs, 3) + WS(rs, 1)]), Ti, ms, &(x[WS(vs, 3) + WS(rs, 1)]));
Chris@82 218 Tt = BYTW(&(W[TWVL * 4]), VSUB(Tp, Ts));
Chris@82 219 ST(&(x[WS(vs, 3) + WS(rs, 2)]), Tt, ms, &(x[WS(vs, 3)]));
Chris@82 220 TE = BYTW(&(W[TWVL * 4]), VSUB(TA, TD));
Chris@82 221 ST(&(x[WS(vs, 3) + WS(rs, 3)]), TE, ms, &(x[WS(vs, 3) + WS(rs, 1)]));
Chris@82 222 }
Chris@82 223 {
Chris@82 224 V T8, Tj, Tu, TF;
Chris@82 225 T8 = BYTW(&(W[0]), VADD(T3, T6));
Chris@82 226 ST(&(x[WS(vs, 1)]), T8, ms, &(x[WS(vs, 1)]));
Chris@82 227 Tj = BYTW(&(W[0]), VADD(Te, Th));
Chris@82 228 ST(&(x[WS(vs, 1) + WS(rs, 1)]), Tj, ms, &(x[WS(vs, 1) + WS(rs, 1)]));
Chris@82 229 Tu = BYTW(&(W[0]), VADD(Tp, Ts));
Chris@82 230 ST(&(x[WS(vs, 1) + WS(rs, 2)]), Tu, ms, &(x[WS(vs, 1)]));
Chris@82 231 TF = BYTW(&(W[0]), VADD(TA, TD));
Chris@82 232 ST(&(x[WS(vs, 1) + WS(rs, 3)]), TF, ms, &(x[WS(vs, 1) + WS(rs, 1)]));
Chris@82 233 }
Chris@82 234 {
Chris@82 235 V Tb, Tm, Tx, TI;
Chris@82 236 Tb = BYTW(&(W[TWVL * 2]), VSUB(T9, Ta));
Chris@82 237 ST(&(x[WS(vs, 2)]), Tb, ms, &(x[WS(vs, 2)]));
Chris@82 238 Tm = BYTW(&(W[TWVL * 2]), VSUB(Tk, Tl));
Chris@82 239 ST(&(x[WS(vs, 2) + WS(rs, 1)]), Tm, ms, &(x[WS(vs, 2) + WS(rs, 1)]));
Chris@82 240 Tx = BYTW(&(W[TWVL * 2]), VSUB(Tv, Tw));
Chris@82 241 ST(&(x[WS(vs, 2) + WS(rs, 2)]), Tx, ms, &(x[WS(vs, 2)]));
Chris@82 242 TI = BYTW(&(W[TWVL * 2]), VSUB(TG, TH));
Chris@82 243 ST(&(x[WS(vs, 2) + WS(rs, 3)]), TI, ms, &(x[WS(vs, 2) + WS(rs, 1)]));
Chris@82 244 }
Chris@82 245 }
Chris@82 246 }
Chris@82 247 VLEAVE();
Chris@82 248 }
Chris@82 249
Chris@82 250 static const tw_instr twinstr[] = {
Chris@82 251 VTW(0, 1),
Chris@82 252 VTW(0, 2),
Chris@82 253 VTW(0, 3),
Chris@82 254 {TW_NEXT, VL, 0}
Chris@82 255 };
Chris@82 256
Chris@82 257 static const ct_desc desc = { 4, XSIMD_STRING("q1bv_4"), twinstr, &GENUS, {44, 24, 0, 0}, 0, 0, 0 };
Chris@82 258
Chris@82 259 void XSIMD(codelet_q1bv_4) (planner *p) {
Chris@82 260 X(kdft_difsq_register) (p, q1bv_4, &desc);
Chris@82 261 }
Chris@82 262 #endif