Mercurial > hg > sv-dependency-builds

annotate src/fftw-3.3.8/dft/simd/common/t3bv_10.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 08:06:09 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 -twiddle-log3 -precompute-twiddles -no-generate-bytw -n 10 -name t3bv_10 -include dft/simd/t3b.h -sign 1 */
Chris@82 29
Chris@82 30 /*
Chris@82 31 * This function contains 57 FP additions, 52 FP multiplications,
Chris@82 32 * (or, 39 additions, 34 multiplications, 18 fused multiply/add),
Chris@82 33 * 41 stack variables, 4 constants, and 20 memory accesses
Chris@82 34 */
Chris@82 35 #include "dft/simd/t3b.h"
Chris@82 36
Chris@82 37 static void t3bv_10(R *ri, R *ii, const R *W, stride rs, INT mb, INT me, INT ms)
Chris@82 38 {
Chris@82 39 DVK(KP559016994, +0.559016994374947424102293417182819058860154590);
Chris@82 40 DVK(KP618033988, +0.618033988749894848204586834365638117720309180);
Chris@82 41 DVK(KP951056516, +0.951056516295153572116439333379382143405698634);
Chris@82 42 DVK(KP250000000, +0.250000000000000000000000000000000000000000000);
Chris@82 43 {
Chris@82 44 INT m;
Chris@82 45 R *x;
Chris@82 46 x = ii;
Chris@82 47 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(10, rs)) {
Chris@82 48 V T2, T3, T4, Ta, T5, T6, Tt, Td, Th;
Chris@82 49 T2 = LDW(&(W[0]));
Chris@82 50 T3 = LDW(&(W[TWVL * 2]));
Chris@82 51 T4 = VZMUL(T2, T3);
Chris@82 52 Ta = VZMULJ(T2, T3);
Chris@82 53 T5 = LDW(&(W[TWVL * 4]));
Chris@82 54 T6 = VZMULJ(T4, T5);
Chris@82 55 Tt = VZMULJ(T3, T5);
Chris@82 56 Td = VZMULJ(Ta, T5);
Chris@82 57 Th = VZMULJ(T2, T5);
Chris@82 58 {
Chris@82 59 V T9, TJ, Ts, Ty, Tz, TN, TO, TP, Tg, Tm, Tn, TK, TL, TM, T1;
Chris@82 60 V T8, T7;
Chris@82 61 T1 = LD(&(x[0]), ms, &(x[0]));
Chris@82 62 T7 = LD(&(x[WS(rs, 5)]), ms, &(x[WS(rs, 1)]));
Chris@82 63 T8 = VZMUL(T6, T7);
Chris@82 64 T9 = VSUB(T1, T8);
Chris@82 65 TJ = VADD(T1, T8);
Chris@82 66 {
Chris@82 67 V Tp, Tx, Tr, Tv;
Chris@82 68 {
Chris@82 69 V To, Tw, Tq, Tu;
Chris@82 70 To = LD(&(x[WS(rs, 4)]), ms, &(x[0]));
Chris@82 71 Tp = VZMUL(T4, To);
Chris@82 72 Tw = LD(&(x[WS(rs, 1)]), ms, &(x[WS(rs, 1)]));
Chris@82 73 Tx = VZMUL(T2, Tw);
Chris@82 74 Tq = LD(&(x[WS(rs, 9)]), ms, &(x[WS(rs, 1)]));
Chris@82 75 Tr = VZMUL(T5, Tq);
Chris@82 76 Tu = LD(&(x[WS(rs, 6)]), ms, &(x[0]));
Chris@82 77 Tv = VZMUL(Tt, Tu);
Chris@82 78 }
Chris@82 79 Ts = VSUB(Tp, Tr);
Chris@82 80 Ty = VSUB(Tv, Tx);
Chris@82 81 Tz = VADD(Ts, Ty);
Chris@82 82 TN = VADD(Tp, Tr);
Chris@82 83 TO = VADD(Tv, Tx);
Chris@82 84 TP = VADD(TN, TO);
Chris@82 85 }
Chris@82 86 {
Chris@82 87 V Tc, Tl, Tf, Tj;
Chris@82 88 {
Chris@82 89 V Tb, Tk, Te, Ti;
Chris@82 90 Tb = LD(&(x[WS(rs, 2)]), ms, &(x[0]));
Chris@82 91 Tc = VZMUL(Ta, Tb);
Chris@82 92 Tk = LD(&(x[WS(rs, 3)]), ms, &(x[WS(rs, 1)]));
Chris@82 93 Tl = VZMUL(T3, Tk);
Chris@82 94 Te = LD(&(x[WS(rs, 7)]), ms, &(x[WS(rs, 1)]));
Chris@82 95 Tf = VZMUL(Td, Te);
Chris@82 96 Ti = LD(&(x[WS(rs, 8)]), ms, &(x[0]));
Chris@82 97 Tj = VZMUL(Th, Ti);
Chris@82 98 }
Chris@82 99 Tg = VSUB(Tc, Tf);
Chris@82 100 Tm = VSUB(Tj, Tl);
Chris@82 101 Tn = VADD(Tg, Tm);
Chris@82 102 TK = VADD(Tc, Tf);
Chris@82 103 TL = VADD(Tj, Tl);
Chris@82 104 TM = VADD(TK, TL);
Chris@82 105 }
Chris@82 106 {
Chris@82 107 V TC, TA, TB, TG, TI, TE, TF, TH, TD;
Chris@82 108 TC = VSUB(Tn, Tz);
Chris@82 109 TA = VADD(Tn, Tz);
Chris@82 110 TB = VFNMS(LDK(KP250000000), TA, T9);
Chris@82 111 TE = VSUB(Tg, Tm);
Chris@82 112 TF = VSUB(Ts, Ty);
Chris@82 113 TG = VMUL(LDK(KP951056516), VFMA(LDK(KP618033988), TF, TE));
Chris@82 114 TI = VMUL(LDK(KP951056516), VFNMS(LDK(KP618033988), TE, TF));
Chris@82 115 ST(&(x[WS(rs, 5)]), VADD(T9, TA), ms, &(x[WS(rs, 1)]));
Chris@82 116 TH = VFNMS(LDK(KP559016994), TC, TB);
Chris@82 117 ST(&(x[WS(rs, 3)]), VFMAI(TI, TH), ms, &(x[WS(rs, 1)]));
Chris@82 118 ST(&(x[WS(rs, 7)]), VFNMSI(TI, TH), ms, &(x[WS(rs, 1)]));
Chris@82 119 TD = VFMA(LDK(KP559016994), TC, TB);
Chris@82 120 ST(&(x[WS(rs, 1)]), VFMAI(TG, TD), ms, &(x[WS(rs, 1)]));
Chris@82 121 ST(&(x[WS(rs, 9)]), VFNMSI(TG, TD), ms, &(x[WS(rs, 1)]));
Chris@82 122 }
Chris@82 123 {
Chris@82 124 V TS, TQ, TR, TW, TY, TU, TV, TX, TT;
Chris@82 125 TS = VSUB(TM, TP);
Chris@82 126 TQ = VADD(TM, TP);
Chris@82 127 TR = VFNMS(LDK(KP250000000), TQ, TJ);
Chris@82 128 TU = VSUB(TN, TO);
Chris@82 129 TV = VSUB(TK, TL);
Chris@82 130 TW = VMUL(LDK(KP951056516), VFNMS(LDK(KP618033988), TV, TU));
Chris@82 131 TY = VMUL(LDK(KP951056516), VFMA(LDK(KP618033988), TU, TV));
Chris@82 132 ST(&(x[0]), VADD(TJ, TQ), ms, &(x[0]));
Chris@82 133 TX = VFMA(LDK(KP559016994), TS, TR);
Chris@82 134 ST(&(x[WS(rs, 4)]), VFNMSI(TY, TX), ms, &(x[0]));
Chris@82 135 ST(&(x[WS(rs, 6)]), VFMAI(TY, TX), ms, &(x[0]));
Chris@82 136 TT = VFNMS(LDK(KP559016994), TS, TR);
Chris@82 137 ST(&(x[WS(rs, 2)]), VFNMSI(TW, TT), ms, &(x[0]));
Chris@82 138 ST(&(x[WS(rs, 8)]), VFMAI(TW, TT), ms, &(x[0]));
Chris@82 139 }
Chris@82 140 }
Chris@82 141 }
Chris@82 142 }
Chris@82 143 VLEAVE();
Chris@82 144 }
Chris@82 145
Chris@82 146 static const tw_instr twinstr[] = {
Chris@82 147 VTW(0, 1),
Chris@82 148 VTW(0, 3),
Chris@82 149 VTW(0, 9),
Chris@82 150 {TW_NEXT, VL, 0}
Chris@82 151 };
Chris@82 152
Chris@82 153 static const ct_desc desc = { 10, XSIMD_STRING("t3bv_10"), twinstr, &GENUS, {39, 34, 18, 0}, 0, 0, 0 };
Chris@82 154
Chris@82 155 void XSIMD(codelet_t3bv_10) (planner *p) {
Chris@82 156 X(kdft_dit_register) (p, t3bv_10, &desc);
Chris@82 157 }
Chris@82 158 #else
Chris@82 159
Chris@82 160 /* Generated by: ../../../genfft/gen_twiddle_c.native -simd -compact -variables 4 -pipeline-latency 8 -twiddle-log3 -precompute-twiddles -no-generate-bytw -n 10 -name t3bv_10 -include dft/simd/t3b.h -sign 1 */
Chris@82 161
Chris@82 162 /*
Chris@82 163 * This function contains 57 FP additions, 42 FP multiplications,
Chris@82 164 * (or, 51 additions, 36 multiplications, 6 fused multiply/add),
Chris@82 165 * 41 stack variables, 4 constants, and 20 memory accesses
Chris@82 166 */
Chris@82 167 #include "dft/simd/t3b.h"
Chris@82 168
Chris@82 169 static void t3bv_10(R *ri, R *ii, const R *W, stride rs, INT mb, INT me, INT ms)
Chris@82 170 {
Chris@82 171 DVK(KP587785252, +0.587785252292473129168705954639072768597652438);
Chris@82 172 DVK(KP951056516, +0.951056516295153572116439333379382143405698634);
Chris@82 173 DVK(KP250000000, +0.250000000000000000000000000000000000000000000);
Chris@82 174 DVK(KP559016994, +0.559016994374947424102293417182819058860154590);
Chris@82 175 {
Chris@82 176 INT m;
Chris@82 177 R *x;
Chris@82 178 x = ii;
Chris@82 179 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(10, rs)) {
Chris@82 180 V T1, T2, T3, Ti, T6, T7, TA, Tb, To;
Chris@82 181 T1 = LDW(&(W[0]));
Chris@82 182 T2 = LDW(&(W[TWVL * 2]));
Chris@82 183 T3 = VZMULJ(T1, T2);
Chris@82 184 Ti = VZMUL(T1, T2);
Chris@82 185 T6 = LDW(&(W[TWVL * 4]));
Chris@82 186 T7 = VZMULJ(T3, T6);
Chris@82 187 TA = VZMULJ(Ti, T6);
Chris@82 188 Tb = VZMULJ(T1, T6);
Chris@82 189 To = VZMULJ(T2, T6);
Chris@82 190 {
Chris@82 191 V TD, TQ, Tn, Tt, Tx, TM, TN, TS, Ta, Tg, Tw, TJ, TK, TR, Tz;
Chris@82 192 V TC, TB;
Chris@82 193 Tz = LD(&(x[0]), ms, &(x[0]));
Chris@82 194 TB = LD(&(x[WS(rs, 5)]), ms, &(x[WS(rs, 1)]));
Chris@82 195 TC = VZMUL(TA, TB);
Chris@82 196 TD = VSUB(Tz, TC);
Chris@82 197 TQ = VADD(Tz, TC);
Chris@82 198 {
Chris@82 199 V Tk, Ts, Tm, Tq;
Chris@82 200 {
Chris@82 201 V Tj, Tr, Tl, Tp;
Chris@82 202 Tj = LD(&(x[WS(rs, 4)]), ms, &(x[0]));
Chris@82 203 Tk = VZMUL(Ti, Tj);
Chris@82 204 Tr = LD(&(x[WS(rs, 1)]), ms, &(x[WS(rs, 1)]));
Chris@82 205 Ts = VZMUL(T1, Tr);
Chris@82 206 Tl = LD(&(x[WS(rs, 9)]), ms, &(x[WS(rs, 1)]));
Chris@82 207 Tm = VZMUL(T6, Tl);
Chris@82 208 Tp = LD(&(x[WS(rs, 6)]), ms, &(x[0]));
Chris@82 209 Tq = VZMUL(To, Tp);
Chris@82 210 }
Chris@82 211 Tn = VSUB(Tk, Tm);
Chris@82 212 Tt = VSUB(Tq, Ts);
Chris@82 213 Tx = VADD(Tn, Tt);
Chris@82 214 TM = VADD(Tk, Tm);
Chris@82 215 TN = VADD(Tq, Ts);
Chris@82 216 TS = VADD(TM, TN);
Chris@82 217 }
Chris@82 218 {
Chris@82 219 V T5, Tf, T9, Td;
Chris@82 220 {
Chris@82 221 V T4, Te, T8, Tc;
Chris@82 222 T4 = LD(&(x[WS(rs, 2)]), ms, &(x[0]));
Chris@82 223 T5 = VZMUL(T3, T4);
Chris@82 224 Te = LD(&(x[WS(rs, 3)]), ms, &(x[WS(rs, 1)]));
Chris@82 225 Tf = VZMUL(T2, Te);
Chris@82 226 T8 = LD(&(x[WS(rs, 7)]), ms, &(x[WS(rs, 1)]));
Chris@82 227 T9 = VZMUL(T7, T8);
Chris@82 228 Tc = LD(&(x[WS(rs, 8)]), ms, &(x[0]));
Chris@82 229 Td = VZMUL(Tb, Tc);
Chris@82 230 }
Chris@82 231 Ta = VSUB(T5, T9);
Chris@82 232 Tg = VSUB(Td, Tf);
Chris@82 233 Tw = VADD(Ta, Tg);
Chris@82 234 TJ = VADD(T5, T9);
Chris@82 235 TK = VADD(Td, Tf);
Chris@82 236 TR = VADD(TJ, TK);
Chris@82 237 }
Chris@82 238 {
Chris@82 239 V Ty, TE, TF, Tv, TI, Th, Tu, TH, TG;
Chris@82 240 Ty = VMUL(LDK(KP559016994), VSUB(Tw, Tx));
Chris@82 241 TE = VADD(Tw, Tx);
Chris@82 242 TF = VFNMS(LDK(KP250000000), TE, TD);
Chris@82 243 Th = VSUB(Ta, Tg);
Chris@82 244 Tu = VSUB(Tn, Tt);
Chris@82 245 Tv = VBYI(VFMA(LDK(KP951056516), Th, VMUL(LDK(KP587785252), Tu)));
Chris@82 246 TI = VBYI(VFNMS(LDK(KP951056516), Tu, VMUL(LDK(KP587785252), Th)));
Chris@82 247 ST(&(x[WS(rs, 5)]), VADD(TD, TE), ms, &(x[WS(rs, 1)]));
Chris@82 248 TH = VSUB(TF, Ty);
Chris@82 249 ST(&(x[WS(rs, 3)]), VSUB(TH, TI), ms, &(x[WS(rs, 1)]));
Chris@82 250 ST(&(x[WS(rs, 7)]), VADD(TI, TH), ms, &(x[WS(rs, 1)]));
Chris@82 251 TG = VADD(Ty, TF);
Chris@82 252 ST(&(x[WS(rs, 1)]), VADD(Tv, TG), ms, &(x[WS(rs, 1)]));
Chris@82 253 ST(&(x[WS(rs, 9)]), VSUB(TG, Tv), ms, &(x[WS(rs, 1)]));
Chris@82 254 }
Chris@82 255 {
Chris@82 256 V TV, TT, TU, TP, TY, TL, TO, TX, TW;
Chris@82 257 TV = VMUL(LDK(KP559016994), VSUB(TR, TS));
Chris@82 258 TT = VADD(TR, TS);
Chris@82 259 TU = VFNMS(LDK(KP250000000), TT, TQ);
Chris@82 260 TL = VSUB(TJ, TK);
Chris@82 261 TO = VSUB(TM, TN);
Chris@82 262 TP = VBYI(VFNMS(LDK(KP951056516), TO, VMUL(LDK(KP587785252), TL)));
Chris@82 263 TY = VBYI(VFMA(LDK(KP951056516), TL, VMUL(LDK(KP587785252), TO)));
Chris@82 264 ST(&(x[0]), VADD(TQ, TT), ms, &(x[0]));
Chris@82 265 TX = VADD(TV, TU);
Chris@82 266 ST(&(x[WS(rs, 4)]), VSUB(TX, TY), ms, &(x[0]));
Chris@82 267 ST(&(x[WS(rs, 6)]), VADD(TY, TX), ms, &(x[0]));
Chris@82 268 TW = VSUB(TU, TV);
Chris@82 269 ST(&(x[WS(rs, 2)]), VADD(TP, TW), ms, &(x[0]));
Chris@82 270 ST(&(x[WS(rs, 8)]), VSUB(TW, TP), ms, &(x[0]));
Chris@82 271 }
Chris@82 272 }
Chris@82 273 }
Chris@82 274 }
Chris@82 275 VLEAVE();
Chris@82 276 }
Chris@82 277
Chris@82 278 static const tw_instr twinstr[] = {
Chris@82 279 VTW(0, 1),
Chris@82 280 VTW(0, 3),
Chris@82 281 VTW(0, 9),
Chris@82 282 {TW_NEXT, VL, 0}
Chris@82 283 };
Chris@82 284
Chris@82 285 static const ct_desc desc = { 10, XSIMD_STRING("t3bv_10"), twinstr, &GENUS, {51, 36, 6, 0}, 0, 0, 0 };
Chris@82 286
Chris@82 287 void XSIMD(codelet_t3bv_10) (planner *p) {
Chris@82 288 X(kdft_dit_register) (p, t3bv_10, &desc);
Chris@82 289 }
Chris@82 290 #endif