Mercurial > hg > sv-dependency-builds

annotate src/fftw-3.3.8/dft/simd/common/t1bv_12.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 08:05:58 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 12 -name t1bv_12 -include dft/simd/t1b.h -sign 1 */
Chris@82 29
Chris@82 30 /*
Chris@82 31 * This function contains 59 FP additions, 42 FP multiplications,
Chris@82 32 * (or, 41 additions, 24 multiplications, 18 fused multiply/add),
Chris@82 33 * 28 stack variables, 2 constants, and 24 memory accesses
Chris@82 34 */
Chris@82 35 #include "dft/simd/t1b.h"
Chris@82 36
Chris@82 37 static void t1bv_12(R *ri, R *ii, const R *W, stride rs, INT mb, INT me, INT ms)
Chris@82 38 {
Chris@82 39 DVK(KP866025403, +0.866025403784438646763723170752936183471402627);
Chris@82 40 DVK(KP500000000, +0.500000000000000000000000000000000000000000000);
Chris@82 41 {
Chris@82 42 INT m;
Chris@82 43 R *x;
Chris@82 44 x = ii;
Chris@82 45 for (m = mb, W = W + (mb * ((TWVL / VL) * 22)); m < me; m = m + VL, x = x + (VL * ms), W = W + (TWVL * 22), MAKE_VOLATILE_STRIDE(12, rs)) {
Chris@82 46 V T1, TK, T6, TA, Tq, TI, Tv, TE, T9, TL, Te, TB, Ti, TH, Tn;
Chris@82 47 V TD;
Chris@82 48 {
Chris@82 49 V T5, T3, T4, T2;
Chris@82 50 T1 = LD(&(x[0]), ms, &(x[0]));
Chris@82 51 T4 = LD(&(x[WS(rs, 8)]), ms, &(x[0]));
Chris@82 52 T5 = BYTW(&(W[TWVL * 14]), T4);
Chris@82 53 T2 = LD(&(x[WS(rs, 4)]), ms, &(x[0]));
Chris@82 54 T3 = BYTW(&(W[TWVL * 6]), T2);
Chris@82 55 TK = VSUB(T3, T5);
Chris@82 56 T6 = VADD(T3, T5);
Chris@82 57 TA = VFNMS(LDK(KP500000000), T6, T1);
Chris@82 58 }
Chris@82 59 {
Chris@82 60 V Tu, Ts, Tp, Tt, Tr;
Chris@82 61 Tp = LD(&(x[WS(rs, 9)]), ms, &(x[WS(rs, 1)]));
Chris@82 62 Tq = BYTW(&(W[TWVL * 16]), Tp);
Chris@82 63 Tt = LD(&(x[WS(rs, 5)]), ms, &(x[WS(rs, 1)]));
Chris@82 64 Tu = BYTW(&(W[TWVL * 8]), Tt);
Chris@82 65 Tr = LD(&(x[WS(rs, 1)]), ms, &(x[WS(rs, 1)]));
Chris@82 66 Ts = BYTW(&(W[0]), Tr);
Chris@82 67 TI = VSUB(Tu, Ts);
Chris@82 68 Tv = VADD(Ts, Tu);
Chris@82 69 TE = VFNMS(LDK(KP500000000), Tv, Tq);
Chris@82 70 }
Chris@82 71 {
Chris@82 72 V Td, Tb, T8, Tc, Ta;
Chris@82 73 T8 = LD(&(x[WS(rs, 6)]), ms, &(x[0]));
Chris@82 74 T9 = BYTW(&(W[TWVL * 10]), T8);
Chris@82 75 Tc = LD(&(x[WS(rs, 2)]), ms, &(x[0]));
Chris@82 76 Td = BYTW(&(W[TWVL * 2]), Tc);
Chris@82 77 Ta = LD(&(x[WS(rs, 10)]), ms, &(x[0]));
Chris@82 78 Tb = BYTW(&(W[TWVL * 18]), Ta);
Chris@82 79 TL = VSUB(Tb, Td);
Chris@82 80 Te = VADD(Tb, Td);
Chris@82 81 TB = VFNMS(LDK(KP500000000), Te, T9);
Chris@82 82 }
Chris@82 83 {
Chris@82 84 V Tm, Tk, Th, Tl, Tj;
Chris@82 85 Th = LD(&(x[WS(rs, 3)]), ms, &(x[WS(rs, 1)]));
Chris@82 86 Ti = BYTW(&(W[TWVL * 4]), Th);
Chris@82 87 Tl = LD(&(x[WS(rs, 11)]), ms, &(x[WS(rs, 1)]));
Chris@82 88 Tm = BYTW(&(W[TWVL * 20]), Tl);
Chris@82 89 Tj = LD(&(x[WS(rs, 7)]), ms, &(x[WS(rs, 1)]));
Chris@82 90 Tk = BYTW(&(W[TWVL * 12]), Tj);
Chris@82 91 TH = VSUB(Tk, Tm);
Chris@82 92 Tn = VADD(Tk, Tm);
Chris@82 93 TD = VFNMS(LDK(KP500000000), Tn, Ti);
Chris@82 94 }
Chris@82 95 {
Chris@82 96 V Tg, Ty, Tx, Tz;
Chris@82 97 {
Chris@82 98 V T7, Tf, To, Tw;
Chris@82 99 T7 = VADD(T1, T6);
Chris@82 100 Tf = VADD(T9, Te);
Chris@82 101 Tg = VSUB(T7, Tf);
Chris@82 102 Ty = VADD(T7, Tf);
Chris@82 103 To = VADD(Ti, Tn);
Chris@82 104 Tw = VADD(Tq, Tv);
Chris@82 105 Tx = VSUB(To, Tw);
Chris@82 106 Tz = VADD(To, Tw);
Chris@82 107 }
Chris@82 108 ST(&(x[WS(rs, 3)]), VFNMSI(Tx, Tg), ms, &(x[WS(rs, 1)]));
Chris@82 109 ST(&(x[0]), VADD(Ty, Tz), ms, &(x[0]));
Chris@82 110 ST(&(x[WS(rs, 9)]), VFMAI(Tx, Tg), ms, &(x[WS(rs, 1)]));
Chris@82 111 ST(&(x[WS(rs, 6)]), VSUB(Ty, Tz), ms, &(x[0]));
Chris@82 112 }
Chris@82 113 {
Chris@82 114 V TS, TW, TV, TX;
Chris@82 115 {
Chris@82 116 V TQ, TR, TT, TU;
Chris@82 117 TQ = VSUB(TA, TB);
Chris@82 118 TR = VADD(TH, TI);
Chris@82 119 TS = VFNMS(LDK(KP866025403), TR, TQ);
Chris@82 120 TW = VFMA(LDK(KP866025403), TR, TQ);
Chris@82 121 TT = VSUB(TD, TE);
Chris@82 122 TU = VSUB(TK, TL);
Chris@82 123 TV = VFMA(LDK(KP866025403), TU, TT);
Chris@82 124 TX = VFNMS(LDK(KP866025403), TU, TT);
Chris@82 125 }
Chris@82 126 ST(&(x[WS(rs, 1)]), VFMAI(TV, TS), ms, &(x[WS(rs, 1)]));
Chris@82 127 ST(&(x[WS(rs, 7)]), VFNMSI(TX, TW), ms, &(x[WS(rs, 1)]));
Chris@82 128 ST(&(x[WS(rs, 11)]), VFNMSI(TV, TS), ms, &(x[WS(rs, 1)]));
Chris@82 129 ST(&(x[WS(rs, 5)]), VFMAI(TX, TW), ms, &(x[WS(rs, 1)]));
Chris@82 130 }
Chris@82 131 {
Chris@82 132 V TG, TO, TN, TP;
Chris@82 133 {
Chris@82 134 V TC, TF, TJ, TM;
Chris@82 135 TC = VADD(TA, TB);
Chris@82 136 TF = VADD(TD, TE);
Chris@82 137 TG = VSUB(TC, TF);
Chris@82 138 TO = VADD(TC, TF);
Chris@82 139 TJ = VSUB(TH, TI);
Chris@82 140 TM = VADD(TK, TL);
Chris@82 141 TN = VMUL(LDK(KP866025403), VSUB(TJ, TM));
Chris@82 142 TP = VMUL(LDK(KP866025403), VADD(TM, TJ));
Chris@82 143 }
Chris@82 144 ST(&(x[WS(rs, 10)]), VFNMSI(TN, TG), ms, &(x[0]));
Chris@82 145 ST(&(x[WS(rs, 4)]), VFMAI(TP, TO), ms, &(x[0]));
Chris@82 146 ST(&(x[WS(rs, 2)]), VFMAI(TN, TG), ms, &(x[0]));
Chris@82 147 ST(&(x[WS(rs, 8)]), VFNMSI(TP, TO), ms, &(x[0]));
Chris@82 148 }
Chris@82 149 }
Chris@82 150 }
Chris@82 151 VLEAVE();
Chris@82 152 }
Chris@82 153
Chris@82 154 static const tw_instr twinstr[] = {
Chris@82 155 VTW(0, 1),
Chris@82 156 VTW(0, 2),
Chris@82 157 VTW(0, 3),
Chris@82 158 VTW(0, 4),
Chris@82 159 VTW(0, 5),
Chris@82 160 VTW(0, 6),
Chris@82 161 VTW(0, 7),
Chris@82 162 VTW(0, 8),
Chris@82 163 VTW(0, 9),
Chris@82 164 VTW(0, 10),
Chris@82 165 VTW(0, 11),
Chris@82 166 {TW_NEXT, VL, 0}
Chris@82 167 };
Chris@82 168
Chris@82 169 static const ct_desc desc = { 12, XSIMD_STRING("t1bv_12"), twinstr, &GENUS, {41, 24, 18, 0}, 0, 0, 0 };
Chris@82 170
Chris@82 171 void XSIMD(codelet_t1bv_12) (planner *p) {
Chris@82 172 X(kdft_dit_register) (p, t1bv_12, &desc);
Chris@82 173 }
Chris@82 174 #else
Chris@82 175
Chris@82 176 /* Generated by: ../../../genfft/gen_twiddle_c.native -simd -compact -variables 4 -pipeline-latency 8 -n 12 -name t1bv_12 -include dft/simd/t1b.h -sign 1 */
Chris@82 177
Chris@82 178 /*
Chris@82 179 * This function contains 59 FP additions, 30 FP multiplications,
Chris@82 180 * (or, 55 additions, 26 multiplications, 4 fused multiply/add),
Chris@82 181 * 28 stack variables, 2 constants, and 24 memory accesses
Chris@82 182 */
Chris@82 183 #include "dft/simd/t1b.h"
Chris@82 184
Chris@82 185 static void t1bv_12(R *ri, R *ii, const R *W, stride rs, INT mb, INT me, INT ms)
Chris@82 186 {
Chris@82 187 DVK(KP866025403, +0.866025403784438646763723170752936183471402627);
Chris@82 188 DVK(KP500000000, +0.500000000000000000000000000000000000000000000);
Chris@82 189 {
Chris@82 190 INT m;
Chris@82 191 R *x;
Chris@82 192 x = ii;
Chris@82 193 for (m = mb, W = W + (mb * ((TWVL / VL) * 22)); m < me; m = m + VL, x = x + (VL * ms), W = W + (TWVL * 22), MAKE_VOLATILE_STRIDE(12, rs)) {
Chris@82 194 V T1, Tt, T6, T7, TB, Tq, TC, TD, T9, Tu, Te, Tf, Tx, Tl, Ty;
Chris@82 195 V Tz;
Chris@82 196 {
Chris@82 197 V T5, T3, T4, T2;
Chris@82 198 T1 = LD(&(x[0]), ms, &(x[0]));
Chris@82 199 T4 = LD(&(x[WS(rs, 8)]), ms, &(x[0]));
Chris@82 200 T5 = BYTW(&(W[TWVL * 14]), T4);
Chris@82 201 T2 = LD(&(x[WS(rs, 4)]), ms, &(x[0]));
Chris@82 202 T3 = BYTW(&(W[TWVL * 6]), T2);
Chris@82 203 Tt = VSUB(T3, T5);
Chris@82 204 T6 = VADD(T3, T5);
Chris@82 205 T7 = VFNMS(LDK(KP500000000), T6, T1);
Chris@82 206 }
Chris@82 207 {
Chris@82 208 V Tn, Tp, Tm, TA, To;
Chris@82 209 Tm = LD(&(x[WS(rs, 1)]), ms, &(x[WS(rs, 1)]));
Chris@82 210 Tn = BYTW(&(W[0]), Tm);
Chris@82 211 TA = LD(&(x[WS(rs, 9)]), ms, &(x[WS(rs, 1)]));
Chris@82 212 TB = BYTW(&(W[TWVL * 16]), TA);
Chris@82 213 To = LD(&(x[WS(rs, 5)]), ms, &(x[WS(rs, 1)]));
Chris@82 214 Tp = BYTW(&(W[TWVL * 8]), To);
Chris@82 215 Tq = VSUB(Tn, Tp);
Chris@82 216 TC = VADD(Tn, Tp);
Chris@82 217 TD = VFNMS(LDK(KP500000000), TC, TB);
Chris@82 218 }
Chris@82 219 {
Chris@82 220 V Td, Tb, T8, Tc, Ta;
Chris@82 221 T8 = LD(&(x[WS(rs, 6)]), ms, &(x[0]));
Chris@82 222 T9 = BYTW(&(W[TWVL * 10]), T8);
Chris@82 223 Tc = LD(&(x[WS(rs, 2)]), ms, &(x[0]));
Chris@82 224 Td = BYTW(&(W[TWVL * 2]), Tc);
Chris@82 225 Ta = LD(&(x[WS(rs, 10)]), ms, &(x[0]));
Chris@82 226 Tb = BYTW(&(W[TWVL * 18]), Ta);
Chris@82 227 Tu = VSUB(Tb, Td);
Chris@82 228 Te = VADD(Tb, Td);
Chris@82 229 Tf = VFNMS(LDK(KP500000000), Te, T9);
Chris@82 230 }
Chris@82 231 {
Chris@82 232 V Ti, Tk, Th, Tw, Tj;
Chris@82 233 Th = LD(&(x[WS(rs, 7)]), ms, &(x[WS(rs, 1)]));
Chris@82 234 Ti = BYTW(&(W[TWVL * 12]), Th);
Chris@82 235 Tw = LD(&(x[WS(rs, 3)]), ms, &(x[WS(rs, 1)]));
Chris@82 236 Tx = BYTW(&(W[TWVL * 4]), Tw);
Chris@82 237 Tj = LD(&(x[WS(rs, 11)]), ms, &(x[WS(rs, 1)]));
Chris@82 238 Tk = BYTW(&(W[TWVL * 20]), Tj);
Chris@82 239 Tl = VSUB(Ti, Tk);
Chris@82 240 Ty = VADD(Ti, Tk);
Chris@82 241 Tz = VFNMS(LDK(KP500000000), Ty, Tx);
Chris@82 242 }
Chris@82 243 {
Chris@82 244 V Ts, TG, TF, TH;
Chris@82 245 {
Chris@82 246 V Tg, Tr, Tv, TE;
Chris@82 247 Tg = VSUB(T7, Tf);
Chris@82 248 Tr = VMUL(LDK(KP866025403), VSUB(Tl, Tq));
Chris@82 249 Ts = VSUB(Tg, Tr);
Chris@82 250 TG = VADD(Tg, Tr);
Chris@82 251 Tv = VMUL(LDK(KP866025403), VSUB(Tt, Tu));
Chris@82 252 TE = VSUB(Tz, TD);
Chris@82 253 TF = VBYI(VADD(Tv, TE));
Chris@82 254 TH = VBYI(VSUB(TE, Tv));
Chris@82 255 }
Chris@82 256 ST(&(x[WS(rs, 11)]), VSUB(Ts, TF), ms, &(x[WS(rs, 1)]));
Chris@82 257 ST(&(x[WS(rs, 5)]), VADD(TG, TH), ms, &(x[WS(rs, 1)]));
Chris@82 258 ST(&(x[WS(rs, 1)]), VADD(Ts, TF), ms, &(x[WS(rs, 1)]));
Chris@82 259 ST(&(x[WS(rs, 7)]), VSUB(TG, TH), ms, &(x[WS(rs, 1)]));
Chris@82 260 }
Chris@82 261 {
Chris@82 262 V TS, TW, TV, TX;
Chris@82 263 {
Chris@82 264 V TQ, TR, TT, TU;
Chris@82 265 TQ = VADD(T1, T6);
Chris@82 266 TR = VADD(T9, Te);
Chris@82 267 TS = VSUB(TQ, TR);
Chris@82 268 TW = VADD(TQ, TR);
Chris@82 269 TT = VADD(Tx, Ty);
Chris@82 270 TU = VADD(TB, TC);
Chris@82 271 TV = VBYI(VSUB(TT, TU));
Chris@82 272 TX = VADD(TT, TU);
Chris@82 273 }
Chris@82 274 ST(&(x[WS(rs, 3)]), VSUB(TS, TV), ms, &(x[WS(rs, 1)]));
Chris@82 275 ST(&(x[0]), VADD(TW, TX), ms, &(x[0]));
Chris@82 276 ST(&(x[WS(rs, 9)]), VADD(TS, TV), ms, &(x[WS(rs, 1)]));
Chris@82 277 ST(&(x[WS(rs, 6)]), VSUB(TW, TX), ms, &(x[0]));
Chris@82 278 }
Chris@82 279 {
Chris@82 280 V TK, TO, TN, TP;
Chris@82 281 {
Chris@82 282 V TI, TJ, TL, TM;
Chris@82 283 TI = VADD(Tl, Tq);
Chris@82 284 TJ = VADD(Tt, Tu);
Chris@82 285 TK = VBYI(VMUL(LDK(KP866025403), VSUB(TI, TJ)));
Chris@82 286 TO = VBYI(VMUL(LDK(KP866025403), VADD(TJ, TI)));
Chris@82 287 TL = VADD(T7, Tf);
Chris@82 288 TM = VADD(Tz, TD);
Chris@82 289 TN = VSUB(TL, TM);
Chris@82 290 TP = VADD(TL, TM);
Chris@82 291 }
Chris@82 292 ST(&(x[WS(rs, 2)]), VADD(TK, TN), ms, &(x[0]));
Chris@82 293 ST(&(x[WS(rs, 8)]), VSUB(TP, TO), ms, &(x[0]));
Chris@82 294 ST(&(x[WS(rs, 10)]), VSUB(TN, TK), ms, &(x[0]));
Chris@82 295 ST(&(x[WS(rs, 4)]), VADD(TO, TP), ms, &(x[0]));
Chris@82 296 }
Chris@82 297 }
Chris@82 298 }
Chris@82 299 VLEAVE();
Chris@82 300 }
Chris@82 301
Chris@82 302 static const tw_instr twinstr[] = {
Chris@82 303 VTW(0, 1),
Chris@82 304 VTW(0, 2),
Chris@82 305 VTW(0, 3),
Chris@82 306 VTW(0, 4),
Chris@82 307 VTW(0, 5),
Chris@82 308 VTW(0, 6),
Chris@82 309 VTW(0, 7),
Chris@82 310 VTW(0, 8),
Chris@82 311 VTW(0, 9),
Chris@82 312 VTW(0, 10),
Chris@82 313 VTW(0, 11),
Chris@82 314 {TW_NEXT, VL, 0}
Chris@82 315 };
Chris@82 316
Chris@82 317 static const ct_desc desc = { 12, XSIMD_STRING("t1bv_12"), twinstr, &GENUS, {55, 26, 4, 0}, 0, 0, 0 };
Chris@82 318
Chris@82 319 void XSIMD(codelet_t1bv_12) (planner *p) {
Chris@82 320 X(kdft_dit_register) (p, t1bv_12, &desc);
Chris@82 321 }
Chris@82 322 #endif