Mercurial > hg > sv-dependency-builds

annotate src/fftw-3.3.5/dft/simd/common/t1fv_16.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 2cd0e3b3e1fd
children
rev   line source
Chris@42 1 /*
Chris@42 2 * Copyright (c) 2003, 2007-14 Matteo Frigo
Chris@42 3 * Copyright (c) 2003, 2007-14 Massachusetts Institute of Technology
Chris@42 4 *
Chris@42 5 * This program is free software; you can redistribute it and/or modify
Chris@42 6 * it under the terms of the GNU General Public License as published by
Chris@42 7 * the Free Software Foundation; either version 2 of the License, or
Chris@42 8 * (at your option) any later version.
Chris@42 9 *
Chris@42 10 * This program is distributed in the hope that it will be useful,
Chris@42 11 * but WITHOUT ANY WARRANTY; without even the implied warranty of
Chris@42 12 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
Chris@42 13 * GNU General Public License for more details.
Chris@42 14 *
Chris@42 15 * You should have received a copy of the GNU General Public License
Chris@42 16 * along with this program; if not, write to the Free Software
Chris@42 17 * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
Chris@42 18 *
Chris@42 19 */
Chris@42 20
Chris@42 21 /* This file was automatically generated --- DO NOT EDIT */
Chris@42 22 /* Generated on Sat Jul 30 16:42:04 EDT 2016 */
Chris@42 23
Chris@42 24 #include "codelet-dft.h"
Chris@42 25
Chris@42 26 #ifdef HAVE_FMA
Chris@42 27
Chris@42 28 /* Generated by: ../../../genfft/gen_twiddle_c.native -fma -reorder-insns -schedule-for-pipeline -simd -compact -variables 4 -pipeline-latency 8 -n 16 -name t1fv_16 -include t1f.h */
Chris@42 29
Chris@42 30 /*
Chris@42 31 * This function contains 87 FP additions, 64 FP multiplications,
Chris@42 32 * (or, 53 additions, 30 multiplications, 34 fused multiply/add),
Chris@42 33 * 61 stack variables, 3 constants, and 32 memory accesses
Chris@42 34 */
Chris@42 35 #include "t1f.h"
Chris@42 36
Chris@42 37 static void t1fv_16(R *ri, R *ii, const R *W, stride rs, INT mb, INT me, INT ms)
Chris@42 38 {
Chris@42 39 DVK(KP923879532, +0.923879532511286756128183189396788286822416626);
Chris@42 40 DVK(KP414213562, +0.414213562373095048801688724209698078569671875);
Chris@42 41 DVK(KP707106781, +0.707106781186547524400844362104849039284835938);
Chris@42 42 {
Chris@42 43 INT m;
Chris@42 44 R *x;
Chris@42 45 x = ri;
Chris@42 46 for (m = mb, W = W + (mb * ((TWVL / VL) * 30)); m < me; m = m + VL, x = x + (VL * ms), W = W + (TWVL * 30), MAKE_VOLATILE_STRIDE(16, rs)) {
Chris@42 47 V TO, Ta, TJ, TP, T14, Tq, T1i, T10, T1b, T1l, T13, T1c, TR, Tl, T15;
Chris@42 48 V Tv;
Chris@42 49 {
Chris@42 50 V Tc, TW, T4, T19, T9, TD, TI, Tj, TZ, T1a, Te, Th, Tn, Tr, Tu;
Chris@42 51 V Tp;
Chris@42 52 {
Chris@42 53 V T1, T2, T5, T7;
Chris@42 54 T1 = LD(&(x[0]), ms, &(x[0]));
Chris@42 55 T2 = LD(&(x[WS(rs, 8)]), ms, &(x[0]));
Chris@42 56 T5 = LD(&(x[WS(rs, 4)]), ms, &(x[0]));
Chris@42 57 T7 = LD(&(x[WS(rs, 12)]), ms, &(x[0]));
Chris@42 58 {
Chris@42 59 V Tz, TG, TB, TE;
Chris@42 60 Tz = LD(&(x[WS(rs, 14)]), ms, &(x[0]));
Chris@42 61 TG = LD(&(x[WS(rs, 10)]), ms, &(x[0]));
Chris@42 62 TB = LD(&(x[WS(rs, 6)]), ms, &(x[0]));
Chris@42 63 TE = LD(&(x[WS(rs, 2)]), ms, &(x[0]));
Chris@42 64 {
Chris@42 65 V Ti, TY, TX, Td, Tg, Tm, Tt, To;
Chris@42 66 {
Chris@42 67 V T3, T6, T8, TA, TH, TC, TF, Tb;
Chris@42 68 Tb = LD(&(x[WS(rs, 1)]), ms, &(x[WS(rs, 1)]));
Chris@42 69 T3 = BYTWJ(&(W[TWVL * 14]), T2);
Chris@42 70 T6 = BYTWJ(&(W[TWVL * 6]), T5);
Chris@42 71 T8 = BYTWJ(&(W[TWVL * 22]), T7);
Chris@42 72 TA = BYTWJ(&(W[TWVL * 26]), Tz);
Chris@42 73 TH = BYTWJ(&(W[TWVL * 18]), TG);
Chris@42 74 TC = BYTWJ(&(W[TWVL * 10]), TB);
Chris@42 75 TF = BYTWJ(&(W[TWVL * 2]), TE);
Chris@42 76 Tc = BYTWJ(&(W[0]), Tb);
Chris@42 77 TW = VSUB(T1, T3);
Chris@42 78 T4 = VADD(T1, T3);
Chris@42 79 T19 = VSUB(T6, T8);
Chris@42 80 T9 = VADD(T6, T8);
Chris@42 81 Ti = LD(&(x[WS(rs, 13)]), ms, &(x[WS(rs, 1)]));
Chris@42 82 TD = VADD(TA, TC);
Chris@42 83 TY = VSUB(TA, TC);
Chris@42 84 TI = VADD(TF, TH);
Chris@42 85 TX = VSUB(TF, TH);
Chris@42 86 }
Chris@42 87 Td = LD(&(x[WS(rs, 9)]), ms, &(x[WS(rs, 1)]));
Chris@42 88 Tg = LD(&(x[WS(rs, 5)]), ms, &(x[WS(rs, 1)]));
Chris@42 89 Tm = LD(&(x[WS(rs, 15)]), ms, &(x[WS(rs, 1)]));
Chris@42 90 Tj = BYTWJ(&(W[TWVL * 24]), Ti);
Chris@42 91 Tt = LD(&(x[WS(rs, 11)]), ms, &(x[WS(rs, 1)]));
Chris@42 92 To = LD(&(x[WS(rs, 7)]), ms, &(x[WS(rs, 1)]));
Chris@42 93 TZ = VADD(TX, TY);
Chris@42 94 T1a = VSUB(TY, TX);
Chris@42 95 Te = BYTWJ(&(W[TWVL * 16]), Td);
Chris@42 96 Th = BYTWJ(&(W[TWVL * 8]), Tg);
Chris@42 97 Tn = BYTWJ(&(W[TWVL * 28]), Tm);
Chris@42 98 Tr = LD(&(x[WS(rs, 3)]), ms, &(x[WS(rs, 1)]));
Chris@42 99 Tu = BYTWJ(&(W[TWVL * 20]), Tt);
Chris@42 100 Tp = BYTWJ(&(W[TWVL * 12]), To);
Chris@42 101 }
Chris@42 102 }
Chris@42 103 }
Chris@42 104 {
Chris@42 105 V Tf, T11, Tk, T12, Ts;
Chris@42 106 TO = VADD(T4, T9);
Chris@42 107 Ta = VSUB(T4, T9);
Chris@42 108 TJ = VSUB(TD, TI);
Chris@42 109 TP = VADD(TI, TD);
Chris@42 110 Tf = VADD(Tc, Te);
Chris@42 111 T11 = VSUB(Tc, Te);
Chris@42 112 Tk = VADD(Th, Tj);
Chris@42 113 T12 = VSUB(Th, Tj);
Chris@42 114 Ts = BYTWJ(&(W[TWVL * 4]), Tr);
Chris@42 115 T14 = VSUB(Tn, Tp);
Chris@42 116 Tq = VADD(Tn, Tp);
Chris@42 117 T1i = VFNMS(LDK(KP707106781), TZ, TW);
Chris@42 118 T10 = VFMA(LDK(KP707106781), TZ, TW);
Chris@42 119 T1b = VFNMS(LDK(KP707106781), T1a, T19);
Chris@42 120 T1l = VFMA(LDK(KP707106781), T1a, T19);
Chris@42 121 T13 = VFNMS(LDK(KP414213562), T12, T11);
Chris@42 122 T1c = VFMA(LDK(KP414213562), T11, T12);
Chris@42 123 TR = VADD(Tf, Tk);
Chris@42 124 Tl = VSUB(Tf, Tk);
Chris@42 125 T15 = VSUB(Tu, Ts);
Chris@42 126 Tv = VADD(Ts, Tu);
Chris@42 127 }
Chris@42 128 }
Chris@42 129 {
Chris@42 130 V T1d, T16, TS, Tw, TU, TQ;
Chris@42 131 T1d = VFMA(LDK(KP414213562), T14, T15);
Chris@42 132 T16 = VFNMS(LDK(KP414213562), T15, T14);
Chris@42 133 TS = VADD(Tq, Tv);
Chris@42 134 Tw = VSUB(Tq, Tv);
Chris@42 135 TU = VSUB(TO, TP);
Chris@42 136 TQ = VADD(TO, TP);
Chris@42 137 {
Chris@42 138 V T1e, T1j, T17, T1m;
Chris@42 139 T1e = VSUB(T1c, T1d);
Chris@42 140 T1j = VADD(T1c, T1d);
Chris@42 141 T17 = VADD(T13, T16);
Chris@42 142 T1m = VSUB(T16, T13);
Chris@42 143 {
Chris@42 144 V TV, TT, TK, Tx;
Chris@42 145 TV = VSUB(TS, TR);
Chris@42 146 TT = VADD(TR, TS);
Chris@42 147 TK = VSUB(Tw, Tl);
Chris@42 148 Tx = VADD(Tl, Tw);
Chris@42 149 {
Chris@42 150 V T1h, T1f, T1o, T1k;
Chris@42 151 T1h = VFMA(LDK(KP923879532), T1e, T1b);
Chris@42 152 T1f = VFNMS(LDK(KP923879532), T1e, T1b);
Chris@42 153 T1o = VFMA(LDK(KP923879532), T1j, T1i);
Chris@42 154 T1k = VFNMS(LDK(KP923879532), T1j, T1i);
Chris@42 155 {
Chris@42 156 V T1g, T18, T1p, T1n;
Chris@42 157 T1g = VFMA(LDK(KP923879532), T17, T10);
Chris@42 158 T18 = VFNMS(LDK(KP923879532), T17, T10);
Chris@42 159 T1p = VFMA(LDK(KP923879532), T1m, T1l);
Chris@42 160 T1n = VFNMS(LDK(KP923879532), T1m, T1l);
Chris@42 161 ST(&(x[WS(rs, 12)]), VFNMSI(TV, TU), ms, &(x[0]));
Chris@42 162 ST(&(x[WS(rs, 4)]), VFMAI(TV, TU), ms, &(x[0]));
Chris@42 163 ST(&(x[0]), VADD(TQ, TT), ms, &(x[0]));
Chris@42 164 ST(&(x[WS(rs, 8)]), VSUB(TQ, TT), ms, &(x[0]));
Chris@42 165 {
Chris@42 166 V TN, TL, TM, Ty;
Chris@42 167 TN = VFMA(LDK(KP707106781), TK, TJ);
Chris@42 168 TL = VFNMS(LDK(KP707106781), TK, TJ);
Chris@42 169 TM = VFMA(LDK(KP707106781), Tx, Ta);
Chris@42 170 Ty = VFNMS(LDK(KP707106781), Tx, Ta);
Chris@42 171 ST(&(x[WS(rs, 1)]), VFNMSI(T1h, T1g), ms, &(x[WS(rs, 1)]));
Chris@42 172 ST(&(x[WS(rs, 15)]), VFMAI(T1h, T1g), ms, &(x[WS(rs, 1)]));
Chris@42 173 ST(&(x[WS(rs, 7)]), VFMAI(T1f, T18), ms, &(x[WS(rs, 1)]));
Chris@42 174 ST(&(x[WS(rs, 9)]), VFNMSI(T1f, T18), ms, &(x[WS(rs, 1)]));
Chris@42 175 ST(&(x[WS(rs, 3)]), VFMAI(T1p, T1o), ms, &(x[WS(rs, 1)]));
Chris@42 176 ST(&(x[WS(rs, 13)]), VFNMSI(T1p, T1o), ms, &(x[WS(rs, 1)]));
Chris@42 177 ST(&(x[WS(rs, 11)]), VFMAI(T1n, T1k), ms, &(x[WS(rs, 1)]));
Chris@42 178 ST(&(x[WS(rs, 5)]), VFNMSI(T1n, T1k), ms, &(x[WS(rs, 1)]));
Chris@42 179 ST(&(x[WS(rs, 14)]), VFNMSI(TN, TM), ms, &(x[0]));
Chris@42 180 ST(&(x[WS(rs, 2)]), VFMAI(TN, TM), ms, &(x[0]));
Chris@42 181 ST(&(x[WS(rs, 10)]), VFMAI(TL, Ty), ms, &(x[0]));
Chris@42 182 ST(&(x[WS(rs, 6)]), VFNMSI(TL, Ty), ms, &(x[0]));
Chris@42 183 }
Chris@42 184 }
Chris@42 185 }
Chris@42 186 }
Chris@42 187 }
Chris@42 188 }
Chris@42 189 }
Chris@42 190 }
Chris@42 191 VLEAVE();
Chris@42 192 }
Chris@42 193
Chris@42 194 static const tw_instr twinstr[] = {
Chris@42 195 VTW(0, 1),
Chris@42 196 VTW(0, 2),
Chris@42 197 VTW(0, 3),
Chris@42 198 VTW(0, 4),
Chris@42 199 VTW(0, 5),
Chris@42 200 VTW(0, 6),
Chris@42 201 VTW(0, 7),
Chris@42 202 VTW(0, 8),
Chris@42 203 VTW(0, 9),
Chris@42 204 VTW(0, 10),
Chris@42 205 VTW(0, 11),
Chris@42 206 VTW(0, 12),
Chris@42 207 VTW(0, 13),
Chris@42 208 VTW(0, 14),
Chris@42 209 VTW(0, 15),
Chris@42 210 {TW_NEXT, VL, 0}
Chris@42 211 };
Chris@42 212
Chris@42 213 static const ct_desc desc = { 16, XSIMD_STRING("t1fv_16"), twinstr, &GENUS, {53, 30, 34, 0}, 0, 0, 0 };
Chris@42 214
Chris@42 215 void XSIMD(codelet_t1fv_16) (planner *p) {
Chris@42 216 X(kdft_dit_register) (p, t1fv_16, &desc);
Chris@42 217 }
Chris@42 218 #else /* HAVE_FMA */
Chris@42 219
Chris@42 220 /* Generated by: ../../../genfft/gen_twiddle_c.native -simd -compact -variables 4 -pipeline-latency 8 -n 16 -name t1fv_16 -include t1f.h */
Chris@42 221
Chris@42 222 /*
Chris@42 223 * This function contains 87 FP additions, 42 FP multiplications,
Chris@42 224 * (or, 83 additions, 38 multiplications, 4 fused multiply/add),
Chris@42 225 * 36 stack variables, 3 constants, and 32 memory accesses
Chris@42 226 */
Chris@42 227 #include "t1f.h"
Chris@42 228
Chris@42 229 static void t1fv_16(R *ri, R *ii, const R *W, stride rs, INT mb, INT me, INT ms)
Chris@42 230 {
Chris@42 231 DVK(KP923879532, +0.923879532511286756128183189396788286822416626);
Chris@42 232 DVK(KP382683432, +0.382683432365089771728459984030398866761344562);
Chris@42 233 DVK(KP707106781, +0.707106781186547524400844362104849039284835938);
Chris@42 234 {
Chris@42 235 INT m;
Chris@42 236 R *x;
Chris@42 237 x = ri;
Chris@42 238 for (m = mb, W = W + (mb * ((TWVL / VL) * 30)); m < me; m = m + VL, x = x + (VL * ms), W = W + (TWVL * 30), MAKE_VOLATILE_STRIDE(16, rs)) {
Chris@42 239 V TJ, T10, TD, T11, T1b, T1c, Ty, TK, T16, T17, T18, Tb, TN, T13, T14;
Chris@42 240 V T15, Tm, TM, TG, TI, TH;
Chris@42 241 TG = LD(&(x[0]), ms, &(x[0]));
Chris@42 242 TH = LD(&(x[WS(rs, 8)]), ms, &(x[0]));
Chris@42 243 TI = BYTWJ(&(W[TWVL * 14]), TH);
Chris@42 244 TJ = VSUB(TG, TI);
Chris@42 245 T10 = VADD(TG, TI);
Chris@42 246 {
Chris@42 247 V TA, TC, Tz, TB;
Chris@42 248 Tz = LD(&(x[WS(rs, 4)]), ms, &(x[0]));
Chris@42 249 TA = BYTWJ(&(W[TWVL * 6]), Tz);
Chris@42 250 TB = LD(&(x[WS(rs, 12)]), ms, &(x[0]));
Chris@42 251 TC = BYTWJ(&(W[TWVL * 22]), TB);
Chris@42 252 TD = VSUB(TA, TC);
Chris@42 253 T11 = VADD(TA, TC);
Chris@42 254 }
Chris@42 255 {
Chris@42 256 V Tp, Tw, Tr, Tu, Ts, Tx;
Chris@42 257 {
Chris@42 258 V To, Tv, Tq, Tt;
Chris@42 259 To = LD(&(x[WS(rs, 14)]), ms, &(x[0]));
Chris@42 260 Tp = BYTWJ(&(W[TWVL * 26]), To);
Chris@42 261 Tv = LD(&(x[WS(rs, 10)]), ms, &(x[0]));
Chris@42 262 Tw = BYTWJ(&(W[TWVL * 18]), Tv);
Chris@42 263 Tq = LD(&(x[WS(rs, 6)]), ms, &(x[0]));
Chris@42 264 Tr = BYTWJ(&(W[TWVL * 10]), Tq);
Chris@42 265 Tt = LD(&(x[WS(rs, 2)]), ms, &(x[0]));
Chris@42 266 Tu = BYTWJ(&(W[TWVL * 2]), Tt);
Chris@42 267 }
Chris@42 268 T1b = VADD(Tp, Tr);
Chris@42 269 T1c = VADD(Tu, Tw);
Chris@42 270 Ts = VSUB(Tp, Tr);
Chris@42 271 Tx = VSUB(Tu, Tw);
Chris@42 272 Ty = VMUL(LDK(KP707106781), VSUB(Ts, Tx));
Chris@42 273 TK = VMUL(LDK(KP707106781), VADD(Tx, Ts));
Chris@42 274 }
Chris@42 275 {
Chris@42 276 V T2, T9, T4, T7, T5, Ta;
Chris@42 277 {
Chris@42 278 V T1, T8, T3, T6;
Chris@42 279 T1 = LD(&(x[WS(rs, 15)]), ms, &(x[WS(rs, 1)]));
Chris@42 280 T2 = BYTWJ(&(W[TWVL * 28]), T1);
Chris@42 281 T8 = LD(&(x[WS(rs, 11)]), ms, &(x[WS(rs, 1)]));
Chris@42 282 T9 = BYTWJ(&(W[TWVL * 20]), T8);
Chris@42 283 T3 = LD(&(x[WS(rs, 7)]), ms, &(x[WS(rs, 1)]));
Chris@42 284 T4 = BYTWJ(&(W[TWVL * 12]), T3);
Chris@42 285 T6 = LD(&(x[WS(rs, 3)]), ms, &(x[WS(rs, 1)]));
Chris@42 286 T7 = BYTWJ(&(W[TWVL * 4]), T6);
Chris@42 287 }
Chris@42 288 T16 = VADD(T2, T4);
Chris@42 289 T17 = VADD(T7, T9);
Chris@42 290 T18 = VSUB(T16, T17);
Chris@42 291 T5 = VSUB(T2, T4);
Chris@42 292 Ta = VSUB(T7, T9);
Chris@42 293 Tb = VFNMS(LDK(KP923879532), Ta, VMUL(LDK(KP382683432), T5));
Chris@42 294 TN = VFMA(LDK(KP923879532), T5, VMUL(LDK(KP382683432), Ta));
Chris@42 295 }
Chris@42 296 {
Chris@42 297 V Td, Tk, Tf, Ti, Tg, Tl;
Chris@42 298 {
Chris@42 299 V Tc, Tj, Te, Th;
Chris@42 300 Tc = LD(&(x[WS(rs, 1)]), ms, &(x[WS(rs, 1)]));
Chris@42 301 Td = BYTWJ(&(W[0]), Tc);
Chris@42 302 Tj = LD(&(x[WS(rs, 13)]), ms, &(x[WS(rs, 1)]));
Chris@42 303 Tk = BYTWJ(&(W[TWVL * 24]), Tj);
Chris@42 304 Te = LD(&(x[WS(rs, 9)]), ms, &(x[WS(rs, 1)]));
Chris@42 305 Tf = BYTWJ(&(W[TWVL * 16]), Te);
Chris@42 306 Th = LD(&(x[WS(rs, 5)]), ms, &(x[WS(rs, 1)]));
Chris@42 307 Ti = BYTWJ(&(W[TWVL * 8]), Th);
Chris@42 308 }
Chris@42 309 T13 = VADD(Td, Tf);
Chris@42 310 T14 = VADD(Ti, Tk);
Chris@42 311 T15 = VSUB(T13, T14);
Chris@42 312 Tg = VSUB(Td, Tf);
Chris@42 313 Tl = VSUB(Ti, Tk);
Chris@42 314 Tm = VFMA(LDK(KP382683432), Tg, VMUL(LDK(KP923879532), Tl));
Chris@42 315 TM = VFNMS(LDK(KP382683432), Tl, VMUL(LDK(KP923879532), Tg));
Chris@42 316 }
Chris@42 317 {
Chris@42 318 V T1a, T1g, T1f, T1h;
Chris@42 319 {
Chris@42 320 V T12, T19, T1d, T1e;
Chris@42 321 T12 = VSUB(T10, T11);
Chris@42 322 T19 = VMUL(LDK(KP707106781), VADD(T15, T18));
Chris@42 323 T1a = VADD(T12, T19);
Chris@42 324 T1g = VSUB(T12, T19);
Chris@42 325 T1d = VSUB(T1b, T1c);
Chris@42 326 T1e = VMUL(LDK(KP707106781), VSUB(T18, T15));
Chris@42 327 T1f = VBYI(VADD(T1d, T1e));
Chris@42 328 T1h = VBYI(VSUB(T1e, T1d));
Chris@42 329 }
Chris@42 330 ST(&(x[WS(rs, 14)]), VSUB(T1a, T1f), ms, &(x[0]));
Chris@42 331 ST(&(x[WS(rs, 6)]), VADD(T1g, T1h), ms, &(x[0]));
Chris@42 332 ST(&(x[WS(rs, 2)]), VADD(T1a, T1f), ms, &(x[0]));
Chris@42 333 ST(&(x[WS(rs, 10)]), VSUB(T1g, T1h), ms, &(x[0]));
Chris@42 334 }
Chris@42 335 {
Chris@42 336 V T1k, T1o, T1n, T1p;
Chris@42 337 {
Chris@42 338 V T1i, T1j, T1l, T1m;
Chris@42 339 T1i = VADD(T10, T11);
Chris@42 340 T1j = VADD(T1c, T1b);
Chris@42 341 T1k = VADD(T1i, T1j);
Chris@42 342 T1o = VSUB(T1i, T1j);
Chris@42 343 T1l = VADD(T13, T14);
Chris@42 344 T1m = VADD(T16, T17);
Chris@42 345 T1n = VADD(T1l, T1m);
Chris@42 346 T1p = VBYI(VSUB(T1m, T1l));
Chris@42 347 }
Chris@42 348 ST(&(x[WS(rs, 8)]), VSUB(T1k, T1n), ms, &(x[0]));
Chris@42 349 ST(&(x[WS(rs, 4)]), VADD(T1o, T1p), ms, &(x[0]));
Chris@42 350 ST(&(x[0]), VADD(T1k, T1n), ms, &(x[0]));
Chris@42 351 ST(&(x[WS(rs, 12)]), VSUB(T1o, T1p), ms, &(x[0]));
Chris@42 352 }
Chris@42 353 {
Chris@42 354 V TF, TQ, TP, TR;
Chris@42 355 {
Chris@42 356 V Tn, TE, TL, TO;
Chris@42 357 Tn = VSUB(Tb, Tm);
Chris@42 358 TE = VSUB(Ty, TD);
Chris@42 359 TF = VBYI(VSUB(Tn, TE));
Chris@42 360 TQ = VBYI(VADD(TE, Tn));
Chris@42 361 TL = VADD(TJ, TK);
Chris@42 362 TO = VADD(TM, TN);
Chris@42 363 TP = VSUB(TL, TO);
Chris@42 364 TR = VADD(TL, TO);
Chris@42 365 }
Chris@42 366 ST(&(x[WS(rs, 7)]), VADD(TF, TP), ms, &(x[WS(rs, 1)]));
Chris@42 367 ST(&(x[WS(rs, 15)]), VSUB(TR, TQ), ms, &(x[WS(rs, 1)]));
Chris@42 368 ST(&(x[WS(rs, 9)]), VSUB(TP, TF), ms, &(x[WS(rs, 1)]));
Chris@42 369 ST(&(x[WS(rs, 1)]), VADD(TQ, TR), ms, &(x[WS(rs, 1)]));
Chris@42 370 }
Chris@42 371 {
Chris@42 372 V TU, TY, TX, TZ;
Chris@42 373 {
Chris@42 374 V TS, TT, TV, TW;
Chris@42 375 TS = VSUB(TJ, TK);
Chris@42 376 TT = VADD(Tm, Tb);
Chris@42 377 TU = VADD(TS, TT);
Chris@42 378 TY = VSUB(TS, TT);
Chris@42 379 TV = VADD(TD, Ty);
Chris@42 380 TW = VSUB(TN, TM);
Chris@42 381 TX = VBYI(VADD(TV, TW));
Chris@42 382 TZ = VBYI(VSUB(TW, TV));
Chris@42 383 }
Chris@42 384 ST(&(x[WS(rs, 13)]), VSUB(TU, TX), ms, &(x[WS(rs, 1)]));
Chris@42 385 ST(&(x[WS(rs, 5)]), VADD(TY, TZ), ms, &(x[WS(rs, 1)]));
Chris@42 386 ST(&(x[WS(rs, 3)]), VADD(TU, TX), ms, &(x[WS(rs, 1)]));
Chris@42 387 ST(&(x[WS(rs, 11)]), VSUB(TY, TZ), ms, &(x[WS(rs, 1)]));
Chris@42 388 }
Chris@42 389 }
Chris@42 390 }
Chris@42 391 VLEAVE();
Chris@42 392 }
Chris@42 393
Chris@42 394 static const tw_instr twinstr[] = {
Chris@42 395 VTW(0, 1),
Chris@42 396 VTW(0, 2),
Chris@42 397 VTW(0, 3),
Chris@42 398 VTW(0, 4),
Chris@42 399 VTW(0, 5),
Chris@42 400 VTW(0, 6),
Chris@42 401 VTW(0, 7),
Chris@42 402 VTW(0, 8),
Chris@42 403 VTW(0, 9),
Chris@42 404 VTW(0, 10),
Chris@42 405 VTW(0, 11),
Chris@42 406 VTW(0, 12),
Chris@42 407 VTW(0, 13),
Chris@42 408 VTW(0, 14),
Chris@42 409 VTW(0, 15),
Chris@42 410 {TW_NEXT, VL, 0}
Chris@42 411 };
Chris@42 412
Chris@42 413 static const ct_desc desc = { 16, XSIMD_STRING("t1fv_16"), twinstr, &GENUS, {83, 38, 4, 0}, 0, 0, 0 };
Chris@42 414
Chris@42 415 void XSIMD(codelet_t1fv_16) (planner *p) {
Chris@42 416 X(kdft_dit_register) (p, t1fv_16, &desc);
Chris@42 417 }
Chris@42 418 #endif /* HAVE_FMA */