Mercurial > hg > sv-dependency-builds

annotate src/fftw-3.3.5/dft/simd/common/n2fv_12.c @ 42:2cd0e3b3e1fd

Find changesets by keywords (author, files, the commit message), revision number or hash, or revset expression.
Current fftw source
author Chris Cannam
date Tue, 18 Oct 2016 13:40:26 +0100
parents
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 16:40:09 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_notw_c.native -fma -reorder-insns -schedule-for-pipeline -simd -compact -variables 4 -pipeline-latency 8 -n 12 -name n2fv_12 -with-ostride 2 -include n2f.h -store-multiple 2 */
Chris@42 29
Chris@42 30 /*
Chris@42 31 * This function contains 48 FP additions, 20 FP multiplications,
Chris@42 32 * (or, 30 additions, 2 multiplications, 18 fused multiply/add),
Chris@42 33 * 61 stack variables, 2 constants, and 30 memory accesses
Chris@42 34 */
Chris@42 35 #include "n2f.h"
Chris@42 36
Chris@42 37 static void n2fv_12(const R *ri, const R *ii, R *ro, R *io, stride is, stride os, INT v, INT ivs, INT ovs)
Chris@42 38 {
Chris@42 39 DVK(KP866025403, +0.866025403784438646763723170752936183471402627);
Chris@42 40 DVK(KP500000000, +0.500000000000000000000000000000000000000000000);
Chris@42 41 {
Chris@42 42 INT i;
Chris@42 43 const R *xi;
Chris@42 44 R *xo;
Chris@42 45 xi = ri;
Chris@42 46 xo = ro;
Chris@42 47 for (i = v; i > 0; i = i - VL, xi = xi + (VL * ivs), xo = xo + (VL * ovs), MAKE_VOLATILE_STRIDE(24, is), MAKE_VOLATILE_STRIDE(24, os)) {
Chris@42 48 V T1, T6, Tk, Tn, Tc, Td, Tf, Tr, T4, Ts, T9, Tg, Te, Tl;
Chris@42 49 {
Chris@42 50 V T2, T3, T7, T8;
Chris@42 51 T1 = LD(&(xi[0]), ivs, &(xi[0]));
Chris@42 52 T6 = LD(&(xi[WS(is, 6)]), ivs, &(xi[0]));
Chris@42 53 T2 = LD(&(xi[WS(is, 4)]), ivs, &(xi[0]));
Chris@42 54 T3 = LD(&(xi[WS(is, 8)]), ivs, &(xi[0]));
Chris@42 55 T7 = LD(&(xi[WS(is, 10)]), ivs, &(xi[0]));
Chris@42 56 T8 = LD(&(xi[WS(is, 2)]), ivs, &(xi[0]));
Chris@42 57 Tk = LD(&(xi[WS(is, 3)]), ivs, &(xi[WS(is, 1)]));
Chris@42 58 Tn = LD(&(xi[WS(is, 9)]), ivs, &(xi[WS(is, 1)]));
Chris@42 59 Tc = LD(&(xi[WS(is, 11)]), ivs, &(xi[WS(is, 1)]));
Chris@42 60 Td = LD(&(xi[WS(is, 7)]), ivs, &(xi[WS(is, 1)]));
Chris@42 61 Tf = LD(&(xi[WS(is, 1)]), ivs, &(xi[WS(is, 1)]));
Chris@42 62 Tr = VSUB(T3, T2);
Chris@42 63 T4 = VADD(T2, T3);
Chris@42 64 Ts = VSUB(T8, T7);
Chris@42 65 T9 = VADD(T7, T8);
Chris@42 66 Tg = LD(&(xi[WS(is, 5)]), ivs, &(xi[WS(is, 1)]));
Chris@42 67 }
Chris@42 68 Te = VSUB(Tc, Td);
Chris@42 69 Tl = VADD(Td, Tc);
Chris@42 70 {
Chris@42 71 V T5, TF, TB, Tt, Ta, TG, Th, To, Tm, TI;
Chris@42 72 T5 = VFNMS(LDK(KP500000000), T4, T1);
Chris@42 73 TF = VADD(T1, T4);
Chris@42 74 TB = VADD(Tr, Ts);
Chris@42 75 Tt = VSUB(Tr, Ts);
Chris@42 76 Ta = VFNMS(LDK(KP500000000), T9, T6);
Chris@42 77 TG = VADD(T6, T9);
Chris@42 78 Th = VSUB(Tf, Tg);
Chris@42 79 To = VADD(Tf, Tg);
Chris@42 80 Tm = VFNMS(LDK(KP500000000), Tl, Tk);
Chris@42 81 TI = VADD(Tk, Tl);
Chris@42 82 {
Chris@42 83 V TH, TL, Tb, Tx, TJ, Tp, Ti, TA;
Chris@42 84 TH = VSUB(TF, TG);
Chris@42 85 TL = VADD(TF, TG);
Chris@42 86 Tb = VSUB(T5, Ta);
Chris@42 87 Tx = VADD(T5, Ta);
Chris@42 88 TJ = VADD(Tn, To);
Chris@42 89 Tp = VFNMS(LDK(KP500000000), To, Tn);
Chris@42 90 Ti = VADD(Te, Th);
Chris@42 91 TA = VSUB(Te, Th);
Chris@42 92 {
Chris@42 93 V Tq, Ty, TK, TM;
Chris@42 94 Tq = VSUB(Tm, Tp);
Chris@42 95 Ty = VADD(Tm, Tp);
Chris@42 96 TK = VSUB(TI, TJ);
Chris@42 97 TM = VADD(TI, TJ);
Chris@42 98 {
Chris@42 99 V TC, TE, Tj, Tv;
Chris@42 100 TC = VMUL(LDK(KP866025403), VSUB(TA, TB));
Chris@42 101 TE = VMUL(LDK(KP866025403), VADD(TB, TA));
Chris@42 102 Tj = VFMA(LDK(KP866025403), Ti, Tb);
Chris@42 103 Tv = VFNMS(LDK(KP866025403), Ti, Tb);
Chris@42 104 {
Chris@42 105 V Tz, TD, Tu, Tw;
Chris@42 106 Tz = VSUB(Tx, Ty);
Chris@42 107 TD = VADD(Tx, Ty);
Chris@42 108 Tu = VFNMS(LDK(KP866025403), Tt, Tq);
Chris@42 109 Tw = VFMA(LDK(KP866025403), Tt, Tq);
Chris@42 110 {
Chris@42 111 V TN, TO, TP, TQ;
Chris@42 112 TN = VADD(TL, TM);
Chris@42 113 STM2(&(xo[0]), TN, ovs, &(xo[0]));
Chris@42 114 TO = VSUB(TL, TM);
Chris@42 115 STM2(&(xo[12]), TO, ovs, &(xo[0]));
Chris@42 116 TP = VFMAI(TK, TH);
Chris@42 117 STM2(&(xo[6]), TP, ovs, &(xo[2]));
Chris@42 118 TQ = VFNMSI(TK, TH);
Chris@42 119 STM2(&(xo[18]), TQ, ovs, &(xo[2]));
Chris@42 120 {
Chris@42 121 V TR, TS, TT, TU;
Chris@42 122 TR = VFMAI(TE, TD);
Chris@42 123 STM2(&(xo[8]), TR, ovs, &(xo[0]));
Chris@42 124 TS = VFNMSI(TE, TD);
Chris@42 125 STM2(&(xo[16]), TS, ovs, &(xo[0]));
Chris@42 126 STN2(&(xo[16]), TS, TQ, ovs);
Chris@42 127 TT = VFNMSI(TC, Tz);
Chris@42 128 STM2(&(xo[20]), TT, ovs, &(xo[0]));
Chris@42 129 TU = VFMAI(TC, Tz);
Chris@42 130 STM2(&(xo[4]), TU, ovs, &(xo[0]));
Chris@42 131 STN2(&(xo[4]), TU, TP, ovs);
Chris@42 132 {
Chris@42 133 V TV, TW, TX, TY;
Chris@42 134 TV = VFNMSI(Tw, Tv);
Chris@42 135 STM2(&(xo[10]), TV, ovs, &(xo[2]));
Chris@42 136 STN2(&(xo[8]), TR, TV, ovs);
Chris@42 137 TW = VFMAI(Tw, Tv);
Chris@42 138 STM2(&(xo[14]), TW, ovs, &(xo[2]));
Chris@42 139 STN2(&(xo[12]), TO, TW, ovs);
Chris@42 140 TX = VFMAI(Tu, Tj);
Chris@42 141 STM2(&(xo[22]), TX, ovs, &(xo[2]));
Chris@42 142 STN2(&(xo[20]), TT, TX, ovs);
Chris@42 143 TY = VFNMSI(Tu, Tj);
Chris@42 144 STM2(&(xo[2]), TY, ovs, &(xo[2]));
Chris@42 145 STN2(&(xo[0]), TN, TY, ovs);
Chris@42 146 }
Chris@42 147 }
Chris@42 148 }
Chris@42 149 }
Chris@42 150 }
Chris@42 151 }
Chris@42 152 }
Chris@42 153 }
Chris@42 154 }
Chris@42 155 }
Chris@42 156 VLEAVE();
Chris@42 157 }
Chris@42 158
Chris@42 159 static const kdft_desc desc = { 12, XSIMD_STRING("n2fv_12"), {30, 2, 18, 0}, &GENUS, 0, 2, 0, 0 };
Chris@42 160
Chris@42 161 void XSIMD(codelet_n2fv_12) (planner *p) {
Chris@42 162 X(kdft_register) (p, n2fv_12, &desc);
Chris@42 163 }
Chris@42 164
Chris@42 165 #else /* HAVE_FMA */
Chris@42 166
Chris@42 167 /* Generated by: ../../../genfft/gen_notw_c.native -simd -compact -variables 4 -pipeline-latency 8 -n 12 -name n2fv_12 -with-ostride 2 -include n2f.h -store-multiple 2 */
Chris@42 168
Chris@42 169 /*
Chris@42 170 * This function contains 48 FP additions, 8 FP multiplications,
Chris@42 171 * (or, 44 additions, 4 multiplications, 4 fused multiply/add),
Chris@42 172 * 33 stack variables, 2 constants, and 30 memory accesses
Chris@42 173 */
Chris@42 174 #include "n2f.h"
Chris@42 175
Chris@42 176 static void n2fv_12(const R *ri, const R *ii, R *ro, R *io, stride is, stride os, INT v, INT ivs, INT ovs)
Chris@42 177 {
Chris@42 178 DVK(KP500000000, +0.500000000000000000000000000000000000000000000);
Chris@42 179 DVK(KP866025403, +0.866025403784438646763723170752936183471402627);
Chris@42 180 {
Chris@42 181 INT i;
Chris@42 182 const R *xi;
Chris@42 183 R *xo;
Chris@42 184 xi = ri;
Chris@42 185 xo = ro;
Chris@42 186 for (i = v; i > 0; i = i - VL, xi = xi + (VL * ivs), xo = xo + (VL * ovs), MAKE_VOLATILE_STRIDE(24, is), MAKE_VOLATILE_STRIDE(24, os)) {
Chris@42 187 V T5, Ta, TJ, Ty, Tq, Tp, Tg, Tl, TI, TA, Tz, Tu;
Chris@42 188 {
Chris@42 189 V T1, T6, T4, Tw, T9, Tx;
Chris@42 190 T1 = LD(&(xi[0]), ivs, &(xi[0]));
Chris@42 191 T6 = LD(&(xi[WS(is, 6)]), ivs, &(xi[0]));
Chris@42 192 {
Chris@42 193 V T2, T3, T7, T8;
Chris@42 194 T2 = LD(&(xi[WS(is, 4)]), ivs, &(xi[0]));
Chris@42 195 T3 = LD(&(xi[WS(is, 8)]), ivs, &(xi[0]));
Chris@42 196 T4 = VADD(T2, T3);
Chris@42 197 Tw = VSUB(T3, T2);
Chris@42 198 T7 = LD(&(xi[WS(is, 10)]), ivs, &(xi[0]));
Chris@42 199 T8 = LD(&(xi[WS(is, 2)]), ivs, &(xi[0]));
Chris@42 200 T9 = VADD(T7, T8);
Chris@42 201 Tx = VSUB(T8, T7);
Chris@42 202 }
Chris@42 203 T5 = VADD(T1, T4);
Chris@42 204 Ta = VADD(T6, T9);
Chris@42 205 TJ = VADD(Tw, Tx);
Chris@42 206 Ty = VMUL(LDK(KP866025403), VSUB(Tw, Tx));
Chris@42 207 Tq = VFNMS(LDK(KP500000000), T9, T6);
Chris@42 208 Tp = VFNMS(LDK(KP500000000), T4, T1);
Chris@42 209 }
Chris@42 210 {
Chris@42 211 V Tc, Th, Tf, Ts, Tk, Tt;
Chris@42 212 Tc = LD(&(xi[WS(is, 3)]), ivs, &(xi[WS(is, 1)]));
Chris@42 213 Th = LD(&(xi[WS(is, 9)]), ivs, &(xi[WS(is, 1)]));
Chris@42 214 {
Chris@42 215 V Td, Te, Ti, Tj;
Chris@42 216 Td = LD(&(xi[WS(is, 7)]), ivs, &(xi[WS(is, 1)]));
Chris@42 217 Te = LD(&(xi[WS(is, 11)]), ivs, &(xi[WS(is, 1)]));
Chris@42 218 Tf = VADD(Td, Te);
Chris@42 219 Ts = VSUB(Te, Td);
Chris@42 220 Ti = LD(&(xi[WS(is, 1)]), ivs, &(xi[WS(is, 1)]));
Chris@42 221 Tj = LD(&(xi[WS(is, 5)]), ivs, &(xi[WS(is, 1)]));
Chris@42 222 Tk = VADD(Ti, Tj);
Chris@42 223 Tt = VSUB(Tj, Ti);
Chris@42 224 }
Chris@42 225 Tg = VADD(Tc, Tf);
Chris@42 226 Tl = VADD(Th, Tk);
Chris@42 227 TI = VADD(Ts, Tt);
Chris@42 228 TA = VFNMS(LDK(KP500000000), Tk, Th);
Chris@42 229 Tz = VFNMS(LDK(KP500000000), Tf, Tc);
Chris@42 230 Tu = VMUL(LDK(KP866025403), VSUB(Ts, Tt));
Chris@42 231 }
Chris@42 232 {
Chris@42 233 V TN, TO, TP, TQ, TR, TS;
Chris@42 234 {
Chris@42 235 V Tb, Tm, Tn, To;
Chris@42 236 Tb = VSUB(T5, Ta);
Chris@42 237 Tm = VBYI(VSUB(Tg, Tl));
Chris@42 238 TN = VSUB(Tb, Tm);
Chris@42 239 STM2(&(xo[18]), TN, ovs, &(xo[2]));
Chris@42 240 TO = VADD(Tb, Tm);
Chris@42 241 STM2(&(xo[6]), TO, ovs, &(xo[2]));
Chris@42 242 Tn = VADD(T5, Ta);
Chris@42 243 To = VADD(Tg, Tl);
Chris@42 244 TP = VSUB(Tn, To);
Chris@42 245 STM2(&(xo[12]), TP, ovs, &(xo[0]));
Chris@42 246 TQ = VADD(Tn, To);
Chris@42 247 STM2(&(xo[0]), TQ, ovs, &(xo[0]));
Chris@42 248 }
Chris@42 249 {
Chris@42 250 V Tv, TE, TC, TD, Tr, TB, TT, TU;
Chris@42 251 Tr = VSUB(Tp, Tq);
Chris@42 252 Tv = VSUB(Tr, Tu);
Chris@42 253 TE = VADD(Tr, Tu);
Chris@42 254 TB = VSUB(Tz, TA);
Chris@42 255 TC = VBYI(VADD(Ty, TB));
Chris@42 256 TD = VBYI(VSUB(Ty, TB));
Chris@42 257 TR = VSUB(Tv, TC);
Chris@42 258 STM2(&(xo[10]), TR, ovs, &(xo[2]));
Chris@42 259 TS = VSUB(TE, TD);
Chris@42 260 STM2(&(xo[22]), TS, ovs, &(xo[2]));
Chris@42 261 TT = VADD(TC, Tv);
Chris@42 262 STM2(&(xo[14]), TT, ovs, &(xo[2]));
Chris@42 263 STN2(&(xo[12]), TP, TT, ovs);
Chris@42 264 TU = VADD(TD, TE);
Chris@42 265 STM2(&(xo[2]), TU, ovs, &(xo[2]));
Chris@42 266 STN2(&(xo[0]), TQ, TU, ovs);
Chris@42 267 }
Chris@42 268 {
Chris@42 269 V TK, TM, TH, TL, TF, TG;
Chris@42 270 TK = VBYI(VMUL(LDK(KP866025403), VSUB(TI, TJ)));
Chris@42 271 TM = VBYI(VMUL(LDK(KP866025403), VADD(TJ, TI)));
Chris@42 272 TF = VADD(Tp, Tq);
Chris@42 273 TG = VADD(Tz, TA);
Chris@42 274 TH = VSUB(TF, TG);
Chris@42 275 TL = VADD(TF, TG);
Chris@42 276 {
Chris@42 277 V TV, TW, TX, TY;
Chris@42 278 TV = VSUB(TH, TK);
Chris@42 279 STM2(&(xo[20]), TV, ovs, &(xo[0]));
Chris@42 280 STN2(&(xo[20]), TV, TS, ovs);
Chris@42 281 TW = VADD(TL, TM);
Chris@42 282 STM2(&(xo[8]), TW, ovs, &(xo[0]));
Chris@42 283 STN2(&(xo[8]), TW, TR, ovs);
Chris@42 284 TX = VADD(TH, TK);
Chris@42 285 STM2(&(xo[4]), TX, ovs, &(xo[0]));
Chris@42 286 STN2(&(xo[4]), TX, TO, ovs);
Chris@42 287 TY = VSUB(TL, TM);
Chris@42 288 STM2(&(xo[16]), TY, ovs, &(xo[0]));
Chris@42 289 STN2(&(xo[16]), TY, TN, ovs);
Chris@42 290 }
Chris@42 291 }
Chris@42 292 }
Chris@42 293 }
Chris@42 294 }
Chris@42 295 VLEAVE();
Chris@42 296 }
Chris@42 297
Chris@42 298 static const kdft_desc desc = { 12, XSIMD_STRING("n2fv_12"), {44, 4, 4, 0}, &GENUS, 0, 2, 0, 0 };
Chris@42 299
Chris@42 300 void XSIMD(codelet_n2fv_12) (planner *p) {
Chris@42 301 X(kdft_register) (p, n2fv_12, &desc);
Chris@42 302 }
Chris@42 303
Chris@42 304 #endif /* HAVE_FMA */