Mercurial > hg > sv-dependency-builds

annotate src/fftw-3.3.8/dft/simd/common/n2fv_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 08:05:07 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_notw_c.native -fma -simd -compact -variables 4 -pipeline-latency 8 -n 12 -name n2fv_12 -with-ostride 2 -include dft/simd/n2f.h -store-multiple 2 */
Chris@82 29
Chris@82 30 /*
Chris@82 31 * This function contains 48 FP additions, 20 FP multiplications,
Chris@82 32 * (or, 30 additions, 2 multiplications, 18 fused multiply/add),
Chris@82 33 * 33 stack variables, 2 constants, and 30 memory accesses
Chris@82 34 */
Chris@82 35 #include "dft/simd/n2f.h"
Chris@82 36
Chris@82 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@82 38 {
Chris@82 39 DVK(KP866025403, +0.866025403784438646763723170752936183471402627);
Chris@82 40 DVK(KP500000000, +0.500000000000000000000000000000000000000000000);
Chris@82 41 {
Chris@82 42 INT i;
Chris@82 43 const R *xi;
Chris@82 44 R *xo;
Chris@82 45 xi = ri;
Chris@82 46 xo = ro;
Chris@82 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@82 48 V T5, Ta, TG, TF, TB, Tt, Ti, Tm, TJ, TI, TA, Tp;
Chris@82 49 {
Chris@82 50 V T1, T6, T4, Tr, T9, Ts;
Chris@82 51 T1 = LD(&(xi[0]), ivs, &(xi[0]));
Chris@82 52 T6 = LD(&(xi[WS(is, 6)]), ivs, &(xi[0]));
Chris@82 53 {
Chris@82 54 V T2, T3, T7, T8;
Chris@82 55 T2 = LD(&(xi[WS(is, 4)]), ivs, &(xi[0]));
Chris@82 56 T3 = LD(&(xi[WS(is, 8)]), ivs, &(xi[0]));
Chris@82 57 T4 = VADD(T2, T3);
Chris@82 58 Tr = VSUB(T3, T2);
Chris@82 59 T7 = LD(&(xi[WS(is, 10)]), ivs, &(xi[0]));
Chris@82 60 T8 = LD(&(xi[WS(is, 2)]), ivs, &(xi[0]));
Chris@82 61 T9 = VADD(T7, T8);
Chris@82 62 Ts = VSUB(T8, T7);
Chris@82 63 }
Chris@82 64 T5 = VFNMS(LDK(KP500000000), T4, T1);
Chris@82 65 Ta = VFNMS(LDK(KP500000000), T9, T6);
Chris@82 66 TG = VADD(T6, T9);
Chris@82 67 TF = VADD(T1, T4);
Chris@82 68 TB = VADD(Tr, Ts);
Chris@82 69 Tt = VSUB(Tr, Ts);
Chris@82 70 }
Chris@82 71 {
Chris@82 72 V Tk, Tn, Te, Tl, Th, To;
Chris@82 73 Tk = LD(&(xi[WS(is, 3)]), ivs, &(xi[WS(is, 1)]));
Chris@82 74 Tn = LD(&(xi[WS(is, 9)]), ivs, &(xi[WS(is, 1)]));
Chris@82 75 {
Chris@82 76 V Tc, Td, Tf, Tg;
Chris@82 77 Tc = LD(&(xi[WS(is, 11)]), ivs, &(xi[WS(is, 1)]));
Chris@82 78 Td = LD(&(xi[WS(is, 7)]), ivs, &(xi[WS(is, 1)]));
Chris@82 79 Te = VSUB(Tc, Td);
Chris@82 80 Tl = VADD(Td, Tc);
Chris@82 81 Tf = LD(&(xi[WS(is, 1)]), ivs, &(xi[WS(is, 1)]));
Chris@82 82 Tg = LD(&(xi[WS(is, 5)]), ivs, &(xi[WS(is, 1)]));
Chris@82 83 Th = VSUB(Tf, Tg);
Chris@82 84 To = VADD(Tf, Tg);
Chris@82 85 }
Chris@82 86 Ti = VADD(Te, Th);
Chris@82 87 Tm = VFNMS(LDK(KP500000000), Tl, Tk);
Chris@82 88 TJ = VADD(Tn, To);
Chris@82 89 TI = VADD(Tk, Tl);
Chris@82 90 TA = VSUB(Te, Th);
Chris@82 91 Tp = VFNMS(LDK(KP500000000), To, Tn);
Chris@82 92 }
Chris@82 93 {
Chris@82 94 V TN, TO, TP, TQ, TT, TU;
Chris@82 95 {
Chris@82 96 V TH, TK, TL, TM;
Chris@82 97 TH = VSUB(TF, TG);
Chris@82 98 TK = VSUB(TI, TJ);
Chris@82 99 TN = VFNMSI(TK, TH);
Chris@82 100 STM2(&(xo[18]), TN, ovs, &(xo[2]));
Chris@82 101 TO = VFMAI(TK, TH);
Chris@82 102 STM2(&(xo[6]), TO, ovs, &(xo[2]));
Chris@82 103 TL = VADD(TF, TG);
Chris@82 104 TM = VADD(TI, TJ);
Chris@82 105 TP = VSUB(TL, TM);
Chris@82 106 STM2(&(xo[12]), TP, ovs, &(xo[0]));
Chris@82 107 TQ = VADD(TL, TM);
Chris@82 108 STM2(&(xo[0]), TQ, ovs, &(xo[0]));
Chris@82 109 }
Chris@82 110 {
Chris@82 111 V Tj, Tv, Tu, Tw, Tb, Tq, TR, TS;
Chris@82 112 Tb = VSUB(T5, Ta);
Chris@82 113 Tj = VFMA(LDK(KP866025403), Ti, Tb);
Chris@82 114 Tv = VFNMS(LDK(KP866025403), Ti, Tb);
Chris@82 115 Tq = VSUB(Tm, Tp);
Chris@82 116 Tu = VFNMS(LDK(KP866025403), Tt, Tq);
Chris@82 117 Tw = VFMA(LDK(KP866025403), Tt, Tq);
Chris@82 118 TR = VFNMSI(Tu, Tj);
Chris@82 119 STM2(&(xo[2]), TR, ovs, &(xo[2]));
Chris@82 120 STN2(&(xo[0]), TQ, TR, ovs);
Chris@82 121 TS = VFMAI(Tw, Tv);
Chris@82 122 STM2(&(xo[14]), TS, ovs, &(xo[2]));
Chris@82 123 STN2(&(xo[12]), TP, TS, ovs);
Chris@82 124 TT = VFMAI(Tu, Tj);
Chris@82 125 STM2(&(xo[22]), TT, ovs, &(xo[2]));
Chris@82 126 TU = VFNMSI(Tw, Tv);
Chris@82 127 STM2(&(xo[10]), TU, ovs, &(xo[2]));
Chris@82 128 }
Chris@82 129 {
Chris@82 130 V TC, TE, Tz, TD, Tx, Ty;
Chris@82 131 TC = VMUL(LDK(KP866025403), VSUB(TA, TB));
Chris@82 132 TE = VMUL(LDK(KP866025403), VADD(TB, TA));
Chris@82 133 Tx = VADD(T5, Ta);
Chris@82 134 Ty = VADD(Tm, Tp);
Chris@82 135 Tz = VSUB(Tx, Ty);
Chris@82 136 TD = VADD(Tx, Ty);
Chris@82 137 {
Chris@82 138 V TV, TW, TX, TY;
Chris@82 139 TV = VFMAI(TC, Tz);
Chris@82 140 STM2(&(xo[4]), TV, ovs, &(xo[0]));
Chris@82 141 STN2(&(xo[4]), TV, TO, ovs);
Chris@82 142 TW = VFNMSI(TE, TD);
Chris@82 143 STM2(&(xo[16]), TW, ovs, &(xo[0]));
Chris@82 144 STN2(&(xo[16]), TW, TN, ovs);
Chris@82 145 TX = VFNMSI(TC, Tz);
Chris@82 146 STM2(&(xo[20]), TX, ovs, &(xo[0]));
Chris@82 147 STN2(&(xo[20]), TX, TT, ovs);
Chris@82 148 TY = VFMAI(TE, TD);
Chris@82 149 STM2(&(xo[8]), TY, ovs, &(xo[0]));
Chris@82 150 STN2(&(xo[8]), TY, TU, ovs);
Chris@82 151 }
Chris@82 152 }
Chris@82 153 }
Chris@82 154 }
Chris@82 155 }
Chris@82 156 VLEAVE();
Chris@82 157 }
Chris@82 158
Chris@82 159 static const kdft_desc desc = { 12, XSIMD_STRING("n2fv_12"), {30, 2, 18, 0}, &GENUS, 0, 2, 0, 0 };
Chris@82 160
Chris@82 161 void XSIMD(codelet_n2fv_12) (planner *p) {
Chris@82 162 X(kdft_register) (p, n2fv_12, &desc);
Chris@82 163 }
Chris@82 164
Chris@82 165 #else
Chris@82 166
Chris@82 167 /* Generated by: ../../../genfft/gen_notw_c.native -simd -compact -variables 4 -pipeline-latency 8 -n 12 -name n2fv_12 -with-ostride 2 -include dft/simd/n2f.h -store-multiple 2 */
Chris@82 168
Chris@82 169 /*
Chris@82 170 * This function contains 48 FP additions, 8 FP multiplications,
Chris@82 171 * (or, 44 additions, 4 multiplications, 4 fused multiply/add),
Chris@82 172 * 33 stack variables, 2 constants, and 30 memory accesses
Chris@82 173 */
Chris@82 174 #include "dft/simd/n2f.h"
Chris@82 175
Chris@82 176 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@82 177 {
Chris@82 178 DVK(KP500000000, +0.500000000000000000000000000000000000000000000);
Chris@82 179 DVK(KP866025403, +0.866025403784438646763723170752936183471402627);
Chris@82 180 {
Chris@82 181 INT i;
Chris@82 182 const R *xi;
Chris@82 183 R *xo;
Chris@82 184 xi = ri;
Chris@82 185 xo = ro;
Chris@82 186 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@82 187 V T5, Ta, TJ, Ty, Tq, Tp, Tg, Tl, TI, TA, Tz, Tu;
Chris@82 188 {
Chris@82 189 V T1, T6, T4, Tw, T9, Tx;
Chris@82 190 T1 = LD(&(xi[0]), ivs, &(xi[0]));
Chris@82 191 T6 = LD(&(xi[WS(is, 6)]), ivs, &(xi[0]));
Chris@82 192 {
Chris@82 193 V T2, T3, T7, T8;
Chris@82 194 T2 = LD(&(xi[WS(is, 4)]), ivs, &(xi[0]));
Chris@82 195 T3 = LD(&(xi[WS(is, 8)]), ivs, &(xi[0]));
Chris@82 196 T4 = VADD(T2, T3);
Chris@82 197 Tw = VSUB(T3, T2);
Chris@82 198 T7 = LD(&(xi[WS(is, 10)]), ivs, &(xi[0]));
Chris@82 199 T8 = LD(&(xi[WS(is, 2)]), ivs, &(xi[0]));
Chris@82 200 T9 = VADD(T7, T8);
Chris@82 201 Tx = VSUB(T8, T7);
Chris@82 202 }
Chris@82 203 T5 = VADD(T1, T4);
Chris@82 204 Ta = VADD(T6, T9);
Chris@82 205 TJ = VADD(Tw, Tx);
Chris@82 206 Ty = VMUL(LDK(KP866025403), VSUB(Tw, Tx));
Chris@82 207 Tq = VFNMS(LDK(KP500000000), T9, T6);
Chris@82 208 Tp = VFNMS(LDK(KP500000000), T4, T1);
Chris@82 209 }
Chris@82 210 {
Chris@82 211 V Tc, Th, Tf, Ts, Tk, Tt;
Chris@82 212 Tc = LD(&(xi[WS(is, 3)]), ivs, &(xi[WS(is, 1)]));
Chris@82 213 Th = LD(&(xi[WS(is, 9)]), ivs, &(xi[WS(is, 1)]));
Chris@82 214 {
Chris@82 215 V Td, Te, Ti, Tj;
Chris@82 216 Td = LD(&(xi[WS(is, 7)]), ivs, &(xi[WS(is, 1)]));
Chris@82 217 Te = LD(&(xi[WS(is, 11)]), ivs, &(xi[WS(is, 1)]));
Chris@82 218 Tf = VADD(Td, Te);
Chris@82 219 Ts = VSUB(Te, Td);
Chris@82 220 Ti = LD(&(xi[WS(is, 1)]), ivs, &(xi[WS(is, 1)]));
Chris@82 221 Tj = LD(&(xi[WS(is, 5)]), ivs, &(xi[WS(is, 1)]));
Chris@82 222 Tk = VADD(Ti, Tj);
Chris@82 223 Tt = VSUB(Tj, Ti);
Chris@82 224 }
Chris@82 225 Tg = VADD(Tc, Tf);
Chris@82 226 Tl = VADD(Th, Tk);
Chris@82 227 TI = VADD(Ts, Tt);
Chris@82 228 TA = VFNMS(LDK(KP500000000), Tk, Th);
Chris@82 229 Tz = VFNMS(LDK(KP500000000), Tf, Tc);
Chris@82 230 Tu = VMUL(LDK(KP866025403), VSUB(Ts, Tt));
Chris@82 231 }
Chris@82 232 {
Chris@82 233 V TN, TO, TP, TQ, TR, TS;
Chris@82 234 {
Chris@82 235 V Tb, Tm, Tn, To;
Chris@82 236 Tb = VSUB(T5, Ta);
Chris@82 237 Tm = VBYI(VSUB(Tg, Tl));
Chris@82 238 TN = VSUB(Tb, Tm);
Chris@82 239 STM2(&(xo[18]), TN, ovs, &(xo[2]));
Chris@82 240 TO = VADD(Tb, Tm);
Chris@82 241 STM2(&(xo[6]), TO, ovs, &(xo[2]));
Chris@82 242 Tn = VADD(T5, Ta);
Chris@82 243 To = VADD(Tg, Tl);
Chris@82 244 TP = VSUB(Tn, To);
Chris@82 245 STM2(&(xo[12]), TP, ovs, &(xo[0]));
Chris@82 246 TQ = VADD(Tn, To);
Chris@82 247 STM2(&(xo[0]), TQ, ovs, &(xo[0]));
Chris@82 248 }
Chris@82 249 {
Chris@82 250 V Tv, TE, TC, TD, Tr, TB, TT, TU;
Chris@82 251 Tr = VSUB(Tp, Tq);
Chris@82 252 Tv = VSUB(Tr, Tu);
Chris@82 253 TE = VADD(Tr, Tu);
Chris@82 254 TB = VSUB(Tz, TA);
Chris@82 255 TC = VBYI(VADD(Ty, TB));
Chris@82 256 TD = VBYI(VSUB(Ty, TB));
Chris@82 257 TR = VSUB(Tv, TC);
Chris@82 258 STM2(&(xo[10]), TR, ovs, &(xo[2]));
Chris@82 259 TS = VSUB(TE, TD);
Chris@82 260 STM2(&(xo[22]), TS, ovs, &(xo[2]));
Chris@82 261 TT = VADD(TC, Tv);
Chris@82 262 STM2(&(xo[14]), TT, ovs, &(xo[2]));
Chris@82 263 STN2(&(xo[12]), TP, TT, ovs);
Chris@82 264 TU = VADD(TD, TE);
Chris@82 265 STM2(&(xo[2]), TU, ovs, &(xo[2]));
Chris@82 266 STN2(&(xo[0]), TQ, TU, ovs);
Chris@82 267 }
Chris@82 268 {
Chris@82 269 V TK, TM, TH, TL, TF, TG;
Chris@82 270 TK = VBYI(VMUL(LDK(KP866025403), VSUB(TI, TJ)));
Chris@82 271 TM = VBYI(VMUL(LDK(KP866025403), VADD(TJ, TI)));
Chris@82 272 TF = VADD(Tp, Tq);
Chris@82 273 TG = VADD(Tz, TA);
Chris@82 274 TH = VSUB(TF, TG);
Chris@82 275 TL = VADD(TF, TG);
Chris@82 276 {
Chris@82 277 V TV, TW, TX, TY;
Chris@82 278 TV = VSUB(TH, TK);
Chris@82 279 STM2(&(xo[20]), TV, ovs, &(xo[0]));
Chris@82 280 STN2(&(xo[20]), TV, TS, ovs);
Chris@82 281 TW = VADD(TL, TM);
Chris@82 282 STM2(&(xo[8]), TW, ovs, &(xo[0]));
Chris@82 283 STN2(&(xo[8]), TW, TR, ovs);
Chris@82 284 TX = VADD(TH, TK);
Chris@82 285 STM2(&(xo[4]), TX, ovs, &(xo[0]));
Chris@82 286 STN2(&(xo[4]), TX, TO, ovs);
Chris@82 287 TY = VSUB(TL, TM);
Chris@82 288 STM2(&(xo[16]), TY, ovs, &(xo[0]));
Chris@82 289 STN2(&(xo[16]), TY, TN, ovs);
Chris@82 290 }
Chris@82 291 }
Chris@82 292 }
Chris@82 293 }
Chris@82 294 }
Chris@82 295 VLEAVE();
Chris@82 296 }
Chris@82 297
Chris@82 298 static const kdft_desc desc = { 12, XSIMD_STRING("n2fv_12"), {44, 4, 4, 0}, &GENUS, 0, 2, 0, 0 };
Chris@82 299
Chris@82 300 void XSIMD(codelet_n2fv_12) (planner *p) {
Chris@82 301 X(kdft_register) (p, n2fv_12, &desc);
Chris@82 302 }
Chris@82 303
Chris@82 304 #endif