Mercurial > hg > sv-dependency-builds

annotate src/fftw-3.3.8/dft/simd/common/n2bv_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 08:05:11 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 -sign 1 -n 10 -name n2bv_10 -with-ostride 2 -include dft/simd/n2b.h -store-multiple 2 */
Chris@82 29
Chris@82 30 /*
Chris@82 31 * This function contains 42 FP additions, 22 FP multiplications,
Chris@82 32 * (or, 24 additions, 4 multiplications, 18 fused multiply/add),
Chris@82 33 * 36 stack variables, 4 constants, and 25 memory accesses
Chris@82 34 */
Chris@82 35 #include "dft/simd/n2b.h"
Chris@82 36
Chris@82 37 static void n2bv_10(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(KP559016994, +0.559016994374947424102293417182819058860154590);
Chris@82 40 DVK(KP250000000, +0.250000000000000000000000000000000000000000000);
Chris@82 41 DVK(KP618033988, +0.618033988749894848204586834365638117720309180);
Chris@82 42 DVK(KP951056516, +0.951056516295153572116439333379382143405698634);
Chris@82 43 {
Chris@82 44 INT i;
Chris@82 45 const R *xi;
Chris@82 46 R *xo;
Chris@82 47 xi = ii;
Chris@82 48 xo = io;
Chris@82 49 for (i = v; i > 0; i = i - VL, xi = xi + (VL * ivs), xo = xo + (VL * ovs), MAKE_VOLATILE_STRIDE(20, is), MAKE_VOLATILE_STRIDE(20, os)) {
Chris@82 50 V T3, Tr, Tm, Tn, TD, TC, Tu, Tx, Ty, Ta, Th, Ti, T1, T2;
Chris@82 51 T1 = LD(&(xi[0]), ivs, &(xi[0]));
Chris@82 52 T2 = LD(&(xi[WS(is, 5)]), ivs, &(xi[WS(is, 1)]));
Chris@82 53 T3 = VSUB(T1, T2);
Chris@82 54 Tr = VADD(T1, T2);
Chris@82 55 {
Chris@82 56 V T6, Ts, Tg, Tw, T9, Tt, Td, Tv;
Chris@82 57 {
Chris@82 58 V T4, T5, Te, Tf;
Chris@82 59 T4 = LD(&(xi[WS(is, 2)]), ivs, &(xi[0]));
Chris@82 60 T5 = LD(&(xi[WS(is, 7)]), ivs, &(xi[WS(is, 1)]));
Chris@82 61 T6 = VSUB(T4, T5);
Chris@82 62 Ts = VADD(T4, T5);
Chris@82 63 Te = LD(&(xi[WS(is, 6)]), ivs, &(xi[0]));
Chris@82 64 Tf = LD(&(xi[WS(is, 1)]), ivs, &(xi[WS(is, 1)]));
Chris@82 65 Tg = VSUB(Te, Tf);
Chris@82 66 Tw = VADD(Te, Tf);
Chris@82 67 }
Chris@82 68 {
Chris@82 69 V T7, T8, Tb, Tc;
Chris@82 70 T7 = LD(&(xi[WS(is, 8)]), ivs, &(xi[0]));
Chris@82 71 T8 = LD(&(xi[WS(is, 3)]), ivs, &(xi[WS(is, 1)]));
Chris@82 72 T9 = VSUB(T7, T8);
Chris@82 73 Tt = VADD(T7, T8);
Chris@82 74 Tb = LD(&(xi[WS(is, 4)]), ivs, &(xi[0]));
Chris@82 75 Tc = LD(&(xi[WS(is, 9)]), ivs, &(xi[WS(is, 1)]));
Chris@82 76 Td = VSUB(Tb, Tc);
Chris@82 77 Tv = VADD(Tb, Tc);
Chris@82 78 }
Chris@82 79 Tm = VSUB(T6, T9);
Chris@82 80 Tn = VSUB(Td, Tg);
Chris@82 81 TD = VSUB(Ts, Tt);
Chris@82 82 TC = VSUB(Tv, Tw);
Chris@82 83 Tu = VADD(Ts, Tt);
Chris@82 84 Tx = VADD(Tv, Tw);
Chris@82 85 Ty = VADD(Tu, Tx);
Chris@82 86 Ta = VADD(T6, T9);
Chris@82 87 Th = VADD(Td, Tg);
Chris@82 88 Ti = VADD(Ta, Th);
Chris@82 89 }
Chris@82 90 {
Chris@82 91 V TH, TI, TK, TL, TM;
Chris@82 92 TH = VADD(T3, Ti);
Chris@82 93 STM2(&(xo[10]), TH, ovs, &(xo[2]));
Chris@82 94 TI = VADD(Tr, Ty);
Chris@82 95 STM2(&(xo[0]), TI, ovs, &(xo[0]));
Chris@82 96 {
Chris@82 97 V To, Tq, Tl, Tp, Tj, Tk, TJ;
Chris@82 98 To = VMUL(LDK(KP951056516), VFMA(LDK(KP618033988), Tn, Tm));
Chris@82 99 Tq = VMUL(LDK(KP951056516), VFNMS(LDK(KP618033988), Tm, Tn));
Chris@82 100 Tj = VFNMS(LDK(KP250000000), Ti, T3);
Chris@82 101 Tk = VSUB(Ta, Th);
Chris@82 102 Tl = VFMA(LDK(KP559016994), Tk, Tj);
Chris@82 103 Tp = VFNMS(LDK(KP559016994), Tk, Tj);
Chris@82 104 TJ = VFMAI(To, Tl);
Chris@82 105 STM2(&(xo[2]), TJ, ovs, &(xo[2]));
Chris@82 106 STN2(&(xo[0]), TI, TJ, ovs);
Chris@82 107 TK = VFNMSI(Tq, Tp);
Chris@82 108 STM2(&(xo[14]), TK, ovs, &(xo[2]));
Chris@82 109 TL = VFNMSI(To, Tl);
Chris@82 110 STM2(&(xo[18]), TL, ovs, &(xo[2]));
Chris@82 111 TM = VFMAI(Tq, Tp);
Chris@82 112 STM2(&(xo[6]), TM, ovs, &(xo[2]));
Chris@82 113 }
Chris@82 114 {
Chris@82 115 V TE, TG, TB, TF, Tz, TA;
Chris@82 116 TE = VMUL(LDK(KP951056516), VFNMS(LDK(KP618033988), TD, TC));
Chris@82 117 TG = VMUL(LDK(KP951056516), VFMA(LDK(KP618033988), TC, TD));
Chris@82 118 Tz = VFNMS(LDK(KP250000000), Ty, Tr);
Chris@82 119 TA = VSUB(Tu, Tx);
Chris@82 120 TB = VFNMS(LDK(KP559016994), TA, Tz);
Chris@82 121 TF = VFMA(LDK(KP559016994), TA, Tz);
Chris@82 122 {
Chris@82 123 V TN, TO, TP, TQ;
Chris@82 124 TN = VFNMSI(TE, TB);
Chris@82 125 STM2(&(xo[4]), TN, ovs, &(xo[0]));
Chris@82 126 STN2(&(xo[4]), TN, TM, ovs);
Chris@82 127 TO = VFMAI(TG, TF);
Chris@82 128 STM2(&(xo[12]), TO, ovs, &(xo[0]));
Chris@82 129 STN2(&(xo[12]), TO, TK, ovs);
Chris@82 130 TP = VFMAI(TE, TB);
Chris@82 131 STM2(&(xo[16]), TP, ovs, &(xo[0]));
Chris@82 132 STN2(&(xo[16]), TP, TL, ovs);
Chris@82 133 TQ = VFNMSI(TG, TF);
Chris@82 134 STM2(&(xo[8]), TQ, ovs, &(xo[0]));
Chris@82 135 STN2(&(xo[8]), TQ, TH, ovs);
Chris@82 136 }
Chris@82 137 }
Chris@82 138 }
Chris@82 139 }
Chris@82 140 }
Chris@82 141 VLEAVE();
Chris@82 142 }
Chris@82 143
Chris@82 144 static const kdft_desc desc = { 10, XSIMD_STRING("n2bv_10"), {24, 4, 18, 0}, &GENUS, 0, 2, 0, 0 };
Chris@82 145
Chris@82 146 void XSIMD(codelet_n2bv_10) (planner *p) {
Chris@82 147 X(kdft_register) (p, n2bv_10, &desc);
Chris@82 148 }
Chris@82 149
Chris@82 150 #else
Chris@82 151
Chris@82 152 /* Generated by: ../../../genfft/gen_notw_c.native -simd -compact -variables 4 -pipeline-latency 8 -sign 1 -n 10 -name n2bv_10 -with-ostride 2 -include dft/simd/n2b.h -store-multiple 2 */
Chris@82 153
Chris@82 154 /*
Chris@82 155 * This function contains 42 FP additions, 12 FP multiplications,
Chris@82 156 * (or, 36 additions, 6 multiplications, 6 fused multiply/add),
Chris@82 157 * 36 stack variables, 4 constants, and 25 memory accesses
Chris@82 158 */
Chris@82 159 #include "dft/simd/n2b.h"
Chris@82 160
Chris@82 161 static void n2bv_10(const R *ri, const R *ii, R *ro, R *io, stride is, stride os, INT v, INT ivs, INT ovs)
Chris@82 162 {
Chris@82 163 DVK(KP250000000, +0.250000000000000000000000000000000000000000000);
Chris@82 164 DVK(KP559016994, +0.559016994374947424102293417182819058860154590);
Chris@82 165 DVK(KP587785252, +0.587785252292473129168705954639072768597652438);
Chris@82 166 DVK(KP951056516, +0.951056516295153572116439333379382143405698634);
Chris@82 167 {
Chris@82 168 INT i;
Chris@82 169 const R *xi;
Chris@82 170 R *xo;
Chris@82 171 xi = ii;
Chris@82 172 xo = io;
Chris@82 173 for (i = v; i > 0; i = i - VL, xi = xi + (VL * ivs), xo = xo + (VL * ovs), MAKE_VOLATILE_STRIDE(20, is), MAKE_VOLATILE_STRIDE(20, os)) {
Chris@82 174 V Tl, Ty, T7, Te, Tw, Tt, Tz, TA, TB, Tg, Th, Tm, Tj, Tk;
Chris@82 175 Tj = LD(&(xi[0]), ivs, &(xi[0]));
Chris@82 176 Tk = LD(&(xi[WS(is, 5)]), ivs, &(xi[WS(is, 1)]));
Chris@82 177 Tl = VSUB(Tj, Tk);
Chris@82 178 Ty = VADD(Tj, Tk);
Chris@82 179 {
Chris@82 180 V T3, Tr, Td, Tv, T6, Ts, Ta, Tu;
Chris@82 181 {
Chris@82 182 V T1, T2, Tb, Tc;
Chris@82 183 T1 = LD(&(xi[WS(is, 2)]), ivs, &(xi[0]));
Chris@82 184 T2 = LD(&(xi[WS(is, 7)]), ivs, &(xi[WS(is, 1)]));
Chris@82 185 T3 = VSUB(T1, T2);
Chris@82 186 Tr = VADD(T1, T2);
Chris@82 187 Tb = LD(&(xi[WS(is, 6)]), ivs, &(xi[0]));
Chris@82 188 Tc = LD(&(xi[WS(is, 1)]), ivs, &(xi[WS(is, 1)]));
Chris@82 189 Td = VSUB(Tb, Tc);
Chris@82 190 Tv = VADD(Tb, Tc);
Chris@82 191 }
Chris@82 192 {
Chris@82 193 V T4, T5, T8, T9;
Chris@82 194 T4 = LD(&(xi[WS(is, 8)]), ivs, &(xi[0]));
Chris@82 195 T5 = LD(&(xi[WS(is, 3)]), ivs, &(xi[WS(is, 1)]));
Chris@82 196 T6 = VSUB(T4, T5);
Chris@82 197 Ts = VADD(T4, T5);
Chris@82 198 T8 = LD(&(xi[WS(is, 4)]), ivs, &(xi[0]));
Chris@82 199 T9 = LD(&(xi[WS(is, 9)]), ivs, &(xi[WS(is, 1)]));
Chris@82 200 Ta = VSUB(T8, T9);
Chris@82 201 Tu = VADD(T8, T9);
Chris@82 202 }
Chris@82 203 T7 = VSUB(T3, T6);
Chris@82 204 Te = VSUB(Ta, Td);
Chris@82 205 Tw = VSUB(Tu, Tv);
Chris@82 206 Tt = VSUB(Tr, Ts);
Chris@82 207 Tz = VADD(Tr, Ts);
Chris@82 208 TA = VADD(Tu, Tv);
Chris@82 209 TB = VADD(Tz, TA);
Chris@82 210 Tg = VADD(T3, T6);
Chris@82 211 Th = VADD(Ta, Td);
Chris@82 212 Tm = VADD(Tg, Th);
Chris@82 213 }
Chris@82 214 {
Chris@82 215 V TH, TI, TK, TL, TM;
Chris@82 216 TH = VADD(Tl, Tm);
Chris@82 217 STM2(&(xo[10]), TH, ovs, &(xo[2]));
Chris@82 218 TI = VADD(Ty, TB);
Chris@82 219 STM2(&(xo[0]), TI, ovs, &(xo[0]));
Chris@82 220 {
Chris@82 221 V Tf, Tq, To, Tp, Ti, Tn, TJ;
Chris@82 222 Tf = VBYI(VFMA(LDK(KP951056516), T7, VMUL(LDK(KP587785252), Te)));
Chris@82 223 Tq = VBYI(VFNMS(LDK(KP951056516), Te, VMUL(LDK(KP587785252), T7)));
Chris@82 224 Ti = VMUL(LDK(KP559016994), VSUB(Tg, Th));
Chris@82 225 Tn = VFNMS(LDK(KP250000000), Tm, Tl);
Chris@82 226 To = VADD(Ti, Tn);
Chris@82 227 Tp = VSUB(Tn, Ti);
Chris@82 228 TJ = VADD(Tf, To);
Chris@82 229 STM2(&(xo[2]), TJ, ovs, &(xo[2]));
Chris@82 230 STN2(&(xo[0]), TI, TJ, ovs);
Chris@82 231 TK = VADD(Tq, Tp);
Chris@82 232 STM2(&(xo[14]), TK, ovs, &(xo[2]));
Chris@82 233 TL = VSUB(To, Tf);
Chris@82 234 STM2(&(xo[18]), TL, ovs, &(xo[2]));
Chris@82 235 TM = VSUB(Tp, Tq);
Chris@82 236 STM2(&(xo[6]), TM, ovs, &(xo[2]));
Chris@82 237 }
Chris@82 238 {
Chris@82 239 V Tx, TG, TE, TF, TC, TD;
Chris@82 240 Tx = VBYI(VFNMS(LDK(KP951056516), Tw, VMUL(LDK(KP587785252), Tt)));
Chris@82 241 TG = VBYI(VFMA(LDK(KP951056516), Tt, VMUL(LDK(KP587785252), Tw)));
Chris@82 242 TC = VFNMS(LDK(KP250000000), TB, Ty);
Chris@82 243 TD = VMUL(LDK(KP559016994), VSUB(Tz, TA));
Chris@82 244 TE = VSUB(TC, TD);
Chris@82 245 TF = VADD(TD, TC);
Chris@82 246 {
Chris@82 247 V TN, TO, TP, TQ;
Chris@82 248 TN = VADD(Tx, TE);
Chris@82 249 STM2(&(xo[4]), TN, ovs, &(xo[0]));
Chris@82 250 STN2(&(xo[4]), TN, TM, ovs);
Chris@82 251 TO = VADD(TG, TF);
Chris@82 252 STM2(&(xo[12]), TO, ovs, &(xo[0]));
Chris@82 253 STN2(&(xo[12]), TO, TK, ovs);
Chris@82 254 TP = VSUB(TE, Tx);
Chris@82 255 STM2(&(xo[16]), TP, ovs, &(xo[0]));
Chris@82 256 STN2(&(xo[16]), TP, TL, ovs);
Chris@82 257 TQ = VSUB(TF, TG);
Chris@82 258 STM2(&(xo[8]), TQ, ovs, &(xo[0]));
Chris@82 259 STN2(&(xo[8]), TQ, TH, ovs);
Chris@82 260 }
Chris@82 261 }
Chris@82 262 }
Chris@82 263 }
Chris@82 264 }
Chris@82 265 VLEAVE();
Chris@82 266 }
Chris@82 267
Chris@82 268 static const kdft_desc desc = { 10, XSIMD_STRING("n2bv_10"), {36, 6, 6, 0}, &GENUS, 0, 2, 0, 0 };
Chris@82 269
Chris@82 270 void XSIMD(codelet_n2bv_10) (planner *p) {
Chris@82 271 X(kdft_register) (p, n2bv_10, &desc);
Chris@82 272 }
Chris@82 273
Chris@82 274 #endif