Mercurial > hg > sv-dependency-builds

annotate src/fftw-3.3.8/rdft/scalar/r2cb/r2cb_16.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 08:07:29 EDT 2018 */
Chris@82 23
Chris@82 24 #include "rdft/codelet-rdft.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_r2cb.native -fma -compact -variables 4 -pipeline-latency 4 -sign 1 -n 16 -name r2cb_16 -include rdft/scalar/r2cb.h */
Chris@82 29
Chris@82 30 /*
Chris@82 31 * This function contains 58 FP additions, 32 FP multiplications,
Chris@82 32 * (or, 26 additions, 0 multiplications, 32 fused multiply/add),
Chris@82 33 * 31 stack variables, 4 constants, and 32 memory accesses
Chris@82 34 */
Chris@82 35 #include "rdft/scalar/r2cb.h"
Chris@82 36
Chris@82 37 static void r2cb_16(R *R0, R *R1, R *Cr, R *Ci, stride rs, stride csr, stride csi, INT v, INT ivs, INT ovs)
Chris@82 38 {
Chris@82 39 DK(KP1_847759065, +1.847759065022573512256366378793576573644833252);
Chris@82 40 DK(KP414213562, +0.414213562373095048801688724209698078569671875);
Chris@82 41 DK(KP1_414213562, +1.414213562373095048801688724209698078569671875);
Chris@82 42 DK(KP2_000000000, +2.000000000000000000000000000000000000000000000);
Chris@82 43 {
Chris@82 44 INT i;
Chris@82 45 for (i = v; i > 0; i = i - 1, R0 = R0 + ovs, R1 = R1 + ovs, Cr = Cr + ivs, Ci = Ci + ivs, MAKE_VOLATILE_STRIDE(64, rs), MAKE_VOLATILE_STRIDE(64, csr), MAKE_VOLATILE_STRIDE(64, csi)) {
Chris@82 46 E T5, TL, Tj, TD, T8, TM, To, TE, Tc, TP, Tf, TQ, Tu, Tz, TR;
Chris@82 47 E TO, TH, TG;
Chris@82 48 {
Chris@82 49 E T4, Ti, T3, Th, T1, T2;
Chris@82 50 T4 = Cr[WS(csr, 4)];
Chris@82 51 Ti = Ci[WS(csi, 4)];
Chris@82 52 T1 = Cr[0];
Chris@82 53 T2 = Cr[WS(csr, 8)];
Chris@82 54 T3 = T1 + T2;
Chris@82 55 Th = T1 - T2;
Chris@82 56 T5 = FMA(KP2_000000000, T4, T3);
Chris@82 57 TL = FNMS(KP2_000000000, T4, T3);
Chris@82 58 Tj = FNMS(KP2_000000000, Ti, Th);
Chris@82 59 TD = FMA(KP2_000000000, Ti, Th);
Chris@82 60 }
Chris@82 61 {
Chris@82 62 E T6, T7, Tk, Tl, Tm, Tn;
Chris@82 63 T6 = Cr[WS(csr, 2)];
Chris@82 64 T7 = Cr[WS(csr, 6)];
Chris@82 65 Tk = T6 - T7;
Chris@82 66 Tl = Ci[WS(csi, 2)];
Chris@82 67 Tm = Ci[WS(csi, 6)];
Chris@82 68 Tn = Tl + Tm;
Chris@82 69 T8 = T6 + T7;
Chris@82 70 TM = Tl - Tm;
Chris@82 71 To = Tk - Tn;
Chris@82 72 TE = Tk + Tn;
Chris@82 73 }
Chris@82 74 {
Chris@82 75 E Tq, Ty, Tv, Tt;
Chris@82 76 {
Chris@82 77 E Ta, Tb, Tw, Tx;
Chris@82 78 Ta = Cr[WS(csr, 1)];
Chris@82 79 Tb = Cr[WS(csr, 7)];
Chris@82 80 Tc = Ta + Tb;
Chris@82 81 Tq = Ta - Tb;
Chris@82 82 Tw = Ci[WS(csi, 1)];
Chris@82 83 Tx = Ci[WS(csi, 7)];
Chris@82 84 Ty = Tw + Tx;
Chris@82 85 TP = Tw - Tx;
Chris@82 86 }
Chris@82 87 {
Chris@82 88 E Td, Te, Tr, Ts;
Chris@82 89 Td = Cr[WS(csr, 5)];
Chris@82 90 Te = Cr[WS(csr, 3)];
Chris@82 91 Tf = Td + Te;
Chris@82 92 Tv = Td - Te;
Chris@82 93 Tr = Ci[WS(csi, 5)];
Chris@82 94 Ts = Ci[WS(csi, 3)];
Chris@82 95 Tt = Tr + Ts;
Chris@82 96 TQ = Tr - Ts;
Chris@82 97 }
Chris@82 98 Tu = Tq - Tt;
Chris@82 99 Tz = Tv + Ty;
Chris@82 100 TR = TP - TQ;
Chris@82 101 TO = Tc - Tf;
Chris@82 102 TH = Tq + Tt;
Chris@82 103 TG = Ty - Tv;
Chris@82 104 }
Chris@82 105 {
Chris@82 106 E T9, Tg, TT, TU;
Chris@82 107 T9 = FMA(KP2_000000000, T8, T5);
Chris@82 108 Tg = Tc + Tf;
Chris@82 109 R0[WS(rs, 4)] = FNMS(KP2_000000000, Tg, T9);
Chris@82 110 R0[0] = FMA(KP2_000000000, Tg, T9);
Chris@82 111 TT = FMA(KP2_000000000, TM, TL);
Chris@82 112 TU = TO + TR;
Chris@82 113 R0[WS(rs, 3)] = FNMS(KP1_414213562, TU, TT);
Chris@82 114 R0[WS(rs, 7)] = FMA(KP1_414213562, TU, TT);
Chris@82 115 }
Chris@82 116 {
Chris@82 117 E TV, TW, Tp, TA;
Chris@82 118 TV = FNMS(KP2_000000000, T8, T5);
Chris@82 119 TW = TQ + TP;
Chris@82 120 R0[WS(rs, 2)] = FNMS(KP2_000000000, TW, TV);
Chris@82 121 R0[WS(rs, 6)] = FMA(KP2_000000000, TW, TV);
Chris@82 122 Tp = FMA(KP1_414213562, To, Tj);
Chris@82 123 TA = FNMS(KP414213562, Tz, Tu);
Chris@82 124 R1[WS(rs, 4)] = FNMS(KP1_847759065, TA, Tp);
Chris@82 125 R1[0] = FMA(KP1_847759065, TA, Tp);
Chris@82 126 }
Chris@82 127 {
Chris@82 128 E TB, TC, TJ, TK;
Chris@82 129 TB = FNMS(KP1_414213562, To, Tj);
Chris@82 130 TC = FMA(KP414213562, Tu, Tz);
Chris@82 131 R1[WS(rs, 2)] = FNMS(KP1_847759065, TC, TB);
Chris@82 132 R1[WS(rs, 6)] = FMA(KP1_847759065, TC, TB);
Chris@82 133 TJ = FMA(KP1_414213562, TE, TD);
Chris@82 134 TK = FMA(KP414213562, TG, TH);
Chris@82 135 R1[WS(rs, 3)] = FNMS(KP1_847759065, TK, TJ);
Chris@82 136 R1[WS(rs, 7)] = FMA(KP1_847759065, TK, TJ);
Chris@82 137 }
Chris@82 138 {
Chris@82 139 E TN, TS, TF, TI;
Chris@82 140 TN = FNMS(KP2_000000000, TM, TL);
Chris@82 141 TS = TO - TR;
Chris@82 142 R0[WS(rs, 5)] = FNMS(KP1_414213562, TS, TN);
Chris@82 143 R0[WS(rs, 1)] = FMA(KP1_414213562, TS, TN);
Chris@82 144 TF = FNMS(KP1_414213562, TE, TD);
Chris@82 145 TI = FNMS(KP414213562, TH, TG);
Chris@82 146 R1[WS(rs, 1)] = FNMS(KP1_847759065, TI, TF);
Chris@82 147 R1[WS(rs, 5)] = FMA(KP1_847759065, TI, TF);
Chris@82 148 }
Chris@82 149 }
Chris@82 150 }
Chris@82 151 }
Chris@82 152
Chris@82 153 static const kr2c_desc desc = { 16, "r2cb_16", {26, 0, 32, 0}, &GENUS };
Chris@82 154
Chris@82 155 void X(codelet_r2cb_16) (planner *p) {
Chris@82 156 X(kr2c_register) (p, r2cb_16, &desc);
Chris@82 157 }
Chris@82 158
Chris@82 159 #else
Chris@82 160
Chris@82 161 /* Generated by: ../../../genfft/gen_r2cb.native -compact -variables 4 -pipeline-latency 4 -sign 1 -n 16 -name r2cb_16 -include rdft/scalar/r2cb.h */
Chris@82 162
Chris@82 163 /*
Chris@82 164 * This function contains 58 FP additions, 18 FP multiplications,
Chris@82 165 * (or, 54 additions, 14 multiplications, 4 fused multiply/add),
Chris@82 166 * 31 stack variables, 4 constants, and 32 memory accesses
Chris@82 167 */
Chris@82 168 #include "rdft/scalar/r2cb.h"
Chris@82 169
Chris@82 170 static void r2cb_16(R *R0, R *R1, R *Cr, R *Ci, stride rs, stride csr, stride csi, INT v, INT ivs, INT ovs)
Chris@82 171 {
Chris@82 172 DK(KP1_847759065, +1.847759065022573512256366378793576573644833252);
Chris@82 173 DK(KP765366864, +0.765366864730179543456919968060797733522689125);
Chris@82 174 DK(KP1_414213562, +1.414213562373095048801688724209698078569671875);
Chris@82 175 DK(KP2_000000000, +2.000000000000000000000000000000000000000000000);
Chris@82 176 {
Chris@82 177 INT i;
Chris@82 178 for (i = v; i > 0; i = i - 1, R0 = R0 + ovs, R1 = R1 + ovs, Cr = Cr + ivs, Ci = Ci + ivs, MAKE_VOLATILE_STRIDE(64, rs), MAKE_VOLATILE_STRIDE(64, csr), MAKE_VOLATILE_STRIDE(64, csi)) {
Chris@82 179 E T9, TS, Tl, TG, T6, TR, Ti, TD, Td, Tq, Tg, Tt, Tn, Tu, TV;
Chris@82 180 E TU, TN, TK;
Chris@82 181 {
Chris@82 182 E T7, T8, TE, Tj, Tk, TF;
Chris@82 183 T7 = Cr[WS(csr, 2)];
Chris@82 184 T8 = Cr[WS(csr, 6)];
Chris@82 185 TE = T7 - T8;
Chris@82 186 Tj = Ci[WS(csi, 2)];
Chris@82 187 Tk = Ci[WS(csi, 6)];
Chris@82 188 TF = Tj + Tk;
Chris@82 189 T9 = KP2_000000000 * (T7 + T8);
Chris@82 190 TS = KP1_414213562 * (TE + TF);
Chris@82 191 Tl = KP2_000000000 * (Tj - Tk);
Chris@82 192 TG = KP1_414213562 * (TE - TF);
Chris@82 193 }
Chris@82 194 {
Chris@82 195 E T5, TC, T3, TA;
Chris@82 196 {
Chris@82 197 E T4, TB, T1, T2;
Chris@82 198 T4 = Cr[WS(csr, 4)];
Chris@82 199 T5 = KP2_000000000 * T4;
Chris@82 200 TB = Ci[WS(csi, 4)];
Chris@82 201 TC = KP2_000000000 * TB;
Chris@82 202 T1 = Cr[0];
Chris@82 203 T2 = Cr[WS(csr, 8)];
Chris@82 204 T3 = T1 + T2;
Chris@82 205 TA = T1 - T2;
Chris@82 206 }
Chris@82 207 T6 = T3 + T5;
Chris@82 208 TR = TA + TC;
Chris@82 209 Ti = T3 - T5;
Chris@82 210 TD = TA - TC;
Chris@82 211 }
Chris@82 212 {
Chris@82 213 E TI, TM, TL, TJ;
Chris@82 214 {
Chris@82 215 E Tb, Tc, To, Tp;
Chris@82 216 Tb = Cr[WS(csr, 1)];
Chris@82 217 Tc = Cr[WS(csr, 7)];
Chris@82 218 Td = Tb + Tc;
Chris@82 219 TI = Tb - Tc;
Chris@82 220 To = Ci[WS(csi, 1)];
Chris@82 221 Tp = Ci[WS(csi, 7)];
Chris@82 222 Tq = To - Tp;
Chris@82 223 TM = To + Tp;
Chris@82 224 }
Chris@82 225 {
Chris@82 226 E Te, Tf, Tr, Ts;
Chris@82 227 Te = Cr[WS(csr, 5)];
Chris@82 228 Tf = Cr[WS(csr, 3)];
Chris@82 229 Tg = Te + Tf;
Chris@82 230 TL = Te - Tf;
Chris@82 231 Tr = Ci[WS(csi, 5)];
Chris@82 232 Ts = Ci[WS(csi, 3)];
Chris@82 233 Tt = Tr - Ts;
Chris@82 234 TJ = Tr + Ts;
Chris@82 235 }
Chris@82 236 Tn = Td - Tg;
Chris@82 237 Tu = Tq - Tt;
Chris@82 238 TV = TM - TL;
Chris@82 239 TU = TI + TJ;
Chris@82 240 TN = TL + TM;
Chris@82 241 TK = TI - TJ;
Chris@82 242 }
Chris@82 243 {
Chris@82 244 E Ta, Th, TT, TW;
Chris@82 245 Ta = T6 + T9;
Chris@82 246 Th = KP2_000000000 * (Td + Tg);
Chris@82 247 R0[WS(rs, 4)] = Ta - Th;
Chris@82 248 R0[0] = Ta + Th;
Chris@82 249 TT = TR - TS;
Chris@82 250 TW = FNMS(KP1_847759065, TV, KP765366864 * TU);
Chris@82 251 R1[WS(rs, 5)] = TT - TW;
Chris@82 252 R1[WS(rs, 1)] = TT + TW;
Chris@82 253 }
Chris@82 254 {
Chris@82 255 E TX, TY, Tm, Tv;
Chris@82 256 TX = TR + TS;
Chris@82 257 TY = FMA(KP1_847759065, TU, KP765366864 * TV);
Chris@82 258 R1[WS(rs, 3)] = TX - TY;
Chris@82 259 R1[WS(rs, 7)] = TX + TY;
Chris@82 260 Tm = Ti - Tl;
Chris@82 261 Tv = KP1_414213562 * (Tn - Tu);
Chris@82 262 R0[WS(rs, 5)] = Tm - Tv;
Chris@82 263 R0[WS(rs, 1)] = Tm + Tv;
Chris@82 264 }
Chris@82 265 {
Chris@82 266 E Tw, Tx, TH, TO;
Chris@82 267 Tw = Ti + Tl;
Chris@82 268 Tx = KP1_414213562 * (Tn + Tu);
Chris@82 269 R0[WS(rs, 3)] = Tw - Tx;
Chris@82 270 R0[WS(rs, 7)] = Tw + Tx;
Chris@82 271 TH = TD + TG;
Chris@82 272 TO = FNMS(KP765366864, TN, KP1_847759065 * TK);
Chris@82 273 R1[WS(rs, 4)] = TH - TO;
Chris@82 274 R1[0] = TH + TO;
Chris@82 275 }
Chris@82 276 {
Chris@82 277 E TP, TQ, Ty, Tz;
Chris@82 278 TP = TD - TG;
Chris@82 279 TQ = FMA(KP765366864, TK, KP1_847759065 * TN);
Chris@82 280 R1[WS(rs, 2)] = TP - TQ;
Chris@82 281 R1[WS(rs, 6)] = TP + TQ;
Chris@82 282 Ty = T6 - T9;
Chris@82 283 Tz = KP2_000000000 * (Tt + Tq);
Chris@82 284 R0[WS(rs, 2)] = Ty - Tz;
Chris@82 285 R0[WS(rs, 6)] = Ty + Tz;
Chris@82 286 }
Chris@82 287 }
Chris@82 288 }
Chris@82 289 }
Chris@82 290
Chris@82 291 static const kr2c_desc desc = { 16, "r2cb_16", {54, 14, 4, 0}, &GENUS };
Chris@82 292
Chris@82 293 void X(codelet_r2cb_16) (planner *p) {
Chris@82 294 X(kr2c_register) (p, r2cb_16, &desc);
Chris@82 295 }
Chris@82 296
Chris@82 297 #endif