Mercurial > hg > sv-dependency-builds

annotate src/fftw-3.3.8/dft/simd/common/n1bv_12.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 08:04:57 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 12 -name n1bv_12 -include dft/simd/n1b.h */
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 * 27 stack variables, 2 constants, and 24 memory accesses
Chris@82 34 */
Chris@82 35 #include "dft/simd/n1b.h"
Chris@82 36
Chris@82 37 static void n1bv_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 = ii;
Chris@82 46 xo = io;
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, TJ, TB, Tq, Tp, Tg, Tl, TG, Ty, Tt, Ts;
Chris@82 49 {
Chris@82 50 V T1, T6, T4, Tz, T9, TA;
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 Tz = VSUB(T2, T3);
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 TA = VSUB(T7, T8);
Chris@82 63 }
Chris@82 64 T5 = VADD(T1, T4);
Chris@82 65 Ta = VADD(T6, T9);
Chris@82 66 TJ = VSUB(Tz, TA);
Chris@82 67 TB = VADD(Tz, TA);
Chris@82 68 Tq = VFNMS(LDK(KP500000000), T9, T6);
Chris@82 69 Tp = VFNMS(LDK(KP500000000), T4, T1);
Chris@82 70 }
Chris@82 71 {
Chris@82 72 V Tc, Th, Tf, Tw, Tk, Tx;
Chris@82 73 Tc = LD(&(xi[WS(is, 3)]), ivs, &(xi[WS(is, 1)]));
Chris@82 74 Th = LD(&(xi[WS(is, 9)]), ivs, &(xi[WS(is, 1)]));
Chris@82 75 {
Chris@82 76 V Td, Te, Ti, Tj;
Chris@82 77 Td = LD(&(xi[WS(is, 7)]), ivs, &(xi[WS(is, 1)]));
Chris@82 78 Te = LD(&(xi[WS(is, 11)]), ivs, &(xi[WS(is, 1)]));
Chris@82 79 Tf = VADD(Td, Te);
Chris@82 80 Tw = VSUB(Td, Te);
Chris@82 81 Ti = LD(&(xi[WS(is, 1)]), ivs, &(xi[WS(is, 1)]));
Chris@82 82 Tj = LD(&(xi[WS(is, 5)]), ivs, &(xi[WS(is, 1)]));
Chris@82 83 Tk = VADD(Ti, Tj);
Chris@82 84 Tx = VSUB(Tj, Ti);
Chris@82 85 }
Chris@82 86 Tg = VADD(Tc, Tf);
Chris@82 87 Tl = VADD(Th, Tk);
Chris@82 88 TG = VADD(Tw, Tx);
Chris@82 89 Ty = VSUB(Tw, Tx);
Chris@82 90 Tt = VFNMS(LDK(KP500000000), Tk, Th);
Chris@82 91 Ts = VFNMS(LDK(KP500000000), Tf, Tc);
Chris@82 92 }
Chris@82 93 {
Chris@82 94 V Tb, Tm, Tn, To;
Chris@82 95 Tb = VSUB(T5, Ta);
Chris@82 96 Tm = VSUB(Tg, Tl);
Chris@82 97 ST(&(xo[WS(os, 3)]), VFNMSI(Tm, Tb), ovs, &(xo[WS(os, 1)]));
Chris@82 98 ST(&(xo[WS(os, 9)]), VFMAI(Tm, Tb), ovs, &(xo[WS(os, 1)]));
Chris@82 99 Tn = VADD(T5, Ta);
Chris@82 100 To = VADD(Tg, Tl);
Chris@82 101 ST(&(xo[WS(os, 6)]), VSUB(Tn, To), ovs, &(xo[0]));
Chris@82 102 ST(&(xo[0]), VADD(Tn, To), ovs, &(xo[0]));
Chris@82 103 }
Chris@82 104 {
Chris@82 105 V TC, TE, Tv, TD, Tr, Tu;
Chris@82 106 TC = VMUL(LDK(KP866025403), VSUB(Ty, TB));
Chris@82 107 TE = VMUL(LDK(KP866025403), VADD(TB, Ty));
Chris@82 108 Tr = VADD(Tp, Tq);
Chris@82 109 Tu = VADD(Ts, Tt);
Chris@82 110 Tv = VSUB(Tr, Tu);
Chris@82 111 TD = VADD(Tr, Tu);
Chris@82 112 ST(&(xo[WS(os, 10)]), VFNMSI(TC, Tv), ovs, &(xo[0]));
Chris@82 113 ST(&(xo[WS(os, 4)]), VFMAI(TE, TD), ovs, &(xo[0]));
Chris@82 114 ST(&(xo[WS(os, 2)]), VFMAI(TC, Tv), ovs, &(xo[0]));
Chris@82 115 ST(&(xo[WS(os, 8)]), VFNMSI(TE, TD), ovs, &(xo[0]));
Chris@82 116 }
Chris@82 117 {
Chris@82 118 V TH, TL, TK, TM, TF, TI;
Chris@82 119 TF = VSUB(Tp, Tq);
Chris@82 120 TH = VFNMS(LDK(KP866025403), TG, TF);
Chris@82 121 TL = VFMA(LDK(KP866025403), TG, TF);
Chris@82 122 TI = VSUB(Ts, Tt);
Chris@82 123 TK = VFMA(LDK(KP866025403), TJ, TI);
Chris@82 124 TM = VFNMS(LDK(KP866025403), TJ, TI);
Chris@82 125 ST(&(xo[WS(os, 1)]), VFMAI(TK, TH), ovs, &(xo[WS(os, 1)]));
Chris@82 126 ST(&(xo[WS(os, 7)]), VFNMSI(TM, TL), ovs, &(xo[WS(os, 1)]));
Chris@82 127 ST(&(xo[WS(os, 11)]), VFNMSI(TK, TH), ovs, &(xo[WS(os, 1)]));
Chris@82 128 ST(&(xo[WS(os, 5)]), VFMAI(TM, TL), ovs, &(xo[WS(os, 1)]));
Chris@82 129 }
Chris@82 130 }
Chris@82 131 }
Chris@82 132 VLEAVE();
Chris@82 133 }
Chris@82 134
Chris@82 135 static const kdft_desc desc = { 12, XSIMD_STRING("n1bv_12"), {30, 2, 18, 0}, &GENUS, 0, 0, 0, 0 };
Chris@82 136
Chris@82 137 void XSIMD(codelet_n1bv_12) (planner *p) {
Chris@82 138 X(kdft_register) (p, n1bv_12, &desc);
Chris@82 139 }
Chris@82 140
Chris@82 141 #else
Chris@82 142
Chris@82 143 /* Generated by: ../../../genfft/gen_notw_c.native -simd -compact -variables 4 -pipeline-latency 8 -sign 1 -n 12 -name n1bv_12 -include dft/simd/n1b.h */
Chris@82 144
Chris@82 145 /*
Chris@82 146 * This function contains 48 FP additions, 8 FP multiplications,
Chris@82 147 * (or, 44 additions, 4 multiplications, 4 fused multiply/add),
Chris@82 148 * 27 stack variables, 2 constants, and 24 memory accesses
Chris@82 149 */
Chris@82 150 #include "dft/simd/n1b.h"
Chris@82 151
Chris@82 152 static void n1bv_12(const R *ri, const R *ii, R *ro, R *io, stride is, stride os, INT v, INT ivs, INT ovs)
Chris@82 153 {
Chris@82 154 DVK(KP866025403, +0.866025403784438646763723170752936183471402627);
Chris@82 155 DVK(KP500000000, +0.500000000000000000000000000000000000000000000);
Chris@82 156 {
Chris@82 157 INT i;
Chris@82 158 const R *xi;
Chris@82 159 R *xo;
Chris@82 160 xi = ii;
Chris@82 161 xo = io;
Chris@82 162 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@82 163 V T5, Ta, TG, TF, Ty, Tm, Ti, Tp, TJ, TI, Tx, Ts;
Chris@82 164 {
Chris@82 165 V T1, T6, T4, Tk, T9, Tl;
Chris@82 166 T1 = LD(&(xi[0]), ivs, &(xi[0]));
Chris@82 167 T6 = LD(&(xi[WS(is, 6)]), ivs, &(xi[0]));
Chris@82 168 {
Chris@82 169 V T2, T3, T7, T8;
Chris@82 170 T2 = LD(&(xi[WS(is, 4)]), ivs, &(xi[0]));
Chris@82 171 T3 = LD(&(xi[WS(is, 8)]), ivs, &(xi[0]));
Chris@82 172 T4 = VADD(T2, T3);
Chris@82 173 Tk = VSUB(T2, T3);
Chris@82 174 T7 = LD(&(xi[WS(is, 10)]), ivs, &(xi[0]));
Chris@82 175 T8 = LD(&(xi[WS(is, 2)]), ivs, &(xi[0]));
Chris@82 176 T9 = VADD(T7, T8);
Chris@82 177 Tl = VSUB(T7, T8);
Chris@82 178 }
Chris@82 179 T5 = VFNMS(LDK(KP500000000), T4, T1);
Chris@82 180 Ta = VFNMS(LDK(KP500000000), T9, T6);
Chris@82 181 TG = VADD(T6, T9);
Chris@82 182 TF = VADD(T1, T4);
Chris@82 183 Ty = VADD(Tk, Tl);
Chris@82 184 Tm = VMUL(LDK(KP866025403), VSUB(Tk, Tl));
Chris@82 185 }
Chris@82 186 {
Chris@82 187 V Tn, Tq, Te, To, Th, Tr;
Chris@82 188 Tn = LD(&(xi[WS(is, 3)]), ivs, &(xi[WS(is, 1)]));
Chris@82 189 Tq = LD(&(xi[WS(is, 9)]), ivs, &(xi[WS(is, 1)]));
Chris@82 190 {
Chris@82 191 V Tc, Td, Tf, Tg;
Chris@82 192 Tc = LD(&(xi[WS(is, 7)]), ivs, &(xi[WS(is, 1)]));
Chris@82 193 Td = LD(&(xi[WS(is, 11)]), ivs, &(xi[WS(is, 1)]));
Chris@82 194 Te = VSUB(Tc, Td);
Chris@82 195 To = VADD(Tc, Td);
Chris@82 196 Tf = LD(&(xi[WS(is, 1)]), ivs, &(xi[WS(is, 1)]));
Chris@82 197 Tg = LD(&(xi[WS(is, 5)]), ivs, &(xi[WS(is, 1)]));
Chris@82 198 Th = VSUB(Tf, Tg);
Chris@82 199 Tr = VADD(Tf, Tg);
Chris@82 200 }
Chris@82 201 Ti = VMUL(LDK(KP866025403), VSUB(Te, Th));
Chris@82 202 Tp = VFNMS(LDK(KP500000000), To, Tn);
Chris@82 203 TJ = VADD(Tq, Tr);
Chris@82 204 TI = VADD(Tn, To);
Chris@82 205 Tx = VADD(Te, Th);
Chris@82 206 Ts = VFNMS(LDK(KP500000000), Tr, Tq);
Chris@82 207 }
Chris@82 208 {
Chris@82 209 V TH, TK, TL, TM;
Chris@82 210 TH = VSUB(TF, TG);
Chris@82 211 TK = VBYI(VSUB(TI, TJ));
Chris@82 212 ST(&(xo[WS(os, 3)]), VSUB(TH, TK), ovs, &(xo[WS(os, 1)]));
Chris@82 213 ST(&(xo[WS(os, 9)]), VADD(TH, TK), ovs, &(xo[WS(os, 1)]));
Chris@82 214 TL = VADD(TF, TG);
Chris@82 215 TM = VADD(TI, TJ);
Chris@82 216 ST(&(xo[WS(os, 6)]), VSUB(TL, TM), ovs, &(xo[0]));
Chris@82 217 ST(&(xo[0]), VADD(TL, TM), ovs, &(xo[0]));
Chris@82 218 }
Chris@82 219 {
Chris@82 220 V Tj, Tv, Tu, Tw, Tb, Tt;
Chris@82 221 Tb = VSUB(T5, Ta);
Chris@82 222 Tj = VSUB(Tb, Ti);
Chris@82 223 Tv = VADD(Tb, Ti);
Chris@82 224 Tt = VSUB(Tp, Ts);
Chris@82 225 Tu = VBYI(VADD(Tm, Tt));
Chris@82 226 Tw = VBYI(VSUB(Tt, Tm));
Chris@82 227 ST(&(xo[WS(os, 11)]), VSUB(Tj, Tu), ovs, &(xo[WS(os, 1)]));
Chris@82 228 ST(&(xo[WS(os, 5)]), VADD(Tv, Tw), ovs, &(xo[WS(os, 1)]));
Chris@82 229 ST(&(xo[WS(os, 1)]), VADD(Tj, Tu), ovs, &(xo[WS(os, 1)]));
Chris@82 230 ST(&(xo[WS(os, 7)]), VSUB(Tv, Tw), ovs, &(xo[WS(os, 1)]));
Chris@82 231 }
Chris@82 232 {
Chris@82 233 V Tz, TD, TC, TE, TA, TB;
Chris@82 234 Tz = VBYI(VMUL(LDK(KP866025403), VSUB(Tx, Ty)));
Chris@82 235 TD = VBYI(VMUL(LDK(KP866025403), VADD(Ty, Tx)));
Chris@82 236 TA = VADD(T5, Ta);
Chris@82 237 TB = VADD(Tp, Ts);
Chris@82 238 TC = VSUB(TA, TB);
Chris@82 239 TE = VADD(TA, TB);
Chris@82 240 ST(&(xo[WS(os, 2)]), VADD(Tz, TC), ovs, &(xo[0]));
Chris@82 241 ST(&(xo[WS(os, 8)]), VSUB(TE, TD), ovs, &(xo[0]));
Chris@82 242 ST(&(xo[WS(os, 10)]), VSUB(TC, Tz), ovs, &(xo[0]));
Chris@82 243 ST(&(xo[WS(os, 4)]), VADD(TD, TE), ovs, &(xo[0]));
Chris@82 244 }
Chris@82 245 }
Chris@82 246 }
Chris@82 247 VLEAVE();
Chris@82 248 }
Chris@82 249
Chris@82 250 static const kdft_desc desc = { 12, XSIMD_STRING("n1bv_12"), {44, 4, 4, 0}, &GENUS, 0, 0, 0, 0 };
Chris@82 251
Chris@82 252 void XSIMD(codelet_n1bv_12) (planner *p) {
Chris@82 253 X(kdft_register) (p, n1bv_12, &desc);
Chris@82 254 }
Chris@82 255
Chris@82 256 #endif