195 lines
5.2 KiB
C
195 lines
5.2 KiB
C
/*
|
|
* Copyright (c) 2003, 2007-14 Matteo Frigo
|
|
* Copyright (c) 2003, 2007-14 Massachusetts Institute of Technology
|
|
*
|
|
* This program is free software; you can redistribute it and/or modify
|
|
* it under the terms of the GNU General Public License as published by
|
|
* the Free Software Foundation; either version 2 of the License, or
|
|
* (at your option) any later version.
|
|
*
|
|
* This program is distributed in the hope that it will be useful,
|
|
* but WITHOUT ANY WARRANTY; without even the implied warranty of
|
|
* MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
|
|
* GNU General Public License for more details.
|
|
*
|
|
* You should have received a copy of the GNU General Public License
|
|
* along with this program; if not, write to the Free Software
|
|
* Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
|
|
*
|
|
*/
|
|
|
|
/* This file was automatically generated --- DO NOT EDIT */
|
|
/* Generated on Tue Sep 14 10:46:55 EDT 2021 */
|
|
|
|
#include "rdft/codelet-rdft.h"
|
|
|
|
#if defined(ARCH_PREFERS_FMA) || defined(ISA_EXTENSION_PREFERS_FMA)
|
|
|
|
/* Generated by: ../../../genfft/gen_hc2hc.native -fma -compact -variables 4 -pipeline-latency 4 -sign 1 -twiddle-log3 -precompute-twiddles -n 4 -dif -name hb2_4 -include rdft/scalar/hb.h */
|
|
|
|
/*
|
|
* This function contains 24 FP additions, 16 FP multiplications,
|
|
* (or, 16 additions, 8 multiplications, 8 fused multiply/add),
|
|
* 33 stack variables, 0 constants, and 16 memory accesses
|
|
*/
|
|
#include "rdft/scalar/hb.h"
|
|
|
|
static void hb2_4(R *cr, R *ci, const R *W, stride rs, INT mb, INT me, INT ms)
|
|
{
|
|
{
|
|
INT m;
|
|
for (m = mb, W = W + ((mb - 1) * 4); m < me; m = m + 1, cr = cr + ms, ci = ci - ms, W = W + 4, MAKE_VOLATILE_STRIDE(8, rs)) {
|
|
E T7, Tb, T8, Ta, Tc, Tg, T9, Tf;
|
|
T7 = W[0];
|
|
Tb = W[3];
|
|
T8 = W[2];
|
|
T9 = T7 * T8;
|
|
Tf = T7 * Tb;
|
|
Ta = W[1];
|
|
Tc = FMA(Ta, Tb, T9);
|
|
Tg = FNMS(Ta, T8, Tf);
|
|
{
|
|
E T3, T6, Td, Tj, Tz, Tx, Tr, Tm, Tv, Ts, Tw, TA;
|
|
{
|
|
E Th, Ti, Tu, Tk, Tl, Tq, Tp, Tt;
|
|
Th = ci[WS(rs, 3)];
|
|
Ti = cr[WS(rs, 2)];
|
|
Tu = Th + Ti;
|
|
Tk = ci[WS(rs, 2)];
|
|
Tl = cr[WS(rs, 3)];
|
|
Tq = Tk + Tl;
|
|
{
|
|
E T1, T2, T4, T5;
|
|
T1 = cr[0];
|
|
T2 = ci[WS(rs, 1)];
|
|
T3 = T1 + T2;
|
|
Tp = T1 - T2;
|
|
T4 = cr[WS(rs, 1)];
|
|
T5 = ci[0];
|
|
T6 = T4 + T5;
|
|
Tt = T4 - T5;
|
|
}
|
|
Td = T3 - T6;
|
|
Tj = Th - Ti;
|
|
Tz = Tu - Tt;
|
|
Tx = Tp + Tq;
|
|
Tr = Tp - Tq;
|
|
Tm = Tk - Tl;
|
|
Tv = Tt + Tu;
|
|
}
|
|
cr[0] = T3 + T6;
|
|
ci[0] = Tj + Tm;
|
|
Ts = T7 * Tr;
|
|
cr[WS(rs, 1)] = FNMS(Ta, Tv, Ts);
|
|
Tw = T7 * Tv;
|
|
ci[WS(rs, 1)] = FMA(Ta, Tr, Tw);
|
|
TA = T8 * Tz;
|
|
ci[WS(rs, 3)] = FMA(Tb, Tx, TA);
|
|
{
|
|
E Ty, Te, To, Tn;
|
|
Ty = T8 * Tx;
|
|
cr[WS(rs, 3)] = FNMS(Tb, Tz, Ty);
|
|
Te = Tc * Td;
|
|
To = Tg * Td;
|
|
Tn = Tj - Tm;
|
|
cr[WS(rs, 2)] = FNMS(Tg, Tn, Te);
|
|
ci[WS(rs, 2)] = FMA(Tc, Tn, To);
|
|
}
|
|
}
|
|
}
|
|
}
|
|
}
|
|
|
|
static const tw_instr twinstr[] = {
|
|
{ TW_CEXP, 1, 1 },
|
|
{ TW_CEXP, 1, 3 },
|
|
{ TW_NEXT, 1, 0 }
|
|
};
|
|
|
|
static const hc2hc_desc desc = { 4, "hb2_4", twinstr, &GENUS, { 16, 8, 8, 0 } };
|
|
|
|
void X(codelet_hb2_4) (planner *p) {
|
|
X(khc2hc_register) (p, hb2_4, &desc);
|
|
}
|
|
#else
|
|
|
|
/* Generated by: ../../../genfft/gen_hc2hc.native -compact -variables 4 -pipeline-latency 4 -sign 1 -twiddle-log3 -precompute-twiddles -n 4 -dif -name hb2_4 -include rdft/scalar/hb.h */
|
|
|
|
/*
|
|
* This function contains 24 FP additions, 16 FP multiplications,
|
|
* (or, 16 additions, 8 multiplications, 8 fused multiply/add),
|
|
* 21 stack variables, 0 constants, and 16 memory accesses
|
|
*/
|
|
#include "rdft/scalar/hb.h"
|
|
|
|
static void hb2_4(R *cr, R *ci, const R *W, stride rs, INT mb, INT me, INT ms)
|
|
{
|
|
{
|
|
INT m;
|
|
for (m = mb, W = W + ((mb - 1) * 4); m < me; m = m + 1, cr = cr + ms, ci = ci - ms, W = W + 4, MAKE_VOLATILE_STRIDE(8, rs)) {
|
|
E T7, T9, T8, Ta, Tb, Td;
|
|
T7 = W[0];
|
|
T9 = W[1];
|
|
T8 = W[2];
|
|
Ta = W[3];
|
|
Tb = FMA(T7, T8, T9 * Ta);
|
|
Td = FNMS(T9, T8, T7 * Ta);
|
|
{
|
|
E T3, Tl, T6, To, Tg, Tp, Tj, Tm, Tc, Tk;
|
|
{
|
|
E T1, T2, T4, T5;
|
|
T1 = cr[0];
|
|
T2 = ci[WS(rs, 1)];
|
|
T3 = T1 + T2;
|
|
Tl = T1 - T2;
|
|
T4 = cr[WS(rs, 1)];
|
|
T5 = ci[0];
|
|
T6 = T4 + T5;
|
|
To = T4 - T5;
|
|
}
|
|
{
|
|
E Te, Tf, Th, Ti;
|
|
Te = ci[WS(rs, 3)];
|
|
Tf = cr[WS(rs, 2)];
|
|
Tg = Te - Tf;
|
|
Tp = Te + Tf;
|
|
Th = ci[WS(rs, 2)];
|
|
Ti = cr[WS(rs, 3)];
|
|
Tj = Th - Ti;
|
|
Tm = Th + Ti;
|
|
}
|
|
cr[0] = T3 + T6;
|
|
ci[0] = Tg + Tj;
|
|
Tc = T3 - T6;
|
|
Tk = Tg - Tj;
|
|
cr[WS(rs, 2)] = FNMS(Td, Tk, Tb * Tc);
|
|
ci[WS(rs, 2)] = FMA(Td, Tc, Tb * Tk);
|
|
{
|
|
E Tn, Tq, Tr, Ts;
|
|
Tn = Tl - Tm;
|
|
Tq = To + Tp;
|
|
cr[WS(rs, 1)] = FNMS(T9, Tq, T7 * Tn);
|
|
ci[WS(rs, 1)] = FMA(T7, Tq, T9 * Tn);
|
|
Tr = Tl + Tm;
|
|
Ts = Tp - To;
|
|
cr[WS(rs, 3)] = FNMS(Ta, Ts, T8 * Tr);
|
|
ci[WS(rs, 3)] = FMA(T8, Ts, Ta * Tr);
|
|
}
|
|
}
|
|
}
|
|
}
|
|
}
|
|
|
|
static const tw_instr twinstr[] = {
|
|
{ TW_CEXP, 1, 1 },
|
|
{ TW_CEXP, 1, 3 },
|
|
{ TW_NEXT, 1, 0 }
|
|
};
|
|
|
|
static const hc2hc_desc desc = { 4, "hb2_4", twinstr, &GENUS, { 16, 8, 8, 0 } };
|
|
|
|
void X(codelet_hb2_4) (planner *p) {
|
|
X(khc2hc_register) (p, hb2_4, &desc);
|
|
}
|
|
#endif
|