iup-stack/fftw/dft/scalar/codelets/q1_2.c

150 lines
4.5 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 10:44:41 EDT 2021 */
#include "dft/codelet-dft.h"
#if defined(ARCH_PREFERS_FMA) || defined(ISA_EXTENSION_PREFERS_FMA)
/* Generated by: ../../../genfft/gen_twidsq.native -fma -compact -variables 4 -pipeline-latency 4 -reload-twiddle -dif -n 2 -name q1_2 -include dft/scalar/q.h */
/*
* This function contains 12 FP additions, 8 FP multiplications,
* (or, 8 additions, 4 multiplications, 4 fused multiply/add),
* 17 stack variables, 0 constants, and 16 memory accesses
*/
#include "dft/scalar/q.h"
static void q1_2(R *rio, R *iio, const R *W, stride rs, stride vs, INT mb, INT me, INT ms)
{
{
INT m;
for (m = mb, W = W + (mb * 2); m < me; m = m + 1, rio = rio + ms, iio = iio + ms, W = W + 2, MAKE_VOLATILE_STRIDE(4, rs), MAKE_VOLATILE_STRIDE(0, vs)) {
E T1, T2, T4, T7, T8, T9, Tb, Tc, Te, Th, Ti, Tj;
T1 = rio[0];
T2 = rio[WS(rs, 1)];
T4 = T1 - T2;
T7 = iio[0];
T8 = iio[WS(rs, 1)];
T9 = T7 - T8;
Tb = rio[WS(vs, 1)];
Tc = rio[WS(vs, 1) + WS(rs, 1)];
Te = Tb - Tc;
Th = iio[WS(vs, 1)];
Ti = iio[WS(vs, 1) + WS(rs, 1)];
Tj = Th - Ti;
rio[0] = T1 + T2;
iio[0] = T7 + T8;
rio[WS(rs, 1)] = Tb + Tc;
iio[WS(rs, 1)] = Th + Ti;
{
E Tf, Tk, Td, Tg;
Td = W[0];
Tf = Td * Te;
Tk = Td * Tj;
Tg = W[1];
rio[WS(vs, 1) + WS(rs, 1)] = FMA(Tg, Tj, Tf);
iio[WS(vs, 1) + WS(rs, 1)] = FNMS(Tg, Te, Tk);
}
{
E T5, Ta, T3, T6;
T3 = W[0];
T5 = T3 * T4;
Ta = T3 * T9;
T6 = W[1];
rio[WS(vs, 1)] = FMA(T6, T9, T5);
iio[WS(vs, 1)] = FNMS(T6, T4, Ta);
}
}
}
}
static const tw_instr twinstr[] = {
{ TW_FULL, 0, 2 },
{ TW_NEXT, 1, 0 }
};
static const ct_desc desc = { 2, "q1_2", twinstr, &GENUS, { 8, 4, 4, 0 }, 0, 0, 0 };
void X(codelet_q1_2) (planner *p) {
X(kdft_difsq_register) (p, q1_2, &desc);
}
#else
/* Generated by: ../../../genfft/gen_twidsq.native -compact -variables 4 -pipeline-latency 4 -reload-twiddle -dif -n 2 -name q1_2 -include dft/scalar/q.h */
/*
* This function contains 12 FP additions, 8 FP multiplications,
* (or, 8 additions, 4 multiplications, 4 fused multiply/add),
* 17 stack variables, 0 constants, and 16 memory accesses
*/
#include "dft/scalar/q.h"
static void q1_2(R *rio, R *iio, const R *W, stride rs, stride vs, INT mb, INT me, INT ms)
{
{
INT m;
for (m = mb, W = W + (mb * 2); m < me; m = m + 1, rio = rio + ms, iio = iio + ms, W = W + 2, MAKE_VOLATILE_STRIDE(4, rs), MAKE_VOLATILE_STRIDE(0, vs)) {
E T1, T2, T4, T6, T7, T8, T9, Ta, Tc, Te, Tf, Tg;
T1 = rio[0];
T2 = rio[WS(rs, 1)];
T4 = T1 - T2;
T6 = iio[0];
T7 = iio[WS(rs, 1)];
T8 = T6 - T7;
T9 = rio[WS(vs, 1)];
Ta = rio[WS(vs, 1) + WS(rs, 1)];
Tc = T9 - Ta;
Te = iio[WS(vs, 1)];
Tf = iio[WS(vs, 1) + WS(rs, 1)];
Tg = Te - Tf;
rio[0] = T1 + T2;
iio[0] = T6 + T7;
rio[WS(rs, 1)] = T9 + Ta;
iio[WS(rs, 1)] = Te + Tf;
{
E Tb, Td, T3, T5;
Tb = W[0];
Td = W[1];
rio[WS(vs, 1) + WS(rs, 1)] = FMA(Tb, Tc, Td * Tg);
iio[WS(vs, 1) + WS(rs, 1)] = FNMS(Td, Tc, Tb * Tg);
T3 = W[0];
T5 = W[1];
rio[WS(vs, 1)] = FMA(T3, T4, T5 * T8);
iio[WS(vs, 1)] = FNMS(T5, T4, T3 * T8);
}
}
}
}
static const tw_instr twinstr[] = {
{ TW_FULL, 0, 2 },
{ TW_NEXT, 1, 0 }
};
static const ct_desc desc = { 2, "q1_2", twinstr, &GENUS, { 8, 4, 4, 0 }, 0, 0, 0 };
void X(codelet_q1_2) (planner *p) {
X(kdft_difsq_register) (p, q1_2, &desc);
}
#endif