197 lines
5.2 KiB
C
197 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 10:46:31 EDT 2021 */
|
|
|
|
#include "rdft/codelet-rdft.h"
|
|
|
|
#if defined(ARCH_PREFERS_FMA) || defined(ISA_EXTENSION_PREFERS_FMA)
|
|
|
|
/* Generated by: ../../../genfft/gen_hc2c.native -fma -compact -variables 4 -pipeline-latency 4 -n 4 -dit -name hc2cf_4 -include rdft/scalar/hc2cf.h */
|
|
|
|
/*
|
|
* This function contains 22 FP additions, 12 FP multiplications,
|
|
* (or, 16 additions, 6 multiplications, 6 fused multiply/add),
|
|
* 15 stack variables, 0 constants, and 16 memory accesses
|
|
*/
|
|
#include "rdft/scalar/hc2cf.h"
|
|
|
|
static void hc2cf_4(R *Rp, R *Ip, R *Rm, R *Im, const R *W, stride rs, INT mb, INT me, INT ms)
|
|
{
|
|
{
|
|
INT m;
|
|
for (m = mb, W = W + ((mb - 1) * 6); m < me; m = m + 1, Rp = Rp + ms, Ip = Ip + ms, Rm = Rm - ms, Im = Im - ms, W = W + 6, MAKE_VOLATILE_STRIDE(16, rs)) {
|
|
E T1, Tv, T7, Tu, Te, To, Tk, Tq;
|
|
T1 = Rp[0];
|
|
Tv = Rm[0];
|
|
{
|
|
E T3, T6, T4, Tt, T2, T5;
|
|
T3 = Rp[WS(rs, 1)];
|
|
T6 = Rm[WS(rs, 1)];
|
|
T2 = W[2];
|
|
T4 = T2 * T3;
|
|
Tt = T2 * T6;
|
|
T5 = W[3];
|
|
T7 = FMA(T5, T6, T4);
|
|
Tu = FNMS(T5, T3, Tt);
|
|
}
|
|
{
|
|
E Ta, Td, Tb, Tn, T9, Tc;
|
|
Ta = Ip[0];
|
|
Td = Im[0];
|
|
T9 = W[0];
|
|
Tb = T9 * Ta;
|
|
Tn = T9 * Td;
|
|
Tc = W[1];
|
|
Te = FMA(Tc, Td, Tb);
|
|
To = FNMS(Tc, Ta, Tn);
|
|
}
|
|
{
|
|
E Tg, Tj, Th, Tp, Tf, Ti;
|
|
Tg = Ip[WS(rs, 1)];
|
|
Tj = Im[WS(rs, 1)];
|
|
Tf = W[4];
|
|
Th = Tf * Tg;
|
|
Tp = Tf * Tj;
|
|
Ti = W[5];
|
|
Tk = FMA(Ti, Tj, Th);
|
|
Tq = FNMS(Ti, Tg, Tp);
|
|
}
|
|
{
|
|
E T8, Tl, Ts, Tw;
|
|
T8 = T1 + T7;
|
|
Tl = Te + Tk;
|
|
Rm[WS(rs, 1)] = T8 - Tl;
|
|
Rp[0] = T8 + Tl;
|
|
Ts = To + Tq;
|
|
Tw = Tu + Tv;
|
|
Im[WS(rs, 1)] = Ts - Tw;
|
|
Ip[0] = Ts + Tw;
|
|
}
|
|
{
|
|
E Tm, Tr, Tx, Ty;
|
|
Tm = T1 - T7;
|
|
Tr = To - Tq;
|
|
Rm[0] = Tm - Tr;
|
|
Rp[WS(rs, 1)] = Tm + Tr;
|
|
Tx = Tk - Te;
|
|
Ty = Tv - Tu;
|
|
Im[0] = Tx - Ty;
|
|
Ip[WS(rs, 1)] = Tx + Ty;
|
|
}
|
|
}
|
|
}
|
|
}
|
|
|
|
static const tw_instr twinstr[] = {
|
|
{ TW_FULL, 1, 4 },
|
|
{ TW_NEXT, 1, 0 }
|
|
};
|
|
|
|
static const hc2c_desc desc = { 4, "hc2cf_4", twinstr, &GENUS, { 16, 6, 6, 0 } };
|
|
|
|
void X(codelet_hc2cf_4) (planner *p) {
|
|
X(khc2c_register) (p, hc2cf_4, &desc, HC2C_VIA_RDFT);
|
|
}
|
|
#else
|
|
|
|
/* Generated by: ../../../genfft/gen_hc2c.native -compact -variables 4 -pipeline-latency 4 -n 4 -dit -name hc2cf_4 -include rdft/scalar/hc2cf.h */
|
|
|
|
/*
|
|
* This function contains 22 FP additions, 12 FP multiplications,
|
|
* (or, 16 additions, 6 multiplications, 6 fused multiply/add),
|
|
* 13 stack variables, 0 constants, and 16 memory accesses
|
|
*/
|
|
#include "rdft/scalar/hc2cf.h"
|
|
|
|
static void hc2cf_4(R *Rp, R *Ip, R *Rm, R *Im, const R *W, stride rs, INT mb, INT me, INT ms)
|
|
{
|
|
{
|
|
INT m;
|
|
for (m = mb, W = W + ((mb - 1) * 6); m < me; m = m + 1, Rp = Rp + ms, Ip = Ip + ms, Rm = Rm - ms, Im = Im - ms, W = W + 6, MAKE_VOLATILE_STRIDE(16, rs)) {
|
|
E T1, Tp, T6, To, Tc, Tk, Th, Tl;
|
|
T1 = Rp[0];
|
|
Tp = Rm[0];
|
|
{
|
|
E T3, T5, T2, T4;
|
|
T3 = Rp[WS(rs, 1)];
|
|
T5 = Rm[WS(rs, 1)];
|
|
T2 = W[2];
|
|
T4 = W[3];
|
|
T6 = FMA(T2, T3, T4 * T5);
|
|
To = FNMS(T4, T3, T2 * T5);
|
|
}
|
|
{
|
|
E T9, Tb, T8, Ta;
|
|
T9 = Ip[0];
|
|
Tb = Im[0];
|
|
T8 = W[0];
|
|
Ta = W[1];
|
|
Tc = FMA(T8, T9, Ta * Tb);
|
|
Tk = FNMS(Ta, T9, T8 * Tb);
|
|
}
|
|
{
|
|
E Te, Tg, Td, Tf;
|
|
Te = Ip[WS(rs, 1)];
|
|
Tg = Im[WS(rs, 1)];
|
|
Td = W[4];
|
|
Tf = W[5];
|
|
Th = FMA(Td, Te, Tf * Tg);
|
|
Tl = FNMS(Tf, Te, Td * Tg);
|
|
}
|
|
{
|
|
E T7, Ti, Tn, Tq;
|
|
T7 = T1 + T6;
|
|
Ti = Tc + Th;
|
|
Rm[WS(rs, 1)] = T7 - Ti;
|
|
Rp[0] = T7 + Ti;
|
|
Tn = Tk + Tl;
|
|
Tq = To + Tp;
|
|
Im[WS(rs, 1)] = Tn - Tq;
|
|
Ip[0] = Tn + Tq;
|
|
}
|
|
{
|
|
E Tj, Tm, Tr, Ts;
|
|
Tj = T1 - T6;
|
|
Tm = Tk - Tl;
|
|
Rm[0] = Tj - Tm;
|
|
Rp[WS(rs, 1)] = Tj + Tm;
|
|
Tr = Th - Tc;
|
|
Ts = Tp - To;
|
|
Im[0] = Tr - Ts;
|
|
Ip[WS(rs, 1)] = Tr + Ts;
|
|
}
|
|
}
|
|
}
|
|
}
|
|
|
|
static const tw_instr twinstr[] = {
|
|
{ TW_FULL, 1, 4 },
|
|
{ TW_NEXT, 1, 0 }
|
|
};
|
|
|
|
static const hc2c_desc desc = { 4, "hc2cf_4", twinstr, &GENUS, { 16, 6, 6, 0 } };
|
|
|
|
void X(codelet_hc2cf_4) (planner *p) {
|
|
X(khc2c_register) (p, hc2cf_4, &desc, HC2C_VIA_RDFT);
|
|
}
|
|
#endif
|