alexander
/
iup-stack
1
0
Fork 0
Code Issues Pull Requests Packages Projects Releases 1 Wiki Activity
iup-stack/fftw/dft/simd/common/t1bv_12.c

323 lines
9.9 KiB
C
Raw Permalink Blame History

/*
* 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 10:45:49 EDT 2021 */
#include "dft/codelet-dft.h"
#if defined(ARCH_PREFERS_FMA) || defined(ISA_EXTENSION_PREFERS_FMA)
/* Generated by: ../../../genfft/gen_twiddle_c.native -fma -simd -compact -variables 4 -pipeline-latency 8 -n 12 -name t1bv_12 -include dft/simd/t1b.h -sign 1 */
/*
* This function contains 59 FP additions, 42 FP multiplications,
* (or, 41 additions, 24 multiplications, 18 fused multiply/add),
* 28 stack variables, 2 constants, and 24 memory accesses
*/
#include "dft/simd/t1b.h"
static void t1bv_12(R *ri, R *ii, const R *W, stride rs, INT mb, INT me, INT ms)
{
DVK(KP866025403, +0.866025403784438646763723170752936183471402627);
DVK(KP500000000, +0.500000000000000000000000000000000000000000000);
{
INT m;
R *x;
x = ii;
for (m = mb, W = W + (mb * ((TWVL / VL) * 22)); m < me; m = m + VL, x = x + (VL * ms), W = W + (TWVL * 22), MAKE_VOLATILE_STRIDE(12, rs)) {
V T1, TK, T6, TA, Tq, TI, Tv, TE, T9, TL, Te, TB, Ti, TH, Tn;
V TD;
{
V T5, T3, T4, T2;
T1 = LD(&(x[0]), ms, &(x[0]));
T4 = LD(&(x[WS(rs, 8)]), ms, &(x[0]));
T5 = BYTW(&(W[TWVL * 14]), T4);
T2 = LD(&(x[WS(rs, 4)]), ms, &(x[0]));
T3 = BYTW(&(W[TWVL * 6]), T2);
TK = VSUB(T3, T5);
T6 = VADD(T3, T5);
TA = VFNMS(LDK(KP500000000), T6, T1);
}
{
V Tu, Ts, Tp, Tt, Tr;
Tp = LD(&(x[WS(rs, 9)]), ms, &(x[WS(rs, 1)]));
Tq = BYTW(&(W[TWVL * 16]), Tp);
Tt = LD(&(x[WS(rs, 5)]), ms, &(x[WS(rs, 1)]));
Tu = BYTW(&(W[TWVL * 8]), Tt);
Tr = LD(&(x[WS(rs, 1)]), ms, &(x[WS(rs, 1)]));
Ts = BYTW(&(W[0]), Tr);
TI = VSUB(Tu, Ts);
Tv = VADD(Ts, Tu);
TE = VFNMS(LDK(KP500000000), Tv, Tq);
}
{
V Td, Tb, T8, Tc, Ta;
T8 = LD(&(x[WS(rs, 6)]), ms, &(x[0]));
T9 = BYTW(&(W[TWVL * 10]), T8);
Tc = LD(&(x[WS(rs, 2)]), ms, &(x[0]));
Td = BYTW(&(W[TWVL * 2]), Tc);
Ta = LD(&(x[WS(rs, 10)]), ms, &(x[0]));
Tb = BYTW(&(W[TWVL * 18]), Ta);
TL = VSUB(Tb, Td);
Te = VADD(Tb, Td);
TB = VFNMS(LDK(KP500000000), Te, T9);
}
{
V Tm, Tk, Th, Tl, Tj;
Th = LD(&(x[WS(rs, 3)]), ms, &(x[WS(rs, 1)]));
Ti = BYTW(&(W[TWVL * 4]), Th);
Tl = LD(&(x[WS(rs, 11)]), ms, &(x[WS(rs, 1)]));
Tm = BYTW(&(W[TWVL * 20]), Tl);
Tj = LD(&(x[WS(rs, 7)]), ms, &(x[WS(rs, 1)]));
Tk = BYTW(&(W[TWVL * 12]), Tj);
TH = VSUB(Tk, Tm);
Tn = VADD(Tk, Tm);
TD = VFNMS(LDK(KP500000000), Tn, Ti);
}
{
V Tg, Ty, Tx, Tz;
{
V T7, Tf, To, Tw;
T7 = VADD(T1, T6);
Tf = VADD(T9, Te);
Tg = VSUB(T7, Tf);
Ty = VADD(T7, Tf);
To = VADD(Ti, Tn);
Tw = VADD(Tq, Tv);
Tx = VSUB(To, Tw);
Tz = VADD(To, Tw);
}
ST(&(x[WS(rs, 3)]), VFNMSI(Tx, Tg), ms, &(x[WS(rs, 1)]));
ST(&(x[0]), VADD(Ty, Tz), ms, &(x[0]));
ST(&(x[WS(rs, 9)]), VFMAI(Tx, Tg), ms, &(x[WS(rs, 1)]));
ST(&(x[WS(rs, 6)]), VSUB(Ty, Tz), ms, &(x[0]));
}
{
V TS, TW, TV, TX;
{
V TQ, TR, TT, TU;
TQ = VSUB(TA, TB);
TR = VADD(TH, TI);
TS = VFNMS(LDK(KP866025403), TR, TQ);
TW = VFMA(LDK(KP866025403), TR, TQ);
TT = VSUB(TD, TE);
TU = VSUB(TK, TL);
TV = VFMA(LDK(KP866025403), TU, TT);
TX = VFNMS(LDK(KP866025403), TU, TT);
}
ST(&(x[WS(rs, 1)]), VFMAI(TV, TS), ms, &(x[WS(rs, 1)]));
ST(&(x[WS(rs, 7)]), VFNMSI(TX, TW), ms, &(x[WS(rs, 1)]));
ST(&(x[WS(rs, 11)]), VFNMSI(TV, TS), ms, &(x[WS(rs, 1)]));
ST(&(x[WS(rs, 5)]), VFMAI(TX, TW), ms, &(x[WS(rs, 1)]));
}
{
V TG, TO, TN, TP;
{
V TC, TF, TJ, TM;
TC = VADD(TA, TB);
TF = VADD(TD, TE);
TG = VSUB(TC, TF);
TO = VADD(TC, TF);
TJ = VSUB(TH, TI);
TM = VADD(TK, TL);
TN = VMUL(LDK(KP866025403), VSUB(TJ, TM));
TP = VMUL(LDK(KP866025403), VADD(TM, TJ));
}
ST(&(x[WS(rs, 10)]), VFNMSI(TN, TG), ms, &(x[0]));
ST(&(x[WS(rs, 4)]), VFMAI(TP, TO), ms, &(x[0]));
ST(&(x[WS(rs, 2)]), VFMAI(TN, TG), ms, &(x[0]));
ST(&(x[WS(rs, 8)]), VFNMSI(TP, TO), ms, &(x[0]));
}
}
}
VLEAVE();
}
static const tw_instr twinstr[] = {
VTW(0, 1),
VTW(0, 2),
VTW(0, 3),
VTW(0, 4),
VTW(0, 5),
VTW(0, 6),
VTW(0, 7),
VTW(0, 8),
VTW(0, 9),
VTW(0, 10),
VTW(0, 11),
{ TW_NEXT, VL, 0 }
};
static const ct_desc desc = { 12, XSIMD_STRING("t1bv_12"), twinstr, &GENUS, { 41, 24, 18, 0 }, 0, 0, 0 };
void XSIMD(codelet_t1bv_12) (planner *p) {
X(kdft_dit_register) (p, t1bv_12, &desc);
}
#else
/* Generated by: ../../../genfft/gen_twiddle_c.native -simd -compact -variables 4 -pipeline-latency 8 -n 12 -name t1bv_12 -include dft/simd/t1b.h -sign 1 */
/*
* This function contains 59 FP additions, 30 FP multiplications,
* (or, 55 additions, 26 multiplications, 4 fused multiply/add),
* 28 stack variables, 2 constants, and 24 memory accesses
*/
#include "dft/simd/t1b.h"
static void t1bv_12(R *ri, R *ii, const R *W, stride rs, INT mb, INT me, INT ms)
{
DVK(KP866025403, +0.866025403784438646763723170752936183471402627);
DVK(KP500000000, +0.500000000000000000000000000000000000000000000);
{
INT m;
R *x;
x = ii;
for (m = mb, W = W + (mb * ((TWVL / VL) * 22)); m < me; m = m + VL, x = x + (VL * ms), W = W + (TWVL * 22), MAKE_VOLATILE_STRIDE(12, rs)) {
V T1, Tt, T6, T7, TB, Tq, TC, TD, T9, Tu, Te, Tf, Tx, Tl, Ty;
V Tz;
{
V T5, T3, T4, T2;
T1 = LD(&(x[0]), ms, &(x[0]));
T4 = LD(&(x[WS(rs, 8)]), ms, &(x[0]));
T5 = BYTW(&(W[TWVL * 14]), T4);
T2 = LD(&(x[WS(rs, 4)]), ms, &(x[0]));
T3 = BYTW(&(W[TWVL * 6]), T2);
Tt = VSUB(T3, T5);
T6 = VADD(T3, T5);
T7 = VFNMS(LDK(KP500000000), T6, T1);
}
{
V Tn, Tp, Tm, TA, To;
Tm = LD(&(x[WS(rs, 1)]), ms, &(x[WS(rs, 1)]));
Tn = BYTW(&(W[0]), Tm);
TA = LD(&(x[WS(rs, 9)]), ms, &(x[WS(rs, 1)]));
TB = BYTW(&(W[TWVL * 16]), TA);
To = LD(&(x[WS(rs, 5)]), ms, &(x[WS(rs, 1)]));
Tp = BYTW(&(W[TWVL * 8]), To);
Tq = VSUB(Tn, Tp);
TC = VADD(Tn, Tp);
TD = VFNMS(LDK(KP500000000), TC, TB);
}
{
V Td, Tb, T8, Tc, Ta;
T8 = LD(&(x[WS(rs, 6)]), ms, &(x[0]));
T9 = BYTW(&(W[TWVL * 10]), T8);
Tc = LD(&(x[WS(rs, 2)]), ms, &(x[0]));
Td = BYTW(&(W[TWVL * 2]), Tc);
Ta = LD(&(x[WS(rs, 10)]), ms, &(x[0]));
Tb = BYTW(&(W[TWVL * 18]), Ta);
Tu = VSUB(Tb, Td);
Te = VADD(Tb, Td);
Tf = VFNMS(LDK(KP500000000), Te, T9);
}
{
V Ti, Tk, Th, Tw, Tj;
Th = LD(&(x[WS(rs, 7)]), ms, &(x[WS(rs, 1)]));
Ti = BYTW(&(W[TWVL * 12]), Th);
Tw = LD(&(x[WS(rs, 3)]), ms, &(x[WS(rs, 1)]));
Tx = BYTW(&(W[TWVL * 4]), Tw);
Tj = LD(&(x[WS(rs, 11)]), ms, &(x[WS(rs, 1)]));
Tk = BYTW(&(W[TWVL * 20]), Tj);
Tl = VSUB(Ti, Tk);
Ty = VADD(Ti, Tk);
Tz = VFNMS(LDK(KP500000000), Ty, Tx);
}
{
V Ts, TG, TF, TH;
{
V Tg, Tr, Tv, TE;
Tg = VSUB(T7, Tf);
Tr = VMUL(LDK(KP866025403), VSUB(Tl, Tq));
Ts = VSUB(Tg, Tr);
TG = VADD(Tg, Tr);
Tv = VMUL(LDK(KP866025403), VSUB(Tt, Tu));
TE = VSUB(Tz, TD);
TF = VBYI(VADD(Tv, TE));
TH = VBYI(VSUB(TE, Tv));
}
ST(&(x[WS(rs, 11)]), VSUB(Ts, TF), ms, &(x[WS(rs, 1)]));
ST(&(x[WS(rs, 5)]), VADD(TG, TH), ms, &(x[WS(rs, 1)]));
ST(&(x[WS(rs, 1)]), VADD(Ts, TF), ms, &(x[WS(rs, 1)]));
ST(&(x[WS(rs, 7)]), VSUB(TG, TH), ms, &(x[WS(rs, 1)]));
}
{
V TS, TW, TV, TX;
{
V TQ, TR, TT, TU;
TQ = VADD(T1, T6);
TR = VADD(T9, Te);
TS = VSUB(TQ, TR);
TW = VADD(TQ, TR);
TT = VADD(Tx, Ty);
TU = VADD(TB, TC);
TV = VBYI(VSUB(TT, TU));
TX = VADD(TT, TU);
}
ST(&(x[WS(rs, 3)]), VSUB(TS, TV), ms, &(x[WS(rs, 1)]));
ST(&(x[0]), VADD(TW, TX), ms, &(x[0]));
ST(&(x[WS(rs, 9)]), VADD(TS, TV), ms, &(x[WS(rs, 1)]));
ST(&(x[WS(rs, 6)]), VSUB(TW, TX), ms, &(x[0]));
}
{
V TK, TO, TN, TP;
{
V TI, TJ, TL, TM;
TI = VADD(Tl, Tq);
TJ = VADD(Tt, Tu);
TK = VBYI(VMUL(LDK(KP866025403), VSUB(TI, TJ)));
TO = VBYI(VMUL(LDK(KP866025403), VADD(TJ, TI)));
TL = VADD(T7, Tf);
TM = VADD(Tz, TD);
TN = VSUB(TL, TM);
TP = VADD(TL, TM);
}
ST(&(x[WS(rs, 2)]), VADD(TK, TN), ms, &(x[0]));
ST(&(x[WS(rs, 8)]), VSUB(TP, TO), ms, &(x[0]));
ST(&(x[WS(rs, 10)]), VSUB(TN, TK), ms, &(x[0]));
ST(&(x[WS(rs, 4)]), VADD(TO, TP), ms, &(x[0]));
}
}
}
VLEAVE();
}
static const tw_instr twinstr[] = {
VTW(0, 1),
VTW(0, 2),
VTW(0, 3),
VTW(0, 4),
VTW(0, 5),
VTW(0, 6),
VTW(0, 7),
VTW(0, 8),
VTW(0, 9),
VTW(0, 10),
VTW(0, 11),
{ TW_NEXT, VL, 0 }
};
static const ct_desc desc = { 12, XSIMD_STRING("t1bv_12"), twinstr, &GENUS, { 55, 26, 4, 0 }, 0, 0, 0 };
void XSIMD(codelet_t1bv_12) (planner *p) {
X(kdft_dit_register) (p, t1bv_12, &desc);
}
#endif
Reference in New Issue View Git Blame Copy Permalink