iup-stack/fftw/dft/simd/common/n2bv_20.c

505 lines
16 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 10:45:22 EDT 2021 */
#include "dft/codelet-dft.h"
#if defined(ARCH_PREFERS_FMA) || defined(ISA_EXTENSION_PREFERS_FMA)
/* Generated by: ../../../genfft/gen_notw_c.native -fma -simd -compact -variables 4 -pipeline-latency 8 -sign 1 -n 20 -name n2bv_20 -with-ostride 2 -include dft/simd/n2b.h -store-multiple 2 */
/*
* This function contains 104 FP additions, 50 FP multiplications,
* (or, 58 additions, 4 multiplications, 46 fused multiply/add),
* 57 stack variables, 4 constants, and 50 memory accesses
*/
#include "dft/simd/n2b.h"
static void n2bv_20(const R *ri, const R *ii, R *ro, R *io, stride is, stride os, INT v, INT ivs, INT ovs)
{
DVK(KP559016994, +0.559016994374947424102293417182819058860154590);
DVK(KP618033988, +0.618033988749894848204586834365638117720309180);
DVK(KP951056516, +0.951056516295153572116439333379382143405698634);
DVK(KP250000000, +0.250000000000000000000000000000000000000000000);
{
INT i;
const R *xi;
R *xo;
xi = ii;
xo = io;
for (i = v; i > 0; i = i - VL, xi = xi + (VL * ivs), xo = xo + (VL * ovs), MAKE_VOLATILE_STRIDE(40, is), MAKE_VOLATILE_STRIDE(40, os)) {
V T3, T1r, TE, T13, Ts, TL, TM, Tz, T16, T19, T1a, T1v, T1w, T1x, T1s;
V T1t, T1u, T1d, T1g, T1h, Ti, Tk, TH, TJ;
{
V T1, T2, T11, TC, TD, T12;
T1 = LD(&(xi[0]), ivs, &(xi[0]));
T2 = LD(&(xi[WS(is, 10)]), ivs, &(xi[0]));
T11 = VADD(T1, T2);
TC = LD(&(xi[WS(is, 5)]), ivs, &(xi[WS(is, 1)]));
TD = LD(&(xi[WS(is, 15)]), ivs, &(xi[WS(is, 1)]));
T12 = VADD(TC, TD);
T3 = VSUB(T1, T2);
T1r = VADD(T11, T12);
TE = VSUB(TC, TD);
T13 = VSUB(T11, T12);
}
{
V T6, T14, Tv, T1c, Ty, T1f, T9, T17, Td, T1b, To, T15, Tr, T18, Tg;
V T1e;
{
V T4, T5, Tt, Tu;
T4 = LD(&(xi[WS(is, 4)]), ivs, &(xi[0]));
T5 = LD(&(xi[WS(is, 14)]), ivs, &(xi[0]));
T6 = VSUB(T4, T5);
T14 = VADD(T4, T5);
Tt = LD(&(xi[WS(is, 13)]), ivs, &(xi[WS(is, 1)]));
Tu = LD(&(xi[WS(is, 3)]), ivs, &(xi[WS(is, 1)]));
Tv = VSUB(Tt, Tu);
T1c = VADD(Tt, Tu);
}
{
V Tw, Tx, T7, T8;
Tw = LD(&(xi[WS(is, 17)]), ivs, &(xi[WS(is, 1)]));
Tx = LD(&(xi[WS(is, 7)]), ivs, &(xi[WS(is, 1)]));
Ty = VSUB(Tw, Tx);
T1f = VADD(Tw, Tx);
T7 = LD(&(xi[WS(is, 16)]), ivs, &(xi[0]));
T8 = LD(&(xi[WS(is, 6)]), ivs, &(xi[0]));
T9 = VSUB(T7, T8);
T17 = VADD(T7, T8);
}
{
V Tb, Tc, Tm, Tn;
Tb = LD(&(xi[WS(is, 8)]), ivs, &(xi[0]));
Tc = LD(&(xi[WS(is, 18)]), ivs, &(xi[0]));
Td = VSUB(Tb, Tc);
T1b = VADD(Tb, Tc);
Tm = LD(&(xi[WS(is, 9)]), ivs, &(xi[WS(is, 1)]));
Tn = LD(&(xi[WS(is, 19)]), ivs, &(xi[WS(is, 1)]));
To = VSUB(Tm, Tn);
T15 = VADD(Tm, Tn);
}
{
V Tp, Tq, Te, Tf;
Tp = LD(&(xi[WS(is, 1)]), ivs, &(xi[WS(is, 1)]));
Tq = LD(&(xi[WS(is, 11)]), ivs, &(xi[WS(is, 1)]));
Tr = VSUB(Tp, Tq);
T18 = VADD(Tp, Tq);
Te = LD(&(xi[WS(is, 12)]), ivs, &(xi[0]));
Tf = LD(&(xi[WS(is, 2)]), ivs, &(xi[0]));
Tg = VSUB(Te, Tf);
T1e = VADD(Te, Tf);
}
Ts = VSUB(To, Tr);
TL = VSUB(T6, T9);
TM = VSUB(Td, Tg);
Tz = VSUB(Tv, Ty);
T16 = VSUB(T14, T15);
T19 = VSUB(T17, T18);
T1a = VADD(T16, T19);
T1v = VADD(T1b, T1c);
T1w = VADD(T1e, T1f);
T1x = VADD(T1v, T1w);
T1s = VADD(T14, T15);
T1t = VADD(T17, T18);
T1u = VADD(T1s, T1t);
T1d = VSUB(T1b, T1c);
T1g = VSUB(T1e, T1f);
T1h = VADD(T1d, T1g);
{
V Ta, Th, TF, TG;
Ta = VADD(T6, T9);
Th = VADD(Td, Tg);
Ti = VADD(Ta, Th);
Tk = VSUB(Ta, Th);
TF = VADD(To, Tr);
TG = VADD(Tv, Ty);
TH = VADD(TF, TG);
TJ = VSUB(TF, TG);
}
}
{
V T1H, T1J, T1K, T1L, T1N, T1I, TZ, T10;
TZ = VADD(T3, Ti);
T10 = VADD(TE, TH);
T1H = VFNMSI(T10, TZ);
STM2(&(xo[30]), T1H, ovs, &(xo[2]));
T1I = VFMAI(T10, TZ);
STM2(&(xo[10]), T1I, ovs, &(xo[2]));
{
V T1A, T1y, T1z, T1E, T1G, T1C, T1D, T1F, T1B, T1M;
T1A = VSUB(T1u, T1x);
T1y = VADD(T1u, T1x);
T1z = VFNMS(LDK(KP250000000), T1y, T1r);
T1C = VSUB(T1s, T1t);
T1D = VSUB(T1v, T1w);
T1E = VMUL(LDK(KP951056516), VFMA(LDK(KP618033988), T1D, T1C));
T1G = VMUL(LDK(KP951056516), VFNMS(LDK(KP618033988), T1C, T1D));
T1J = VADD(T1r, T1y);
STM2(&(xo[0]), T1J, ovs, &(xo[0]));
T1F = VFNMS(LDK(KP559016994), T1A, T1z);
T1K = VFMAI(T1G, T1F);
STM2(&(xo[16]), T1K, ovs, &(xo[0]));
T1L = VFNMSI(T1G, T1F);
STM2(&(xo[24]), T1L, ovs, &(xo[0]));
T1B = VFMA(LDK(KP559016994), T1A, T1z);
T1M = VFNMSI(T1E, T1B);
STM2(&(xo[8]), T1M, ovs, &(xo[0]));
STN2(&(xo[8]), T1M, T1I, ovs);
T1N = VFMAI(T1E, T1B);
STM2(&(xo[32]), T1N, ovs, &(xo[0]));
}
{
V T1O, T1P, T1R, T1S;
{
V T1k, T1i, T1j, T1o, T1q, T1m, T1n, T1p, T1Q, T1l;
T1k = VSUB(T1a, T1h);
T1i = VADD(T1a, T1h);
T1j = VFNMS(LDK(KP250000000), T1i, T13);
T1m = VSUB(T1d, T1g);
T1n = VSUB(T16, T19);
T1o = VMUL(LDK(KP951056516), VFNMS(LDK(KP618033988), T1n, T1m));
T1q = VMUL(LDK(KP951056516), VFMA(LDK(KP618033988), T1m, T1n));
T1O = VADD(T13, T1i);
STM2(&(xo[20]), T1O, ovs, &(xo[0]));
T1p = VFMA(LDK(KP559016994), T1k, T1j);
T1P = VFMAI(T1q, T1p);
STM2(&(xo[12]), T1P, ovs, &(xo[0]));
T1Q = VFNMSI(T1q, T1p);
STM2(&(xo[28]), T1Q, ovs, &(xo[0]));
STN2(&(xo[28]), T1Q, T1H, ovs);
T1l = VFNMS(LDK(KP559016994), T1k, T1j);
T1R = VFNMSI(T1o, T1l);
STM2(&(xo[4]), T1R, ovs, &(xo[0]));
T1S = VFMAI(T1o, T1l);
STM2(&(xo[36]), T1S, ovs, &(xo[0]));
}
{
V TA, TN, TV, TS, TK, TU, Tl, TR, TI, Tj;
TA = VFMA(LDK(KP618033988), Tz, Ts);
TN = VFMA(LDK(KP618033988), TM, TL);
TV = VFNMS(LDK(KP618033988), TL, TM);
TS = VFNMS(LDK(KP618033988), Ts, Tz);
TI = VFNMS(LDK(KP250000000), TH, TE);
TK = VFMA(LDK(KP559016994), TJ, TI);
TU = VFNMS(LDK(KP559016994), TJ, TI);
Tj = VFNMS(LDK(KP250000000), Ti, T3);
Tl = VFMA(LDK(KP559016994), Tk, Tj);
TR = VFNMS(LDK(KP559016994), Tk, Tj);
{
V TB, TO, T1T, T1U;
TB = VFNMS(LDK(KP951056516), TA, Tl);
TO = VFMA(LDK(KP951056516), TN, TK);
T1T = VFNMSI(TO, TB);
STM2(&(xo[38]), T1T, ovs, &(xo[2]));
STN2(&(xo[36]), T1S, T1T, ovs);
T1U = VFMAI(TO, TB);
STM2(&(xo[2]), T1U, ovs, &(xo[2]));
STN2(&(xo[0]), T1J, T1U, ovs);
}
{
V TX, TY, T1V, T1W;
TX = VFNMS(LDK(KP951056516), TS, TR);
TY = VFMA(LDK(KP951056516), TV, TU);
T1V = VFNMSI(TY, TX);
STM2(&(xo[14]), T1V, ovs, &(xo[2]));
STN2(&(xo[12]), T1P, T1V, ovs);
T1W = VFMAI(TY, TX);
STM2(&(xo[26]), T1W, ovs, &(xo[2]));
STN2(&(xo[24]), T1L, T1W, ovs);
}
{
V TP, TQ, T1X, T1Y;
TP = VFMA(LDK(KP951056516), TA, Tl);
TQ = VFNMS(LDK(KP951056516), TN, TK);
T1X = VFNMSI(TQ, TP);
STM2(&(xo[22]), T1X, ovs, &(xo[2]));
STN2(&(xo[20]), T1O, T1X, ovs);
T1Y = VFMAI(TQ, TP);
STM2(&(xo[18]), T1Y, ovs, &(xo[2]));
STN2(&(xo[16]), T1K, T1Y, ovs);
}
{
V TT, TW, T1Z, T20;
TT = VFMA(LDK(KP951056516), TS, TR);
TW = VFNMS(LDK(KP951056516), TV, TU);
T1Z = VFNMSI(TW, TT);
STM2(&(xo[6]), T1Z, ovs, &(xo[2]));
STN2(&(xo[4]), T1R, T1Z, ovs);
T20 = VFMAI(TW, TT);
STM2(&(xo[34]), T20, ovs, &(xo[2]));
STN2(&(xo[32]), T1N, T20, ovs);
}
}
}
}
}
}
VLEAVE();
}
static const kdft_desc desc = { 20, XSIMD_STRING("n2bv_20"), { 58, 4, 46, 0 }, &GENUS, 0, 2, 0, 0 };
void XSIMD(codelet_n2bv_20) (planner *p) { X(kdft_register) (p, n2bv_20, &desc);
}
#else
/* Generated by: ../../../genfft/gen_notw_c.native -simd -compact -variables 4 -pipeline-latency 8 -sign 1 -n 20 -name n2bv_20 -with-ostride 2 -include dft/simd/n2b.h -store-multiple 2 */
/*
* This function contains 104 FP additions, 24 FP multiplications,
* (or, 92 additions, 12 multiplications, 12 fused multiply/add),
* 57 stack variables, 4 constants, and 50 memory accesses
*/
#include "dft/simd/n2b.h"
static void n2bv_20(const R *ri, const R *ii, R *ro, R *io, stride is, stride os, INT v, INT ivs, INT ovs)
{
DVK(KP587785252, +0.587785252292473129168705954639072768597652438);
DVK(KP951056516, +0.951056516295153572116439333379382143405698634);
DVK(KP250000000, +0.250000000000000000000000000000000000000000000);
DVK(KP559016994, +0.559016994374947424102293417182819058860154590);
{
INT i;
const R *xi;
R *xo;
xi = ii;
xo = io;
for (i = v; i > 0; i = i - VL, xi = xi + (VL * ivs), xo = xo + (VL * ovs), MAKE_VOLATILE_STRIDE(40, is), MAKE_VOLATILE_STRIDE(40, os)) {
V T3, T1y, TH, T1i, Ts, TL, TM, Tz, T13, T16, T1j, T1u, T1v, T1w, T1r;
V T1s, T1t, T1a, T1d, T1k, Ti, Tk, TE, TI;
{
V T1, T2, T1g, TF, TG, T1h;
T1 = LD(&(xi[0]), ivs, &(xi[0]));
T2 = LD(&(xi[WS(is, 10)]), ivs, &(xi[0]));
T1g = VADD(T1, T2);
TF = LD(&(xi[WS(is, 5)]), ivs, &(xi[WS(is, 1)]));
TG = LD(&(xi[WS(is, 15)]), ivs, &(xi[WS(is, 1)]));
T1h = VADD(TF, TG);
T3 = VSUB(T1, T2);
T1y = VADD(T1g, T1h);
TH = VSUB(TF, TG);
T1i = VSUB(T1g, T1h);
}
{
V T6, T11, Tv, T19, Ty, T1c, T9, T14, Td, T18, To, T12, Tr, T15, Tg;
V T1b;
{
V T4, T5, Tt, Tu;
T4 = LD(&(xi[WS(is, 4)]), ivs, &(xi[0]));
T5 = LD(&(xi[WS(is, 14)]), ivs, &(xi[0]));
T6 = VSUB(T4, T5);
T11 = VADD(T4, T5);
Tt = LD(&(xi[WS(is, 13)]), ivs, &(xi[WS(is, 1)]));
Tu = LD(&(xi[WS(is, 3)]), ivs, &(xi[WS(is, 1)]));
Tv = VSUB(Tt, Tu);
T19 = VADD(Tt, Tu);
}
{
V Tw, Tx, T7, T8;
Tw = LD(&(xi[WS(is, 17)]), ivs, &(xi[WS(is, 1)]));
Tx = LD(&(xi[WS(is, 7)]), ivs, &(xi[WS(is, 1)]));
Ty = VSUB(Tw, Tx);
T1c = VADD(Tw, Tx);
T7 = LD(&(xi[WS(is, 16)]), ivs, &(xi[0]));
T8 = LD(&(xi[WS(is, 6)]), ivs, &(xi[0]));
T9 = VSUB(T7, T8);
T14 = VADD(T7, T8);
}
{
V Tb, Tc, Tm, Tn;
Tb = LD(&(xi[WS(is, 8)]), ivs, &(xi[0]));
Tc = LD(&(xi[WS(is, 18)]), ivs, &(xi[0]));
Td = VSUB(Tb, Tc);
T18 = VADD(Tb, Tc);
Tm = LD(&(xi[WS(is, 9)]), ivs, &(xi[WS(is, 1)]));
Tn = LD(&(xi[WS(is, 19)]), ivs, &(xi[WS(is, 1)]));
To = VSUB(Tm, Tn);
T12 = VADD(Tm, Tn);
}
{
V Tp, Tq, Te, Tf;
Tp = LD(&(xi[WS(is, 1)]), ivs, &(xi[WS(is, 1)]));
Tq = LD(&(xi[WS(is, 11)]), ivs, &(xi[WS(is, 1)]));
Tr = VSUB(Tp, Tq);
T15 = VADD(Tp, Tq);
Te = LD(&(xi[WS(is, 12)]), ivs, &(xi[0]));
Tf = LD(&(xi[WS(is, 2)]), ivs, &(xi[0]));
Tg = VSUB(Te, Tf);
T1b = VADD(Te, Tf);
}
Ts = VSUB(To, Tr);
TL = VSUB(T6, T9);
TM = VSUB(Td, Tg);
Tz = VSUB(Tv, Ty);
T13 = VSUB(T11, T12);
T16 = VSUB(T14, T15);
T1j = VADD(T13, T16);
T1u = VADD(T18, T19);
T1v = VADD(T1b, T1c);
T1w = VADD(T1u, T1v);
T1r = VADD(T11, T12);
T1s = VADD(T14, T15);
T1t = VADD(T1r, T1s);
T1a = VSUB(T18, T19);
T1d = VSUB(T1b, T1c);
T1k = VADD(T1a, T1d);
{
V Ta, Th, TC, TD;
Ta = VADD(T6, T9);
Th = VADD(Td, Tg);
Ti = VADD(Ta, Th);
Tk = VMUL(LDK(KP559016994), VSUB(Ta, Th));
TC = VADD(To, Tr);
TD = VADD(Tv, Ty);
TE = VMUL(LDK(KP559016994), VSUB(TC, TD));
TI = VADD(TC, TD);
}
}
{
V T1H, T1J, T1K, T1L, T1N, T1I, TZ, T10;
TZ = VADD(T3, Ti);
T10 = VBYI(VADD(TH, TI));
T1H = VSUB(TZ, T10);
STM2(&(xo[30]), T1H, ovs, &(xo[2]));
T1I = VADD(TZ, T10);
STM2(&(xo[10]), T1I, ovs, &(xo[2]));
{
V T1x, T1z, T1A, T1E, T1G, T1C, T1D, T1F, T1B, T1M;
T1x = VMUL(LDK(KP559016994), VSUB(T1t, T1w));
T1z = VADD(T1t, T1w);
T1A = VFNMS(LDK(KP250000000), T1z, T1y);
T1C = VSUB(T1r, T1s);
T1D = VSUB(T1u, T1v);
T1E = VBYI(VFMA(LDK(KP951056516), T1C, VMUL(LDK(KP587785252), T1D)));
T1G = VBYI(VFNMS(LDK(KP951056516), T1D, VMUL(LDK(KP587785252), T1C)));
T1J = VADD(T1y, T1z);
STM2(&(xo[0]), T1J, ovs, &(xo[0]));
T1F = VSUB(T1A, T1x);
T1K = VSUB(T1F, T1G);
STM2(&(xo[16]), T1K, ovs, &(xo[0]));
T1L = VADD(T1G, T1F);
STM2(&(xo[24]), T1L, ovs, &(xo[0]));
T1B = VADD(T1x, T1A);
T1M = VSUB(T1B, T1E);
STM2(&(xo[8]), T1M, ovs, &(xo[0]));
STN2(&(xo[8]), T1M, T1I, ovs);
T1N = VADD(T1E, T1B);
STM2(&(xo[32]), T1N, ovs, &(xo[0]));
}
{
V T1O, T1P, T1R, T1S;
{
V T1n, T1l, T1m, T1f, T1p, T17, T1e, T1q, T1Q, T1o;
T1n = VMUL(LDK(KP559016994), VSUB(T1j, T1k));
T1l = VADD(T1j, T1k);
T1m = VFNMS(LDK(KP250000000), T1l, T1i);
T17 = VSUB(T13, T16);
T1e = VSUB(T1a, T1d);
T1f = VBYI(VFNMS(LDK(KP951056516), T1e, VMUL(LDK(KP587785252), T17)));
T1p = VBYI(VFMA(LDK(KP951056516), T17, VMUL(LDK(KP587785252), T1e)));
T1O = VADD(T1i, T1l);
STM2(&(xo[20]), T1O, ovs, &(xo[0]));
T1q = VADD(T1n, T1m);
T1P = VADD(T1p, T1q);
STM2(&(xo[12]), T1P, ovs, &(xo[0]));
T1Q = VSUB(T1q, T1p);
STM2(&(xo[28]), T1Q, ovs, &(xo[0]));
STN2(&(xo[28]), T1Q, T1H, ovs);
T1o = VSUB(T1m, T1n);
T1R = VADD(T1f, T1o);
STM2(&(xo[4]), T1R, ovs, &(xo[0]));
T1S = VSUB(T1o, T1f);
STM2(&(xo[36]), T1S, ovs, &(xo[0]));
}
{
V TA, TN, TU, TS, TK, TV, Tl, TR, TJ, Tj;
TA = VFNMS(LDK(KP951056516), Tz, VMUL(LDK(KP587785252), Ts));
TN = VFNMS(LDK(KP951056516), TM, VMUL(LDK(KP587785252), TL));
TU = VFMA(LDK(KP951056516), TL, VMUL(LDK(KP587785252), TM));
TS = VFMA(LDK(KP951056516), Ts, VMUL(LDK(KP587785252), Tz));
TJ = VFNMS(LDK(KP250000000), TI, TH);
TK = VSUB(TE, TJ);
TV = VADD(TE, TJ);
Tj = VFNMS(LDK(KP250000000), Ti, T3);
Tl = VSUB(Tj, Tk);
TR = VADD(Tk, Tj);
{
V TB, TO, T1T, T1U;
TB = VSUB(Tl, TA);
TO = VBYI(VSUB(TK, TN));
T1T = VSUB(TB, TO);
STM2(&(xo[34]), T1T, ovs, &(xo[2]));
STN2(&(xo[32]), T1N, T1T, ovs);
T1U = VADD(TB, TO);
STM2(&(xo[6]), T1U, ovs, &(xo[2]));
STN2(&(xo[4]), T1R, T1U, ovs);
}
{
V TX, TY, T1V, T1W;
TX = VADD(TR, TS);
TY = VBYI(VSUB(TV, TU));
T1V = VSUB(TX, TY);
STM2(&(xo[22]), T1V, ovs, &(xo[2]));
STN2(&(xo[20]), T1O, T1V, ovs);
T1W = VADD(TX, TY);
STM2(&(xo[18]), T1W, ovs, &(xo[2]));
STN2(&(xo[16]), T1K, T1W, ovs);
}
{
V TP, TQ, T1X, T1Y;
TP = VADD(Tl, TA);
TQ = VBYI(VADD(TN, TK));
T1X = VSUB(TP, TQ);
STM2(&(xo[26]), T1X, ovs, &(xo[2]));
STN2(&(xo[24]), T1L, T1X, ovs);
T1Y = VADD(TP, TQ);
STM2(&(xo[14]), T1Y, ovs, &(xo[2]));
STN2(&(xo[12]), T1P, T1Y, ovs);
}
{
V TT, TW, T1Z, T20;
TT = VSUB(TR, TS);
TW = VBYI(VADD(TU, TV));
T1Z = VSUB(TT, TW);
STM2(&(xo[38]), T1Z, ovs, &(xo[2]));
STN2(&(xo[36]), T1S, T1Z, ovs);
T20 = VADD(TT, TW);
STM2(&(xo[2]), T20, ovs, &(xo[2]));
STN2(&(xo[0]), T1J, T20, ovs);
}
}
}
}
}
}
VLEAVE();
}
static const kdft_desc desc = { 20, XSIMD_STRING("n2bv_20"), { 92, 12, 12, 0 }, &GENUS, 0, 2, 0, 0 };
void XSIMD(codelet_n2bv_20) (planner *p) { X(kdft_register) (p, n2bv_20, &desc);
}
#endif