541 lines
18 KiB
C
541 lines
18 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 10:45:57 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 -twiddle-log3 -precompute-twiddles -no-generate-bytw -n 20 -name t3bv_20 -include dft/simd/t3b.h -sign 1 */
|
|
|
|
/*
|
|
* This function contains 138 FP additions, 118 FP multiplications,
|
|
* (or, 92 additions, 72 multiplications, 46 fused multiply/add),
|
|
* 73 stack variables, 4 constants, and 40 memory accesses
|
|
*/
|
|
#include "dft/simd/t3b.h"
|
|
|
|
static void t3bv_20(R *ri, R *ii, const R *W, stride rs, INT mb, INT me, INT ms)
|
|
{
|
|
DVK(KP559016994, +0.559016994374947424102293417182819058860154590);
|
|
DVK(KP618033988, +0.618033988749894848204586834365638117720309180);
|
|
DVK(KP951056516, +0.951056516295153572116439333379382143405698634);
|
|
DVK(KP250000000, +0.250000000000000000000000000000000000000000000);
|
|
{
|
|
INT m;
|
|
R *x;
|
|
x = ii;
|
|
for (m = mb, W = W + (mb * ((TWVL / VL) * 8)); m < me; m = m + VL, x = x + (VL * ms), W = W + (TWVL * 8), MAKE_VOLATILE_STRIDE(20, rs)) {
|
|
V T2, T8, T9, TA, T3, Tc, T4, TV, T14, Tl, Tq, Tx, TQ, Td, Te;
|
|
V T1d, Ti, Tt, T11;
|
|
T2 = LDW(&(W[0]));
|
|
T8 = LDW(&(W[TWVL * 2]));
|
|
T9 = VZMUL(T2, T8);
|
|
TA = VZMULJ(T2, T8);
|
|
T3 = LDW(&(W[TWVL * 4]));
|
|
Tc = VZMULJ(T9, T3);
|
|
T4 = VZMUL(T2, T3);
|
|
TV = VZMUL(T9, T3);
|
|
T14 = VZMULJ(TA, T3);
|
|
Tl = VZMULJ(T8, T3);
|
|
Tq = VZMULJ(T2, T3);
|
|
Tx = VZMUL(T8, T3);
|
|
TQ = VZMUL(TA, T3);
|
|
Td = LDW(&(W[TWVL * 6]));
|
|
Te = VZMULJ(Tc, Td);
|
|
T1d = VZMULJ(T9, Td);
|
|
Ti = VZMULJ(T8, Td);
|
|
Tt = VZMULJ(T2, Td);
|
|
T11 = VZMULJ(TA, Td);
|
|
{
|
|
V T7, T1g, T1F, T23, TU, T1n, T1o, T18, Tp, TE, TF, T27, T28, T29, T1P;
|
|
V T1S, T1T, T1h, T1i, T1j, T24, T25, T26, T1I, T1L, T1M, T1B, T1C;
|
|
{
|
|
V T1, T1f, T6, T1c, T1e, T5, T1b, T1D, T1E;
|
|
T1 = LD(&(x[0]), ms, &(x[0]));
|
|
T1e = LD(&(x[WS(rs, 15)]), ms, &(x[WS(rs, 1)]));
|
|
T1f = VZMUL(T1d, T1e);
|
|
T5 = LD(&(x[WS(rs, 10)]), ms, &(x[0]));
|
|
T6 = VZMUL(T4, T5);
|
|
T1b = LD(&(x[WS(rs, 5)]), ms, &(x[WS(rs, 1)]));
|
|
T1c = VZMUL(Tc, T1b);
|
|
T7 = VSUB(T1, T6);
|
|
T1g = VSUB(T1c, T1f);
|
|
T1D = VADD(T1, T6);
|
|
T1E = VADD(T1c, T1f);
|
|
T1F = VSUB(T1D, T1E);
|
|
T23 = VADD(T1D, T1E);
|
|
}
|
|
{
|
|
V Th, T1G, T10, T1O, T17, T1R, To, T1J, Tw, T1N, TN, T1H, TT, T1K, TD;
|
|
V T1Q;
|
|
{
|
|
V Tb, Tg, Ta, Tf;
|
|
Ta = LD(&(x[WS(rs, 4)]), ms, &(x[0]));
|
|
Tb = VZMUL(T9, Ta);
|
|
Tf = LD(&(x[WS(rs, 14)]), ms, &(x[0]));
|
|
Tg = VZMUL(Te, Tf);
|
|
Th = VSUB(Tb, Tg);
|
|
T1G = VADD(Tb, Tg);
|
|
}
|
|
{
|
|
V TX, TZ, TW, TY;
|
|
TW = LD(&(x[WS(rs, 13)]), ms, &(x[WS(rs, 1)]));
|
|
TX = VZMUL(TV, TW);
|
|
TY = LD(&(x[WS(rs, 3)]), ms, &(x[WS(rs, 1)]));
|
|
TZ = VZMUL(T8, TY);
|
|
T10 = VSUB(TX, TZ);
|
|
T1O = VADD(TX, TZ);
|
|
}
|
|
{
|
|
V T13, T16, T12, T15;
|
|
T12 = LD(&(x[WS(rs, 17)]), ms, &(x[WS(rs, 1)]));
|
|
T13 = VZMUL(T11, T12);
|
|
T15 = LD(&(x[WS(rs, 7)]), ms, &(x[WS(rs, 1)]));
|
|
T16 = VZMUL(T14, T15);
|
|
T17 = VSUB(T13, T16);
|
|
T1R = VADD(T13, T16);
|
|
}
|
|
{
|
|
V Tk, Tn, Tj, Tm;
|
|
Tj = LD(&(x[WS(rs, 16)]), ms, &(x[0]));
|
|
Tk = VZMUL(Ti, Tj);
|
|
Tm = LD(&(x[WS(rs, 6)]), ms, &(x[0]));
|
|
Tn = VZMUL(Tl, Tm);
|
|
To = VSUB(Tk, Tn);
|
|
T1J = VADD(Tk, Tn);
|
|
}
|
|
{
|
|
V Ts, Tv, Tr, Tu;
|
|
Tr = LD(&(x[WS(rs, 8)]), ms, &(x[0]));
|
|
Ts = VZMUL(Tq, Tr);
|
|
Tu = LD(&(x[WS(rs, 18)]), ms, &(x[0]));
|
|
Tv = VZMUL(Tt, Tu);
|
|
Tw = VSUB(Ts, Tv);
|
|
T1N = VADD(Ts, Tv);
|
|
}
|
|
{
|
|
V TK, TM, TJ, TL;
|
|
TJ = LD(&(x[WS(rs, 9)]), ms, &(x[WS(rs, 1)]));
|
|
TK = VZMUL(T3, TJ);
|
|
TL = LD(&(x[WS(rs, 19)]), ms, &(x[WS(rs, 1)]));
|
|
TM = VZMUL(Td, TL);
|
|
TN = VSUB(TK, TM);
|
|
T1H = VADD(TK, TM);
|
|
}
|
|
{
|
|
V TP, TS, TO, TR;
|
|
TO = LD(&(x[WS(rs, 1)]), ms, &(x[WS(rs, 1)]));
|
|
TP = VZMUL(T2, TO);
|
|
TR = LD(&(x[WS(rs, 11)]), ms, &(x[WS(rs, 1)]));
|
|
TS = VZMUL(TQ, TR);
|
|
TT = VSUB(TP, TS);
|
|
T1K = VADD(TP, TS);
|
|
}
|
|
{
|
|
V Tz, TC, Ty, TB;
|
|
Ty = LD(&(x[WS(rs, 12)]), ms, &(x[0]));
|
|
Tz = VZMUL(Tx, Ty);
|
|
TB = LD(&(x[WS(rs, 2)]), ms, &(x[0]));
|
|
TC = VZMUL(TA, TB);
|
|
TD = VSUB(Tz, TC);
|
|
T1Q = VADD(Tz, TC);
|
|
}
|
|
TU = VSUB(TN, TT);
|
|
T1n = VSUB(Th, To);
|
|
T1o = VSUB(Tw, TD);
|
|
T18 = VSUB(T10, T17);
|
|
Tp = VADD(Th, To);
|
|
TE = VADD(Tw, TD);
|
|
TF = VADD(Tp, TE);
|
|
T27 = VADD(T1N, T1O);
|
|
T28 = VADD(T1Q, T1R);
|
|
T29 = VADD(T27, T28);
|
|
T1P = VSUB(T1N, T1O);
|
|
T1S = VSUB(T1Q, T1R);
|
|
T1T = VADD(T1P, T1S);
|
|
T1h = VADD(TN, TT);
|
|
T1i = VADD(T10, T17);
|
|
T1j = VADD(T1h, T1i);
|
|
T24 = VADD(T1G, T1H);
|
|
T25 = VADD(T1J, T1K);
|
|
T26 = VADD(T24, T25);
|
|
T1I = VSUB(T1G, T1H);
|
|
T1L = VSUB(T1J, T1K);
|
|
T1M = VADD(T1I, T1L);
|
|
}
|
|
T1B = VADD(T7, TF);
|
|
T1C = VADD(T1g, T1j);
|
|
ST(&(x[WS(rs, 15)]), VFNMSI(T1C, T1B), ms, &(x[WS(rs, 1)]));
|
|
ST(&(x[WS(rs, 5)]), VFMAI(T1C, T1B), ms, &(x[WS(rs, 1)]));
|
|
{
|
|
V T2c, T2a, T2b, T2g, T2i, T2e, T2f, T2h, T2d;
|
|
T2c = VSUB(T26, T29);
|
|
T2a = VADD(T26, T29);
|
|
T2b = VFNMS(LDK(KP250000000), T2a, T23);
|
|
T2e = VSUB(T24, T25);
|
|
T2f = VSUB(T27, T28);
|
|
T2g = VMUL(LDK(KP951056516), VFMA(LDK(KP618033988), T2f, T2e));
|
|
T2i = VMUL(LDK(KP951056516), VFNMS(LDK(KP618033988), T2e, T2f));
|
|
ST(&(x[0]), VADD(T23, T2a), ms, &(x[0]));
|
|
T2h = VFNMS(LDK(KP559016994), T2c, T2b);
|
|
ST(&(x[WS(rs, 8)]), VFMAI(T2i, T2h), ms, &(x[0]));
|
|
ST(&(x[WS(rs, 12)]), VFNMSI(T2i, T2h), ms, &(x[0]));
|
|
T2d = VFMA(LDK(KP559016994), T2c, T2b);
|
|
ST(&(x[WS(rs, 4)]), VFNMSI(T2g, T2d), ms, &(x[0]));
|
|
ST(&(x[WS(rs, 16)]), VFMAI(T2g, T2d), ms, &(x[0]));
|
|
}
|
|
{
|
|
V T1W, T1U, T1V, T20, T22, T1Y, T1Z, T21, T1X;
|
|
T1W = VSUB(T1M, T1T);
|
|
T1U = VADD(T1M, T1T);
|
|
T1V = VFNMS(LDK(KP250000000), T1U, T1F);
|
|
T1Y = VSUB(T1P, T1S);
|
|
T1Z = VSUB(T1I, T1L);
|
|
T20 = VMUL(LDK(KP951056516), VFNMS(LDK(KP618033988), T1Z, T1Y));
|
|
T22 = VMUL(LDK(KP951056516), VFMA(LDK(KP618033988), T1Y, T1Z));
|
|
ST(&(x[WS(rs, 10)]), VADD(T1F, T1U), ms, &(x[0]));
|
|
T21 = VFMA(LDK(KP559016994), T1W, T1V);
|
|
ST(&(x[WS(rs, 6)]), VFMAI(T22, T21), ms, &(x[0]));
|
|
ST(&(x[WS(rs, 14)]), VFNMSI(T22, T21), ms, &(x[0]));
|
|
T1X = VFNMS(LDK(KP559016994), T1W, T1V);
|
|
ST(&(x[WS(rs, 2)]), VFNMSI(T20, T1X), ms, &(x[0]));
|
|
ST(&(x[WS(rs, 18)]), VFMAI(T20, T1X), ms, &(x[0]));
|
|
}
|
|
{
|
|
V T19, T1p, T1x, T1u, T1m, T1w, TI, T1t;
|
|
T19 = VFMA(LDK(KP618033988), T18, TU);
|
|
T1p = VFMA(LDK(KP618033988), T1o, T1n);
|
|
T1x = VFNMS(LDK(KP618033988), T1n, T1o);
|
|
T1u = VFNMS(LDK(KP618033988), TU, T18);
|
|
{
|
|
V T1k, T1l, TG, TH;
|
|
T1k = VFNMS(LDK(KP250000000), T1j, T1g);
|
|
T1l = VSUB(T1h, T1i);
|
|
T1m = VFMA(LDK(KP559016994), T1l, T1k);
|
|
T1w = VFNMS(LDK(KP559016994), T1l, T1k);
|
|
TG = VFNMS(LDK(KP250000000), TF, T7);
|
|
TH = VSUB(Tp, TE);
|
|
TI = VFMA(LDK(KP559016994), TH, TG);
|
|
T1t = VFNMS(LDK(KP559016994), TH, TG);
|
|
}
|
|
{
|
|
V T1a, T1q, T1z, T1A;
|
|
T1a = VFNMS(LDK(KP951056516), T19, TI);
|
|
T1q = VFMA(LDK(KP951056516), T1p, T1m);
|
|
ST(&(x[WS(rs, 19)]), VFNMSI(T1q, T1a), ms, &(x[WS(rs, 1)]));
|
|
ST(&(x[WS(rs, 1)]), VFMAI(T1q, T1a), ms, &(x[WS(rs, 1)]));
|
|
T1z = VFNMS(LDK(KP951056516), T1u, T1t);
|
|
T1A = VFMA(LDK(KP951056516), T1x, T1w);
|
|
ST(&(x[WS(rs, 7)]), VFNMSI(T1A, T1z), ms, &(x[WS(rs, 1)]));
|
|
ST(&(x[WS(rs, 13)]), VFMAI(T1A, T1z), ms, &(x[WS(rs, 1)]));
|
|
}
|
|
{
|
|
V T1r, T1s, T1v, T1y;
|
|
T1r = VFMA(LDK(KP951056516), T19, TI);
|
|
T1s = VFNMS(LDK(KP951056516), T1p, T1m);
|
|
ST(&(x[WS(rs, 11)]), VFNMSI(T1s, T1r), ms, &(x[WS(rs, 1)]));
|
|
ST(&(x[WS(rs, 9)]), VFMAI(T1s, T1r), ms, &(x[WS(rs, 1)]));
|
|
T1v = VFMA(LDK(KP951056516), T1u, T1t);
|
|
T1y = VFNMS(LDK(KP951056516), T1x, T1w);
|
|
ST(&(x[WS(rs, 3)]), VFNMSI(T1y, T1v), ms, &(x[WS(rs, 1)]));
|
|
ST(&(x[WS(rs, 17)]), VFMAI(T1y, T1v), ms, &(x[WS(rs, 1)]));
|
|
}
|
|
}
|
|
}
|
|
}
|
|
}
|
|
VLEAVE();
|
|
}
|
|
|
|
static const tw_instr twinstr[] = {
|
|
VTW(0, 1),
|
|
VTW(0, 3),
|
|
VTW(0, 9),
|
|
VTW(0, 19),
|
|
{ TW_NEXT, VL, 0 }
|
|
};
|
|
|
|
static const ct_desc desc = { 20, XSIMD_STRING("t3bv_20"), twinstr, &GENUS, { 92, 72, 46, 0 }, 0, 0, 0 };
|
|
|
|
void XSIMD(codelet_t3bv_20) (planner *p) {
|
|
X(kdft_dit_register) (p, t3bv_20, &desc);
|
|
}
|
|
#else
|
|
|
|
/* Generated by: ../../../genfft/gen_twiddle_c.native -simd -compact -variables 4 -pipeline-latency 8 -twiddle-log3 -precompute-twiddles -no-generate-bytw -n 20 -name t3bv_20 -include dft/simd/t3b.h -sign 1 */
|
|
|
|
/*
|
|
* This function contains 138 FP additions, 92 FP multiplications,
|
|
* (or, 126 additions, 80 multiplications, 12 fused multiply/add),
|
|
* 73 stack variables, 4 constants, and 40 memory accesses
|
|
*/
|
|
#include "dft/simd/t3b.h"
|
|
|
|
static void t3bv_20(R *ri, R *ii, const R *W, stride rs, INT mb, INT me, INT ms)
|
|
{
|
|
DVK(KP587785252, +0.587785252292473129168705954639072768597652438);
|
|
DVK(KP951056516, +0.951056516295153572116439333379382143405698634);
|
|
DVK(KP250000000, +0.250000000000000000000000000000000000000000000);
|
|
DVK(KP559016994, +0.559016994374947424102293417182819058860154590);
|
|
{
|
|
INT m;
|
|
R *x;
|
|
x = ii;
|
|
for (m = mb, W = W + (mb * ((TWVL / VL) * 8)); m < me; m = m + VL, x = x + (VL * ms), W = W + (TWVL * 8), MAKE_VOLATILE_STRIDE(20, rs)) {
|
|
V T2, T8, T9, TA, T3, Tc, T4, TV, T14, Tl, Tq, Tx, TQ, Td, Te;
|
|
V T1g, Ti, Tt, T11;
|
|
T2 = LDW(&(W[0]));
|
|
T8 = LDW(&(W[TWVL * 2]));
|
|
T9 = VZMUL(T2, T8);
|
|
TA = VZMULJ(T2, T8);
|
|
T3 = LDW(&(W[TWVL * 4]));
|
|
Tc = VZMULJ(T9, T3);
|
|
T4 = VZMUL(T2, T3);
|
|
TV = VZMUL(T9, T3);
|
|
T14 = VZMULJ(TA, T3);
|
|
Tl = VZMULJ(T8, T3);
|
|
Tq = VZMULJ(T2, T3);
|
|
Tx = VZMUL(T8, T3);
|
|
TQ = VZMUL(TA, T3);
|
|
Td = LDW(&(W[TWVL * 6]));
|
|
Te = VZMULJ(Tc, Td);
|
|
T1g = VZMULJ(T9, Td);
|
|
Ti = VZMULJ(T8, Td);
|
|
Tt = VZMULJ(T2, Td);
|
|
T11 = VZMULJ(TA, Td);
|
|
{
|
|
V T7, T1j, T1U, T2a, TU, T1n, T1o, T18, Tp, TE, TF, T26, T27, T28, T1M;
|
|
V T1P, T1W, T1b, T1c, T1k, T23, T24, T25, T1F, T1I, T1V, T1B, T1C;
|
|
{
|
|
V T1, T1i, T6, T1f, T1h, T5, T1e, T1S, T1T;
|
|
T1 = LD(&(x[0]), ms, &(x[0]));
|
|
T1h = LD(&(x[WS(rs, 15)]), ms, &(x[WS(rs, 1)]));
|
|
T1i = VZMUL(T1g, T1h);
|
|
T5 = LD(&(x[WS(rs, 10)]), ms, &(x[0]));
|
|
T6 = VZMUL(T4, T5);
|
|
T1e = LD(&(x[WS(rs, 5)]), ms, &(x[WS(rs, 1)]));
|
|
T1f = VZMUL(Tc, T1e);
|
|
T7 = VSUB(T1, T6);
|
|
T1j = VSUB(T1f, T1i);
|
|
T1S = VADD(T1, T6);
|
|
T1T = VADD(T1f, T1i);
|
|
T1U = VSUB(T1S, T1T);
|
|
T2a = VADD(T1S, T1T);
|
|
}
|
|
{
|
|
V Th, T1D, T10, T1L, T17, T1O, To, T1G, Tw, T1K, TN, T1E, TT, T1H, TD;
|
|
V T1N;
|
|
{
|
|
V Tb, Tg, Ta, Tf;
|
|
Ta = LD(&(x[WS(rs, 4)]), ms, &(x[0]));
|
|
Tb = VZMUL(T9, Ta);
|
|
Tf = LD(&(x[WS(rs, 14)]), ms, &(x[0]));
|
|
Tg = VZMUL(Te, Tf);
|
|
Th = VSUB(Tb, Tg);
|
|
T1D = VADD(Tb, Tg);
|
|
}
|
|
{
|
|
V TX, TZ, TW, TY;
|
|
TW = LD(&(x[WS(rs, 13)]), ms, &(x[WS(rs, 1)]));
|
|
TX = VZMUL(TV, TW);
|
|
TY = LD(&(x[WS(rs, 3)]), ms, &(x[WS(rs, 1)]));
|
|
TZ = VZMUL(T8, TY);
|
|
T10 = VSUB(TX, TZ);
|
|
T1L = VADD(TX, TZ);
|
|
}
|
|
{
|
|
V T13, T16, T12, T15;
|
|
T12 = LD(&(x[WS(rs, 17)]), ms, &(x[WS(rs, 1)]));
|
|
T13 = VZMUL(T11, T12);
|
|
T15 = LD(&(x[WS(rs, 7)]), ms, &(x[WS(rs, 1)]));
|
|
T16 = VZMUL(T14, T15);
|
|
T17 = VSUB(T13, T16);
|
|
T1O = VADD(T13, T16);
|
|
}
|
|
{
|
|
V Tk, Tn, Tj, Tm;
|
|
Tj = LD(&(x[WS(rs, 16)]), ms, &(x[0]));
|
|
Tk = VZMUL(Ti, Tj);
|
|
Tm = LD(&(x[WS(rs, 6)]), ms, &(x[0]));
|
|
Tn = VZMUL(Tl, Tm);
|
|
To = VSUB(Tk, Tn);
|
|
T1G = VADD(Tk, Tn);
|
|
}
|
|
{
|
|
V Ts, Tv, Tr, Tu;
|
|
Tr = LD(&(x[WS(rs, 8)]), ms, &(x[0]));
|
|
Ts = VZMUL(Tq, Tr);
|
|
Tu = LD(&(x[WS(rs, 18)]), ms, &(x[0]));
|
|
Tv = VZMUL(Tt, Tu);
|
|
Tw = VSUB(Ts, Tv);
|
|
T1K = VADD(Ts, Tv);
|
|
}
|
|
{
|
|
V TK, TM, TJ, TL;
|
|
TJ = LD(&(x[WS(rs, 9)]), ms, &(x[WS(rs, 1)]));
|
|
TK = VZMUL(T3, TJ);
|
|
TL = LD(&(x[WS(rs, 19)]), ms, &(x[WS(rs, 1)]));
|
|
TM = VZMUL(Td, TL);
|
|
TN = VSUB(TK, TM);
|
|
T1E = VADD(TK, TM);
|
|
}
|
|
{
|
|
V TP, TS, TO, TR;
|
|
TO = LD(&(x[WS(rs, 1)]), ms, &(x[WS(rs, 1)]));
|
|
TP = VZMUL(T2, TO);
|
|
TR = LD(&(x[WS(rs, 11)]), ms, &(x[WS(rs, 1)]));
|
|
TS = VZMUL(TQ, TR);
|
|
TT = VSUB(TP, TS);
|
|
T1H = VADD(TP, TS);
|
|
}
|
|
{
|
|
V Tz, TC, Ty, TB;
|
|
Ty = LD(&(x[WS(rs, 12)]), ms, &(x[0]));
|
|
Tz = VZMUL(Tx, Ty);
|
|
TB = LD(&(x[WS(rs, 2)]), ms, &(x[0]));
|
|
TC = VZMUL(TA, TB);
|
|
TD = VSUB(Tz, TC);
|
|
T1N = VADD(Tz, TC);
|
|
}
|
|
TU = VSUB(TN, TT);
|
|
T1n = VSUB(Th, To);
|
|
T1o = VSUB(Tw, TD);
|
|
T18 = VSUB(T10, T17);
|
|
Tp = VADD(Th, To);
|
|
TE = VADD(Tw, TD);
|
|
TF = VADD(Tp, TE);
|
|
T26 = VADD(T1K, T1L);
|
|
T27 = VADD(T1N, T1O);
|
|
T28 = VADD(T26, T27);
|
|
T1M = VSUB(T1K, T1L);
|
|
T1P = VSUB(T1N, T1O);
|
|
T1W = VADD(T1M, T1P);
|
|
T1b = VADD(TN, TT);
|
|
T1c = VADD(T10, T17);
|
|
T1k = VADD(T1b, T1c);
|
|
T23 = VADD(T1D, T1E);
|
|
T24 = VADD(T1G, T1H);
|
|
T25 = VADD(T23, T24);
|
|
T1F = VSUB(T1D, T1E);
|
|
T1I = VSUB(T1G, T1H);
|
|
T1V = VADD(T1F, T1I);
|
|
}
|
|
T1B = VADD(T7, TF);
|
|
T1C = VBYI(VADD(T1j, T1k));
|
|
ST(&(x[WS(rs, 15)]), VSUB(T1B, T1C), ms, &(x[WS(rs, 1)]));
|
|
ST(&(x[WS(rs, 5)]), VADD(T1B, T1C), ms, &(x[WS(rs, 1)]));
|
|
{
|
|
V T29, T2b, T2c, T2g, T2i, T2e, T2f, T2h, T2d;
|
|
T29 = VMUL(LDK(KP559016994), VSUB(T25, T28));
|
|
T2b = VADD(T25, T28);
|
|
T2c = VFNMS(LDK(KP250000000), T2b, T2a);
|
|
T2e = VSUB(T23, T24);
|
|
T2f = VSUB(T26, T27);
|
|
T2g = VBYI(VFMA(LDK(KP951056516), T2e, VMUL(LDK(KP587785252), T2f)));
|
|
T2i = VBYI(VFNMS(LDK(KP951056516), T2f, VMUL(LDK(KP587785252), T2e)));
|
|
ST(&(x[0]), VADD(T2a, T2b), ms, &(x[0]));
|
|
T2h = VSUB(T2c, T29);
|
|
ST(&(x[WS(rs, 8)]), VSUB(T2h, T2i), ms, &(x[0]));
|
|
ST(&(x[WS(rs, 12)]), VADD(T2i, T2h), ms, &(x[0]));
|
|
T2d = VADD(T29, T2c);
|
|
ST(&(x[WS(rs, 4)]), VSUB(T2d, T2g), ms, &(x[0]));
|
|
ST(&(x[WS(rs, 16)]), VADD(T2g, T2d), ms, &(x[0]));
|
|
}
|
|
{
|
|
V T1Z, T1X, T1Y, T1R, T21, T1J, T1Q, T22, T20;
|
|
T1Z = VMUL(LDK(KP559016994), VSUB(T1V, T1W));
|
|
T1X = VADD(T1V, T1W);
|
|
T1Y = VFNMS(LDK(KP250000000), T1X, T1U);
|
|
T1J = VSUB(T1F, T1I);
|
|
T1Q = VSUB(T1M, T1P);
|
|
T1R = VBYI(VFNMS(LDK(KP951056516), T1Q, VMUL(LDK(KP587785252), T1J)));
|
|
T21 = VBYI(VFMA(LDK(KP951056516), T1J, VMUL(LDK(KP587785252), T1Q)));
|
|
ST(&(x[WS(rs, 10)]), VADD(T1U, T1X), ms, &(x[0]));
|
|
T22 = VADD(T1Z, T1Y);
|
|
ST(&(x[WS(rs, 6)]), VADD(T21, T22), ms, &(x[0]));
|
|
ST(&(x[WS(rs, 14)]), VSUB(T22, T21), ms, &(x[0]));
|
|
T20 = VSUB(T1Y, T1Z);
|
|
ST(&(x[WS(rs, 2)]), VADD(T1R, T20), ms, &(x[0]));
|
|
ST(&(x[WS(rs, 18)]), VSUB(T20, T1R), ms, &(x[0]));
|
|
}
|
|
{
|
|
V T19, T1p, T1w, T1u, T1m, T1x, TI, T1t;
|
|
T19 = VFNMS(LDK(KP951056516), T18, VMUL(LDK(KP587785252), TU));
|
|
T1p = VFNMS(LDK(KP951056516), T1o, VMUL(LDK(KP587785252), T1n));
|
|
T1w = VFMA(LDK(KP951056516), T1n, VMUL(LDK(KP587785252), T1o));
|
|
T1u = VFMA(LDK(KP951056516), TU, VMUL(LDK(KP587785252), T18));
|
|
{
|
|
V T1d, T1l, TG, TH;
|
|
T1d = VMUL(LDK(KP559016994), VSUB(T1b, T1c));
|
|
T1l = VFNMS(LDK(KP250000000), T1k, T1j);
|
|
T1m = VSUB(T1d, T1l);
|
|
T1x = VADD(T1d, T1l);
|
|
TG = VFNMS(LDK(KP250000000), TF, T7);
|
|
TH = VMUL(LDK(KP559016994), VSUB(Tp, TE));
|
|
TI = VSUB(TG, TH);
|
|
T1t = VADD(TH, TG);
|
|
}
|
|
{
|
|
V T1a, T1q, T1z, T1A;
|
|
T1a = VSUB(TI, T19);
|
|
T1q = VBYI(VSUB(T1m, T1p));
|
|
ST(&(x[WS(rs, 17)]), VSUB(T1a, T1q), ms, &(x[WS(rs, 1)]));
|
|
ST(&(x[WS(rs, 3)]), VADD(T1a, T1q), ms, &(x[WS(rs, 1)]));
|
|
T1z = VADD(T1t, T1u);
|
|
T1A = VBYI(VSUB(T1x, T1w));
|
|
ST(&(x[WS(rs, 11)]), VSUB(T1z, T1A), ms, &(x[WS(rs, 1)]));
|
|
ST(&(x[WS(rs, 9)]), VADD(T1z, T1A), ms, &(x[WS(rs, 1)]));
|
|
}
|
|
{
|
|
V T1r, T1s, T1v, T1y;
|
|
T1r = VADD(TI, T19);
|
|
T1s = VBYI(VADD(T1p, T1m));
|
|
ST(&(x[WS(rs, 13)]), VSUB(T1r, T1s), ms, &(x[WS(rs, 1)]));
|
|
ST(&(x[WS(rs, 7)]), VADD(T1r, T1s), ms, &(x[WS(rs, 1)]));
|
|
T1v = VSUB(T1t, T1u);
|
|
T1y = VBYI(VADD(T1w, T1x));
|
|
ST(&(x[WS(rs, 19)]), VSUB(T1v, T1y), ms, &(x[WS(rs, 1)]));
|
|
ST(&(x[WS(rs, 1)]), VADD(T1v, T1y), ms, &(x[WS(rs, 1)]));
|
|
}
|
|
}
|
|
}
|
|
}
|
|
}
|
|
VLEAVE();
|
|
}
|
|
|
|
static const tw_instr twinstr[] = {
|
|
VTW(0, 1),
|
|
VTW(0, 3),
|
|
VTW(0, 9),
|
|
VTW(0, 19),
|
|
{ TW_NEXT, VL, 0 }
|
|
};
|
|
|
|
static const ct_desc desc = { 20, XSIMD_STRING("t3bv_20"), twinstr, &GENUS, { 126, 80, 12, 0 }, 0, 0, 0 };
|
|
|
|
void XSIMD(codelet_t3bv_20) (planner *p) {
|
|
X(kdft_dit_register) (p, t3bv_20, &desc);
|
|
}
|
|
#endif
|