388 lines
9.9 KiB
C
388 lines
9.9 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 10:46:55 EDT 2021 */
|
|
|
|
#include "rdft/codelet-rdft.h"
|
|
|
|
#if defined(ARCH_PREFERS_FMA) || defined(ISA_EXTENSION_PREFERS_FMA)
|
|
|
|
/* Generated by: ../../../genfft/gen_hc2hc.native -fma -compact -variables 4 -pipeline-latency 4 -sign 1 -twiddle-log3 -precompute-twiddles -n 8 -dif -name hb2_8 -include rdft/scalar/hb.h */
|
|
|
|
/*
|
|
* This function contains 74 FP additions, 50 FP multiplications,
|
|
* (or, 44 additions, 20 multiplications, 30 fused multiply/add),
|
|
* 47 stack variables, 1 constants, and 32 memory accesses
|
|
*/
|
|
#include "rdft/scalar/hb.h"
|
|
|
|
static void hb2_8(R *cr, R *ci, const R *W, stride rs, INT mb, INT me, INT ms)
|
|
{
|
|
DK(KP707106781, +0.707106781186547524400844362104849039284835938);
|
|
{
|
|
INT m;
|
|
for (m = mb, W = W + ((mb - 1) * 6); m < me; m = m + 1, cr = cr + ms, ci = ci - ms, W = W + 6, MAKE_VOLATILE_STRIDE(16, rs)) {
|
|
E Tf, Tg, Tl, Tp, Ti, Tj, Tk, T1b, T1u, T1e, T1o, To, Tq, TK;
|
|
{
|
|
E Th, T1n, T1t, Tn, Tm, TJ;
|
|
Tf = W[0];
|
|
Tg = W[2];
|
|
Th = Tf * Tg;
|
|
Tl = W[4];
|
|
T1n = Tf * Tl;
|
|
Tp = W[5];
|
|
T1t = Tf * Tp;
|
|
Ti = W[1];
|
|
Tj = W[3];
|
|
Tn = Tf * Tj;
|
|
Tk = FMA(Ti, Tj, Th);
|
|
T1b = FNMS(Ti, Tj, Th);
|
|
T1u = FNMS(Ti, Tl, T1t);
|
|
T1e = FMA(Ti, Tg, Tn);
|
|
T1o = FMA(Ti, Tp, T1n);
|
|
Tm = Tk * Tl;
|
|
TJ = Tk * Tp;
|
|
To = FNMS(Ti, Tg, Tn);
|
|
Tq = FMA(To, Tp, Tm);
|
|
TK = FNMS(To, Tl, TJ);
|
|
}
|
|
{
|
|
E T7, T1p, T1v, Tv, TP, T13, T1h, TZ, Te, T1k, T1w, T1q, TQ, TR, T10;
|
|
E TG, T14;
|
|
{
|
|
E T3, Tr, TO, T1f, T6, TL, Tu, T1g;
|
|
{
|
|
E T1, T2, TM, TN;
|
|
T1 = cr[0];
|
|
T2 = ci[WS(rs, 3)];
|
|
T3 = T1 + T2;
|
|
Tr = T1 - T2;
|
|
TM = ci[WS(rs, 7)];
|
|
TN = cr[WS(rs, 4)];
|
|
TO = TM + TN;
|
|
T1f = TM - TN;
|
|
}
|
|
{
|
|
E T4, T5, Ts, Tt;
|
|
T4 = cr[WS(rs, 2)];
|
|
T5 = ci[WS(rs, 1)];
|
|
T6 = T4 + T5;
|
|
TL = T4 - T5;
|
|
Ts = ci[WS(rs, 5)];
|
|
Tt = cr[WS(rs, 6)];
|
|
Tu = Ts + Tt;
|
|
T1g = Ts - Tt;
|
|
}
|
|
T7 = T3 + T6;
|
|
T1p = T3 - T6;
|
|
T1v = T1f - T1g;
|
|
Tv = Tr - Tu;
|
|
TP = TL + TO;
|
|
T13 = TO - TL;
|
|
T1h = T1f + T1g;
|
|
TZ = Tr + Tu;
|
|
}
|
|
{
|
|
E Ta, Tw, TE, T1j, Td, TB, Tz, T1i, TA, TF;
|
|
{
|
|
E T8, T9, TC, TD;
|
|
T8 = cr[WS(rs, 1)];
|
|
T9 = ci[WS(rs, 2)];
|
|
Ta = T8 + T9;
|
|
Tw = T8 - T9;
|
|
TC = ci[WS(rs, 4)];
|
|
TD = cr[WS(rs, 7)];
|
|
TE = TC + TD;
|
|
T1j = TC - TD;
|
|
}
|
|
{
|
|
E Tb, Tc, Tx, Ty;
|
|
Tb = ci[0];
|
|
Tc = cr[WS(rs, 3)];
|
|
Td = Tb + Tc;
|
|
TB = Tb - Tc;
|
|
Tx = ci[WS(rs, 6)];
|
|
Ty = cr[WS(rs, 5)];
|
|
Tz = Tx + Ty;
|
|
T1i = Tx - Ty;
|
|
}
|
|
Te = Ta + Td;
|
|
T1k = T1i + T1j;
|
|
T1w = Ta - Td;
|
|
T1q = T1j - T1i;
|
|
TQ = Tw + Tz;
|
|
TR = TB + TE;
|
|
T10 = TQ + TR;
|
|
TA = Tw - Tz;
|
|
TF = TB - TE;
|
|
TG = TA + TF;
|
|
T14 = TA - TF;
|
|
}
|
|
cr[0] = T7 + Te;
|
|
ci[0] = T1h + T1k;
|
|
{
|
|
E T11, T12, T15, T16;
|
|
T11 = FNMS(KP707106781, T10, TZ);
|
|
T12 = Tg * T11;
|
|
T15 = FMA(KP707106781, T14, T13);
|
|
T16 = Tg * T15;
|
|
cr[WS(rs, 3)] = FNMS(Tj, T15, T12);
|
|
ci[WS(rs, 3)] = FMA(Tj, T11, T16);
|
|
}
|
|
{
|
|
E T1z, T1A, T1B, T1C;
|
|
T1z = T1p + T1q;
|
|
T1A = Tk * T1z;
|
|
T1B = T1w + T1v;
|
|
T1C = Tk * T1B;
|
|
cr[WS(rs, 2)] = FNMS(To, T1B, T1A);
|
|
ci[WS(rs, 2)] = FMA(To, T1z, T1C);
|
|
}
|
|
{
|
|
E T17, T18, T19, T1a;
|
|
T17 = FMA(KP707106781, T10, TZ);
|
|
T18 = Tl * T17;
|
|
T19 = FNMS(KP707106781, T14, T13);
|
|
T1a = Tl * T19;
|
|
cr[WS(rs, 7)] = FNMS(Tp, T19, T18);
|
|
ci[WS(rs, 7)] = FMA(Tp, T17, T1a);
|
|
}
|
|
{
|
|
E T1l, T1d, T1m, T1c;
|
|
T1l = T1h - T1k;
|
|
T1c = T7 - Te;
|
|
T1d = T1b * T1c;
|
|
T1m = T1e * T1c;
|
|
cr[WS(rs, 4)] = FNMS(T1e, T1l, T1d);
|
|
ci[WS(rs, 4)] = FMA(T1b, T1l, T1m);
|
|
}
|
|
{
|
|
E T1r, T1s, T1x, T1y;
|
|
T1r = T1p - T1q;
|
|
T1s = T1o * T1r;
|
|
T1x = T1v - T1w;
|
|
T1y = T1o * T1x;
|
|
cr[WS(rs, 6)] = FNMS(T1u, T1x, T1s);
|
|
ci[WS(rs, 6)] = FMA(T1u, T1r, T1y);
|
|
}
|
|
{
|
|
E TT, TX, TW, TY, TI, TU, TS, TV, TH;
|
|
TS = TQ - TR;
|
|
TT = FNMS(KP707106781, TS, TP);
|
|
TX = FMA(KP707106781, TS, TP);
|
|
TV = FMA(KP707106781, TG, Tv);
|
|
TW = Tf * TV;
|
|
TY = Ti * TV;
|
|
TH = FNMS(KP707106781, TG, Tv);
|
|
TI = Tq * TH;
|
|
TU = TK * TH;
|
|
cr[WS(rs, 5)] = FNMS(TK, TT, TI);
|
|
ci[WS(rs, 5)] = FMA(Tq, TT, TU);
|
|
cr[WS(rs, 1)] = FNMS(Ti, TX, TW);
|
|
ci[WS(rs, 1)] = FMA(Tf, TX, TY);
|
|
}
|
|
}
|
|
}
|
|
}
|
|
}
|
|
|
|
static const tw_instr twinstr[] = {
|
|
{ TW_CEXP, 1, 1 },
|
|
{ TW_CEXP, 1, 3 },
|
|
{ TW_CEXP, 1, 7 },
|
|
{ TW_NEXT, 1, 0 }
|
|
};
|
|
|
|
static const hc2hc_desc desc = { 8, "hb2_8", twinstr, &GENUS, { 44, 20, 30, 0 } };
|
|
|
|
void X(codelet_hb2_8) (planner *p) {
|
|
X(khc2hc_register) (p, hb2_8, &desc);
|
|
}
|
|
#else
|
|
|
|
/* Generated by: ../../../genfft/gen_hc2hc.native -compact -variables 4 -pipeline-latency 4 -sign 1 -twiddle-log3 -precompute-twiddles -n 8 -dif -name hb2_8 -include rdft/scalar/hb.h */
|
|
|
|
/*
|
|
* This function contains 74 FP additions, 44 FP multiplications,
|
|
* (or, 56 additions, 26 multiplications, 18 fused multiply/add),
|
|
* 46 stack variables, 1 constants, and 32 memory accesses
|
|
*/
|
|
#include "rdft/scalar/hb.h"
|
|
|
|
static void hb2_8(R *cr, R *ci, const R *W, stride rs, INT mb, INT me, INT ms)
|
|
{
|
|
DK(KP707106781, +0.707106781186547524400844362104849039284835938);
|
|
{
|
|
INT m;
|
|
for (m = mb, W = W + ((mb - 1) * 6); m < me; m = m + 1, cr = cr + ms, ci = ci - ms, W = W + 6, MAKE_VOLATILE_STRIDE(16, rs)) {
|
|
E Tf, Ti, Tg, Tj, Tl, Tp, TP, TR, TF, TG, TH, T15, TL, TT;
|
|
{
|
|
E Th, To, Tk, Tn;
|
|
Tf = W[0];
|
|
Ti = W[1];
|
|
Tg = W[2];
|
|
Tj = W[3];
|
|
Th = Tf * Tg;
|
|
To = Ti * Tg;
|
|
Tk = Ti * Tj;
|
|
Tn = Tf * Tj;
|
|
Tl = Th - Tk;
|
|
Tp = Tn + To;
|
|
TP = Th + Tk;
|
|
TR = Tn - To;
|
|
TF = W[4];
|
|
TG = W[5];
|
|
TH = FMA(Tf, TF, Ti * TG);
|
|
T15 = FNMS(TR, TF, TP * TG);
|
|
TL = FNMS(Ti, TF, Tf * TG);
|
|
TT = FMA(TP, TF, TR * TG);
|
|
}
|
|
{
|
|
E T7, T1f, T1i, Tw, TI, TW, T18, TM, Te, T19, T1a, TD, TJ, TZ, T12;
|
|
E TN, Tm, TE;
|
|
{
|
|
E T3, TU, Tv, TV, T6, T16, Ts, T17;
|
|
{
|
|
E T1, T2, Tt, Tu;
|
|
T1 = cr[0];
|
|
T2 = ci[WS(rs, 3)];
|
|
T3 = T1 + T2;
|
|
TU = T1 - T2;
|
|
Tt = ci[WS(rs, 5)];
|
|
Tu = cr[WS(rs, 6)];
|
|
Tv = Tt - Tu;
|
|
TV = Tt + Tu;
|
|
}
|
|
{
|
|
E T4, T5, Tq, Tr;
|
|
T4 = cr[WS(rs, 2)];
|
|
T5 = ci[WS(rs, 1)];
|
|
T6 = T4 + T5;
|
|
T16 = T4 - T5;
|
|
Tq = ci[WS(rs, 7)];
|
|
Tr = cr[WS(rs, 4)];
|
|
Ts = Tq - Tr;
|
|
T17 = Tq + Tr;
|
|
}
|
|
T7 = T3 + T6;
|
|
T1f = TU + TV;
|
|
T1i = T17 - T16;
|
|
Tw = Ts + Tv;
|
|
TI = T3 - T6;
|
|
TW = TU - TV;
|
|
T18 = T16 + T17;
|
|
TM = Ts - Tv;
|
|
}
|
|
{
|
|
E Ta, TX, TC, T11, Td, T10, Tz, TY;
|
|
{
|
|
E T8, T9, TA, TB;
|
|
T8 = cr[WS(rs, 1)];
|
|
T9 = ci[WS(rs, 2)];
|
|
Ta = T8 + T9;
|
|
TX = T8 - T9;
|
|
TA = ci[WS(rs, 4)];
|
|
TB = cr[WS(rs, 7)];
|
|
TC = TA - TB;
|
|
T11 = TA + TB;
|
|
}
|
|
{
|
|
E Tb, Tc, Tx, Ty;
|
|
Tb = ci[0];
|
|
Tc = cr[WS(rs, 3)];
|
|
Td = Tb + Tc;
|
|
T10 = Tb - Tc;
|
|
Tx = ci[WS(rs, 6)];
|
|
Ty = cr[WS(rs, 5)];
|
|
Tz = Tx - Ty;
|
|
TY = Tx + Ty;
|
|
}
|
|
Te = Ta + Td;
|
|
T19 = TX + TY;
|
|
T1a = T10 + T11;
|
|
TD = Tz + TC;
|
|
TJ = TC - Tz;
|
|
TZ = TX - TY;
|
|
T12 = T10 - T11;
|
|
TN = Ta - Td;
|
|
}
|
|
cr[0] = T7 + Te;
|
|
ci[0] = Tw + TD;
|
|
Tm = T7 - Te;
|
|
TE = Tw - TD;
|
|
cr[WS(rs, 4)] = FNMS(Tp, TE, Tl * Tm);
|
|
ci[WS(rs, 4)] = FMA(Tp, Tm, Tl * TE);
|
|
{
|
|
E TQ, TS, TK, TO;
|
|
TQ = TI + TJ;
|
|
TS = TN + TM;
|
|
cr[WS(rs, 2)] = FNMS(TR, TS, TP * TQ);
|
|
ci[WS(rs, 2)] = FMA(TP, TS, TR * TQ);
|
|
TK = TI - TJ;
|
|
TO = TM - TN;
|
|
cr[WS(rs, 6)] = FNMS(TL, TO, TH * TK);
|
|
ci[WS(rs, 6)] = FMA(TH, TO, TL * TK);
|
|
}
|
|
{
|
|
E T1h, T1l, T1k, T1m, T1g, T1j;
|
|
T1g = KP707106781 * (T19 + T1a);
|
|
T1h = T1f - T1g;
|
|
T1l = T1f + T1g;
|
|
T1j = KP707106781 * (TZ - T12);
|
|
T1k = T1i + T1j;
|
|
T1m = T1i - T1j;
|
|
cr[WS(rs, 3)] = FNMS(Tj, T1k, Tg * T1h);
|
|
ci[WS(rs, 3)] = FMA(Tg, T1k, Tj * T1h);
|
|
cr[WS(rs, 7)] = FNMS(TG, T1m, TF * T1l);
|
|
ci[WS(rs, 7)] = FMA(TF, T1m, TG * T1l);
|
|
}
|
|
{
|
|
E T14, T1d, T1c, T1e, T13, T1b;
|
|
T13 = KP707106781 * (TZ + T12);
|
|
T14 = TW - T13;
|
|
T1d = TW + T13;
|
|
T1b = KP707106781 * (T19 - T1a);
|
|
T1c = T18 - T1b;
|
|
T1e = T18 + T1b;
|
|
cr[WS(rs, 5)] = FNMS(T15, T1c, TT * T14);
|
|
ci[WS(rs, 5)] = FMA(T15, T14, TT * T1c);
|
|
cr[WS(rs, 1)] = FNMS(Ti, T1e, Tf * T1d);
|
|
ci[WS(rs, 1)] = FMA(Ti, T1d, Tf * T1e);
|
|
}
|
|
}
|
|
}
|
|
}
|
|
}
|
|
|
|
static const tw_instr twinstr[] = {
|
|
{ TW_CEXP, 1, 1 },
|
|
{ TW_CEXP, 1, 3 },
|
|
{ TW_CEXP, 1, 7 },
|
|
{ TW_NEXT, 1, 0 }
|
|
};
|
|
|
|
static const hc2hc_desc desc = { 8, "hb2_8", twinstr, &GENUS, { 56, 26, 18, 0 } };
|
|
|
|
void X(codelet_hb2_8) (planner *p) {
|
|
X(khc2hc_register) (p, hb2_8, &desc);
|
|
}
|
|
#endif
|