furnace/extern/fftw/rdft/scalar/r2cf/hc2cf_8.c

377 lines
9.2 KiB
C
Raw Normal View 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 10:46:31 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 8 -dit -name hc2cf_8 -include rdft/scalar/hc2cf.h */
/*
* This function contains 66 FP additions, 36 FP multiplications,
* (or, 44 additions, 14 multiplications, 22 fused multiply/add),
* 34 stack variables, 1 constants, and 32 memory accesses
*/
#include "rdft/scalar/hc2cf.h"
static void hc2cf_8(R *Rp, R *Ip, R *Rm, R *Im, const R *W, stride rs, INT mb, INT me, INT ms)
{
DK(KP707106781, +0.707106781186547524400844362104849039284835938);
{
INT m;
for (m = mb, W = W + ((mb - 1) * 14); m < me; m = m + 1, Rp = Rp + ms, Ip = Ip + ms, Rm = Rm - ms, Im = Im - ms, W = W + 14, MAKE_VOLATILE_STRIDE(32, rs)) {
E T1, T1m, T7, T1l, Tk, TS, Te, TQ, TF, T14, TL, T16, T12, T17, Ts;
E TX, Ty, TZ, TV, T10;
T1 = Rp[0];
T1m = Rm[0];
{
E T3, T6, T4, T1k, T2, T5;
T3 = Rp[WS(rs, 2)];
T6 = Rm[WS(rs, 2)];
T2 = W[6];
T4 = T2 * T3;
T1k = T2 * T6;
T5 = W[7];
T7 = FMA(T5, T6, T4);
T1l = FNMS(T5, T3, T1k);
}
{
E Tg, Tj, Th, TR, Tf, Ti;
Tg = Rp[WS(rs, 3)];
Tj = Rm[WS(rs, 3)];
Tf = W[10];
Th = Tf * Tg;
TR = Tf * Tj;
Ti = W[11];
Tk = FMA(Ti, Tj, Th);
TS = FNMS(Ti, Tg, TR);
}
{
E Ta, Td, Tb, TP, T9, Tc;
Ta = Rp[WS(rs, 1)];
Td = Rm[WS(rs, 1)];
T9 = W[2];
Tb = T9 * Ta;
TP = T9 * Td;
Tc = W[3];
Te = FMA(Tc, Td, Tb);
TQ = FNMS(Tc, Ta, TP);
}
{
E TB, TE, TC, T13, TH, TK, TI, T15, TA, TG, TD, TJ;
TB = Ip[WS(rs, 3)];
TE = Im[WS(rs, 3)];
TA = W[12];
TC = TA * TB;
T13 = TA * TE;
TH = Ip[WS(rs, 1)];
TK = Im[WS(rs, 1)];
TG = W[4];
TI = TG * TH;
T15 = TG * TK;
TD = W[13];
TF = FMA(TD, TE, TC);
T14 = FNMS(TD, TB, T13);
TJ = W[5];
TL = FMA(TJ, TK, TI);
T16 = FNMS(TJ, TH, T15);
T12 = TF - TL;
T17 = T14 - T16;
}
{
E To, Tr, Tp, TW, Tu, Tx, Tv, TY, Tn, Tt, Tq, Tw;
To = Ip[0];
Tr = Im[0];
Tn = W[0];
Tp = Tn * To;
TW = Tn * Tr;
Tu = Ip[WS(rs, 2)];
Tx = Im[WS(rs, 2)];
Tt = W[8];
Tv = Tt * Tu;
TY = Tt * Tx;
Tq = W[1];
Ts = FMA(Tq, Tr, Tp);
TX = FNMS(Tq, To, TW);
Tw = W[9];
Ty = FMA(Tw, Tx, Tv);
TZ = FNMS(Tw, Tu, TY);
TV = Ts - Ty;
T10 = TX - TZ;
}
{
E TU, T1a, T1t, T1v, T19, T1w, T1d, T1u;
{
E TO, TT, T1r, T1s;
TO = T1 - T7;
TT = TQ - TS;
TU = TO + TT;
T1a = TO - TT;
T1r = T1m - T1l;
T1s = Te - Tk;
T1t = T1r - T1s;
T1v = T1s + T1r;
}
{
E T11, T18, T1b, T1c;
T11 = TV + T10;
T18 = T12 - T17;
T19 = T11 + T18;
T1w = T18 - T11;
T1b = T10 - TV;
T1c = T12 + T17;
T1d = T1b - T1c;
T1u = T1b + T1c;
}
Rm[WS(rs, 2)] = FNMS(KP707106781, T19, TU);
Im[WS(rs, 2)] = FMS(KP707106781, T1u, T1t);
Rp[WS(rs, 1)] = FMA(KP707106781, T19, TU);
Ip[WS(rs, 1)] = FMA(KP707106781, T1u, T1t);
Rm[0] = FNMS(KP707106781, T1d, T1a);
Im[0] = FMS(KP707106781, T1w, T1v);
Rp[WS(rs, 3)] = FMA(KP707106781, T1d, T1a);
Ip[WS(rs, 3)] = FMA(KP707106781, T1w, T1v);
}
{
E Tm, T1e, T1o, T1q, TN, T1p, T1h, T1i;
{
E T8, Tl, T1j, T1n;
T8 = T1 + T7;
Tl = Te + Tk;
Tm = T8 + Tl;
T1e = T8 - Tl;
T1j = TQ + TS;
T1n = T1l + T1m;
T1o = T1j + T1n;
T1q = T1n - T1j;
}
{
E Tz, TM, T1f, T1g;
Tz = Ts + Ty;
TM = TF + TL;
TN = Tz + TM;
T1p = TM - Tz;
T1f = TX + TZ;
T1g = T14 + T16;
T1h = T1f - T1g;
T1i = T1f + T1g;
}
Rm[WS(rs, 3)] = Tm - TN;
Im[WS(rs, 3)] = T1i - T1o;
Rp[0] = Tm + TN;
Ip[0] = T1i + T1o;
Rm[WS(rs, 1)] = T1e - T1h;
Im[WS(rs, 1)] = T1p - T1q;
Rp[WS(rs, 2)] = T1e + T1h;
Ip[WS(rs, 2)] = T1p + T1q;
}
}
}
}
static const tw_instr twinstr[] = {
{ TW_FULL, 1, 8 },
{ TW_NEXT, 1, 0 }
};
static const hc2c_desc desc = { 8, "hc2cf_8", twinstr, &GENUS, { 44, 14, 22, 0 } };
void X(codelet_hc2cf_8) (planner *p) {
X(khc2c_register) (p, hc2cf_8, &desc, HC2C_VIA_RDFT);
}
#else
/* Generated by: ../../../genfft/gen_hc2c.native -compact -variables 4 -pipeline-latency 4 -n 8 -dit -name hc2cf_8 -include rdft/scalar/hc2cf.h */
/*
* This function contains 66 FP additions, 32 FP multiplications,
* (or, 52 additions, 18 multiplications, 14 fused multiply/add),
* 28 stack variables, 1 constants, and 32 memory accesses
*/
#include "rdft/scalar/hc2cf.h"
static void hc2cf_8(R *Rp, R *Ip, R *Rm, R *Im, const R *W, stride rs, INT mb, INT me, INT ms)
{
DK(KP707106781, +0.707106781186547524400844362104849039284835938);
{
INT m;
for (m = mb, W = W + ((mb - 1) * 14); m < me; m = m + 1, Rp = Rp + ms, Ip = Ip + ms, Rm = Rm - ms, Im = Im - ms, W = W + 14, MAKE_VOLATILE_STRIDE(32, rs)) {
E T7, T1e, TH, T19, TF, T13, TR, TU, Ti, T1f, TK, T16, Tu, T12, TM;
E TP;
{
E T1, T18, T6, T17;
T1 = Rp[0];
T18 = Rm[0];
{
E T3, T5, T2, T4;
T3 = Rp[WS(rs, 2)];
T5 = Rm[WS(rs, 2)];
T2 = W[6];
T4 = W[7];
T6 = FMA(T2, T3, T4 * T5);
T17 = FNMS(T4, T3, T2 * T5);
}
T7 = T1 + T6;
T1e = T18 - T17;
TH = T1 - T6;
T19 = T17 + T18;
}
{
E Tz, TS, TE, TT;
{
E Tw, Ty, Tv, Tx;
Tw = Ip[WS(rs, 3)];
Ty = Im[WS(rs, 3)];
Tv = W[12];
Tx = W[13];
Tz = FMA(Tv, Tw, Tx * Ty);
TS = FNMS(Tx, Tw, Tv * Ty);
}
{
E TB, TD, TA, TC;
TB = Ip[WS(rs, 1)];
TD = Im[WS(rs, 1)];
TA = W[4];
TC = W[5];
TE = FMA(TA, TB, TC * TD);
TT = FNMS(TC, TB, TA * TD);
}
TF = Tz + TE;
T13 = TS + TT;
TR = Tz - TE;
TU = TS - TT;
}
{
E Tc, TI, Th, TJ;
{
E T9, Tb, T8, Ta;
T9 = Rp[WS(rs, 1)];
Tb = Rm[WS(rs, 1)];
T8 = W[2];
Ta = W[3];
Tc = FMA(T8, T9, Ta * Tb);
TI = FNMS(Ta, T9, T8 * Tb);
}
{
E Te, Tg, Td, Tf;
Te = Rp[WS(rs, 3)];
Tg = Rm[WS(rs, 3)];
Td = W[10];
Tf = W[11];
Th = FMA(Td, Te, Tf * Tg);
TJ = FNMS(Tf, Te, Td * Tg);
}
Ti = Tc + Th;
T1f = Tc - Th;
TK = TI - TJ;
T16 = TI + TJ;
}
{
E To, TN, Tt, TO;
{
E Tl, Tn, Tk, Tm;
Tl = Ip[0];
Tn = Im[0];
Tk = W[0];
Tm = W[1];
To = FMA(Tk, Tl, Tm * Tn);
TN = FNMS(Tm, Tl, Tk * Tn);
}
{
E Tq, Ts, Tp, Tr;
Tq = Ip[WS(rs, 2)];
Ts = Im[WS(rs, 2)];
Tp = W[8];
Tr = W[9];
Tt = FMA(Tp, Tq, Tr * Ts);
TO = FNMS(Tr, Tq, Tp * Ts);
}
Tu = To + Tt;
T12 = TN + TO;
TM = To - Tt;
TP = TN - TO;
}
{
E Tj, TG, T1b, T1c;
Tj = T7 + Ti;
TG = Tu + TF;
Rm[WS(rs, 3)] = Tj - TG;
Rp[0] = Tj + TG;
{
E T15, T1a, T11, T14;
T15 = T12 + T13;
T1a = T16 + T19;
Im[WS(rs, 3)] = T15 - T1a;
Ip[0] = T15 + T1a;
T11 = T7 - Ti;
T14 = T12 - T13;
Rm[WS(rs, 1)] = T11 - T14;
Rp[WS(rs, 2)] = T11 + T14;
}
T1b = TF - Tu;
T1c = T19 - T16;
Im[WS(rs, 1)] = T1b - T1c;
Ip[WS(rs, 2)] = T1b + T1c;
{
E TX, T1g, T10, T1d, TY, TZ;
TX = TH - TK;
T1g = T1e - T1f;
TY = TP - TM;
TZ = TR + TU;
T10 = KP707106781 * (TY - TZ);
T1d = KP707106781 * (TY + TZ);
Rm[0] = TX - T10;
Ip[WS(rs, 1)] = T1d + T1g;
Rp[WS(rs, 3)] = TX + T10;
Im[WS(rs, 2)] = T1d - T1g;
}
{
E TL, T1i, TW, T1h, TQ, TV;
TL = TH + TK;
T1i = T1f + T1e;
TQ = TM + TP;
TV = TR - TU;
TW = KP707106781 * (TQ + TV);
T1h = KP707106781 * (TV - TQ);
Rm[WS(rs, 2)] = TL - TW;
Ip[WS(rs, 3)] = T1h + T1i;
Rp[WS(rs, 1)] = TL + TW;
Im[0] = T1h - T1i;
}
}
}
}
}
static const tw_instr twinstr[] = {
{ TW_FULL, 1, 8 },
{ TW_NEXT, 1, 0 }
};
static const hc2c_desc desc = { 8, "hc2cf_8", twinstr, &GENUS, { 52, 18, 14, 0 } };
void X(codelet_hc2cf_8) (planner *p) {
X(khc2c_register) (p, hc2cf_8, &desc, HC2C_VIA_RDFT);
}
#endif