/*
* Copyright (c) 2003, 2006 Matteo Frigo
* Copyright (c) 2003, 2006 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., 59 Temple Place, Suite 330, Boston, MA 02111-1307 USA
*
*/
/* This file was automatically generated --- DO NOT EDIT */
/* Generated on Sun Jul 2 15:50:22 EDT 2006 */
#include "codelet-rdft.h"
#ifdef HAVE_FMA
/* Generated by: ../../../genfft/gen_hc2hc -fma -reorder-insns -schedule-for-pipeline -compact -variables 4 -pipeline-latency 4 -n 9 -dit -name hf_9 -include hf.h */
/*
* This function contains 96 FP additions, 88 FP multiplications,
* (or, 24 additions, 16 multiplications, 72 fused multiply/add),
* 75 stack variables, and 36 memory accesses
*/
/*
* Generator Id's :
* $Id: algsimp.ml,v 1.9 2006-02-12 23:34:12 athena Exp $
* $Id: fft.ml,v 1.4 2006-01-05 03:04:27 stevenj Exp $
* $Id: gen_hc2hc.ml,v 1.16 2006-02-12 23:34:12 athena Exp $
*/
#include "hf.h"
static const R *hf_9(R *rio, R *iio, const R *W, stride ios, INT m, INT dist)
{
DK(KP777861913, +0.777861913430206160028177977318626690410586096);
DK(KP852868531, +0.852868531952443209628250963940074071936020296);
DK(KP839099631, +0.839099631177280011763127298123181364687434283);
DK(KP492403876, +0.492403876506104029683371512294761506835321626);
DK(KP984807753, +0.984807753012208059366743024589523013670643252);
DK(KP954188894, +0.954188894138671133499268364187245676532219158);
DK(KP363970234, +0.363970234266202361351047882776834043890471784);
DK(KP176326980, +0.176326980708464973471090386868618986121633062);
DK(KP866025403, +0.866025403784438646763723170752936183471402627);
DK(KP500000000, +0.500000000000000000000000000000000000000000000);
INT i;
for (i = m - 2; i > 0; i = i - 2, rio = rio + dist, iio = iio - dist, W = W + 16, MAKE_VOLATILE_STRIDE(ios)) {
E T1K, T20, T1H, T1Z;
{
E T1, T1R, T10, T1Q, Te, T1W, Th, Tk, T1l, T1r, T1q, T1N, Ti, T1g, TT;
E T12, Tn, Tq, Tp, T17, Tx, T14, To, Tj;
{
E T9, Tc, TY, Ta, Tb, TX, T7, TZ, Td;
T1 = rio[0];
{
E T3, T6, T8, TW, T4, T2, T5;
T3 = rio[WS(ios, 3)];
T6 = iio[-WS(ios, 5)];
T2 = W[4];
T9 = rio[WS(ios, 6)];
Tc = iio[-WS(ios, 2)];
T8 = W[10];
TW = T2 * T6;
T4 = T2 * T3;
T5 = W[5];
TY = T8 * Tc;
Ta = T8 * T9;
Tb = W[11];
TX = FNMS(T5, T3, TW);
T7 = FMA(T5, T6, T4);
}
T1R = iio[-WS(ios, 8)];
TZ = FNMS(Tb, T9, TY);
Td = FMA(Tb, Tc, Ta);
{
E TB, TE, T1n, TC, TH, TK, T1k, TR, TG, TJ, TD, TA;
TB = rio[WS(ios, 2)];
TE = iio[-WS(ios, 6)];
T10 = TX - TZ;
T1Q = TX + TZ;
Te = T7 + Td;
T1W = Td - T7;
TA = W[2];
{
E TN, TQ, TP, T1j, TO, TM;
TN = rio[WS(ios, 8)];
TQ = iio[0];
TM = W[14];
T1n = TA * TE;
TC = TA * TB;
TP = W[15];
T1j = TM * TQ;
TO = TM * TN;
TH = rio[WS(ios, 5)];
TK = iio[-WS(ios, 3)];
T1k = FNMS(TP, TN, T1j);
TR = FMA(TP, TQ, TO);
TG = W[8];
TJ = W[9];
}
TD = W[3];
{
E T1o, TF, T1i, TL, T1h, TI, Tg, T1p, TS;
Th = rio[WS(ios, 1)];
T1h = TG * TK;
TI = TG * TH;
T1o = FNMS(TD, TB, T1n);
TF = FMA(TD, TE, TC);
T1i = FNMS(TJ, TH, T1h);
TL = FMA(TJ, TK, TI);
Tk = iio[-WS(ios, 7)];
Tg = W[0];
T1p = T1i + T1k;
T1l = T1i - T1k;
TS = TL + TR;
T1r = TR - TL;
T1q = FNMS(KP500000000, T1p, T1o);
T1N = T1o + T1p;
Ti = Tg * Th;
T1g = FNMS(KP500000000, TS, TF);
TT = TF + TS;
T12 = Tg * Tk;
}
{
E Tt, Tw, Ts, Tv, T16, Tu, Tm;
Tt = rio[WS(ios, 7)];
Tw = iio[-WS(ios, 1)];
Ts = W[12];
Tv = W[13];
Tn = rio[WS(ios, 4)];
Tq = iio[-WS(ios, 4)];
T16 = Ts * Tw;
Tu = Ts * Tt;
Tm = W[6];
Tp = W[7];
T17 = FNMS(Tv, Tt, T16);
Tx = FMA(Tv, Tw, Tu);
T14 = Tm * Tq;
To = Tm * Tn;
}
Tj = W[1];
}
}
{
E TV, Tf, T1d, T1a, T19, T1M, T1V, T1S, T1c, Tz, Tl, Ty;
TV = FNMS(KP500000000, Te, T1);
Tf = T1 + Te;
{
E T15, Tr, T13, T18;
T15 = FNMS(Tp, Tn, T14);
Tr = FMA(Tp, Tq, To);
T13 = FNMS(Tj, Th, T12);
Tl = FMA(Tj, Tk, Ti);
T18 = T15 + T17;
T1d = T15 - T17;
Ty = Tr + Tx;
T1a = Tr - Tx;
T19 = FNMS(KP500000000, T18, T13);
T1M = T13 + T18;
T1V = FNMS(KP500000000, T1Q, T1R);
T1S = T1Q + T1R;
}
T1c = FNMS(KP500000000, Ty, Tl);
Tz = Tl + Ty;
{
E T11, T1z, T1E, T1D, T21, T1X, T1I, T1C, T22, T1y, T24, T1u, T1U, TU;
T1U = TT - Tz;
TU = Tz + TT;
{
E T1P, T1O, T1L, T1T;
T1P = T1M + T1N;
T1O = T1M - T1N;
T11 = FNMS(KP866025403, T10, TV);
T1z = FMA(KP866025403, T10, TV);
T1L = FNMS(KP500000000, TU, Tf);
rio[0] = Tf + TU;
T1T = FNMS(KP500000000, T1P, T1S);
iio[0] = T1P + T1S;
rio[WS(ios, 3)] = FMA(KP866025403, T1O, T1L);
iio[-WS(ios, 6)] = FNMS(KP866025403, T1O, T1L);
iio[-WS(ios, 3)] = FMA(KP866025403, T1U, T1T);
rio[WS(ios, 6)] = FMS(KP866025403, T1U, T1T);
}
{
E T1B, T1m, T1w, T1f, T1s, T1A, T1b, T1e, T1x, T1t;
T1E = FNMS(KP866025403, T1a, T19);
T1b = FMA(KP866025403, T1a, T19);
T1e = FNMS(KP866025403, T1d, T1c);
T1D = FMA(KP866025403, T1d, T1c);
T1B = FMA(KP866025403, T1l, T1g);
T1m = FNMS(KP866025403, T1l, T1g);
T21 = FNMS(KP866025403, T1W, T1V);
T1X = FMA(KP866025403, T1W, T1V);
T1w = FNMS(KP176326980, T1b, T1e);
T1f = FMA(KP176326980, T1e, T1b);
T1s = FNMS(KP866025403, T1r, T1q);
T1A = FMA(KP866025403, T1r, T1q);
T1x = FMA(KP363970234, T1m, T1s);
T1t = FNMS(KP363970234, T1s, T1m);
T1I = FNMS(KP176326980, T1A, T1B);
T1C = FMA(KP176326980, T1B, T1A);
T22 = FMA(KP954188894, T1x, T1w);
T1y = FNMS(KP954188894, T1x, T1w);
T24 = FMA(KP954188894, T1t, T1f);
T1u = FNMS(KP954188894, T1t, T1f);
}
{
E T1Y, T1G, T23, T1F, T1J, T1v;
iio[-WS(ios, 2)] = FNMS(KP984807753, T22, T21);
T1v = FNMS(KP492403876, T1u, T11);
rio[WS(ios, 2)] = FMA(KP984807753, T1u, T11);
T1F = FMA(KP839099631, T1E, T1D);
T1J = FNMS(KP839099631, T1D, T1E);
iio[-WS(ios, 5)] = FNMS(KP852868531, T1y, T1v);
iio[-WS(ios, 8)] = FMA(KP852868531, T1y, T1v);
T1K = FMA(KP777861913, T1J, T1I);
T1Y = FNMS(KP777861913, T1J, T1I);
T1G = FMA(KP777861913, T1F, T1C);
T20 = FNMS(KP777861913, T1F, T1C);
T23 = FMA(KP492403876, T22, T21);
iio[-WS(ios, 1)] = FNMS(KP984807753, T1Y, T1X);
rio[WS(ios, 1)] = FMA(KP984807753, T1G, T1z);
T1H = FNMS(KP492403876, T1G, T1z);
rio[WS(ios, 8)] = -(FMA(KP852868531, T24, T23));
rio[WS(ios, 5)] = FMS(KP852868531, T24, T23);
T1Z = FMA(KP492403876, T1Y, T1X);
}
}
}
}
iio[-WS(ios, 7)] = FNMS(KP852868531, T1K, T1H);
rio[WS(ios, 4)] = FMA(KP852868531, T1K, T1H);
iio[-WS(ios, 4)] = FMA(KP852868531, T20, T1Z);
rio[WS(ios, 7)] = FMS(KP852868531, T20, T1Z);
}
return W;
}
static const tw_instr twinstr[] = {
{TW_FULL, 0, 9},
{TW_NEXT, 1, 0}
};
static const hc2hc_desc desc = { 9, "hf_9", twinstr, &GENUS, {24, 16, 72, 0}, 0, 0, 0 };
void X(codelet_hf_9) (planner *p) {
X(khc2hc_register) (p, hf_9, &desc);
}
#else /* HAVE_FMA */
/* Generated by: ../../../genfft/gen_hc2hc -compact -variables 4 -pipeline-latency 4 -n 9 -dit -name hf_9 -include hf.h */
/*
* This function contains 96 FP additions, 72 FP multiplications,
* (or, 60 additions, 36 multiplications, 36 fused multiply/add),
* 41 stack variables, and 36 memory accesses
*/
/*
* Generator Id's :
* $Id: algsimp.ml,v 1.9 2006-02-12 23:34:12 athena Exp $
* $Id: fft.ml,v 1.4 2006-01-05 03:04:27 stevenj Exp $
* $Id: gen_hc2hc.ml,v 1.16 2006-02-12 23:34:12 athena Exp $
*/
#include "hf.h"
static const R *hf_9(R *rio, R *iio, const R *W, stride ios, INT m, INT dist)
{
DK(KP939692620, +0.939692620785908384054109277324731469936208134);
DK(KP342020143, +0.342020143325668733044099614682259580763083368);
DK(KP984807753, +0.984807753012208059366743024589523013670643252);
DK(KP173648177, +0.173648177666930348851716626769314796000375677);
DK(KP642787609, +0.642787609686539326322643409907263432907559884);
DK(KP766044443, +0.766044443118978035202392650555416673935832457);
DK(KP500000000, +0.500000000000000000000000000000000000000000000);
DK(KP866025403, +0.866025403784438646763723170752936183471402627);
INT i;
for (i = m - 2; i > 0; i = i - 2, rio = rio + dist, iio = iio - dist, W = W + 16, MAKE_VOLATILE_STRIDE(ios)) {
E T1, T1B, TQ, T1G, Tc, TN, T1A, T1H, TL, T1x, T17, T1o, T1c, T1n, Tu;
E T1w, TW, T1k, T11, T1l;
{
E T6, TO, Tb, TP;
T1 = rio[0];
T1B = iio[-WS(ios, 8)];
{
E T3, T5, T2, T4;
T3 = rio[WS(ios, 3)];
T5 = iio[-WS(ios, 5)];
T2 = W[4];
T4 = W[5];
T6 = FMA(T2, T3, T4 * T5);
TO = FNMS(T4, T3, T2 * T5);
}
{
E T8, Ta, T7, T9;
T8 = rio[WS(ios, 6)];
Ta = iio[-WS(ios, 2)];
T7 = W[10];
T9 = W[11];
Tb = FMA(T7, T8, T9 * Ta);
TP = FNMS(T9, T8, T7 * Ta);
}
TQ = KP866025403 * (TO - TP);
T1G = KP866025403 * (Tb - T6);
Tc = T6 + Tb;
TN = FNMS(KP500000000, Tc, T1);
T1A = TO + TP;
T1H = FNMS(KP500000000, T1A, T1B);
}
{
E Tz, T19, TE, T14, TJ, T15, TK, T1a;
{
E Tw, Ty, Tv, Tx;
Tw = rio[WS(ios, 2)];
Ty = iio[-WS(ios, 6)];
Tv = W[2];
Tx = W[3];
Tz = FMA(Tv, Tw, Tx * Ty);
T19 = FNMS(Tx, Tw, Tv * Ty);
}
{
E TB, TD, TA, TC;
TB = rio[WS(ios, 5)];
TD = iio[-WS(ios, 3)];
TA = W[8];
TC = W[9];
TE = FMA(TA, TB, TC * TD);
T14 = FNMS(TC, TB, TA * TD);
}
{
E TG, TI, TF, TH;
TG = rio[WS(ios, 8)];
TI = iio[0];
TF = W[14];
TH = W[15];
TJ = FMA(TF, TG, TH * TI);
T15 = FNMS(TH, TG, TF * TI);
}
TK = TE + TJ;
T1a = T14 + T15;
TL = Tz + TK;
T1x = T19 + T1a;
{
E T13, T16, T18, T1b;
T13 = FNMS(KP500000000, TK, Tz);
T16 = KP866025403 * (T14 - T15);
T17 = T13 + T16;
T1o = T13 - T16;
T18 = KP866025403 * (TJ - TE);
T1b = FNMS(KP500000000, T1a, T19);
T1c = T18 + T1b;
T1n = T1b - T18;
}
}
{
E Ti, TY, Tn, TT, Ts, TU, Tt, TZ;
{
E Tf, Th, Te, Tg;
Tf = rio[WS(ios, 1)];
Th = iio[-WS(ios, 7)];
Te = W[0];
Tg = W[1];
Ti = FMA(Te, Tf, Tg * Th);
TY = FNMS(Tg, Tf, Te * Th);
}
{
E Tk, Tm, Tj, Tl;
Tk = rio[WS(ios, 4)];
Tm = iio[-WS(ios, 4)];
Tj = W[6];
Tl = W[7];
Tn = FMA(Tj, Tk, Tl * Tm);
TT = FNMS(Tl, Tk, Tj * Tm);
}
{
E Tp, Tr, To, Tq;
Tp = rio[WS(ios, 7)];
Tr = iio[-WS(ios, 1)];
To = W[12];
Tq = W[13];
Ts = FMA(To, Tp, Tq * Tr);
TU = FNMS(Tq, Tp, To * Tr);
}
Tt = Tn + Ts;
TZ = TT + TU;
Tu = Ti + Tt;
T1w = TY + TZ;
{
E TS, TV, TX, T10;
TS = FNMS(KP500000000, Tt, Ti);
TV = KP866025403 * (TT - TU);
TW = TS + TV;
T1k = TS - TV;
TX = KP866025403 * (Ts - Tn);
T10 = FNMS(KP500000000, TZ, TY);
T11 = TX + T10;
T1l = T10 - TX;
}
}
{
E T1y, Td, TM, T1v;
T1y = KP866025403 * (T1w - T1x);
Td = T1 + Tc;
TM = Tu + TL;
T1v = FNMS(KP500000000, TM, Td);
rio[0] = Td + TM;
rio[WS(ios, 3)] = T1v + T1y;
iio[-WS(ios, 6)] = T1v - T1y;
}
{
E T1D, T1z, T1C, T1E;
T1D = KP866025403 * (TL - Tu);
T1z = T1w + T1x;
T1C = T1A + T1B;
T1E = FNMS(KP500000000, T1z, T1C);
iio[0] = T1z + T1C;
iio[-WS(ios, 3)] = T1D + T1E;
rio[WS(ios, 6)] = T1D - T1E;
}
{
E TR, T1I, T1e, T1J, T1i, T1F, T1f, T1K;
TR = TN + TQ;
T1I = T1G + T1H;
{
E T12, T1d, T1g, T1h;
T12 = FMA(KP766044443, TW, KP642787609 * T11);
T1d = FMA(KP173648177, T17, KP984807753 * T1c);
T1e = T12 + T1d;
T1J = KP866025403 * (T1d - T12);
T1g = FNMS(KP642787609, TW, KP766044443 * T11);
T1h = FNMS(KP984807753, T17, KP173648177 * T1c);
T1i = KP866025403 * (T1g - T1h);
T1F = T1g + T1h;
}
rio[WS(ios, 1)] = TR + T1e;
iio[-WS(ios, 1)] = T1F + T1I;
T1f = FNMS(KP500000000, T1e, TR);
iio[-WS(ios, 7)] = T1f - T1i;
rio[WS(ios, 4)] = T1f + T1i;
T1K = FNMS(KP500000000, T1F, T1I);
rio[WS(ios, 7)] = T1J - T1K;
iio[-WS(ios, 4)] = T1J + T1K;
}
{
E T1j, T1M, T1q, T1O, T1u, T1L, T1r, T1N;
T1j = TN - TQ;
T1M = T1H - T1G;
{
E T1m, T1p, T1s, T1t;
T1m = FMA(KP173648177, T1k, KP984807753 * T1l);
T1p = FNMS(KP939692620, T1o, KP342020143 * T1n);
T1q = T1m + T1p;
T1O = KP866025403 * (T1p - T1m);
T1s = FNMS(KP984807753, T1k, KP173648177 * T1l);
T1t = FMA(KP342020143, T1o, KP939692620 * T1n);
T1u = KP866025403 * (T1s + T1t);
T1L = T1s - T1t;
}
rio[WS(ios, 2)] = T1j + T1q;
iio[-WS(ios, 2)] = T1L + T1M;
T1r = FNMS(KP500000000, T1q, T1j);
iio[-WS(ios, 8)] = T1r - T1u;
iio[-WS(ios, 5)] = T1r + T1u;
T1N = FMS(KP500000000, T1L, T1M);
rio[WS(ios, 5)] = T1N - T1O;
rio[WS(ios, 8)] = T1O + T1N;
}
}
return W;
}
static const tw_instr twinstr[] = {
{TW_FULL, 0, 9},
{TW_NEXT, 1, 0}
};
static const hc2hc_desc desc = { 9, "hf_9", twinstr, &GENUS, {60, 36, 36, 0}, 0, 0, 0 };
void X(codelet_hf_9) (planner *p) {
X(khc2hc_register) (p, hf_9, &desc);
}
#endif /* HAVE_FMA */
syntax highlighted by Code2HTML, v. 0.9.1