hf_4.c

最新的FFT程序

/*
 * 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 Fri Jan 27 20:18:49 EST 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 4 -dit -name hf_4 -include hf.h */

/*
 * This function contains 22 FP additions, 12 FP multiplications,
 * (or, 16 additions, 6 multiplications, 6 fused multiply/add),
 * 31 stack variables, and 16 memory accesses
 */
/*
 * Generator Id's : 
 * $Id: algsimp.ml,v 1.8 2006-01-05 03:04:27 stevenj Exp $
 * $Id: fft.ml,v 1.4 2006-01-05 03:04:27 stevenj Exp $
 * $Id: gen_hc2hc.ml,v 1.15 2006-01-05 03:04:27 stevenj Exp $
 */

#include "hf.h"

static const R *hf_4(R *rio, R *iio, const R *W, stride ios, INT m, INT dist)
{
     INT i;
     for (i = m - 2; i > 0; i = i - 2, rio = rio + dist, iio = iio - dist, W = W + 6, MAKE_VOLATILE_STRIDE(ios)) {
	  E Tq, Tk, Ty, Tw, T8, Tm, To, Te;
	  {
	       E T1, Tv, Ta, Td, Tu, T7, T9, Tc, Tn, Tb;
	       {
		    E Tg, Tj, Tf, Ti;
		    T1 = rio[0];
		    Tg = rio[WS(ios, 3)];
		    Tj = iio[0];
		    Tf = W[4];
		    Ti = W[5];
		    Tv = iio[-WS(ios, 3)];
		    {
			 E T3, T6, T5, Tt, T4, T2, Tp, Th;
			 T3 = rio[WS(ios, 2)];
			 Tp = Tf * Tj;
			 Th = Tf * Tg;
			 T6 = iio[-WS(ios, 1)];
			 T2 = W[2];
			 Tq = FNMS(Ti, Tg, Tp);
			 Tk = FMA(Ti, Tj, Th);
			 T5 = W[3];
			 Tt = T2 * T6;
			 T4 = T2 * T3;
			 Ta = rio[WS(ios, 1)];
			 Td = iio[-WS(ios, 2)];
			 Tu = FNMS(T5, T3, Tt);
			 T7 = FMA(T5, T6, T4);
			 T9 = W[0];
			 Tc = W[1];
		    }
	       }
	       Ty = Tv - Tu;
	       Tw = Tu + Tv;
	       T8 = T1 + T7;
	       Tm = T1 - T7;
	       Tn = T9 * Td;
	       Tb = T9 * Ta;
	       To = FNMS(Tc, Ta, Tn);
	       Te = FMA(Tc, Td, Tb);
	  }
	  {
	       E Ts, Tr, Tl, Tx;
	       Ts = To + Tq;
	       Tr = To - Tq;
	       Tl = Te + Tk;
	       Tx = Tk - Te;
	       rio[WS(ios, 1)] = Tm + Tr;
	       iio[-WS(ios, 3)] = Tm - Tr;
	       iio[0] = Ts + Tw;
	       rio[WS(ios, 2)] = Ts - Tw;
	       iio[-WS(ios, 1)] = Tx + Ty;
	       rio[WS(ios, 3)] = Tx - Ty;
	       rio[0] = T8 + Tl;
	       iio[-WS(ios, 2)] = T8 - Tl;
	  }
     }
     return W;
}

static const tw_instr twinstr[] = {
     {TW_FULL, 0, 4},
     {TW_NEXT, 1, 0}
};

static const hc2hc_desc desc = { 4, "hf_4", twinstr, &GENUS, {16, 6, 6, 0}, 0, 0, 0 };

void X(codelet_hf_4) (planner *p) {
     X(khc2hc_register) (p, hf_4, &desc);
}
#else				/* HAVE_FMA */

/* Generated by: ../../../genfft/gen_hc2hc -compact -variables 4 -pipeline-latency 4 -n 4 -dit -name hf_4 -include hf.h */

/*
 * This function contains 22 FP additions, 12 FP multiplications,
 * (or, 16 additions, 6 multiplications, 6 fused multiply/add),
 * 13 stack variables, and 16 memory accesses
 */
/*
 * Generator Id's : 
 * $Id: algsimp.ml,v 1.8 2006-01-05 03:04:27 stevenj Exp $
 * $Id: fft.ml,v 1.4 2006-01-05 03:04:27 stevenj Exp $
 * $Id: gen_hc2hc.ml,v 1.15 2006-01-05 03:04:27 stevenj Exp $
 */

#include "hf.h"

static const R *hf_4(R *rio, R *iio, const R *W, stride ios, INT m, INT dist)
{
     INT i;
     for (i = m - 2; i > 0; i = i - 2, rio = rio + dist, iio = iio - dist, W = W + 6, MAKE_VOLATILE_STRIDE(ios)) {
	  E T1, Th, Tl, Tp, T6, To, Tc, Tk;
	  T1 = rio[0];
	  {
	       E Te, Tg, Td, Tf;
	       Te = rio[WS(ios, 3)];
	       Tg = iio[0];
	       Td = W[4];
	       Tf = W[5];
	       Th = FMA(Td, Te, Tf * Tg);
	       Tl = FNMS(Tf, Te, Td * Tg);
	  }
	  Tp = iio[-WS(ios, 3)];
	  {
	       E T3, T5, T2, T4;
	       T3 = rio[WS(ios, 2)];
	       T5 = iio[-WS(ios, 1)];
	       T2 = W[2];
	       T4 = W[3];
	       T6 = FMA(T2, T3, T4 * T5);
	       To = FNMS(T4, T3, T2 * T5);
	  }
	  {
	       E T9, Tb, T8, Ta;
	       T9 = rio[WS(ios, 1)];
	       Tb = iio[-WS(ios, 2)];
	       T8 = W[0];
	       Ta = W[1];
	       Tc = FMA(T8, T9, Ta * Tb);
	       Tk = FNMS(Ta, T9, T8 * Tb);
	  }
	  {
	       E T7, Ti, Tn, Tq;
	       T7 = T1 + T6;
	       Ti = Tc + Th;
	       iio[-WS(ios, 2)] = T7 - Ti;
	       rio[0] = T7 + Ti;
	       Tn = Tk + Tl;
	       Tq = To + Tp;
	       rio[WS(ios, 2)] = Tn - Tq;
	       iio[0] = Tn + Tq;
	  }
	  {
	       E Tj, Tm, Tr, Ts;
	       Tj = T1 - T6;
	       Tm = Tk - Tl;
	       iio[-WS(ios, 3)] = Tj - Tm;
	       rio[WS(ios, 1)] = Tj + Tm;
	       Tr = Th - Tc;
	       Ts = Tp - To;
	       rio[WS(ios, 3)] = Tr - Ts;
	       iio[-WS(ios, 1)] = Tr + Ts;
	  }
     }
     return W;
}

static const tw_instr twinstr[] = {
     {TW_FULL, 0, 4},
     {TW_NEXT, 1, 0}
};

static const hc2hc_desc desc = { 4, "hf_4", twinstr, &GENUS, {16, 6, 6, 0}, 0, 0, 0 };

void X(codelet_hf_4) (planner *p) {
     X(khc2hc_register) (p, hf_4, &desc);
}
#endif				/* HAVE_FMA */