hf_5.c

快速fft变换

/*
 * Copyright (c) 2003, 2007-8 Matteo Frigo
 * Copyright (c) 2003, 2007-8 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 Sat Nov 15 20:56:13 EST 2008 */

#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 5 -dit -name hf_5 -include hf.h */

/*
 * This function contains 40 FP additions, 34 FP multiplications,
 * (or, 14 additions, 8 multiplications, 26 fused multiply/add),
 * 43 stack variables, 4 constants, and 20 memory accesses
 */
#include "hf.h"

static void hf_5(R *cr, R *ci, const R *W, stride rs, INT mb, INT me, INT ms)
{
     DK(KP951056516, +0.951056516295153572116439333379382143405698634);
     DK(KP559016994, +0.559016994374947424102293417182819058860154590);
     DK(KP250000000, +0.250000000000000000000000000000000000000000000);
     DK(KP618033988, +0.618033988749894848204586834365638117720309180);
     INT m;
     for (m = mb, W = W + ((mb - 1) * 8); m < me; m = m + 1, cr = cr + ms, ci = ci - ms, W = W + 8, MAKE_VOLATILE_STRIDE(rs)) {
	  E T1, TJ, TK, TA, TR, Te, TC, Tk, TE, Tq;
	  {
	       E Tg, Tj, Tm, TB, Th, Tp, Tl, Ti, To, TD, Tn;
	       T1 = cr[0];
	       TJ = ci[0];
	       {
		    E T9, Tc, Ty, Ta, Tb, Tx, T7, Tf, Tz, Td;
		    {
			 E T3, T6, T8, Tw, T4, T2, T5;
			 T3 = cr[WS(rs, 1)];
			 T6 = ci[WS(rs, 1)];
			 T2 = W[0];
			 T9 = cr[WS(rs, 4)];
			 Tc = ci[WS(rs, 4)];
			 T8 = W[6];
			 Tw = T2 * T6;
			 T4 = T2 * T3;
			 T5 = W[1];
			 Ty = T8 * Tc;
			 Ta = T8 * T9;
			 Tb = W[7];
			 Tx = FNMS(T5, T3, Tw);
			 T7 = FMA(T5, T6, T4);
		    }
		    Tg = cr[WS(rs, 2)];
		    Tz = FNMS(Tb, T9, Ty);
		    Td = FMA(Tb, Tc, Ta);
		    Tj = ci[WS(rs, 2)];
		    Tf = W[2];
		    TK = Tx + Tz;
		    TA = Tx - Tz;
		    TR = Td - T7;
		    Te = T7 + Td;
		    Tm = cr[WS(rs, 3)];
		    TB = Tf * Tj;
		    Th = Tf * Tg;
		    Tp = ci[WS(rs, 3)];
		    Tl = W[4];
		    Ti = W[3];
		    To = W[5];
	       }
	       TD = Tl * Tp;
	       Tn = Tl * Tm;
	       TC = FNMS(Ti, Tg, TB);
	       Tk = FMA(Ti, Tj, Th);
	       TE = FNMS(To, Tm, TD);
	       Tq = FMA(To, Tp, Tn);
	  }
	  {
	       E TG, TI, TO, TS, TU, Tu, TN, Tt, TL, TF;
	       TL = TC + TE;
	       TF = TC - TE;
	       {
		    E Tr, TQ, TM, Ts;
		    Tr = Tk + Tq;
		    TQ = Tk - Tq;
		    TG = FMA(KP618033988, TF, TA);
		    TI = FNMS(KP618033988, TA, TF);
		    TO = TK - TL;
		    TM = TK + TL;
		    TS = FMA(KP618033988, TR, TQ);
		    TU = FNMS(KP618033988, TQ, TR);
		    Tu = Te - Tr;
		    Ts = Te + Tr;
		    ci[WS(rs, 4)] = TM + TJ;
		    TN = FNMS(KP250000000, TM, TJ);
		    cr[0] = T1 + Ts;
		    Tt = FNMS(KP250000000, Ts, T1);
	       }
	       {
		    E TT, TP, Tv, TH;
		    TT = FMA(KP559016994, TO, TN);
		    TP = FNMS(KP559016994, TO, TN);
		    Tv = FMA(KP559016994, Tu, Tt);
		    TH = FNMS(KP559016994, Tu, Tt);
		    ci[WS(rs, 2)] = FMA(KP951056516, TS, TP);
		    cr[WS(rs, 3)] = FMS(KP951056516, TS, TP);
		    ci[WS(rs, 3)] = FMA(KP951056516, TU, TT);
		    cr[WS(rs, 4)] = FMS(KP951056516, TU, TT);
		    ci[WS(rs, 1)] = FMA(KP951056516, TI, TH);
		    cr[WS(rs, 2)] = FNMS(KP951056516, TI, TH);
		    cr[WS(rs, 1)] = FMA(KP951056516, TG, Tv);
		    ci[0] = FNMS(KP951056516, TG, Tv);
	       }
	  }
     }
}

static const tw_instr twinstr[] = {
     {TW_FULL, 1, 5},
     {TW_NEXT, 1, 0}
};

static const hc2hc_desc desc = { 5, "hf_5", twinstr, &GENUS, {14, 8, 26, 0} };

void X(codelet_hf_5) (planner *p) {
     X(khc2hc_register) (p, hf_5, &desc);
}
#else				/* HAVE_FMA */

/* Generated by: ../../../genfft/gen_hc2hc -compact -variables 4 -pipeline-latency 4 -n 5 -dit -name hf_5 -include hf.h */

/*
 * This function contains 40 FP additions, 28 FP multiplications,
 * (or, 26 additions, 14 multiplications, 14 fused multiply/add),
 * 29 stack variables, 4 constants, and 20 memory accesses
 */
#include "hf.h"

static void hf_5(R *cr, R *ci, const R *W, stride rs, INT mb, INT me, INT ms)
{
     DK(KP250000000, +0.250000000000000000000000000000000000000000000);
     DK(KP559016994, +0.559016994374947424102293417182819058860154590);
     DK(KP587785252, +0.587785252292473129168705954639072768597652438);
     DK(KP951056516, +0.951056516295153572116439333379382143405698634);
     INT m;
     for (m = mb, W = W + ((mb - 1) * 8); m < me; m = m + 1, cr = cr + ms, ci = ci - ms, W = W + 8, MAKE_VOLATILE_STRIDE(rs)) {
	  E T1, TE, Tu, Tx, TC, TB, TF, TG, TH, Tc, Tn, To;
	  T1 = cr[0];
	  TE = ci[0];
	  {
	       E T6, Ts, Tm, Tw, Tb, Tt, Th, Tv;
	       {
		    E T3, T5, T2, T4;
		    T3 = cr[WS(rs, 1)];
		    T5 = ci[WS(rs, 1)];
		    T2 = W[0];
		    T4 = W[1];
		    T6 = FMA(T2, T3, T4 * T5);
		    Ts = FNMS(T4, T3, T2 * T5);
	       }
	       {
		    E Tj, Tl, Ti, Tk;
		    Tj = cr[WS(rs, 3)];
		    Tl = ci[WS(rs, 3)];
		    Ti = W[4];
		    Tk = W[5];
		    Tm = FMA(Ti, Tj, Tk * Tl);
		    Tw = FNMS(Tk, Tj, Ti * Tl);
	       }
	       {
		    E T8, Ta, T7, T9;
		    T8 = cr[WS(rs, 4)];
		    Ta = ci[WS(rs, 4)];
		    T7 = W[6];
		    T9 = W[7];
		    Tb = FMA(T7, T8, T9 * Ta);
		    Tt = FNMS(T9, T8, T7 * Ta);
	       }
	       {
		    E Te, Tg, Td, Tf;
		    Te = cr[WS(rs, 2)];
		    Tg = ci[WS(rs, 2)];
		    Td = W[2];
		    Tf = W[3];
		    Th = FMA(Td, Te, Tf * Tg);
		    Tv = FNMS(Tf, Te, Td * Tg);
	       }
	       Tu = Ts - Tt;
	       Tx = Tv - Tw;
	       TC = Th - Tm;
	       TB = Tb - T6;
	       TF = Ts + Tt;
	       TG = Tv + Tw;
	       TH = TF + TG;
	       Tc = T6 + Tb;
	       Tn = Th + Tm;
	       To = Tc + Tn;
	  }
	  cr[0] = T1 + To;
	  {
	       E Ty, TA, Tr, Tz, Tp, Tq;
	       Ty = FMA(KP951056516, Tu, KP587785252 * Tx);
	       TA = FNMS(KP587785252, Tu, KP951056516 * Tx);
	       Tp = KP559016994 * (Tc - Tn);
	       Tq = FNMS(KP250000000, To, T1);
	       Tr = Tp + Tq;
	       Tz = Tq - Tp;
	       ci[0] = Tr - Ty;
	       ci[WS(rs, 1)] = Tz + TA;
	       cr[WS(rs, 1)] = Tr + Ty;
	       cr[WS(rs, 2)] = Tz - TA;
	  }
	  ci[WS(rs, 4)] = TH + TE;
	  {
	       E TD, TL, TK, TM, TI, TJ;
	       TD = FMA(KP587785252, TB, KP951056516 * TC);
	       TL = FNMS(KP587785252, TC, KP951056516 * TB);
	       TI = FNMS(KP250000000, TH, TE);
	       TJ = KP559016994 * (TF - TG);
	       TK = TI - TJ;
	       TM = TJ + TI;
	       cr[WS(rs, 3)] = TD - TK;
	       ci[WS(rs, 3)] = TL + TM;
	       ci[WS(rs, 2)] = TD + TK;
	       cr[WS(rs, 4)] = TL - TM;
	  }
     }
}

static const tw_instr twinstr[] = {
     {TW_FULL, 1, 5},
     {TW_NEXT, 1, 0}
};

static const hc2hc_desc desc = { 5, "hf_5", twinstr, &GENUS, {26, 14, 14, 0} };

void X(codelet_hf_5) (planner *p) {
     X(khc2hc_register) (p, hf_5, &desc);
}
#endif				/* HAVE_FMA */