t1bv_15.c

快速傅立叶变换库函数

/*
 * 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 Sat Oct  4 10:27:24 EDT 2008 */

#include "codelet-dft.h"

#ifdef HAVE_FMA

/* Generated by: ../../../genfft/gen_twiddle_c -fma -reorder-insns -schedule-for-pipeline -simd -compact -variables 4 -pipeline-latency 8 -n 15 -name t1bv_15 -include t1b.h -sign 1 */

/*
 * This function contains 92 FP additions, 77 FP multiplications,
 * (or, 50 additions, 35 multiplications, 42 fused multiply/add),
 * 81 stack variables, and 30 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_twiddle_c.ml,v 1.14 2006-02-12 23:34:12 athena Exp $
 */

#include "t1b.h"

static const R *t1bv_15(R *ri, R *ii, const R *W, stride ios, INT m, INT dist)
{
     DVK(KP823639103, +0.823639103546331925877420039278190003029660514);
     DVK(KP910592997, +0.910592997310029334643087372129977886038870291);
     DVK(KP559016994, +0.559016994374947424102293417182819058860154590);
     DVK(KP951056516, +0.951056516295153572116439333379382143405698634);
     DVK(KP866025403, +0.866025403784438646763723170752936183471402627);
     DVK(KP250000000, +0.250000000000000000000000000000000000000000000);
     DVK(KP618033988, +0.618033988749894848204586834365638117720309180);
     DVK(KP500000000, +0.500000000000000000000000000000000000000000000);
     INT i;
     R *x;
     x = ii;
     for (i = m; i > 0; i = i - VL, x = x + (VL * dist), W = W + (TWVL * 28), MAKE_VOLATILE_STRIDE(ios)) {
	  V Tq, Ty, Th, TV, TK, Ts, T1f, T7, Tu, TA, TC, Tj, Tk, T1g, Tf;
	  {
	       V T1, T4, T2, T9, Te;
	       T1 = LD(&(x[0]), dist, &(x[0]));
	       T4 = LD(&(x[WS(ios, 10)]), dist, &(x[0]));
	       T2 = LD(&(x[WS(ios, 5)]), dist, &(x[WS(ios, 1)]));
	       {
		    V T8, Tp, Tx, Tg;
		    T8 = LD(&(x[WS(ios, 3)]), dist, &(x[WS(ios, 1)]));
		    Tp = LD(&(x[WS(ios, 6)]), dist, &(x[0]));
		    Tx = LD(&(x[WS(ios, 9)]), dist, &(x[WS(ios, 1)]));
		    Tg = LD(&(x[WS(ios, 12)]), dist, &(x[0]));
		    {
			 V Tb, Td, Tr, T6, Tt, Tz, TB, Ti;
			 {
			      V T5, T3, Ta, Tc;
			      Ta = LD(&(x[WS(ios, 8)]), dist, &(x[0]));
			      Tc = LD(&(x[WS(ios, 13)]), dist, &(x[WS(ios, 1)]));
			      T5 = BYTW(&(W[TWVL * 18]), T4);
			      T3 = BYTW(&(W[TWVL * 8]), T2);
			      T9 = BYTW(&(W[TWVL * 4]), T8);
			      Tq = BYTW(&(W[TWVL * 10]), Tp);
			      Ty = BYTW(&(W[TWVL * 16]), Tx);
			      Th = BYTW(&(W[TWVL * 22]), Tg);
			      Tb = BYTW(&(W[TWVL * 14]), Ta);
			      Td = BYTW(&(W[TWVL * 24]), Tc);
			      Tr = LD(&(x[WS(ios, 11)]), dist, &(x[WS(ios, 1)]));
			      TV = VSUB(T3, T5);
			      T6 = VADD(T3, T5);
			      Tt = LD(&(x[WS(ios, 1)]), dist, &(x[WS(ios, 1)]));
			 }
			 Tz = LD(&(x[WS(ios, 14)]), dist, &(x[0]));
			 TB = LD(&(x[WS(ios, 4)]), dist, &(x[0]));
			 Ti = LD(&(x[WS(ios, 2)]), dist, &(x[0]));
			 Te = VADD(Tb, Td);
			 TK = VSUB(Tb, Td);
			 Ts = BYTW(&(W[TWVL * 20]), Tr);
			 T1f = VADD(T1, T6);
			 T7 = VFNMS(LDK(KP500000000), T6, T1);
			 Tu = BYTW(&(W[0]), Tt);
			 TA = BYTW(&(W[TWVL * 26]), Tz);
			 TC = BYTW(&(W[TWVL * 6]), TB);
			 Tj = BYTW(&(W[TWVL * 2]), Ti);
			 Tk = LD(&(x[WS(ios, 7)]), dist, &(x[WS(ios, 1)]));
		    }
	       }
	       T1g = VADD(T9, Te);
	       Tf = VFNMS(LDK(KP500000000), Te, T9);
	  }
	  {
	       V Tv, TN, TD, TO, Tl;
	       Tv = VADD(Ts, Tu);
	       TN = VSUB(Ts, Tu);
	       TD = VADD(TA, TC);
	       TO = VSUB(TA, TC);
	       Tl = BYTW(&(W[TWVL * 12]), Tk);
	       {
		    V Tw, T1j, TX, TP, TE, T1k, TL, Tm;
		    Tw = VFNMS(LDK(KP500000000), Tv, Tq);
		    T1j = VADD(Tq, Tv);
		    TX = VADD(TN, TO);
		    TP = VSUB(TN, TO);
		    TE = VFNMS(LDK(KP500000000), TD, Ty);
		    T1k = VADD(Ty, TD);
		    TL = VSUB(Tj, Tl);
		    Tm = VADD(Tj, Tl);
		    {
			 V TT, TF, T1q, T1l, TW, TM, T1h, Tn;
			 TT = VSUB(Tw, TE);
			 TF = VADD(Tw, TE);
			 T1q = VSUB(T1j, T1k);
			 T1l = VADD(T1j, T1k);
			 TW = VADD(TK, TL);
			 TM = VSUB(TK, TL);
			 T1h = VADD(Th, Tm);
			 Tn = VFNMS(LDK(KP500000000), Tm, Th);
			 {
			      V T10, TY, T16, TQ, T1r, T1i, TS, To, TZ, T1e;
			      T10 = VSUB(TW, TX);
			      TY = VADD(TW, TX);
			      T16 = VFNMS(LDK(KP618033988), TM, TP);
			      TQ = VFMA(LDK(KP618033988), TP, TM);
			      T1r = VSUB(T1g, T1h);
			      T1i = VADD(T1g, T1h);
			      TS = VSUB(Tf, Tn);
			      To = VADD(Tf, Tn);
			      TZ = VFNMS(LDK(KP250000000), TY, TV);
			      T1e = VMUL(LDK(KP866025403), VADD(TV, TY));
			      {
				   V T1u, T1s, T1o, T18, TU, TG, TI, T19, T11, T1n, T1m;
				   T1u = VMUL(LDK(KP951056516), VFMA(LDK(KP618033988), T1q, T1r));
				   T1s = VMUL(LDK(KP951056516), VFNMS(LDK(KP618033988), T1r, T1q));
				   T1m = VADD(T1i, T1l);
				   T1o = VSUB(T1i, T1l);
				   T18 = VFNMS(LDK(KP618033988), TS, TT);
				   TU = VFMA(LDK(KP618033988), TT, TS);
				   TG = VADD(To, TF);
				   TI = VSUB(To, TF);
				   T19 = VFNMS(LDK(KP559016994), T10, TZ);
				   T11 = VFMA(LDK(KP559016994), T10, TZ);
				   ST(&(x[0]), VADD(T1f, T1m), dist, &(x[0]));
				   T1n = VFNMS(LDK(KP250000000), T1m, T1f);
				   {
					V T1a, T1c, T14, T12, T1p, T1t, T15, TJ, T1d, TH;
					T1d = VADD(T7, TG);
					TH = VFNMS(LDK(KP250000000), TG, T7);
					T1a = VMUL(LDK(KP951056516), VFMA(LDK(KP910592997), T19, T18));
					T1c = VMUL(LDK(KP951056516), VFNMS(LDK(KP910592997), T19, T18));
					T14 = VMUL(LDK(KP951056516), VFNMS(LDK(KP910592997), T11, TU));
					T12 = VMUL(LDK(KP951056516), VFMA(LDK(KP910592997), T11, TU));
					T1p = VFNMS(LDK(KP559016994), T1o, T1n);
					T1t = VFMA(LDK(KP559016994), T1o, T1n);
					ST(&(x[WS(ios, 10)]), VFMAI(T1e, T1d), dist, &(x[0]));
					ST(&(x[WS(ios, 5)]), VFNMSI(T1e, T1d), dist, &(x[WS(ios, 1)]));
					T15 = VFNMS(LDK(KP559016994), TI, TH);
					TJ = VFMA(LDK(KP559016994), TI, TH);
					{
					     V T17, T1b, T13, TR;
					     ST(&(x[WS(ios, 12)]), VFNMSI(T1s, T1p), dist, &(x[0]));
					     ST(&(x[WS(ios, 3)]), VFMAI(T1s, T1p), dist, &(x[WS(ios, 1)]));
					     ST(&(x[WS(ios, 9)]), VFNMSI(T1u, T1t), dist, &(x[WS(ios, 1)]));
					     ST(&(x[WS(ios, 6)]), VFMAI(T1u, T1t), dist, &(x[0]));
					     T17 = VFNMS(LDK(KP823639103), T16, T15);
					     T1b = VFMA(LDK(KP823639103), T16, T15);
					     T13 = VFMA(LDK(KP823639103), TQ, TJ);
					     TR = VFNMS(LDK(KP823639103), TQ, TJ);
					     ST(&(x[WS(ios, 13)]), VFMAI(T1a, T17), dist, &(x[WS(ios, 1)]));
					     ST(&(x[WS(ios, 2)]), VFNMSI(T1a, T17), dist, &(x[0]));
					     ST(&(x[WS(ios, 8)]), VFMAI(T1c, T1b), dist, &(x[0]));
					     ST(&(x[WS(ios, 7)]), VFNMSI(T1c, T1b), dist, &(x[WS(ios, 1)]));
					     ST(&(x[WS(ios, 11)]), VFMAI(T14, T13), dist, &(x[WS(ios, 1)]));
					     ST(&(x[WS(ios, 4)]), VFNMSI(T14, T13), dist, &(x[0]));
					     ST(&(x[WS(ios, 14)]), VFNMSI(T12, TR), dist, &(x[0]));
					     ST(&(x[WS(ios, 1)]), VFMAI(T12, TR), dist, &(x[WS(ios, 1)]));
					}
				   }
			      }
			 }
		    }
	       }
	  }
     }
     return W;
}

static const tw_instr twinstr[] = {
     VTW(1),
     VTW(2),
     VTW(3),
     VTW(4),
     VTW(5),
     VTW(6),
     VTW(7),
     VTW(8),
     VTW(9),
     VTW(10),
     VTW(11),
     VTW(12),
     VTW(13),
     VTW(14),