t3bv_5.c

用于DFT计算的c语言的库的最新版本,包含丰富的函数库.

/*
 * 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 Mon Feb  9 19:53:21 EST 2009 */

#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 -twiddle-log3 -precompute-twiddles -no-generate-bytw -n 5 -name t3bv_5 -include t3b.h -sign 1 */

/*
 * This function contains 22 FP additions, 23 FP multiplications,
 * (or, 13 additions, 14 multiplications, 9 fused multiply/add),
 * 30 stack variables, 4 constants, and 10 memory accesses
 */
#include "t3b.h"

static void t3bv_5(R *ri, R *ii, const R *W, stride rs, INT mb, INT me, INT ms)
{
     DVK(KP559016994, +0.559016994374947424102293417182819058860154590);
     DVK(KP250000000, +0.250000000000000000000000000000000000000000000);
     DVK(KP618033988, +0.618033988749894848204586834365638117720309180);
     DVK(KP951056516, +0.951056516295153572116439333379382143405698634);
     INT m;
     R *x;
     x = ii;
     for (m = mb, W = W + (mb * ((TWVL / VL) * 4)); m < me; m = m + VL, x = x + (VL * ms), W = W + (TWVL * 4), MAKE_VOLATILE_STRIDE(rs)) {
	  V T2, T5, T1, T3, Td, T7, Tb;
	  T2 = LDW(&(W[0]));
	  T5 = LDW(&(W[TWVL * 2]));
	  T1 = LD(&(x[0]), ms, &(x[0]));
	  T3 = LD(&(x[WS(rs, 1)]), ms, &(x[WS(rs, 1)]));
	  Td = LD(&(x[WS(rs, 3)]), ms, &(x[WS(rs, 1)]));
	  T7 = LD(&(x[WS(rs, 4)]), ms, &(x[0]));
	  Tb = LD(&(x[WS(rs, 2)]), ms, &(x[0]));
	  {
	       V Ta, T6, T4, Te, Tc, T8;
	       Ta = VZMULJ(T2, T5);
	       T6 = VZMUL(T2, T5);
	       T4 = VZMUL(T2, T3);
	       Te = VZMUL(T5, Td);
	       Tc = VZMUL(Ta, Tb);
	       T8 = VZMUL(T6, T7);
	       {
		    V Tf, Tl, T9, Tk;
		    Tf = VADD(Tc, Te);
		    Tl = VSUB(Tc, Te);
		    T9 = VADD(T4, T8);
		    Tk = VSUB(T4, T8);
		    {
			 V Ti, Tg, To, Tm, Th, Tn, Tj;
			 Ti = VSUB(T9, Tf);
			 Tg = VADD(T9, Tf);
			 To = VMUL(LDK(KP951056516), VFNMS(LDK(KP618033988), Tk, Tl));
			 Tm = VMUL(LDK(KP951056516), VFMA(LDK(KP618033988), Tl, Tk));
			 Th = VFNMS(LDK(KP250000000), Tg, T1);
			 ST(&(x[0]), VADD(T1, Tg), ms, &(x[0]));
			 Tn = VFNMS(LDK(KP559016994), Ti, Th);
			 Tj = VFMA(LDK(KP559016994), Ti, Th);
			 ST(&(x[WS(rs, 2)]), VFNMSI(To, Tn), ms, &(x[0]));
			 ST(&(x[WS(rs, 3)]), VFMAI(To, Tn), ms, &(x[WS(rs, 1)]));
			 ST(&(x[WS(rs, 4)]), VFNMSI(Tm, Tj), ms, &(x[0]));
			 ST(&(x[WS(rs, 1)]), VFMAI(Tm, Tj), ms, &(x[WS(rs, 1)]));
		    }
	       }
	  }
     }
}

static const tw_instr twinstr[] = {
     VTW(0, 1),
     VTW(0, 3),
     {TW_NEXT, VL, 0}
};

static const ct_desc desc = { 5, "t3bv_5", twinstr, &GENUS, {13, 14, 9, 0}, 0, 0, 0 };

void X(codelet_t3bv_5) (planner *p) {
     X(kdft_dit_register) (p, t3bv_5, &desc);
}
#else				/* HAVE_FMA */

/* Generated by: ../../../genfft/gen_twiddle_c -simd -compact -variables 4 -pipeline-latency 8 -twiddle-log3 -precompute-twiddles -no-generate-bytw -n 5 -name t3bv_5 -include t3b.h -sign 1 */

/*
 * This function contains 22 FP additions, 18 FP multiplications,
 * (or, 19 additions, 15 multiplications, 3 fused multiply/add),
 * 24 stack variables, 4 constants, and 10 memory accesses
 */
#include "t3b.h"

static void t3bv_5(R *ri, R *ii, const R *W, stride rs, INT mb, INT me, INT ms)
{
     DVK(KP250000000, +0.250000000000000000000000000000000000000000000);
     DVK(KP559016994, +0.559016994374947424102293417182819058860154590);
     DVK(KP587785252, +0.587785252292473129168705954639072768597652438);
     DVK(KP951056516, +0.951056516295153572116439333379382143405698634);
     INT m;
     R *x;
     x = ii;
     for (m = mb, W = W + (mb * ((TWVL / VL) * 4)); m < me; m = m + VL, x = x + (VL * ms), W = W + (TWVL * 4), MAKE_VOLATILE_STRIDE(rs)) {
	  V T1, T4, T5, T9;
	  T1 = LDW(&(W[0]));
	  T4 = LDW(&(W[TWVL * 2]));
	  T5 = VZMUL(T1, T4);
	  T9 = VZMULJ(T1, T4);
	  {
	       V Tj, T8, Te, Tg, Th, Tk;
	       Tj = LD(&(x[0]), ms, &(x[0]));
	       {
		    V T3, Td, T7, Tb;
		    {
			 V T2, Tc, T6, Ta;
			 T2 = LD(&(x[WS(rs, 1)]), ms, &(x[WS(rs, 1)]));
			 T3 = VZMUL(T1, T2);
			 Tc = LD(&(x[WS(rs, 3)]), ms, &(x[WS(rs, 1)]));
			 Td = VZMUL(T4, Tc);
			 T6 = LD(&(x[WS(rs, 4)]), ms, &(x[0]));
			 T7 = VZMUL(T5, T6);
			 Ta = LD(&(x[WS(rs, 2)]), ms, &(x[0]));
			 Tb = VZMUL(T9, Ta);
		    }
		    T8 = VSUB(T3, T7);
		    Te = VSUB(Tb, Td);
		    Tg = VADD(T3, T7);
		    Th = VADD(Tb, Td);
		    Tk = VADD(Tg, Th);
	       }
	       ST(&(x[0]), VADD(Tj, Tk), ms, &(x[0]));
	       {
		    V Tf, Tn, Tm, To, Ti, Tl;
		    Tf = VBYI(VFMA(LDK(KP951056516), T8, VMUL(LDK(KP587785252), Te)));
		    Tn = VBYI(VFNMS(LDK(KP951056516), Te, VMUL(LDK(KP587785252), T8)));
		    Ti = VMUL(LDK(KP559016994), VSUB(Tg, Th));
		    Tl = VFNMS(LDK(KP250000000), Tk, Tj);
		    Tm = VADD(Ti, Tl);
		    To = VSUB(Tl, Ti);
		    ST(&(x[WS(rs, 1)]), VADD(Tf, Tm), ms, &(x[WS(rs, 1)]));
		    ST(&(x[WS(rs, 3)]), VSUB(To, Tn), ms, &(x[WS(rs, 1)]));
		    ST(&(x[WS(rs, 4)]), VSUB(Tm, Tf), ms, &(x[0]));
		    ST(&(x[WS(rs, 2)]), VADD(Tn, To), ms, &(x[0]));
	       }
	  }
     }
}

static const tw_instr twinstr[] = {
     VTW(0, 1),
     VTW(0, 3),
     {TW_NEXT, VL, 0}
};

static const ct_desc desc = { 5, "t3bv_5", twinstr, &GENUS, {19, 15, 3, 0}, 0, 0, 0 };

void X(codelet_t3bv_5) (planner *p) {
     X(kdft_dit_register) (p, t3bv_5, &desc);
}
#endif				/* HAVE_FMA */