n2sv_4.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 20:46:23 EST 2008 */

#include "codelet-dft.h"

#ifdef HAVE_FMA

/* Generated by: ../../../genfft/gen_notw -fma -reorder-insns -schedule-for-pipeline -simd -compact -variables 4 -pipeline-latency 8 -n 4 -name n2sv_4 -with-ostride 1 -include n2s.h -store-multiple 4 */

/*
 * This function contains 16 FP additions, 0 FP multiplications,
 * (or, 16 additions, 0 multiplications, 0 fused multiply/add),
 * 25 stack variables, 0 constants, and 18 memory accesses
 */
#include "n2s.h"

static void n2sv_4(const R *ri, const R *ii, R *ro, R *io, stride is, stride os, INT v, INT ivs, INT ovs)
{
     INT i;
     for (i = v; i > 0; i = i - (2 * VL), ri = ri + ((2 * VL) * ivs), ii = ii + ((2 * VL) * ivs), ro = ro + ((2 * VL) * ovs), io = io + ((2 * VL) * ovs), MAKE_VOLATILE_STRIDE(is), MAKE_VOLATILE_STRIDE(os)) {
	  V T1, T2, T7, T8, T4, T5, Tc, Td;
	  T1 = LD(&(ri[0]), ivs, &(ri[0]));
	  T2 = LD(&(ri[WS(is, 2)]), ivs, &(ri[0]));
	  T7 = LD(&(ii[0]), ivs, &(ii[0]));
	  T8 = LD(&(ii[WS(is, 2)]), ivs, &(ii[0]));
	  T4 = LD(&(ri[WS(is, 1)]), ivs, &(ri[WS(is, 1)]));
	  T5 = LD(&(ri[WS(is, 3)]), ivs, &(ri[WS(is, 1)]));
	  Tc = LD(&(ii[WS(is, 1)]), ivs, &(ii[WS(is, 1)]));
	  Td = LD(&(ii[WS(is, 3)]), ivs, &(ii[WS(is, 1)]));
	  {
	       V T3, Tb, T9, Tf, T6, Ta, Te, Tg;
	       T3 = VADD(T1, T2);
	       Tb = VSUB(T1, T2);
	       T9 = VSUB(T7, T8);
	       Tf = VADD(T7, T8);
	       T6 = VADD(T4, T5);
	       Ta = VSUB(T4, T5);
	       Te = VSUB(Tc, Td);
	       Tg = VADD(Tc, Td);
	       {
		    V Th, Ti, Tj, Tk;
		    Th = VADD(Ta, T9);
		    STM4(&(io[3]), Th, ovs, &(io[1]));
		    Ti = VSUB(T9, Ta);
		    STM4(&(io[1]), Ti, ovs, &(io[1]));
		    Tj = VADD(T3, T6);
		    STM4(&(ro[0]), Tj, ovs, &(ro[0]));
		    Tk = VSUB(T3, T6);
		    STM4(&(ro[2]), Tk, ovs, &(ro[0]));
		    {
			 V Tl, Tm, Tn, To;
			 Tl = VADD(Tf, Tg);
			 STM4(&(io[0]), Tl, ovs, &(io[0]));
			 Tm = VSUB(Tf, Tg);
			 STM4(&(io[2]), Tm, ovs, &(io[0]));
			 STN4(&(io[0]), Tl, Ti, Tm, Th, ovs);
			 Tn = VSUB(Tb, Te);
			 STM4(&(ro[3]), Tn, ovs, &(ro[1]));
			 To = VADD(Tb, Te);
			 STM4(&(ro[1]), To, ovs, &(ro[1]));
			 STN4(&(ro[0]), Tj, To, Tk, Tn, ovs);
		    }
	       }
	  }
     }
}

static const kdft_desc desc = { 4, "n2sv_4", {16, 0, 0, 0}, &GENUS, 0, 1, 0, 0 };
void X(codelet_n2sv_4) (planner *p) {
     X(kdft_register) (p, n2sv_4, &desc);
}

#else				/* HAVE_FMA */

/* Generated by: ../../../genfft/gen_notw -simd -compact -variables 4 -pipeline-latency 8 -n 4 -name n2sv_4 -with-ostride 1 -include n2s.h -store-multiple 4 */

/*
 * This function contains 16 FP additions, 0 FP multiplications,
 * (or, 16 additions, 0 multiplications, 0 fused multiply/add),
 * 17 stack variables, 0 constants, and 18 memory accesses
 */
#include "n2s.h"

static void n2sv_4(const R *ri, const R *ii, R *ro, R *io, stride is, stride os, INT v, INT ivs, INT ovs)
{
     INT i;
     for (i = v; i > 0; i = i - (2 * VL), ri = ri + ((2 * VL) * ivs), ii = ii + ((2 * VL) * ivs), ro = ro + ((2 * VL) * ovs), io = io + ((2 * VL) * ovs), MAKE_VOLATILE_STRIDE(is), MAKE_VOLATILE_STRIDE(os)) {
	  V T3, Tb, T9, Tf, T6, Ta, Te, Tg;
	  {
	       V T1, T2, T7, T8;
	       T1 = LD(&(ri[0]), ivs, &(ri[0]));
	       T2 = LD(&(ri[WS(is, 2)]), ivs, &(ri[0]));
	       T3 = VADD(T1, T2);
	       Tb = VSUB(T1, T2);
	       T7 = LD(&(ii[0]), ivs, &(ii[0]));
	       T8 = LD(&(ii[WS(is, 2)]), ivs, &(ii[0]));
	       T9 = VSUB(T7, T8);
	       Tf = VADD(T7, T8);
	  }
	  {
	       V T4, T5, Tc, Td;
	       T4 = LD(&(ri[WS(is, 1)]), ivs, &(ri[WS(is, 1)]));
	       T5 = LD(&(ri[WS(is, 3)]), ivs, &(ri[WS(is, 1)]));
	       T6 = VADD(T4, T5);
	       Ta = VSUB(T4, T5);
	       Tc = LD(&(ii[WS(is, 1)]), ivs, &(ii[WS(is, 1)]));
	       Td = LD(&(ii[WS(is, 3)]), ivs, &(ii[WS(is, 1)]));
	       Te = VSUB(Tc, Td);
	       Tg = VADD(Tc, Td);
	  }
	  {
	       V Th, Ti, Tj, Tk;
	       Th = VSUB(T3, T6);
	       STM4(&(ro[2]), Th, ovs, &(ro[0]));
	       Ti = VSUB(Tf, Tg);
	       STM4(&(io[2]), Ti, ovs, &(io[0]));
	       Tj = VADD(T3, T6);
	       STM4(&(ro[0]), Tj, ovs, &(ro[0]));
	       Tk = VADD(Tf, Tg);
	       STM4(&(io[0]), Tk, ovs, &(io[0]));
	       {
		    V Tl, Tm, Tn, To;
		    Tl = VSUB(T9, Ta);
		    STM4(&(io[1]), Tl, ovs, &(io[1]));
		    Tm = VADD(Tb, Te);
		    STM4(&(ro[1]), Tm, ovs, &(ro[1]));
		    Tn = VADD(Ta, T9);
		    STM4(&(io[3]), Tn, ovs, &(io[1]));
		    STN4(&(io[0]), Tk, Tl, Ti, Tn, ovs);
		    To = VSUB(Tb, Te);
		    STM4(&(ro[3]), To, ovs, &(ro[1]));
		    STN4(&(ro[0]), Tj, Tm, Th, To, ovs);
	       }
	  }
     }
}

static const kdft_desc desc = { 4, "n2sv_4", {16, 0, 0, 0}, &GENUS, 0, 1, 0, 0 };
void X(codelet_n2sv_4) (planner *p) {
     X(kdft_register) (p, n2sv_4, &desc);
}

#endif				/* HAVE_FMA */