hc2cbdftv_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 21:14:14 EST 2008 */

#include "codelet-rdft.h"

#ifdef HAVE_FMA

/* Generated by: ../../../genfft/gen_hc2cdft_c -fma -reorder-insns -schedule-for-pipeline -simd -compact -variables 4 -pipeline-latency 8 -trivial-stores -variables 32 -no-generate-bytw -n 4 -dif -sign 1 -name hc2cbdftv_4 -include hc2cbv.h */

/*
 * This function contains 15 FP additions, 12 FP multiplications,
 * (or, 9 additions, 6 multiplications, 6 fused multiply/add),
 * 20 stack variables, 0 constants, and 8 memory accesses
 */
#include "hc2cbv.h"

static void hc2cbdftv_4(R *Rp, R *Ip, R *Rm, R *Im, const R *W, stride rs, INT mb, INT me, INT ms)
{
     INT m;
     for (m = mb, W = W + ((mb - 1) * ((TWVL / VL) * 6)); m < me; m = m + VL, Rp = Rp + (VL * ms), Ip = Ip + (VL * ms), Rm = Rm - (VL * ms), Im = Im - (VL * ms), W = W + (TWVL * 6), MAKE_VOLATILE_STRIDE(rs)) {
	  V T2, T3, T5, T6, Tf, T1, T9, Ta, T4, Tb, T7, Tc, Th, T8, Tg;
	  V Te, Td, Ti, Tj;
	  T2 = LD(&(Rp[0]), ms, &(Rp[0]));
	  T3 = LD(&(Rm[WS(rs, 1)]), -ms, &(Rm[WS(rs, 1)]));
	  T5 = LD(&(Rp[WS(rs, 1)]), ms, &(Rp[WS(rs, 1)]));
	  T6 = LD(&(Rm[0]), -ms, &(Rm[0]));
	  Tf = LDW(&(W[0]));
	  T1 = LDW(&(W[TWVL * 4]));
	  T9 = LDW(&(W[TWVL * 2]));
	  Ta = VFMACONJ(T3, T2);
	  T4 = VFNMSCONJ(T3, T2);
	  Tb = VFMACONJ(T6, T5);
	  T7 = VFNMSCONJ(T6, T5);
	  Tc = VZMUL(T9, VSUB(Ta, Tb));
	  Th = VADD(Ta, Tb);
	  T8 = VZMULI(T1, VFNMSI(T7, T4));
	  Tg = VZMULI(Tf, VFMAI(T7, T4));
	  Te = VCONJ(VSUB(Tc, T8));
	  Td = VADD(T8, Tc);
	  Ti = VADD(Tg, Th);
	  Tj = VCONJ(VSUB(Th, Tg));
	  ST(&(Rm[WS(rs, 1)]), Te, -ms, &(Rm[WS(rs, 1)]));
	  ST(&(Rp[WS(rs, 1)]), Td, ms, &(Rp[WS(rs, 1)]));
	  ST(&(Rp[0]), Ti, ms, &(Rp[0]));
	  ST(&(Rm[0]), Tj, -ms, &(Rm[0]));
     }
}

static const tw_instr twinstr[] = {
     VTW(1, 1),
     VTW(1, 2),
     VTW(1, 3),
     {TW_NEXT, VL, 0}
};

static const hc2c_desc desc = { 4, "hc2cbdftv_4", twinstr, &GENUS, {9, 6, 6, 0} };

void X(codelet_hc2cbdftv_4) (planner *p) {
     X(khc2c_register) (p, hc2cbdftv_4, &desc, HC2C_VIA_DFT);
}
#else				/* HAVE_FMA */

/* Generated by: ../../../genfft/gen_hc2cdft_c -simd -compact -variables 4 -pipeline-latency 8 -trivial-stores -variables 32 -no-generate-bytw -n 4 -dif -sign 1 -name hc2cbdftv_4 -include hc2cbv.h */

/*
 * This function contains 15 FP additions, 6 FP multiplications,
 * (or, 15 additions, 6 multiplications, 0 fused multiply/add),
 * 22 stack variables, 0 constants, and 8 memory accesses
 */
#include "hc2cbv.h"

static void hc2cbdftv_4(R *Rp, R *Ip, R *Rm, R *Im, const R *W, stride rs, INT mb, INT me, INT ms)
{
     INT m;
     for (m = mb, W = W + ((mb - 1) * ((TWVL / VL) * 6)); m < me; m = m + VL, Rp = Rp + (VL * ms), Ip = Ip + (VL * ms), Rm = Rm - (VL * ms), Im = Im - (VL * ms), W = W + (TWVL * 6), MAKE_VOLATILE_STRIDE(rs)) {
	  V T5, Tc, T9, Td, T2, T4, T3, T6, T8, T7, Tj, Ti, Th, Tk, Tl;
	  V Ta, Te, T1, Tb, Tf, Tg;
	  T2 = LD(&(Rp[0]), ms, &(Rp[0]));
	  T3 = LD(&(Rm[WS(rs, 1)]), -ms, &(Rm[WS(rs, 1)]));
	  T4 = VCONJ(T3);
	  T5 = VSUB(T2, T4);
	  Tc = VADD(T2, T4);
	  T6 = LD(&(Rp[WS(rs, 1)]), ms, &(Rp[WS(rs, 1)]));
	  T7 = LD(&(Rm[0]), -ms, &(Rm[0]));
	  T8 = VCONJ(T7);
	  T9 = VBYI(VSUB(T6, T8));
	  Td = VADD(T6, T8);
	  Tj = VADD(Tc, Td);
	  Th = LDW(&(W[0]));
	  Ti = VZMULI(Th, VADD(T5, T9));
	  Tk = VADD(Ti, Tj);
	  ST(&(Rp[0]), Tk, ms, &(Rp[0]));
	  Tl = VCONJ(VSUB(Tj, Ti));
	  ST(&(Rm[0]), Tl, -ms, &(Rm[0]));
	  T1 = LDW(&(W[TWVL * 4]));
	  Ta = VZMULI(T1, VSUB(T5, T9));
	  Tb = LDW(&(W[TWVL * 2]));
	  Te = VZMUL(Tb, VSUB(Tc, Td));
	  Tf = VADD(Ta, Te);
	  ST(&(Rp[WS(rs, 1)]), Tf, ms, &(Rp[WS(rs, 1)]));
	  Tg = VCONJ(VSUB(Te, Ta));
	  ST(&(Rm[WS(rs, 1)]), Tg, -ms, &(Rm[WS(rs, 1)]));
     }
}

static const tw_instr twinstr[] = {
     VTW(1, 1),
     VTW(1, 2),
     VTW(1, 3),
     {TW_NEXT, VL, 0}
};

static const hc2c_desc desc = { 4, "hc2cbdftv_4", twinstr, &GENUS, {15, 6, 0, 0} };

void X(codelet_hc2cbdftv_4) (planner *p) {
     X(khc2c_register) (p, hc2cbdftv_4, &desc, HC2C_VIA_DFT);
}
#endif				/* HAVE_FMA */