hc2cbdftv_16.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 21:14:18 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 16 -dif -sign 1 -name hc2cbdftv_16 -include hc2cbv.h */

/*
 * This function contains 103 FP additions, 80 FP multiplications,
 * (or, 53 additions, 30 multiplications, 50 fused multiply/add),
 * 123 stack variables, 3 constants, and 32 memory accesses
 */
#include "hc2cbv.h"

static void hc2cbdftv_16(R *Rp, R *Ip, R *Rm, R *Im, const R *W, stride rs, INT mb, INT me, INT ms)
{
     DVK(KP923879532, +0.923879532511286756128183189396788286822416626);
     DVK(KP414213562, +0.414213562373095048801688724209698078569671875);
     DVK(KP707106781, +0.707106781186547524400844362104849039284835938);
     INT m;
     for (m = mb, W = W + ((mb - 1) * ((TWVL / VL) * 30)); m < me; m = m + VL, Rp = Rp + (VL * ms), Ip = Ip + (VL * ms), Rm = Rm - (VL * ms), Im = Im - (VL * ms), W = W + (TWVL * 30), MAKE_VOLATILE_STRIDE(rs)) {
	  V T1D, T1F, TV, TW, T17, T18, T1B, T1A, T1H, T1G;
	  {
	       V T8, Tv, Tb, TF, Tl, TJ, TP, T1w, TE, T1t, T10, T1p, TG, Te, Tg;
	       V Th, T2, T3, Ts, Tt, T5, T6, Tp, Tq, T9, TA, T4, TC, Tu, TN;
	       V T7, TB, Tr, Ta, Tj, Tk, Tc, Td, TY, TD, TO, TZ, T1Q, T19, T1I;
	       V T1d, Tf, T11, TH, TQ, Ti, TI, T1k, T1K, T1S, T1r, T14, T16, TU, Ty;
	       V T1z, TX, T1o, T1, TK, TR, Tm, T12, T1C, Tz, T15;
	       T2 = LD(&(Rp[0]), ms, &(Rp[0]));
	       T3 = LD(&(Rm[WS(rs, 7)]), -ms, &(Rm[WS(rs, 1)]));
	       Ts = LD(&(Rp[WS(rs, 6)]), ms, &(Rp[0]));
	       Tt = LD(&(Rm[WS(rs, 1)]), -ms, &(Rm[WS(rs, 1)]));
	       T5 = LD(&(Rp[WS(rs, 4)]), ms, &(Rp[0]));
	       T6 = LD(&(Rm[WS(rs, 3)]), -ms, &(Rm[WS(rs, 1)]));
	       Tp = LD(&(Rp[WS(rs, 2)]), ms, &(Rp[0]));
	       Tq = LD(&(Rm[WS(rs, 5)]), -ms, &(Rm[WS(rs, 1)]));
	       T9 = LD(&(Rp[WS(rs, 1)]), ms, &(Rp[WS(rs, 1)]));
	       TA = VFNMSCONJ(T3, T2);
	       T4 = VFMACONJ(T3, T2);
	       TC = VFMSCONJ(Tt, Ts);
	       Tu = VFMACONJ(Tt, Ts);
	       TN = VFNMSCONJ(T6, T5);
	       T7 = VFMACONJ(T6, T5);
	       TB = VFNMSCONJ(Tq, Tp);
	       Tr = VFMACONJ(Tq, Tp);
	       Ta = LD(&(Rm[WS(rs, 6)]), -ms, &(Rm[0]));
	       Tj = LD(&(Rp[WS(rs, 3)]), ms, &(Rp[WS(rs, 1)]));
	       Tk = LD(&(Rm[WS(rs, 4)]), -ms, &(Rm[0]));
	       Tc = LD(&(Rp[WS(rs, 5)]), ms, &(Rp[WS(rs, 1)]));
	       Td = LD(&(Rm[WS(rs, 2)]), -ms, &(Rm[0]));
	       T8 = VSUB(T4, T7);
	       TY = VADD(T4, T7);
	       TD = VADD(TB, TC);
	       TO = VSUB(TB, TC);
	       Tv = VSUB(Tr, Tu);
	       TZ = VADD(Tr, Tu);
	       Tb = VFMACONJ(Ta, T9);
	       TF = VFNMSCONJ(Ta, T9);
	       Tl = VFMACONJ(Tk, Tj);
	       TJ = VFNMSCONJ(Tk, Tj);
	       TP = VFMA(LDK(KP707106781), TO, TN);
	       T1w = VFNMS(LDK(KP707106781), TO, TN);
	       TE = VFMA(LDK(KP707106781), TD, TA);
	       T1t = VFNMS(LDK(KP707106781), TD, TA);
	       T10 = VADD(TY, TZ);
	       T1p = VSUB(TY, TZ);
	       TG = VFNMSCONJ(Td, Tc);
	       Te = VFMACONJ(Td, Tc);
	       Tg = LD(&(Rp[WS(rs, 7)]), ms, &(Rp[WS(rs, 1)]));
	       Th = LD(&(Rm[0]), -ms, &(Rm[0]));
	       T1Q = LDW(&(W[TWVL * 22]));
	       T19 = LDW(&(W[TWVL * 26]));
	       T1I = LDW(&(W[TWVL * 2]));
	       T1d = LDW(&(W[TWVL * 28]));
	       Tf = VSUB(Tb, Te);
	       T11 = VADD(Tb, Te);
	       TH = VFNMS(LDK(KP414213562), TG, TF);
	       TQ = VFMA(LDK(KP414213562), TF, TG);
	       Ti = VFMACONJ(Th, Tg);
	       TI = VFMSCONJ(Th, Tg);
	       T1k = LDW(&(W[0]));
	       T1K = LDW(&(W[TWVL * 4]));
	       T1S = LDW(&(W[TWVL * 24]));
	       TX = LDW(&(W[TWVL * 14]));
	       T1o = LDW(&(W[TWVL * 6]));
	       T1 = LDW(&(W[TWVL * 10]));
	       TK = VFMA(LDK(KP414213562), TJ, TI);
	       TR = VFNMS(LDK(KP414213562), TI, TJ);
	       Tm = VSUB(Ti, Tl);
	       T12 = VADD(Ti, Tl);
	       T1C = LDW(&(W[TWVL * 18]));
	       Tz = LDW(&(W[TWVL * 12]));
	       T15 = LDW(&(W[TWVL * 16]));
	       {
		    V T1v, T1y, T1N, T1g, T1J, T1c, T1U, T1V, T1m, T1n, T1s, TS, T1u, TL, T1x;
		    V T13, T1q, Tn, Tw, T1L, T1f, TT, T1M, T1e, TM, T1R, T1j, T1b, Tx, T1a;
		    V To, T1T, T1l, T1E, T1O, T1P, T1h, T1i;
		    T1s = LDW(&(W[TWVL * 8]));
		    TS = VADD(TQ, TR);
		    T1u = VSUB(TQ, TR);
		    TL = VADD(TH, TK);
		    T1x = VSUB(TH, TK);
		    T13 = VADD(T11, T12);
		    T1q = VSUB(T11, T12);
		    Tn = VADD(Tf, Tm);
		    Tw = VSUB(Tf, Tm);
		    T1L = VFMA(LDK(KP923879532), T1u, T1t);
		    T1v = VFNMS(LDK(KP923879532), T1u, T1t);
		    T1f = VFMA(LDK(KP923879532), TS, TP);
		    TT = VFNMS(LDK(KP923879532), TS, TP);
		    T1M = VFNMS(LDK(KP923879532), T1x, T1w);
		    T1y = VFMA(LDK(KP923879532), T1x, T1w);
		    T1e = VFMA(LDK(KP923879532), TL, TE);
		    TM = VFNMS(LDK(KP923879532), TL, TE);
		    T1r = VZMUL(T1o, VFMAI(T1q, T1p));
		    T1R = VZMUL(T1Q, VFNMSI(T1q, T1p));
		    T14 = VZMUL(TX, VSUB(T10, T13));
		    T1j = VADD(T10, T13);
		    T1b = VFMA(LDK(KP707106781), Tw, Tv);
		    Tx = VFNMS(LDK(KP707106781), Tw, Tv);
		    T1a = VFMA(LDK(KP707106781), Tn, T8);
		    To = VFNMS(LDK(KP707106781), Tn, T8);
		    T1T = VZMULI(T1S, VFMAI(T1M, T1L));
		    T1N = VZMULI(T1K, VFNMSI(T1M, T1L));
		    T16 = VZMULI(T15, VFMAI(TT, TM));
		    TU = VZMULI(Tz, VFNMSI(TT, TM));
		    T1l = VZMULI(T1k, VFMAI(T1f, T1e));
		    T1g = VZMULI(T1d, VFNMSI(T1f, T1e));
		    T1D = VZMUL(T1C, VFMAI(Tx, To));
		    Ty = VZMUL(T1, VFNMSI(Tx, To));
		    T1J = VZMUL(T1I, VFMAI(T1b, T1a));
		    T1c = VZMUL(T19, VFNMSI(T1b, T1a));
		    T1U = VCONJ(VSUB(T1R, T1T));
		    T1V = VADD(T1R, T1T);
		    T1m = VCONJ(VSUB(T1j, T1l));
		    T1n = VADD(T1j, T1l);
		    T1z = VZMULI(T1s, VFMAI(T1y, T1v));
		    T1E = LDW(&(W[TWVL * 20]));
		    T1O = VCONJ(VSUB(T1J, T1N));
		    T1P = VADD(T1J, T1N);
		    T1h = VCONJ(VSUB(T1c, T1g));
		    T1i = VADD(T1c, T1g);
		    ST(&(Rp[WS(rs, 6)]), T1V, ms, &(Rp[0]));
		    ST(&(Rm[WS(rs, 6)]), T1U, -ms, &(Rm[0]));
		    ST(&(Rp[0]), T1n, ms, &(Rp[0]));
		    ST(&(Rm[0]), T1m, -ms, &(Rm[0]));
		    ST(&(Rp[WS(rs, 1)]), T1P, ms, &(Rp[WS(rs, 1)]));
		    ST(&(Rm[WS(rs, 1)]), T1O, -ms, &(Rm[WS(rs, 1)]));
		    ST(&(Rp[WS(rs, 7)]), T1i, ms, &(Rp[WS(rs, 1)]));
		    ST(&(Rm[WS(rs, 7)]), T1h, -ms, &(Rm[WS(rs, 1)]));
		    T1F = VZMULI(T1E, VFNMSI(T1y, T1v));
	       }
	       TV = VCONJ(VSUB(Ty, TU));
	       TW = VADD(Ty, TU);
	       T17 = VCONJ(VSUB(T14, T16));
	       T18 = VADD(T14, T16);
	       T1B = VADD(T1r, T1z);
	       T1A = VCONJ(VSUB(T1r, T1z));
	  }
	  T1H = VADD(T1D, T1F);
	  T1G = VCONJ(VSUB(T1D, T1F));
	  ST(&(Rm[WS(rs, 3)]), TV, -ms, &(Rm[WS(rs, 1)]));
	  ST(&(Rp[WS(rs, 3)]), TW, ms, &(Rp[WS(rs, 1)]));
	  ST(&(Rm[WS(rs, 4)]), T17, -ms, &(Rm[0]));
	  ST(&(Rm[WS(rs, 2)]), T1A, -ms, &(Rm[0]));
	  ST(&(Rp[WS(rs, 2)]), T1B, ms, &(Rp[0]));
	  ST(&(Rp[WS(rs, 4)]), T18, ms, &(Rp[0]));
	  ST(&(Rp[WS(rs, 5)]), T1H, ms, &(Rp[WS(rs, 1)]));
	  ST(&(Rm[WS(rs, 5)]), T1G, -ms, &(Rm[WS(rs, 1)]));
     }
}

static const tw_instr twinstr[] = {
     VTW(1, 1),
     VTW(1, 2),
     VTW(1, 3),
     VTW(1, 4),
     VTW(1, 5),
     VTW(1, 6),
     VTW(1, 7),
     VTW(1, 8),
     VTW(1, 9),
     VTW(1, 10),
     VTW(1, 11),
     VTW(1, 12),
     VTW(1, 13),
     VTW(1, 14),
     VTW(1, 15),
     {TW_NEXT, VL, 0}
};