t2fv_20.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 19:52:50 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 -n 20 -name t2fv_20 -include t2f.h */

/*
 * This function contains 123 FP additions, 88 FP multiplications,
 * (or, 77 additions, 42 multiplications, 46 fused multiply/add),
 * 68 stack variables, 4 constants, and 40 memory accesses
 */
#include "t2f.h"

static void t2fv_20(R *ri, R *ii, const R *W, stride rs, INT mb, INT me, INT ms)
{
     DVK(KP951056516, +0.951056516295153572116439333379382143405698634);
     DVK(KP559016994, +0.559016994374947424102293417182819058860154590);
     DVK(KP250000000, +0.250000000000000000000000000000000000000000000);
     DVK(KP618033988, +0.618033988749894848204586834365638117720309180);
     INT m;
     R *x;
     x = ri;
     for (m = mb, W = W + (mb * ((TWVL / VL) * 38)); m < me; m = m + VL, x = x + (VL * ms), W = W + (TWVL * 38), MAKE_VOLATILE_STRIDE(rs)) {
	  V T4, Tx, T1m, T1K, T1y, Tk, Tf, T16, T10, TT, T1O, T1w, T1L, T1p, T1M;
	  V T1s, TZ, TI, T1x, Tp;
	  {
	       V T1, Tv, T2, Tt;
	       T1 = LD(&(x[0]), ms, &(x[0]));
	       Tv = LD(&(x[WS(rs, 15)]), ms, &(x[WS(rs, 1)]));
	       T2 = LD(&(x[WS(rs, 10)]), ms, &(x[0]));
	       Tt = LD(&(x[WS(rs, 5)]), ms, &(x[WS(rs, 1)]));
	       {
		    V T9, T1n, TN, T1v, TS, Te, T1q, T1u, TE, TG, Tm, T1o, TC, Tn, T1r;
		    V TH, To;
		    {
			 V TP, TR, Ta, Tc;
			 {
			      V T5, T7, TJ, TL, T1k, T1l;
			      T5 = LD(&(x[WS(rs, 4)]), ms, &(x[0]));
			      T7 = LD(&(x[WS(rs, 14)]), ms, &(x[0]));
			      TJ = LD(&(x[WS(rs, 13)]), ms, &(x[WS(rs, 1)]));
			      TL = LD(&(x[WS(rs, 3)]), ms, &(x[WS(rs, 1)]));
			      {
				   V Tw, T3, Tu, T6, T8, TK, TM, TO, TQ;
				   TO = LD(&(x[WS(rs, 17)]), ms, &(x[WS(rs, 1)]));
				   Tw = BYTWJ(&(W[TWVL * 28]), Tv);
				   T3 = BYTWJ(&(W[TWVL * 18]), T2);
				   Tu = BYTWJ(&(W[TWVL * 8]), Tt);
				   T6 = BYTWJ(&(W[TWVL * 6]), T5);
				   T8 = BYTWJ(&(W[TWVL * 26]), T7);
				   TK = BYTWJ(&(W[TWVL * 24]), TJ);
				   TM = BYTWJ(&(W[TWVL * 4]), TL);
				   TP = BYTWJ(&(W[TWVL * 32]), TO);
				   TQ = LD(&(x[WS(rs, 7)]), ms, &(x[WS(rs, 1)]));
				   T4 = VSUB(T1, T3);
				   T1k = VADD(T1, T3);
				   Tx = VSUB(Tu, Tw);
				   T1l = VADD(Tu, Tw);
				   T9 = VSUB(T6, T8);
				   T1n = VADD(T6, T8);
				   TN = VSUB(TK, TM);
				   T1v = VADD(TK, TM);
				   TR = BYTWJ(&(W[TWVL * 12]), TQ);
			      }
			      Ta = LD(&(x[WS(rs, 16)]), ms, &(x[0]));
			      T1m = VSUB(T1k, T1l);
			      T1K = VADD(T1k, T1l);
			      Tc = LD(&(x[WS(rs, 6)]), ms, &(x[0]));
			 }
			 {
			      V Tb, TA, Td, Th, Tj, Tz, Tg, Ti, Ty;
			      Tg = LD(&(x[WS(rs, 8)]), ms, &(x[0]));
			      Ti = LD(&(x[WS(rs, 18)]), ms, &(x[0]));
			      Ty = LD(&(x[WS(rs, 9)]), ms, &(x[WS(rs, 1)]));
			      TS = VSUB(TP, TR);
			      T1y = VADD(TP, TR);
			      Tb = BYTWJ(&(W[TWVL * 30]), Ta);
			      TA = LD(&(x[WS(rs, 19)]), ms, &(x[WS(rs, 1)]));
			      Td = BYTWJ(&(W[TWVL * 10]), Tc);
			      Th = BYTWJ(&(W[TWVL * 14]), Tg);
			      Tj = BYTWJ(&(W[TWVL * 34]), Ti);
			      Tz = BYTWJ(&(W[TWVL * 16]), Ty);
			      {
				   V TD, TF, TB, Tl;
				   TD = LD(&(x[WS(rs, 1)]), ms, &(x[WS(rs, 1)]));
				   TF = LD(&(x[WS(rs, 11)]), ms, &(x[WS(rs, 1)]));
				   Tl = LD(&(x[WS(rs, 12)]), ms, &(x[0]));
				   TB = BYTWJ(&(W[TWVL * 36]), TA);
				   Te = VSUB(Tb, Td);
				   T1q = VADD(Tb, Td);
				   Tk = VSUB(Th, Tj);
				   T1u = VADD(Th, Tj);
				   TE = BYTWJ(&(W[0]), TD);
				   TG = BYTWJ(&(W[TWVL * 20]), TF);
				   Tm = BYTWJ(&(W[TWVL * 22]), Tl);
				   T1o = VADD(Tz, TB);
				   TC = VSUB(Tz, TB);
				   Tn = LD(&(x[WS(rs, 2)]), ms, &(x[0]));
			      }
			 }
		    }
		    Tf = VADD(T9, Te);
		    T16 = VSUB(T9, Te);
		    T10 = VSUB(TS, TN);
		    TT = VADD(TN, TS);
		    T1r = VADD(TE, TG);
		    TH = VSUB(TE, TG);
		    T1O = VADD(T1u, T1v);
		    T1w = VSUB(T1u, T1v);
		    To = BYTWJ(&(W[TWVL * 2]), Tn);
		    T1L = VADD(T1n, T1o);
		    T1p = VSUB(T1n, T1o);
		    T1M = VADD(T1q, T1r);
		    T1s = VSUB(T1q, T1r);
		    TZ = VSUB(TH, TC);
		    TI = VADD(TC, TH);
		    T1x = VADD(Tm, To);
		    Tp = VSUB(Tm, To);
	       }
	  }
	  {
	       V T1V, T1N, T14, T1d, T11, T1G, T1t, T1z, T1P, Tq, T17, T13, TV, TU;
	       T1V = VSUB(T1L, T1M);
	       T1N = VADD(T1L, T1M);
	       T14 = VSUB(TT, TI);
	       TU = VADD(TI, TT);
	       T1d = VFNMS(LDK(KP618033988), TZ, T10);
	       T11 = VFMA(LDK(KP618033988), T10, TZ);
	       T1G = VSUB(T1p, T1s);
	       T1t = VADD(T1p, T1s);
	       T1z = VSUB(T1x, T1y);
	       T1P = VADD(T1x, T1y);
	       Tq = VADD(Tk, Tp);
	       T17 = VSUB(Tk, Tp);
	       T13 = VFNMS(LDK(KP250000000), TU, Tx);
	       TV = VADD(Tx, TU);
	       {
		    V T1J, T1H, T1D, T1Z, T1X, T1T, T1h, T1j, T1b, T19, T1C, T1S, T1c, TY, T1F;
		    V T1A;
		    T1F = VSUB(T1w, T1z);
		    T1A = VADD(T1w, T1z);
		    {
			 V T1W, T1Q, TX, Tr;
			 T1W = VSUB(T1O, T1P);
			 T1Q = VADD(T1O, T1P);
			 TX = VSUB(Tf, Tq);
			 Tr = VADD(Tf, Tq);
			 {
			      V T1g, T18, T1f, T15;
			      T1g = VFNMS(LDK(KP618033988), T16, T17);
			      T18 = VFMA(LDK(KP618033988), T17, T16);
			      T1f = VFMA(LDK(KP559016994), T14, T13);
			      T15 = VFNMS(LDK(KP559016994), T14, T13);
			      T1J = VMUL(LDK(KP951056516), VFMA(LDK(KP618033988), T1F, T1G));
			      T1H = VMUL(LDK(KP951056516), VFNMS(LDK(KP618033988), T1G, T1F));
			      {
				   V T1B, T1R, TW, Ts;
				   T1B = VADD(T1t, T1A);
				   T1D = VSUB(T1t, T1A);
				   T1Z = VMUL(LDK(KP951056516), VFNMS(LDK(KP618033988), T1V, T1W));
				   T1X = VMUL(LDK(KP951056516), VFMA(LDK(KP618033988), T1W, T1V));
				   T1R = VADD(T1N, T1Q);
				   T1T = VSUB(T1N, T1Q);
				   TW = VFNMS(LDK(KP250000000), Tr, T4);
				   Ts = VADD(T4, Tr);
				   T1h = VFNMS(LDK(KP951056516), T1g, T1f);
				   T1j = VFMA(LDK(KP951056516), T1g, T1f);
				   T1b = VFNMS(LDK(KP951056516), T18, T15);
				   T19 = VFMA(LDK(KP951056516), T18, T15);
				   ST(&(x[WS(rs, 10)]), VADD(T1m, T1B), ms, &(x[0]));
				   T1C = VFNMS(LDK(KP250000000), T1B, T1m);
				   ST(&(x[0]), VADD(T1K, T1R), ms, &(x[0]));
				   T1S = VFNMS(LDK(KP250000000), T1R, T1K);
				   T1c = VFNMS(LDK(KP559016994), TX, TW);
				   TY = VFMA(LDK(KP559016994), TX, TW);
				   ST(&(x[WS(rs, 15)]), VFMAI(TV, Ts), ms, &(x[WS(rs, 1)]));
				   ST(&(x[WS(rs, 5)]), VFNMSI(TV, Ts), ms, &(x[WS(rs, 1)]));
			      }
			 }
		    }
		    {
			 V T1E, T1I, T1U, T1Y;
			 T1E = VFNMS(LDK(KP559016994), T1D, T1C);
			 T1I = VFMA(LDK(KP559016994), T1D, T1C);
			 T1U = VFMA(LDK(KP559016994), T1T, T1S);
			 T1Y = VFNMS(LDK(KP559016994), T1T, T1S);
			 {
			      V T1e, T1i, T1a, T12;
			      T1e = VFNMS(LDK(KP951056516), T1d, T1c);
			      T1i = VFMA(LDK(KP951056516), T1d, T1c);
			      T1a = VFNMS(LDK(KP951056516), T11, TY);
			      T12 = VFMA(LDK(KP951056516), T11, TY);
			      ST(&(x[WS(rs, 18)]), VFNMSI(T1H, T1E), ms, &(x[0]));
			      ST(&(x[WS(rs, 2)]), VFMAI(T1H, T1E), ms, &(x[0]));
			      ST(&(x[WS(rs, 14)]), VFMAI(T1J, T1I), ms, &(x[0]));
			      ST(&(x[WS(rs, 6)]), VFNMSI(T1J, T1I), ms, &(x[0]));
			      ST(&(x[WS(rs, 16)]), VFNMSI(T1X, T1U), ms, &(x[0]));
			      ST(&(x[WS(rs, 4)]), VFMAI(T1X, T1U), ms, &(x[0]));
			      ST(&(x[WS(rs, 12)]), VFMAI(T1Z, T1Y), ms, &(x[0]));
			      ST(&(x[WS(rs, 8)]), VFNMSI(T1Z, T1Y), ms, &(x[0]));
			      ST(&(x[WS(rs, 3)]), VFMAI(T1h, T1e), ms, &(x[WS(rs, 1)]));
			      ST(&(x[WS(rs, 17)]), VFNMSI(T1h, T1e), ms, &(x[WS(rs, 1)]));
			      ST(&(x[WS(rs, 7)]), VFMAI(T1j, T1i), ms, &(x[WS(rs, 1)]));
			      ST(&(x[WS(rs, 13)]), VFNMSI(T1j, T1i), ms, &(x[WS(rs, 1)]));
			      ST(&(x[WS(rs, 11)]), VFMAI(T1b, T1a), ms, &(x[WS(rs, 1)]));
			      ST(&(x[WS(rs, 9)]), VFNMSI(T1b, T1a), ms, &(x[WS(rs, 1)]));
			      ST(&(x[WS(rs, 19)]), VFMAI(T19, T12), ms, &(x[WS(rs, 1)]));
			      ST(&(x[WS(rs, 1)]), VFNMSI(T19, T12), ms, &(x[WS(rs, 1)]));
			 }
		    }
	       }
	  }
     }
}

static const tw_instr twinstr[] = {
     VTW(0, 1),
     VTW(0, 2),
     VTW(0, 3),
     VTW(0, 4),
     VTW(0, 5),
     VTW(0, 6),
     VTW(0, 7),
     VTW(0, 8),
     VTW(0, 9),
     VTW(0, 10),
     VTW(0, 11),
     VTW(0, 12),
     VTW(0, 13),
     VTW(0, 14),
     VTW(0, 15),
     VTW(0, 16),
     VTW(0, 17),
     VTW(0, 18),