Chris@10: /*
Chris@10:  * Copyright (c) 2003, 2007-11 Matteo Frigo
Chris@10:  * Copyright (c) 2003, 2007-11 Massachusetts Institute of Technology
Chris@10:  *
Chris@10:  * This program is free software; you can redistribute it and/or modify
Chris@10:  * it under the terms of the GNU General Public License as published by
Chris@10:  * the Free Software Foundation; either version 2 of the License, or
Chris@10:  * (at your option) any later version.
Chris@10:  *
Chris@10:  * This program is distributed in the hope that it will be useful,
Chris@10:  * but WITHOUT ANY WARRANTY; without even the implied warranty of
Chris@10:  * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
Chris@10:  * GNU General Public License for more details.
Chris@10:  *
Chris@10:  * You should have received a copy of the GNU General Public License
Chris@10:  * along with this program; if not, write to the Free Software
Chris@10:  * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA  02110-1301  USA
Chris@10:  *
Chris@10:  */
Chris@10: 
Chris@10: /* This file was automatically generated --- DO NOT EDIT */
Chris@10: /* Generated on Sun Nov 25 07:35:59 EST 2012 */
Chris@10: 
Chris@10: #include "codelet-dft.h"
Chris@10: 
Chris@10: #ifdef HAVE_FMA
Chris@10: 
Chris@10: /* Generated by: ../../../genfft/gen_twiddle.native -fma -reorder-insns -schedule-for-pipeline -compact -variables 4 -pipeline-latency 4 -twiddle-log3 -precompute-twiddles -n 4 -name t2_4 -include t.h */
Chris@10: 
Chris@10: /*
Chris@10:  * This function contains 24 FP additions, 16 FP multiplications,
Chris@10:  * (or, 16 additions, 8 multiplications, 8 fused multiply/add),
Chris@10:  * 33 stack variables, 0 constants, and 16 memory accesses
Chris@10:  */
Chris@10: #include "t.h"
Chris@10: 
Chris@10: static void t2_4(R *ri, R *ii, const R *W, stride rs, INT mb, INT me, INT ms)
Chris@10: {
Chris@10:      {
Chris@10: 	  INT m;
Chris@10: 	  for (m = mb, W = W + (mb * 4); m < me; m = m + 1, ri = ri + ms, ii = ii + ms, W = W + 4, MAKE_VOLATILE_STRIDE(8, rs)) {
Chris@10: 	       E Ti, Tq, To, Te, Ty, Tz, Tm, Ts;
Chris@10: 	       {
Chris@10: 		    E T2, T6, T3, T5;
Chris@10: 		    T2 = W[0];
Chris@10: 		    T6 = W[3];
Chris@10: 		    T3 = W[2];
Chris@10: 		    T5 = W[1];
Chris@10: 		    {
Chris@10: 			 E T1, Tx, Td, Tw, Tj, Tl, Ta, T4, Tk, Tr;
Chris@10: 			 T1 = ri[0];
Chris@10: 			 Ta = T2 * T6;
Chris@10: 			 T4 = T2 * T3;
Chris@10: 			 Tx = ii[0];
Chris@10: 			 {
Chris@10: 			      E T8, Tb, T7, Tc;
Chris@10: 			      T8 = ri[WS(rs, 2)];
Chris@10: 			      Tb = FNMS(T5, T3, Ta);
Chris@10: 			      T7 = FMA(T5, T6, T4);
Chris@10: 			      Tc = ii[WS(rs, 2)];
Chris@10: 			      {
Chris@10: 				   E Tf, Th, T9, Tv, Tg, Tp;
Chris@10: 				   Tf = ri[WS(rs, 1)];
Chris@10: 				   Th = ii[WS(rs, 1)];
Chris@10: 				   T9 = T7 * T8;
Chris@10: 				   Tv = T7 * Tc;
Chris@10: 				   Tg = T2 * Tf;
Chris@10: 				   Tp = T2 * Th;
Chris@10: 				   Td = FMA(Tb, Tc, T9);
Chris@10: 				   Tw = FNMS(Tb, T8, Tv);
Chris@10: 				   Ti = FMA(T5, Th, Tg);
Chris@10: 				   Tq = FNMS(T5, Tf, Tp);
Chris@10: 			      }
Chris@10: 			      Tj = ri[WS(rs, 3)];
Chris@10: 			      Tl = ii[WS(rs, 3)];
Chris@10: 			 }
Chris@10: 			 To = T1 - Td;
Chris@10: 			 Te = T1 + Td;
Chris@10: 			 Ty = Tw + Tx;
Chris@10: 			 Tz = Tx - Tw;
Chris@10: 			 Tk = T3 * Tj;
Chris@10: 			 Tr = T3 * Tl;
Chris@10: 			 Tm = FMA(T6, Tl, Tk);
Chris@10: 			 Ts = FNMS(T6, Tj, Tr);
Chris@10: 		    }
Chris@10: 	       }
Chris@10: 	       {
Chris@10: 		    E Tn, TA, Tu, Tt;
Chris@10: 		    Tn = Ti + Tm;
Chris@10: 		    TA = Ti - Tm;
Chris@10: 		    Tu = Tq + Ts;
Chris@10: 		    Tt = Tq - Ts;
Chris@10: 		    ii[WS(rs, 3)] = TA + Tz;
Chris@10: 		    ii[WS(rs, 1)] = Tz - TA;
Chris@10: 		    ri[0] = Te + Tn;
Chris@10: 		    ri[WS(rs, 2)] = Te - Tn;
Chris@10: 		    ri[WS(rs, 1)] = To + Tt;
Chris@10: 		    ri[WS(rs, 3)] = To - Tt;
Chris@10: 		    ii[WS(rs, 2)] = Ty - Tu;
Chris@10: 		    ii[0] = Tu + Ty;
Chris@10: 	       }
Chris@10: 	  }
Chris@10:      }
Chris@10: }
Chris@10: 
Chris@10: static const tw_instr twinstr[] = {
Chris@10:      {TW_CEXP, 0, 1},
Chris@10:      {TW_CEXP, 0, 3},
Chris@10:      {TW_NEXT, 1, 0}
Chris@10: };
Chris@10: 
Chris@10: static const ct_desc desc = { 4, "t2_4", twinstr, &GENUS, {16, 8, 8, 0}, 0, 0, 0 };
Chris@10: 
Chris@10: void X(codelet_t2_4) (planner *p) {
Chris@10:      X(kdft_dit_register) (p, t2_4, &desc);
Chris@10: }
Chris@10: #else				/* HAVE_FMA */
Chris@10: 
Chris@10: /* Generated by: ../../../genfft/gen_twiddle.native -compact -variables 4 -pipeline-latency 4 -twiddle-log3 -precompute-twiddles -n 4 -name t2_4 -include t.h */
Chris@10: 
Chris@10: /*
Chris@10:  * This function contains 24 FP additions, 16 FP multiplications,
Chris@10:  * (or, 16 additions, 8 multiplications, 8 fused multiply/add),
Chris@10:  * 21 stack variables, 0 constants, and 16 memory accesses
Chris@10:  */
Chris@10: #include "t.h"
Chris@10: 
Chris@10: static void t2_4(R *ri, R *ii, const R *W, stride rs, INT mb, INT me, INT ms)
Chris@10: {
Chris@10:      {
Chris@10: 	  INT m;
Chris@10: 	  for (m = mb, W = W + (mb * 4); m < me; m = m + 1, ri = ri + ms, ii = ii + ms, W = W + 4, MAKE_VOLATILE_STRIDE(8, rs)) {
Chris@10: 	       E T2, T4, T3, T5, T6, T8;
Chris@10: 	       T2 = W[0];
Chris@10: 	       T4 = W[1];
Chris@10: 	       T3 = W[2];
Chris@10: 	       T5 = W[3];
Chris@10: 	       T6 = FMA(T2, T3, T4 * T5);
Chris@10: 	       T8 = FNMS(T4, T3, T2 * T5);
Chris@10: 	       {
Chris@10: 		    E T1, Tp, Ta, To, Te, Tk, Th, Tl, T7, T9;
Chris@10: 		    T1 = ri[0];
Chris@10: 		    Tp = ii[0];
Chris@10: 		    T7 = ri[WS(rs, 2)];
Chris@10: 		    T9 = ii[WS(rs, 2)];
Chris@10: 		    Ta = FMA(T6, T7, T8 * T9);
Chris@10: 		    To = FNMS(T8, T7, T6 * T9);
Chris@10: 		    {
Chris@10: 			 E Tc, Td, Tf, Tg;
Chris@10: 			 Tc = ri[WS(rs, 1)];
Chris@10: 			 Td = ii[WS(rs, 1)];
Chris@10: 			 Te = FMA(T2, Tc, T4 * Td);
Chris@10: 			 Tk = FNMS(T4, Tc, T2 * Td);
Chris@10: 			 Tf = ri[WS(rs, 3)];
Chris@10: 			 Tg = ii[WS(rs, 3)];
Chris@10: 			 Th = FMA(T3, Tf, T5 * Tg);
Chris@10: 			 Tl = FNMS(T5, Tf, T3 * Tg);
Chris@10: 		    }
Chris@10: 		    {
Chris@10: 			 E Tb, Ti, Tn, Tq;
Chris@10: 			 Tb = T1 + Ta;
Chris@10: 			 Ti = Te + Th;
Chris@10: 			 ri[WS(rs, 2)] = Tb - Ti;
Chris@10: 			 ri[0] = Tb + Ti;
Chris@10: 			 Tn = Tk + Tl;
Chris@10: 			 Tq = To + Tp;
Chris@10: 			 ii[0] = Tn + Tq;
Chris@10: 			 ii[WS(rs, 2)] = Tq - Tn;
Chris@10: 		    }
Chris@10: 		    {
Chris@10: 			 E Tj, Tm, Tr, Ts;
Chris@10: 			 Tj = T1 - Ta;
Chris@10: 			 Tm = Tk - Tl;
Chris@10: 			 ri[WS(rs, 3)] = Tj - Tm;
Chris@10: 			 ri[WS(rs, 1)] = Tj + Tm;
Chris@10: 			 Tr = Tp - To;
Chris@10: 			 Ts = Te - Th;
Chris@10: 			 ii[WS(rs, 1)] = Tr - Ts;
Chris@10: 			 ii[WS(rs, 3)] = Ts + Tr;
Chris@10: 		    }
Chris@10: 	       }
Chris@10: 	  }
Chris@10:      }
Chris@10: }
Chris@10: 
Chris@10: static const tw_instr twinstr[] = {
Chris@10:      {TW_CEXP, 0, 1},
Chris@10:      {TW_CEXP, 0, 3},
Chris@10:      {TW_NEXT, 1, 0}
Chris@10: };
Chris@10: 
Chris@10: static const ct_desc desc = { 4, "t2_4", twinstr, &GENUS, {16, 8, 8, 0}, 0, 0, 0 };
Chris@10: 
Chris@10: void X(codelet_t2_4) (planner *p) {
Chris@10:      X(kdft_dit_register) (p, t2_4, &desc);
Chris@10: }
Chris@10: #endif				/* HAVE_FMA */