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 07:41:36 EST 2012 */
Chris@10: 
Chris@10: #include "codelet-rdft.h"
Chris@10: 
Chris@10: #ifdef HAVE_FMA
Chris@10: 
Chris@10: /* Generated by: ../../../genfft/gen_r2cb.native -fma -reorder-insns -schedule-for-pipeline -compact -variables 4 -pipeline-latency 4 -sign 1 -n 12 -name r2cbIII_12 -dft-III -include r2cbIII.h */
Chris@10: 
Chris@10: /*
Chris@10:  * This function contains 42 FP additions, 20 FP multiplications,
Chris@10:  * (or, 30 additions, 8 multiplications, 12 fused multiply/add),
Chris@10:  * 37 stack variables, 4 constants, and 24 memory accesses
Chris@10:  */
Chris@10: #include "r2cbIII.h"
Chris@10: 
Chris@10: static void r2cbIII_12(R *R0, R *R1, R *Cr, R *Ci, stride rs, stride csr, stride csi, INT v, INT ivs, INT ovs)
Chris@10: {
Chris@10:      DK(KP707106781, +0.707106781186547524400844362104849039284835938);
Chris@10:      DK(KP1_414213562, +1.414213562373095048801688724209698078569671875);
Chris@10:      DK(KP1_732050807, +1.732050807568877293527446341505872366942805254);
Chris@10:      DK(KP2_000000000, +2.000000000000000000000000000000000000000000000);
Chris@10:      {
Chris@10: 	  INT i;
Chris@10: 	  for (i = v; i > 0; i = i - 1, R0 = R0 + ovs, R1 = R1 + ovs, Cr = Cr + ivs, Ci = Ci + ivs, MAKE_VOLATILE_STRIDE(48, rs), MAKE_VOLATILE_STRIDE(48, csr), MAKE_VOLATILE_STRIDE(48, csi)) {
Chris@10: 	       E TE, TD, TF, TG;
Chris@10: 	       {
Chris@10: 		    E Tx, T6, Te, Tb, T5, Tw, Ts, To, Th, Ti, T9, TA;
Chris@10: 		    {
Chris@10: 			 E T1, Tq, Tc, Td, T4, T2, T3, T7, T8, Tr;
Chris@10: 			 T1 = Cr[WS(csr, 1)];
Chris@10: 			 T2 = Cr[WS(csr, 5)];
Chris@10: 			 T3 = Cr[WS(csr, 2)];
Chris@10: 			 Tq = Ci[WS(csi, 1)];
Chris@10: 			 Tc = Ci[WS(csi, 5)];
Chris@10: 			 Td = Ci[WS(csi, 2)];
Chris@10: 			 T4 = T2 + T3;
Chris@10: 			 Tx = T2 - T3;
Chris@10: 			 T6 = Cr[WS(csr, 4)];
Chris@10: 			 Te = Tc + Td;
Chris@10: 			 Tr = Td - Tc;
Chris@10: 			 Tb = FNMS(KP2_000000000, T1, T4);
Chris@10: 			 T5 = T1 + T4;
Chris@10: 			 T7 = Cr[0];
Chris@10: 			 Tw = FMA(KP2_000000000, Tq, Tr);
Chris@10: 			 Ts = Tq - Tr;
Chris@10: 			 T8 = Cr[WS(csr, 3)];
Chris@10: 			 To = Ci[WS(csi, 4)];
Chris@10: 			 Th = Ci[0];
Chris@10: 			 Ti = Ci[WS(csi, 3)];
Chris@10: 			 T9 = T7 + T8;
Chris@10: 			 TA = T7 - T8;
Chris@10: 		    }
Chris@10: 		    {
Chris@10: 			 E Tl, Tm, Tv, TC;
Chris@10: 			 {
Chris@10: 			      E Tf, Ty, Tk, TB;
Chris@10: 			      {
Chris@10: 				   E Tj, Tn, Tg, Ta;
Chris@10: 				   Tl = FNMS(KP1_732050807, Te, Tb);
Chris@10: 				   Tf = FMA(KP1_732050807, Te, Tb);
Chris@10: 				   Tj = Th + Ti;
Chris@10: 				   Tn = Ti - Th;
Chris@10: 				   Tg = FNMS(KP2_000000000, T6, T9);
Chris@10: 				   Ta = T6 + T9;
Chris@10: 				   {
Chris@10: 					E Tu, Tt, Tz, Tp;
Chris@10: 					Ty = FMA(KP1_732050807, Tx, Tw);
Chris@10: 					TE = FNMS(KP1_732050807, Tx, Tw);
Chris@10: 					Tz = FMA(KP2_000000000, To, Tn);
Chris@10: 					Tp = Tn - To;
Chris@10: 					Tm = FMA(KP1_732050807, Tj, Tg);
Chris@10: 					Tk = FNMS(KP1_732050807, Tj, Tg);
Chris@10: 					Tu = T5 - Ta;
Chris@10: 					R0[0] = KP2_000000000 * (T5 + Ta);
Chris@10: 					Tt = Tp - Ts;
Chris@10: 					R0[WS(rs, 3)] = KP2_000000000 * (Ts + Tp);
Chris@10: 					Tv = Tk - Tf;
Chris@10: 					TD = FMA(KP1_732050807, TA, Tz);
Chris@10: 					TB = FNMS(KP1_732050807, TA, Tz);
Chris@10: 					R1[WS(rs, 4)] = KP1_414213562 * (Tu + Tt);
Chris@10: 					R1[WS(rs, 1)] = KP1_414213562 * (Tt - Tu);
Chris@10: 				   }
Chris@10: 			      }
Chris@10: 			      R0[WS(rs, 2)] = Tf + Tk;
Chris@10: 			      TC = Ty + TB;
Chris@10: 			      R0[WS(rs, 5)] = TB - Ty;
Chris@10: 			 }
Chris@10: 			 R1[WS(rs, 3)] = KP707106781 * (Tv + TC);
Chris@10: 			 R1[0] = KP707106781 * (Tv - TC);
Chris@10: 			 TF = Tl - Tm;
Chris@10: 			 R0[WS(rs, 4)] = -(Tl + Tm);
Chris@10: 		    }
Chris@10: 	       }
Chris@10: 	       R0[WS(rs, 1)] = TD - TE;
Chris@10: 	       TG = TE + TD;
Chris@10: 	       R1[WS(rs, 5)] = KP707106781 * (TF - TG);
Chris@10: 	       R1[WS(rs, 2)] = KP707106781 * (TF + TG);
Chris@10: 	  }
Chris@10:      }
Chris@10: }
Chris@10: 
Chris@10: static const kr2c_desc desc = { 12, "r2cbIII_12", {30, 8, 12, 0}, &GENUS };
Chris@10: 
Chris@10: void X(codelet_r2cbIII_12) (planner *p) {
Chris@10:      X(kr2c_register) (p, r2cbIII_12, &desc);
Chris@10: }
Chris@10: 
Chris@10: #else				/* HAVE_FMA */
Chris@10: 
Chris@10: /* Generated by: ../../../genfft/gen_r2cb.native -compact -variables 4 -pipeline-latency 4 -sign 1 -n 12 -name r2cbIII_12 -dft-III -include r2cbIII.h */
Chris@10: 
Chris@10: /*
Chris@10:  * This function contains 42 FP additions, 20 FP multiplications,
Chris@10:  * (or, 38 additions, 16 multiplications, 4 fused multiply/add),
Chris@10:  * 25 stack variables, 4 constants, and 24 memory accesses
Chris@10:  */
Chris@10: #include "r2cbIII.h"
Chris@10: 
Chris@10: static void r2cbIII_12(R *R0, R *R1, R *Cr, R *Ci, stride rs, stride csr, stride csi, INT v, INT ivs, INT ovs)
Chris@10: {
Chris@10:      DK(KP1_414213562, +1.414213562373095048801688724209698078569671875);
Chris@10:      DK(KP2_000000000, +2.000000000000000000000000000000000000000000000);
Chris@10:      DK(KP500000000, +0.500000000000000000000000000000000000000000000);
Chris@10:      DK(KP866025403, +0.866025403784438646763723170752936183471402627);
Chris@10:      {
Chris@10: 	  INT i;
Chris@10: 	  for (i = v; i > 0; i = i - 1, R0 = R0 + ovs, R1 = R1 + ovs, Cr = Cr + ivs, Ci = Ci + ivs, MAKE_VOLATILE_STRIDE(48, rs), MAKE_VOLATILE_STRIDE(48, csr), MAKE_VOLATILE_STRIDE(48, csi)) {
Chris@10: 	       E T5, Tw, Tb, Te, Tx, Ts, Ta, TA, Tg, Tj, Tz, Tp, Tt, Tu;
Chris@10: 	       {
Chris@10: 		    E T1, T2, T3, T4;
Chris@10: 		    T1 = Cr[WS(csr, 1)];
Chris@10: 		    T2 = Cr[WS(csr, 5)];
Chris@10: 		    T3 = Cr[WS(csr, 2)];
Chris@10: 		    T4 = T2 + T3;
Chris@10: 		    T5 = T1 + T4;
Chris@10: 		    Tw = KP866025403 * (T2 - T3);
Chris@10: 		    Tb = FNMS(KP500000000, T4, T1);
Chris@10: 	       }
Chris@10: 	       {
Chris@10: 		    E Tq, Tc, Td, Tr;
Chris@10: 		    Tq = Ci[WS(csi, 1)];
Chris@10: 		    Tc = Ci[WS(csi, 5)];
Chris@10: 		    Td = Ci[WS(csi, 2)];
Chris@10: 		    Tr = Td - Tc;
Chris@10: 		    Te = KP866025403 * (Tc + Td);
Chris@10: 		    Tx = FMA(KP500000000, Tr, Tq);
Chris@10: 		    Ts = Tq - Tr;
Chris@10: 	       }
Chris@10: 	       {
Chris@10: 		    E T6, T7, T8, T9;
Chris@10: 		    T6 = Cr[WS(csr, 4)];
Chris@10: 		    T7 = Cr[0];
Chris@10: 		    T8 = Cr[WS(csr, 3)];
Chris@10: 		    T9 = T7 + T8;
Chris@10: 		    Ta = T6 + T9;
Chris@10: 		    TA = KP866025403 * (T7 - T8);
Chris@10: 		    Tg = FNMS(KP500000000, T9, T6);
Chris@10: 	       }
Chris@10: 	       {
Chris@10: 		    E To, Th, Ti, Tn;
Chris@10: 		    To = Ci[WS(csi, 4)];
Chris@10: 		    Th = Ci[0];
Chris@10: 		    Ti = Ci[WS(csi, 3)];
Chris@10: 		    Tn = Ti - Th;
Chris@10: 		    Tj = KP866025403 * (Th + Ti);
Chris@10: 		    Tz = FMA(KP500000000, Tn, To);
Chris@10: 		    Tp = Tn - To;
Chris@10: 	       }
Chris@10: 	       R0[0] = KP2_000000000 * (T5 + Ta);
Chris@10: 	       R0[WS(rs, 3)] = KP2_000000000 * (Ts + Tp);
Chris@10: 	       Tt = Tp - Ts;
Chris@10: 	       Tu = T5 - Ta;
Chris@10: 	       R1[WS(rs, 1)] = KP1_414213562 * (Tt - Tu);
Chris@10: 	       R1[WS(rs, 4)] = KP1_414213562 * (Tu + Tt);
Chris@10: 	       {
Chris@10: 		    E Tf, Tk, Tv, Ty, TB, TC;
Chris@10: 		    Tf = Tb - Te;
Chris@10: 		    Tk = Tg + Tj;
Chris@10: 		    Tv = Tf - Tk;
Chris@10: 		    Ty = Tw + Tx;
Chris@10: 		    TB = Tz - TA;
Chris@10: 		    TC = Ty + TB;
Chris@10: 		    R0[WS(rs, 2)] = -(KP2_000000000 * (Tf + Tk));
Chris@10: 		    R0[WS(rs, 5)] = KP2_000000000 * (TB - Ty);
Chris@10: 		    R1[0] = KP1_414213562 * (Tv - TC);
Chris@10: 		    R1[WS(rs, 3)] = KP1_414213562 * (Tv + TC);
Chris@10: 	       }
Chris@10: 	       {
Chris@10: 		    E Tl, Tm, TF, TD, TE, TG;
Chris@10: 		    Tl = Tb + Te;
Chris@10: 		    Tm = Tg - Tj;
Chris@10: 		    TF = Tm - Tl;
Chris@10: 		    TD = TA + Tz;
Chris@10: 		    TE = Tx - Tw;
Chris@10: 		    TG = TE + TD;
Chris@10: 		    R0[WS(rs, 4)] = KP2_000000000 * (Tl + Tm);
Chris@10: 		    R1[WS(rs, 2)] = KP1_414213562 * (TF + TG);
Chris@10: 		    R0[WS(rs, 1)] = KP2_000000000 * (TD - TE);
Chris@10: 		    R1[WS(rs, 5)] = KP1_414213562 * (TF - TG);
Chris@10: 	       }
Chris@10: 	  }
Chris@10:      }
Chris@10: }
Chris@10: 
Chris@10: static const kr2c_desc desc = { 12, "r2cbIII_12", {38, 16, 4, 0}, &GENUS };
Chris@10: 
Chris@10: void X(codelet_r2cbIII_12) (planner *p) {
Chris@10:      X(kr2c_register) (p, r2cbIII_12, &desc);
Chris@10: }
Chris@10: 
Chris@10: #endif				/* HAVE_FMA */