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 07:40:15 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_r2cf.native -fma -reorder-insns -schedule-for-pipeline -compact -variables 4 -pipeline-latency 4 -n 12 -name r2cfII_12 -dft-II -include r2cfII.h */
Chris@10: 
Chris@10: /*
Chris@10:  * This function contains 45 FP additions, 24 FP multiplications,
Chris@10:  * (or, 21 additions, 0 multiplications, 24 fused multiply/add),
Chris@10:  * 37 stack variables, 3 constants, and 24 memory accesses
Chris@10:  */
Chris@10: #include "r2cfII.h"
Chris@10: 
Chris@10: static void r2cfII_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(KP866025403, +0.866025403784438646763723170752936183471402627);
Chris@10:      DK(KP500000000, +0.500000000000000000000000000000000000000000000);
Chris@10:      {
Chris@10: 	  INT i;
Chris@10: 	  for (i = v; i > 0; i = i - 1, R0 = R0 + ivs, R1 = R1 + ivs, Cr = Cr + ovs, Ci = Ci + ovs, MAKE_VOLATILE_STRIDE(48, rs), MAKE_VOLATILE_STRIDE(48, csr), MAKE_VOLATILE_STRIDE(48, csi)) {
Chris@10: 	       E TD, TB, Tp, T9, Tq, Tr, TE, To, Ts, TC;
Chris@10: 	       {
Chris@10: 		    E T8, T1, Tv, Tm, TF, Tz, Tl, Ta, Tb, Tt, TA, T4, Tc;
Chris@10: 		    {
Chris@10: 			 E Tx, Th, Ti, Tj, Ty, T6, T7, T2, T3, Tk;
Chris@10: 			 Tx = R0[WS(rs, 3)];
Chris@10: 			 T6 = R0[WS(rs, 5)];
Chris@10: 			 T7 = R0[WS(rs, 1)];
Chris@10: 			 Th = R1[WS(rs, 4)];
Chris@10: 			 Ti = R1[WS(rs, 2)];
Chris@10: 			 Tj = R1[0];
Chris@10: 			 Ty = T6 + T7;
Chris@10: 			 T8 = T6 - T7;
Chris@10: 			 T1 = R0[0];
Chris@10: 			 Tv = Ti - Tj - Th;
Chris@10: 			 Tk = Ti - Tj;
Chris@10: 			 Tm = Ti + Tj;
Chris@10: 			 TF = Tx - Ty;
Chris@10: 			 Tz = FMA(KP500000000, Ty, Tx);
Chris@10: 			 T2 = R0[WS(rs, 2)];
Chris@10: 			 T3 = R0[WS(rs, 4)];
Chris@10: 			 Tl = FMA(KP500000000, Tk, Th);
Chris@10: 			 Ta = R1[WS(rs, 1)];
Chris@10: 			 Tb = R1[WS(rs, 3)];
Chris@10: 			 Tt = T1 + T3 - T2;
Chris@10: 			 TA = T3 + T2;
Chris@10: 			 T4 = T2 - T3;
Chris@10: 			 Tc = R1[WS(rs, 5)];
Chris@10: 		    }
Chris@10: 		    {
Chris@10: 			 E Tn, Tg, T5, Tu;
Chris@10: 			 TD = FNMS(KP866025403, TA, Tz);
Chris@10: 			 TB = FMA(KP866025403, TA, Tz);
Chris@10: 			 T5 = FMA(KP500000000, T4, T1);
Chris@10: 			 Tu = Ta + Tc - Tb;
Chris@10: 			 {
Chris@10: 			      E Td, Tf, TG, Tw, Te;
Chris@10: 			      Td = Tb - Tc;
Chris@10: 			      Tf = Tc + Tb;
Chris@10: 			      Tp = FMA(KP866025403, T8, T5);
Chris@10: 			      T9 = FNMS(KP866025403, T8, T5);
Chris@10: 			      TG = Tv - Tu;
Chris@10: 			      Tw = Tu + Tv;
Chris@10: 			      Te = FMA(KP500000000, Td, Ta);
Chris@10: 			      Tq = FMA(KP866025403, Tm, Tl);
Chris@10: 			      Tn = FNMS(KP866025403, Tm, Tl);
Chris@10: 			      Ci[WS(csi, 1)] = FMA(KP707106781, TG, TF);
Chris@10: 			      Ci[WS(csi, 4)] = FMS(KP707106781, TG, TF);
Chris@10: 			      Cr[WS(csr, 4)] = FMA(KP707106781, Tw, Tt);
Chris@10: 			      Cr[WS(csr, 1)] = FNMS(KP707106781, Tw, Tt);
Chris@10: 			      Tg = FNMS(KP866025403, Tf, Te);
Chris@10: 			      Tr = FMA(KP866025403, Tf, Te);
Chris@10: 			 }
Chris@10: 			 TE = Tg + Tn;
Chris@10: 			 To = Tg - Tn;
Chris@10: 		    }
Chris@10: 	       }
Chris@10: 	       Ci[WS(csi, 2)] = FMS(KP707106781, TE, TD);
Chris@10: 	       Ci[WS(csi, 3)] = FMA(KP707106781, TE, TD);
Chris@10: 	       Cr[0] = FMA(KP707106781, To, T9);
Chris@10: 	       Cr[WS(csr, 5)] = FNMS(KP707106781, To, T9);
Chris@10: 	       Ts = Tq - Tr;
Chris@10: 	       TC = Tr + Tq;
Chris@10: 	       Ci[0] = -(FMA(KP707106781, TC, TB));
Chris@10: 	       Ci[WS(csi, 5)] = FNMS(KP707106781, TC, TB);
Chris@10: 	       Cr[WS(csr, 2)] = FMA(KP707106781, Ts, Tp);
Chris@10: 	       Cr[WS(csr, 3)] = FNMS(KP707106781, Ts, Tp);
Chris@10: 	  }
Chris@10:      }
Chris@10: }
Chris@10: 
Chris@10: static const kr2c_desc desc = { 12, "r2cfII_12", {21, 0, 24, 0}, &GENUS };
Chris@10: 
Chris@10: void X(codelet_r2cfII_12) (planner *p) {
Chris@10:      X(kr2c_register) (p, r2cfII_12, &desc);
Chris@10: }
Chris@10: 
Chris@10: #else				/* HAVE_FMA */
Chris@10: 
Chris@10: /* Generated by: ../../../genfft/gen_r2cf.native -compact -variables 4 -pipeline-latency 4 -n 12 -name r2cfII_12 -dft-II -include r2cfII.h */
Chris@10: 
Chris@10: /*
Chris@10:  * This function contains 43 FP additions, 12 FP multiplications,
Chris@10:  * (or, 39 additions, 8 multiplications, 4 fused multiply/add),
Chris@10:  * 28 stack variables, 5 constants, and 24 memory accesses
Chris@10:  */
Chris@10: #include "r2cfII.h"
Chris@10: 
Chris@10: static void r2cfII_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(KP353553390, +0.353553390593273762200422181052424519642417969);
Chris@10:      DK(KP707106781, +0.707106781186547524400844362104849039284835938);
Chris@10:      DK(KP612372435, +0.612372435695794524549321018676472847991486870);
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 + ivs, R1 = R1 + ivs, Cr = Cr + ovs, Ci = Ci + ovs, MAKE_VOLATILE_STRIDE(48, rs), MAKE_VOLATILE_STRIDE(48, csr), MAKE_VOLATILE_STRIDE(48, csi)) {
Chris@10: 	       E Tx, Tg, T4, Tz, Ty, Tj, TA, T9, Tm, Tl, Te, Tp, To, Tf, TE;
Chris@10: 	       E TF;
Chris@10: 	       {
Chris@10: 		    E T1, T3, T2, Th, Ti;
Chris@10: 		    T1 = R0[0];
Chris@10: 		    T3 = R0[WS(rs, 2)];
Chris@10: 		    T2 = R0[WS(rs, 4)];
Chris@10: 		    Tx = KP866025403 * (T2 + T3);
Chris@10: 		    Tg = FMA(KP500000000, T3 - T2, T1);
Chris@10: 		    T4 = T1 + T2 - T3;
Chris@10: 		    Tz = R0[WS(rs, 3)];
Chris@10: 		    Th = R0[WS(rs, 5)];
Chris@10: 		    Ti = R0[WS(rs, 1)];
Chris@10: 		    Ty = Th + Ti;
Chris@10: 		    Tj = KP866025403 * (Th - Ti);
Chris@10: 		    TA = FMA(KP500000000, Ty, Tz);
Chris@10: 	       }
Chris@10: 	       {
Chris@10: 		    E T5, T6, T7, T8;
Chris@10: 		    T5 = R1[WS(rs, 1)];
Chris@10: 		    T6 = R1[WS(rs, 5)];
Chris@10: 		    T7 = R1[WS(rs, 3)];
Chris@10: 		    T8 = T6 - T7;
Chris@10: 		    T9 = T5 + T8;
Chris@10: 		    Tm = KP612372435 * (T6 + T7);
Chris@10: 		    Tl = FNMS(KP353553390, T8, KP707106781 * T5);
Chris@10: 	       }
Chris@10: 	       {
Chris@10: 		    E Td, Ta, Tb, Tc;
Chris@10: 		    Td = R1[WS(rs, 4)];
Chris@10: 		    Ta = R1[WS(rs, 2)];
Chris@10: 		    Tb = R1[0];
Chris@10: 		    Tc = Ta - Tb;
Chris@10: 		    Te = Tc - Td;
Chris@10: 		    Tp = FMA(KP353553390, Tc, KP707106781 * Td);
Chris@10: 		    To = KP612372435 * (Ta + Tb);
Chris@10: 	       }
Chris@10: 	       Tf = KP707106781 * (T9 + Te);
Chris@10: 	       Cr[WS(csr, 1)] = T4 - Tf;
Chris@10: 	       Cr[WS(csr, 4)] = T4 + Tf;
Chris@10: 	       TE = KP707106781 * (Te - T9);
Chris@10: 	       TF = Tz - Ty;
Chris@10: 	       Ci[WS(csi, 4)] = TE - TF;
Chris@10: 	       Ci[WS(csi, 1)] = TE + TF;
Chris@10: 	       {
Chris@10: 		    E Tk, TB, Tr, Tw, Tn, Tq;
Chris@10: 		    Tk = Tg - Tj;
Chris@10: 		    TB = Tx - TA;
Chris@10: 		    Tn = Tl - Tm;
Chris@10: 		    Tq = To - Tp;
Chris@10: 		    Tr = Tn + Tq;
Chris@10: 		    Tw = Tn - Tq;
Chris@10: 		    Cr[WS(csr, 5)] = Tk - Tr;
Chris@10: 		    Ci[WS(csi, 2)] = Tw + TB;
Chris@10: 		    Cr[0] = Tk + Tr;
Chris@10: 		    Ci[WS(csi, 3)] = Tw - TB;
Chris@10: 	       }
Chris@10: 	       {
Chris@10: 		    E Ts, TD, Tv, TC, Tt, Tu;
Chris@10: 		    Ts = Tg + Tj;
Chris@10: 		    TD = Tx + TA;
Chris@10: 		    Tt = To + Tp;
Chris@10: 		    Tu = Tm + Tl;
Chris@10: 		    Tv = Tt - Tu;
Chris@10: 		    TC = Tu + Tt;
Chris@10: 		    Cr[WS(csr, 3)] = Ts - Tv;
Chris@10: 		    Ci[WS(csi, 5)] = TD - TC;
Chris@10: 		    Cr[WS(csr, 2)] = Ts + Tv;
Chris@10: 		    Ci[0] = -(TC + TD);
Chris@10: 	       }
Chris@10: 	  }
Chris@10:      }
Chris@10: }
Chris@10: 
Chris@10: static const kr2c_desc desc = { 12, "r2cfII_12", {39, 8, 4, 0}, &GENUS };
Chris@10: 
Chris@10: void X(codelet_r2cfII_12) (planner *p) {
Chris@10:      X(kr2c_register) (p, r2cfII_12, &desc);
Chris@10: }
Chris@10: 
Chris@10: #endif				/* HAVE_FMA */