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 07:35:42 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_notw.native -fma -reorder-insns -schedule-for-pipeline -compact -variables 4 -pipeline-latency 4 -n 8 -name n1_8 -include n.h */
Chris@10: 
Chris@10: /*
Chris@10:  * This function contains 52 FP additions, 8 FP multiplications,
Chris@10:  * (or, 44 additions, 0 multiplications, 8 fused multiply/add),
Chris@10:  * 36 stack variables, 1 constants, and 32 memory accesses
Chris@10:  */
Chris@10: #include "n.h"
Chris@10: 
Chris@10: static void n1_8(const R *ri, const R *ii, R *ro, R *io, stride is, stride os, INT v, INT ivs, INT ovs)
Chris@10: {
Chris@10:      DK(KP707106781, +0.707106781186547524400844362104849039284835938);
Chris@10:      {
Chris@10: 	  INT i;
Chris@10: 	  for (i = v; i > 0; i = i - 1, ri = ri + ivs, ii = ii + ivs, ro = ro + ovs, io = io + ovs, MAKE_VOLATILE_STRIDE(32, is), MAKE_VOLATILE_STRIDE(32, os)) {
Chris@10: 	       E TF, TE, TD, TI;
Chris@10: 	       {
Chris@10: 		    E Tn, T3, TC, Ti, TB, T6, To, Tl, Td, TN, Tz, TH, Ta, Tq, Tt;
Chris@10: 		    E TM;
Chris@10: 		    {
Chris@10: 			 E T4, T5, Tj, Tk;
Chris@10: 			 {
Chris@10: 			      E T1, T2, Tg, Th;
Chris@10: 			      T1 = ri[0];
Chris@10: 			      T2 = ri[WS(is, 4)];
Chris@10: 			      Tg = ii[0];
Chris@10: 			      Th = ii[WS(is, 4)];
Chris@10: 			      T4 = ri[WS(is, 2)];
Chris@10: 			      Tn = T1 - T2;
Chris@10: 			      T3 = T1 + T2;
Chris@10: 			      TC = Tg - Th;
Chris@10: 			      Ti = Tg + Th;
Chris@10: 			      T5 = ri[WS(is, 6)];
Chris@10: 			 }
Chris@10: 			 Tj = ii[WS(is, 2)];
Chris@10: 			 Tk = ii[WS(is, 6)];
Chris@10: 			 {
Chris@10: 			      E Tb, Tc, Tw, Tx;
Chris@10: 			      Tb = ri[WS(is, 7)];
Chris@10: 			      TB = T4 - T5;
Chris@10: 			      T6 = T4 + T5;
Chris@10: 			      To = Tj - Tk;
Chris@10: 			      Tl = Tj + Tk;
Chris@10: 			      Tc = ri[WS(is, 3)];
Chris@10: 			      Tw = ii[WS(is, 7)];
Chris@10: 			      Tx = ii[WS(is, 3)];
Chris@10: 			      {
Chris@10: 				   E T8, Tv, Ty, T9, Tr, Ts;
Chris@10: 				   T8 = ri[WS(is, 1)];
Chris@10: 				   Td = Tb + Tc;
Chris@10: 				   Tv = Tb - Tc;
Chris@10: 				   TN = Tw + Tx;
Chris@10: 				   Ty = Tw - Tx;
Chris@10: 				   T9 = ri[WS(is, 5)];
Chris@10: 				   Tr = ii[WS(is, 1)];
Chris@10: 				   Ts = ii[WS(is, 5)];
Chris@10: 				   Tz = Tv - Ty;
Chris@10: 				   TH = Tv + Ty;
Chris@10: 				   Ta = T8 + T9;
Chris@10: 				   Tq = T8 - T9;
Chris@10: 				   Tt = Tr - Ts;
Chris@10: 				   TM = Tr + Ts;
Chris@10: 			      }
Chris@10: 			 }
Chris@10: 		    }
Chris@10: 		    {
Chris@10: 			 E TL, TG, Tu, Tf, Tm, TO;
Chris@10: 			 {
Chris@10: 			      E T7, Te, TP, TQ;
Chris@10: 			      TL = T3 - T6;
Chris@10: 			      T7 = T3 + T6;
Chris@10: 			      TG = Tt - Tq;
Chris@10: 			      Tu = Tq + Tt;
Chris@10: 			      Te = Ta + Td;
Chris@10: 			      Tf = Td - Ta;
Chris@10: 			      Tm = Ti - Tl;
Chris@10: 			      TP = Ti + Tl;
Chris@10: 			      TQ = TM + TN;
Chris@10: 			      TO = TM - TN;
Chris@10: 			      ro[0] = T7 + Te;
Chris@10: 			      ro[WS(os, 4)] = T7 - Te;
Chris@10: 			      io[0] = TP + TQ;
Chris@10: 			      io[WS(os, 4)] = TP - TQ;
Chris@10: 			 }
Chris@10: 			 {
Chris@10: 			      E Tp, TA, TJ, TK;
Chris@10: 			      TF = Tn - To;
Chris@10: 			      Tp = Tn + To;
Chris@10: 			      io[WS(os, 6)] = Tm - Tf;
Chris@10: 			      io[WS(os, 2)] = Tf + Tm;
Chris@10: 			      ro[WS(os, 2)] = TL + TO;
Chris@10: 			      ro[WS(os, 6)] = TL - TO;
Chris@10: 			      TA = Tu + Tz;
Chris@10: 			      TE = Tz - Tu;
Chris@10: 			      TD = TB + TC;
Chris@10: 			      TJ = TC - TB;
Chris@10: 			      TK = TG + TH;
Chris@10: 			      TI = TG - TH;
Chris@10: 			      ro[WS(os, 1)] = FMA(KP707106781, TA, Tp);
Chris@10: 			      ro[WS(os, 5)] = FNMS(KP707106781, TA, Tp);
Chris@10: 			      io[WS(os, 1)] = FMA(KP707106781, TK, TJ);
Chris@10: 			      io[WS(os, 5)] = FNMS(KP707106781, TK, TJ);
Chris@10: 			 }
Chris@10: 		    }
Chris@10: 	       }
Chris@10: 	       io[WS(os, 3)] = FMA(KP707106781, TE, TD);
Chris@10: 	       io[WS(os, 7)] = FNMS(KP707106781, TE, TD);
Chris@10: 	       ro[WS(os, 3)] = FMA(KP707106781, TI, TF);
Chris@10: 	       ro[WS(os, 7)] = FNMS(KP707106781, TI, TF);
Chris@10: 	  }
Chris@10:      }
Chris@10: }
Chris@10: 
Chris@10: static const kdft_desc desc = { 8, "n1_8", {44, 0, 8, 0}, &GENUS, 0, 0, 0, 0 };
Chris@10: 
Chris@10: void X(codelet_n1_8) (planner *p) {
Chris@10:      X(kdft_register) (p, n1_8, &desc);
Chris@10: }
Chris@10: 
Chris@10: #else				/* HAVE_FMA */
Chris@10: 
Chris@10: /* Generated by: ../../../genfft/gen_notw.native -compact -variables 4 -pipeline-latency 4 -n 8 -name n1_8 -include n.h */
Chris@10: 
Chris@10: /*
Chris@10:  * This function contains 52 FP additions, 4 FP multiplications,
Chris@10:  * (or, 52 additions, 4 multiplications, 0 fused multiply/add),
Chris@10:  * 28 stack variables, 1 constants, and 32 memory accesses
Chris@10:  */
Chris@10: #include "n.h"
Chris@10: 
Chris@10: static void n1_8(const R *ri, const R *ii, R *ro, R *io, stride is, stride os, INT v, INT ivs, INT ovs)
Chris@10: {
Chris@10:      DK(KP707106781, +0.707106781186547524400844362104849039284835938);
Chris@10:      {
Chris@10: 	  INT i;
Chris@10: 	  for (i = v; i > 0; i = i - 1, ri = ri + ivs, ii = ii + ivs, ro = ro + ovs, io = io + ovs, MAKE_VOLATILE_STRIDE(32, is), MAKE_VOLATILE_STRIDE(32, os)) {
Chris@10: 	       E T3, Tn, Ti, TC, T6, TB, Tl, To, Td, TN, Tz, TH, Ta, TM, Tu;
Chris@10: 	       E TG;
Chris@10: 	       {
Chris@10: 		    E T1, T2, Tj, Tk;
Chris@10: 		    T1 = ri[0];
Chris@10: 		    T2 = ri[WS(is, 4)];
Chris@10: 		    T3 = T1 + T2;
Chris@10: 		    Tn = T1 - T2;
Chris@10: 		    {
Chris@10: 			 E Tg, Th, T4, T5;
Chris@10: 			 Tg = ii[0];
Chris@10: 			 Th = ii[WS(is, 4)];
Chris@10: 			 Ti = Tg + Th;
Chris@10: 			 TC = Tg - Th;
Chris@10: 			 T4 = ri[WS(is, 2)];
Chris@10: 			 T5 = ri[WS(is, 6)];
Chris@10: 			 T6 = T4 + T5;
Chris@10: 			 TB = T4 - T5;
Chris@10: 		    }
Chris@10: 		    Tj = ii[WS(is, 2)];
Chris@10: 		    Tk = ii[WS(is, 6)];
Chris@10: 		    Tl = Tj + Tk;
Chris@10: 		    To = Tj - Tk;
Chris@10: 		    {
Chris@10: 			 E Tb, Tc, Tv, Tw, Tx, Ty;
Chris@10: 			 Tb = ri[WS(is, 7)];
Chris@10: 			 Tc = ri[WS(is, 3)];
Chris@10: 			 Tv = Tb - Tc;
Chris@10: 			 Tw = ii[WS(is, 7)];
Chris@10: 			 Tx = ii[WS(is, 3)];
Chris@10: 			 Ty = Tw - Tx;
Chris@10: 			 Td = Tb + Tc;
Chris@10: 			 TN = Tw + Tx;
Chris@10: 			 Tz = Tv - Ty;
Chris@10: 			 TH = Tv + Ty;
Chris@10: 		    }
Chris@10: 		    {
Chris@10: 			 E T8, T9, Tq, Tr, Ts, Tt;
Chris@10: 			 T8 = ri[WS(is, 1)];
Chris@10: 			 T9 = ri[WS(is, 5)];
Chris@10: 			 Tq = T8 - T9;
Chris@10: 			 Tr = ii[WS(is, 1)];
Chris@10: 			 Ts = ii[WS(is, 5)];
Chris@10: 			 Tt = Tr - Ts;
Chris@10: 			 Ta = T8 + T9;
Chris@10: 			 TM = Tr + Ts;
Chris@10: 			 Tu = Tq + Tt;
Chris@10: 			 TG = Tt - Tq;
Chris@10: 		    }
Chris@10: 	       }
Chris@10: 	       {
Chris@10: 		    E T7, Te, TP, TQ;
Chris@10: 		    T7 = T3 + T6;
Chris@10: 		    Te = Ta + Td;
Chris@10: 		    ro[WS(os, 4)] = T7 - Te;
Chris@10: 		    ro[0] = T7 + Te;
Chris@10: 		    TP = Ti + Tl;
Chris@10: 		    TQ = TM + TN;
Chris@10: 		    io[WS(os, 4)] = TP - TQ;
Chris@10: 		    io[0] = TP + TQ;
Chris@10: 	       }
Chris@10: 	       {
Chris@10: 		    E Tf, Tm, TL, TO;
Chris@10: 		    Tf = Td - Ta;
Chris@10: 		    Tm = Ti - Tl;
Chris@10: 		    io[WS(os, 2)] = Tf + Tm;
Chris@10: 		    io[WS(os, 6)] = Tm - Tf;
Chris@10: 		    TL = T3 - T6;
Chris@10: 		    TO = TM - TN;
Chris@10: 		    ro[WS(os, 6)] = TL - TO;
Chris@10: 		    ro[WS(os, 2)] = TL + TO;
Chris@10: 	       }
Chris@10: 	       {
Chris@10: 		    E Tp, TA, TJ, TK;
Chris@10: 		    Tp = Tn + To;
Chris@10: 		    TA = KP707106781 * (Tu + Tz);
Chris@10: 		    ro[WS(os, 5)] = Tp - TA;
Chris@10: 		    ro[WS(os, 1)] = Tp + TA;
Chris@10: 		    TJ = TC - TB;
Chris@10: 		    TK = KP707106781 * (TG + TH);
Chris@10: 		    io[WS(os, 5)] = TJ - TK;
Chris@10: 		    io[WS(os, 1)] = TJ + TK;
Chris@10: 	       }
Chris@10: 	       {
Chris@10: 		    E TD, TE, TF, TI;
Chris@10: 		    TD = TB + TC;
Chris@10: 		    TE = KP707106781 * (Tz - Tu);
Chris@10: 		    io[WS(os, 7)] = TD - TE;
Chris@10: 		    io[WS(os, 3)] = TD + TE;
Chris@10: 		    TF = Tn - To;
Chris@10: 		    TI = KP707106781 * (TG - TH);
Chris@10: 		    ro[WS(os, 7)] = TF - TI;
Chris@10: 		    ro[WS(os, 3)] = TF + TI;
Chris@10: 	       }
Chris@10: 	  }
Chris@10:      }
Chris@10: }
Chris@10: 
Chris@10: static const kdft_desc desc = { 8, "n1_8", {52, 4, 0, 0}, &GENUS, 0, 0, 0, 0 };
Chris@10: 
Chris@10: void X(codelet_n1_8) (planner *p) {
Chris@10:      X(kdft_register) (p, n1_8, &desc);
Chris@10: }
Chris@10: 
Chris@10: #endif				/* HAVE_FMA */