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 07:39:47 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 16 -name r2cf_16 -include r2cf.h */
Chris@10: 
Chris@10: /*
Chris@10:  * This function contains 58 FP additions, 20 FP multiplications,
Chris@10:  * (or, 38 additions, 0 multiplications, 20 fused multiply/add),
Chris@10:  * 38 stack variables, 3 constants, and 32 memory accesses
Chris@10:  */
Chris@10: #include "r2cf.h"
Chris@10: 
Chris@10: static void r2cf_16(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(KP923879532, +0.923879532511286756128183189396788286822416626);
Chris@10:      DK(KP707106781, +0.707106781186547524400844362104849039284835938);
Chris@10:      DK(KP414213562, +0.414213562373095048801688724209698078569671875);
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(64, rs), MAKE_VOLATILE_STRIDE(64, csr), MAKE_VOLATILE_STRIDE(64, csi)) {
Chris@10: 	       E TQ, TP;
Chris@10: 	       {
Chris@10: 		    E TB, TN, Tf, T7, Te, Tv, TO, TE, Tq, TJ, Tp, TI, TT, Ty, Tm;
Chris@10: 		    E Tr, TK, Ts;
Chris@10: 		    {
Chris@10: 			 E TC, Ta, Td, TD;
Chris@10: 			 {
Chris@10: 			      E T1, T2, T4, T5;
Chris@10: 			      T1 = R0[0];
Chris@10: 			      T2 = R0[WS(rs, 4)];
Chris@10: 			      T4 = R0[WS(rs, 2)];
Chris@10: 			      T5 = R0[WS(rs, 6)];
Chris@10: 			      {
Chris@10: 				   E T8, T3, T6, T9, Tb, Tc;
Chris@10: 				   T8 = R0[WS(rs, 1)];
Chris@10: 				   TB = T1 - T2;
Chris@10: 				   T3 = T1 + T2;
Chris@10: 				   TN = T4 - T5;
Chris@10: 				   T6 = T4 + T5;
Chris@10: 				   T9 = R0[WS(rs, 5)];
Chris@10: 				   Tb = R0[WS(rs, 7)];
Chris@10: 				   Tc = R0[WS(rs, 3)];
Chris@10: 				   Tf = T3 - T6;
Chris@10: 				   T7 = T3 + T6;
Chris@10: 				   TC = T8 - T9;
Chris@10: 				   Ta = T8 + T9;
Chris@10: 				   Td = Tb + Tc;
Chris@10: 				   TD = Tb - Tc;
Chris@10: 			      }
Chris@10: 			 }
Chris@10: 			 {
Chris@10: 			      E TG, Ti, Tj, Tk, Tg, Th;
Chris@10: 			      Tg = R1[0];
Chris@10: 			      Th = R1[WS(rs, 4)];
Chris@10: 			      Te = Ta + Td;
Chris@10: 			      Tv = Td - Ta;
Chris@10: 			      TO = TD - TC;
Chris@10: 			      TE = TC + TD;
Chris@10: 			      TG = Tg - Th;
Chris@10: 			      Ti = Tg + Th;
Chris@10: 			      Tj = R1[WS(rs, 2)];
Chris@10: 			      Tk = R1[WS(rs, 6)];
Chris@10: 			      {
Chris@10: 				   E Tn, To, TH, Tl;
Chris@10: 				   Tn = R1[WS(rs, 7)];
Chris@10: 				   To = R1[WS(rs, 3)];
Chris@10: 				   Tq = R1[WS(rs, 1)];
Chris@10: 				   TH = Tj - Tk;
Chris@10: 				   Tl = Tj + Tk;
Chris@10: 				   TJ = Tn - To;
Chris@10: 				   Tp = Tn + To;
Chris@10: 				   TI = FNMS(KP414213562, TH, TG);
Chris@10: 				   TT = FMA(KP414213562, TG, TH);
Chris@10: 				   Ty = Ti + Tl;
Chris@10: 				   Tm = Ti - Tl;
Chris@10: 				   Tr = R1[WS(rs, 5)];
Chris@10: 			      }
Chris@10: 			 }
Chris@10: 		    }
Chris@10: 		    Cr[WS(csr, 4)] = T7 - Te;
Chris@10: 		    TK = Tr - Tq;
Chris@10: 		    Ts = Tq + Tr;
Chris@10: 		    {
Chris@10: 			 E Tx, TV, TF, TS, Tz, Tt, TM, TL;
Chris@10: 			 Tx = T7 + Te;
Chris@10: 			 TV = FNMS(KP707106781, TE, TB);
Chris@10: 			 TF = FMA(KP707106781, TE, TB);
Chris@10: 			 TL = FNMS(KP414213562, TK, TJ);
Chris@10: 			 TS = FMA(KP414213562, TJ, TK);
Chris@10: 			 Tz = Tp + Ts;
Chris@10: 			 Tt = Tp - Ts;
Chris@10: 			 TM = TI + TL;
Chris@10: 			 TQ = TL - TI;
Chris@10: 			 {
Chris@10: 			      E TR, TU, TW, TA, Tw, Tu;
Chris@10: 			      TP = FMA(KP707106781, TO, TN);
Chris@10: 			      TR = FNMS(KP707106781, TO, TN);
Chris@10: 			      TA = Ty + Tz;
Chris@10: 			      Ci[WS(csi, 4)] = Tz - Ty;
Chris@10: 			      Tw = Tt - Tm;
Chris@10: 			      Tu = Tm + Tt;
Chris@10: 			      Cr[WS(csr, 1)] = FMA(KP923879532, TM, TF);
Chris@10: 			      Cr[WS(csr, 7)] = FNMS(KP923879532, TM, TF);
Chris@10: 			      Cr[0] = Tx + TA;
Chris@10: 			      Cr[WS(csr, 8)] = Tx - TA;
Chris@10: 			      Ci[WS(csi, 6)] = FMS(KP707106781, Tw, Tv);
Chris@10: 			      Ci[WS(csi, 2)] = FMA(KP707106781, Tw, Tv);
Chris@10: 			      Cr[WS(csr, 2)] = FMA(KP707106781, Tu, Tf);
Chris@10: 			      Cr[WS(csr, 6)] = FNMS(KP707106781, Tu, Tf);
Chris@10: 			      TU = TS - TT;
Chris@10: 			      TW = TT + TS;
Chris@10: 			      Ci[WS(csi, 7)] = FMA(KP923879532, TU, TR);
Chris@10: 			      Ci[WS(csi, 1)] = FMS(KP923879532, TU, TR);
Chris@10: 			      Cr[WS(csr, 3)] = FMA(KP923879532, TW, TV);
Chris@10: 			      Cr[WS(csr, 5)] = FNMS(KP923879532, TW, TV);
Chris@10: 			 }
Chris@10: 		    }
Chris@10: 	       }
Chris@10: 	       Ci[WS(csi, 5)] = FMS(KP923879532, TQ, TP);
Chris@10: 	       Ci[WS(csi, 3)] = FMA(KP923879532, TQ, TP);
Chris@10: 	  }
Chris@10:      }
Chris@10: }
Chris@10: 
Chris@10: static const kr2c_desc desc = { 16, "r2cf_16", {38, 0, 20, 0}, &GENUS };
Chris@10: 
Chris@10: void X(codelet_r2cf_16) (planner *p) {
Chris@10:      X(kr2c_register) (p, r2cf_16, &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 16 -name r2cf_16 -include r2cf.h */
Chris@10: 
Chris@10: /*
Chris@10:  * This function contains 58 FP additions, 12 FP multiplications,
Chris@10:  * (or, 54 additions, 8 multiplications, 4 fused multiply/add),
Chris@10:  * 34 stack variables, 3 constants, and 32 memory accesses
Chris@10:  */
Chris@10: #include "r2cf.h"
Chris@10: 
Chris@10: static void r2cf_16(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(KP923879532, +0.923879532511286756128183189396788286822416626);
Chris@10:      DK(KP382683432, +0.382683432365089771728459984030398866761344562);
Chris@10:      DK(KP707106781, +0.707106781186547524400844362104849039284835938);
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(64, rs), MAKE_VOLATILE_STRIDE(64, csr), MAKE_VOLATILE_STRIDE(64, csi)) {
Chris@10: 	       E T3, T6, T7, Tz, Ti, Ta, Td, Te, TA, Th, Tq, TV, TF, TP, Tx;
Chris@10: 	       E TU, TE, TM, Tg, Tf, TJ, TQ;
Chris@10: 	       {
Chris@10: 		    E T1, T2, T4, T5;
Chris@10: 		    T1 = R0[0];
Chris@10: 		    T2 = R0[WS(rs, 4)];
Chris@10: 		    T3 = T1 + T2;
Chris@10: 		    T4 = R0[WS(rs, 2)];
Chris@10: 		    T5 = R0[WS(rs, 6)];
Chris@10: 		    T6 = T4 + T5;
Chris@10: 		    T7 = T3 + T6;
Chris@10: 		    Tz = T1 - T2;
Chris@10: 		    Ti = T4 - T5;
Chris@10: 	       }
Chris@10: 	       {
Chris@10: 		    E T8, T9, Tb, Tc;
Chris@10: 		    T8 = R0[WS(rs, 1)];
Chris@10: 		    T9 = R0[WS(rs, 5)];
Chris@10: 		    Ta = T8 + T9;
Chris@10: 		    Tg = T8 - T9;
Chris@10: 		    Tb = R0[WS(rs, 7)];
Chris@10: 		    Tc = R0[WS(rs, 3)];
Chris@10: 		    Td = Tb + Tc;
Chris@10: 		    Tf = Tb - Tc;
Chris@10: 	       }
Chris@10: 	       Te = Ta + Td;
Chris@10: 	       TA = KP707106781 * (Tg + Tf);
Chris@10: 	       Th = KP707106781 * (Tf - Tg);
Chris@10: 	       {
Chris@10: 		    E Tm, TN, Tp, TO;
Chris@10: 		    {
Chris@10: 			 E Tk, Tl, Tn, To;
Chris@10: 			 Tk = R1[WS(rs, 7)];
Chris@10: 			 Tl = R1[WS(rs, 3)];
Chris@10: 			 Tm = Tk - Tl;
Chris@10: 			 TN = Tk + Tl;
Chris@10: 			 Tn = R1[WS(rs, 1)];
Chris@10: 			 To = R1[WS(rs, 5)];
Chris@10: 			 Tp = Tn - To;
Chris@10: 			 TO = Tn + To;
Chris@10: 		    }
Chris@10: 		    Tq = FNMS(KP923879532, Tp, KP382683432 * Tm);
Chris@10: 		    TV = TN + TO;
Chris@10: 		    TF = FMA(KP923879532, Tm, KP382683432 * Tp);
Chris@10: 		    TP = TN - TO;
Chris@10: 	       }
Chris@10: 	       {
Chris@10: 		    E Tt, TK, Tw, TL;
Chris@10: 		    {
Chris@10: 			 E Tr, Ts, Tu, Tv;
Chris@10: 			 Tr = R1[0];
Chris@10: 			 Ts = R1[WS(rs, 4)];
Chris@10: 			 Tt = Tr - Ts;
Chris@10: 			 TK = Tr + Ts;
Chris@10: 			 Tu = R1[WS(rs, 2)];
Chris@10: 			 Tv = R1[WS(rs, 6)];
Chris@10: 			 Tw = Tu - Tv;
Chris@10: 			 TL = Tu + Tv;
Chris@10: 		    }
Chris@10: 		    Tx = FMA(KP382683432, Tt, KP923879532 * Tw);
Chris@10: 		    TU = TK + TL;
Chris@10: 		    TE = FNMS(KP382683432, Tw, KP923879532 * Tt);
Chris@10: 		    TM = TK - TL;
Chris@10: 	       }
Chris@10: 	       Cr[WS(csr, 4)] = T7 - Te;
Chris@10: 	       Ci[WS(csi, 4)] = TV - TU;
Chris@10: 	       {
Chris@10: 		    E Tj, Ty, TD, TG;
Chris@10: 		    Tj = Th - Ti;
Chris@10: 		    Ty = Tq - Tx;
Chris@10: 		    Ci[WS(csi, 1)] = Tj + Ty;
Chris@10: 		    Ci[WS(csi, 7)] = Ty - Tj;
Chris@10: 		    TD = Tz + TA;
Chris@10: 		    TG = TE + TF;
Chris@10: 		    Cr[WS(csr, 7)] = TD - TG;
Chris@10: 		    Cr[WS(csr, 1)] = TD + TG;
Chris@10: 	       }
Chris@10: 	       {
Chris@10: 		    E TB, TC, TH, TI;
Chris@10: 		    TB = Tz - TA;
Chris@10: 		    TC = Tx + Tq;
Chris@10: 		    Cr[WS(csr, 5)] = TB - TC;
Chris@10: 		    Cr[WS(csr, 3)] = TB + TC;
Chris@10: 		    TH = Ti + Th;
Chris@10: 		    TI = TF - TE;
Chris@10: 		    Ci[WS(csi, 3)] = TH + TI;
Chris@10: 		    Ci[WS(csi, 5)] = TI - TH;
Chris@10: 	       }
Chris@10: 	       TJ = T3 - T6;
Chris@10: 	       TQ = KP707106781 * (TM + TP);
Chris@10: 	       Cr[WS(csr, 6)] = TJ - TQ;
Chris@10: 	       Cr[WS(csr, 2)] = TJ + TQ;
Chris@10: 	       {
Chris@10: 		    E TR, TS, TT, TW;
Chris@10: 		    TR = Td - Ta;
Chris@10: 		    TS = KP707106781 * (TP - TM);
Chris@10: 		    Ci[WS(csi, 2)] = TR + TS;
Chris@10: 		    Ci[WS(csi, 6)] = TS - TR;
Chris@10: 		    TT = T7 + Te;
Chris@10: 		    TW = TU + TV;
Chris@10: 		    Cr[WS(csr, 8)] = TT - TW;
Chris@10: 		    Cr[0] = TT + TW;
Chris@10: 	       }
Chris@10: 	  }
Chris@10:      }
Chris@10: }
Chris@10: 
Chris@10: static const kr2c_desc desc = { 16, "r2cf_16", {54, 8, 4, 0}, &GENUS };
Chris@10: 
Chris@10: void X(codelet_r2cf_16) (planner *p) {
Chris@10:      X(kr2c_register) (p, r2cf_16, &desc);
Chris@10: }
Chris@10: 
Chris@10: #endif				/* HAVE_FMA */