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 16 -name r2cbIII_16 -dft-III -include r2cbIII.h */
Chris@10: 
Chris@10: /*
Chris@10:  * This function contains 66 FP additions, 36 FP multiplications,
Chris@10:  * (or, 46 additions, 16 multiplications, 20 fused multiply/add),
Chris@10:  * 55 stack variables, 9 constants, and 32 memory accesses
Chris@10:  */
Chris@10: #include "r2cbIII.h"
Chris@10: 
Chris@10: static void r2cbIII_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(KP668178637, +0.668178637919298919997757686523080761552472251);
Chris@10:      DK(KP1_662939224, +1.662939224605090474157576755235811513477121624);
Chris@10:      DK(KP198912367, +0.198912367379658006911597622644676228597850501);
Chris@10:      DK(KP1_961570560, +1.961570560806460898252364472268478073947867462);
Chris@10:      DK(KP707106781, +0.707106781186547524400844362104849039284835938);
Chris@10:      DK(KP1_414213562, +1.414213562373095048801688724209698078569671875);
Chris@10:      DK(KP414213562, +0.414213562373095048801688724209698078569671875);
Chris@10:      DK(KP1_847759065, +1.847759065022573512256366378793576573644833252);
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(64, rs), MAKE_VOLATILE_STRIDE(64, csr), MAKE_VOLATILE_STRIDE(64, csi)) {
Chris@10: 	       E TA, TD, Tv, TG, TE, TF;
Chris@10: 	       {
Chris@10: 		    E TK, TP, T7, T13, TW, TH, Tj, TC, To, Te, TX, TS, T12, Tt, TB;
Chris@10: 		    {
Chris@10: 			 E T4, Tf, T3, TU, Tz, T5, Tg, Th;
Chris@10: 			 {
Chris@10: 			      E T1, T2, Tx, Ty;
Chris@10: 			      T1 = Cr[0];
Chris@10: 			      T2 = Cr[WS(csr, 7)];
Chris@10: 			      Tx = Ci[0];
Chris@10: 			      Ty = Ci[WS(csi, 7)];
Chris@10: 			      T4 = Cr[WS(csr, 4)];
Chris@10: 			      Tf = T1 - T2;
Chris@10: 			      T3 = T1 + T2;
Chris@10: 			      TU = Ty - Tx;
Chris@10: 			      Tz = Tx + Ty;
Chris@10: 			      T5 = Cr[WS(csr, 3)];
Chris@10: 			      Tg = Ci[WS(csi, 4)];
Chris@10: 			      Th = Ci[WS(csi, 3)];
Chris@10: 			 }
Chris@10: 			 {
Chris@10: 			      E Tb, Tk, Ta, TR, Tn, Tc, Tq, Tr;
Chris@10: 			      {
Chris@10: 				   E T8, T9, Tl, Tm;
Chris@10: 				   T8 = Cr[WS(csr, 2)];
Chris@10: 				   {
Chris@10: 					E Tw, T6, TV, Ti;
Chris@10: 					Tw = T4 - T5;
Chris@10: 					T6 = T4 + T5;
Chris@10: 					TV = Th - Tg;
Chris@10: 					Ti = Tg + Th;
Chris@10: 					TK = Tw - Tz;
Chris@10: 					TA = Tw + Tz;
Chris@10: 					TP = T3 - T6;
Chris@10: 					T7 = T3 + T6;
Chris@10: 					T13 = TV + TU;
Chris@10: 					TW = TU - TV;
Chris@10: 					TH = Tf + Ti;
Chris@10: 					Tj = Tf - Ti;
Chris@10: 					T9 = Cr[WS(csr, 5)];
Chris@10: 				   }
Chris@10: 				   Tl = Ci[WS(csi, 2)];
Chris@10: 				   Tm = Ci[WS(csi, 5)];
Chris@10: 				   Tb = Cr[WS(csr, 1)];
Chris@10: 				   Tk = T8 - T9;
Chris@10: 				   Ta = T8 + T9;
Chris@10: 				   TR = Tl - Tm;
Chris@10: 				   Tn = Tl + Tm;
Chris@10: 				   Tc = Cr[WS(csr, 6)];
Chris@10: 				   Tq = Ci[WS(csi, 1)];
Chris@10: 				   Tr = Ci[WS(csi, 6)];
Chris@10: 			      }
Chris@10: 			      TC = Tk + Tn;
Chris@10: 			      To = Tk - Tn;
Chris@10: 			      {
Chris@10: 				   E Tp, Td, TQ, Ts;
Chris@10: 				   Tp = Tb - Tc;
Chris@10: 				   Td = Tb + Tc;
Chris@10: 				   TQ = Tr - Tq;
Chris@10: 				   Ts = Tq + Tr;
Chris@10: 				   Te = Ta + Td;
Chris@10: 				   TX = Ta - Td;
Chris@10: 				   TS = TQ - TR;
Chris@10: 				   T12 = TR + TQ;
Chris@10: 				   Tt = Tp - Ts;
Chris@10: 				   TB = Tp + Ts;
Chris@10: 			      }
Chris@10: 			 }
Chris@10: 		    }
Chris@10: 		    {
Chris@10: 			 E T10, TT, TY, TZ;
Chris@10: 			 R0[0] = KP2_000000000 * (T7 + Te);
Chris@10: 			 R0[WS(rs, 4)] = KP2_000000000 * (T13 - T12);
Chris@10: 			 T10 = TP - TS;
Chris@10: 			 TT = TP + TS;
Chris@10: 			 TY = TW - TX;
Chris@10: 			 TZ = TX + TW;
Chris@10: 			 {
Chris@10: 			      E T11, T14, TI, TL, Tu;
Chris@10: 			      T11 = T7 - Te;
Chris@10: 			      T14 = T12 + T13;
Chris@10: 			      R0[WS(rs, 5)] = KP1_847759065 * (FNMS(KP414213562, TT, TY));
Chris@10: 			      R0[WS(rs, 1)] = KP1_847759065 * (FMA(KP414213562, TY, TT));
Chris@10: 			      R0[WS(rs, 6)] = KP1_414213562 * (T14 - T11);
Chris@10: 			      R0[WS(rs, 2)] = KP1_414213562 * (T11 + T14);
Chris@10: 			      TD = TB - TC;
Chris@10: 			      TI = TC + TB;
Chris@10: 			      TL = To - Tt;
Chris@10: 			      Tu = To + Tt;
Chris@10: 			      {
Chris@10: 				   E TO, TJ, TN, TM;
Chris@10: 				   R0[WS(rs, 7)] = -(KP1_847759065 * (FNMS(KP414213562, TZ, T10)));
Chris@10: 				   R0[WS(rs, 3)] = KP1_847759065 * (FMA(KP414213562, T10, TZ));
Chris@10: 				   TO = FMA(KP707106781, TI, TH);
Chris@10: 				   TJ = FNMS(KP707106781, TI, TH);
Chris@10: 				   TN = FMA(KP707106781, TL, TK);
Chris@10: 				   TM = FNMS(KP707106781, TL, TK);
Chris@10: 				   Tv = FMA(KP707106781, Tu, Tj);
Chris@10: 				   TG = FNMS(KP707106781, Tu, Tj);
Chris@10: 				   R1[WS(rs, 3)] = KP1_961570560 * (FMA(KP198912367, TO, TN));
Chris@10: 				   R1[WS(rs, 7)] = -(KP1_961570560 * (FNMS(KP198912367, TN, TO)));
Chris@10: 				   R1[WS(rs, 5)] = KP1_662939224 * (FNMS(KP668178637, TJ, TM));
Chris@10: 				   R1[WS(rs, 1)] = KP1_662939224 * (FMA(KP668178637, TM, TJ));
Chris@10: 			      }
Chris@10: 			 }
Chris@10: 		    }
Chris@10: 	       }
Chris@10: 	       TE = FNMS(KP707106781, TD, TA);
Chris@10: 	       TF = FMA(KP707106781, TD, TA);
Chris@10: 	       R1[WS(rs, 2)] = -(KP1_662939224 * (FNMS(KP668178637, TG, TF)));
Chris@10: 	       R1[WS(rs, 6)] = -(KP1_662939224 * (FMA(KP668178637, TF, TG)));
Chris@10: 	       R1[WS(rs, 4)] = -(KP1_961570560 * (FMA(KP198912367, Tv, TE)));
Chris@10: 	       R1[0] = KP1_961570560 * (FNMS(KP198912367, TE, Tv));
Chris@10: 	  }
Chris@10:      }
Chris@10: }
Chris@10: 
Chris@10: static const kr2c_desc desc = { 16, "r2cbIII_16", {46, 16, 20, 0}, &GENUS };
Chris@10: 
Chris@10: void X(codelet_r2cbIII_16) (planner *p) {
Chris@10:      X(kr2c_register) (p, r2cbIII_16, &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 16 -name r2cbIII_16 -dft-III -include r2cbIII.h */
Chris@10: 
Chris@10: /*
Chris@10:  * This function contains 66 FP additions, 32 FP multiplications,
Chris@10:  * (or, 54 additions, 20 multiplications, 12 fused multiply/add),
Chris@10:  * 40 stack variables, 9 constants, and 32 memory accesses
Chris@10:  */
Chris@10: #include "r2cbIII.h"
Chris@10: 
Chris@10: static void r2cbIII_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(KP1_961570560, +1.961570560806460898252364472268478073947867462);
Chris@10:      DK(KP390180644, +0.390180644032256535696569736954044481855383236);
Chris@10:      DK(KP1_111140466, +1.111140466039204449485661627897065748749874382);
Chris@10:      DK(KP1_662939224, +1.662939224605090474157576755235811513477121624);
Chris@10:      DK(KP707106781, +0.707106781186547524400844362104849039284835938);
Chris@10:      DK(KP1_414213562, +1.414213562373095048801688724209698078569671875);
Chris@10:      DK(KP765366864, +0.765366864730179543456919968060797733522689125);
Chris@10:      DK(KP1_847759065, +1.847759065022573512256366378793576573644833252);
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(64, rs), MAKE_VOLATILE_STRIDE(64, csr), MAKE_VOLATILE_STRIDE(64, csi)) {
Chris@10: 	       E T7, TW, T13, Tj, TD, TK, TP, TH, Te, TX, T12, To, Tt, Tx, TS;
Chris@10: 	       E Tw, TT, TY;
Chris@10: 	       {
Chris@10: 		    E T3, Tf, TC, TV, T6, Tz, Ti, TU;
Chris@10: 		    {
Chris@10: 			 E T1, T2, TA, TB;
Chris@10: 			 T1 = Cr[0];
Chris@10: 			 T2 = Cr[WS(csr, 7)];
Chris@10: 			 T3 = T1 + T2;
Chris@10: 			 Tf = T1 - T2;
Chris@10: 			 TA = Ci[0];
Chris@10: 			 TB = Ci[WS(csi, 7)];
Chris@10: 			 TC = TA + TB;
Chris@10: 			 TV = TB - TA;
Chris@10: 		    }
Chris@10: 		    {
Chris@10: 			 E T4, T5, Tg, Th;
Chris@10: 			 T4 = Cr[WS(csr, 4)];
Chris@10: 			 T5 = Cr[WS(csr, 3)];
Chris@10: 			 T6 = T4 + T5;
Chris@10: 			 Tz = T4 - T5;
Chris@10: 			 Tg = Ci[WS(csi, 4)];
Chris@10: 			 Th = Ci[WS(csi, 3)];
Chris@10: 			 Ti = Tg + Th;
Chris@10: 			 TU = Tg - Th;
Chris@10: 		    }
Chris@10: 		    T7 = T3 + T6;
Chris@10: 		    TW = TU + TV;
Chris@10: 		    T13 = TV - TU;
Chris@10: 		    Tj = Tf - Ti;
Chris@10: 		    TD = Tz + TC;
Chris@10: 		    TK = Tz - TC;
Chris@10: 		    TP = T3 - T6;
Chris@10: 		    TH = Tf + Ti;
Chris@10: 	       }
Chris@10: 	       {
Chris@10: 		    E Ta, Tk, Tn, TR, Td, Tp, Ts, TQ;
Chris@10: 		    {
Chris@10: 			 E T8, T9, Tl, Tm;
Chris@10: 			 T8 = Cr[WS(csr, 2)];
Chris@10: 			 T9 = Cr[WS(csr, 5)];
Chris@10: 			 Ta = T8 + T9;
Chris@10: 			 Tk = T8 - T9;
Chris@10: 			 Tl = Ci[WS(csi, 2)];
Chris@10: 			 Tm = Ci[WS(csi, 5)];
Chris@10: 			 Tn = Tl + Tm;
Chris@10: 			 TR = Tl - Tm;
Chris@10: 		    }
Chris@10: 		    {
Chris@10: 			 E Tb, Tc, Tq, Tr;
Chris@10: 			 Tb = Cr[WS(csr, 1)];
Chris@10: 			 Tc = Cr[WS(csr, 6)];
Chris@10: 			 Td = Tb + Tc;
Chris@10: 			 Tp = Tb - Tc;
Chris@10: 			 Tq = Ci[WS(csi, 1)];
Chris@10: 			 Tr = Ci[WS(csi, 6)];
Chris@10: 			 Ts = Tq + Tr;
Chris@10: 			 TQ = Tr - Tq;
Chris@10: 		    }
Chris@10: 		    Te = Ta + Td;
Chris@10: 		    TX = Ta - Td;
Chris@10: 		    T12 = TR + TQ;
Chris@10: 		    To = Tk - Tn;
Chris@10: 		    Tt = Tp - Ts;
Chris@10: 		    Tx = Tp + Ts;
Chris@10: 		    TS = TQ - TR;
Chris@10: 		    Tw = Tk + Tn;
Chris@10: 	       }
Chris@10: 	       R0[0] = KP2_000000000 * (T7 + Te);
Chris@10: 	       R0[WS(rs, 4)] = KP2_000000000 * (T13 - T12);
Chris@10: 	       TT = TP + TS;
Chris@10: 	       TY = TW - TX;
Chris@10: 	       R0[WS(rs, 1)] = FMA(KP1_847759065, TT, KP765366864 * TY);
Chris@10: 	       R0[WS(rs, 5)] = FNMS(KP765366864, TT, KP1_847759065 * TY);
Chris@10: 	       {
Chris@10: 		    E T11, T14, TZ, T10;
Chris@10: 		    T11 = T7 - Te;
Chris@10: 		    T14 = T12 + T13;
Chris@10: 		    R0[WS(rs, 2)] = KP1_414213562 * (T11 + T14);
Chris@10: 		    R0[WS(rs, 6)] = KP1_414213562 * (T14 - T11);
Chris@10: 		    TZ = TP - TS;
Chris@10: 		    T10 = TX + TW;
Chris@10: 		    R0[WS(rs, 3)] = FMA(KP765366864, TZ, KP1_847759065 * T10);
Chris@10: 		    R0[WS(rs, 7)] = FNMS(KP1_847759065, TZ, KP765366864 * T10);
Chris@10: 	       }
Chris@10: 	       {
Chris@10: 		    E TJ, TN, TM, TO, TI, TL;
Chris@10: 		    TI = KP707106781 * (Tw + Tx);
Chris@10: 		    TJ = TH - TI;
Chris@10: 		    TN = TH + TI;
Chris@10: 		    TL = KP707106781 * (To - Tt);
Chris@10: 		    TM = TK - TL;
Chris@10: 		    TO = TL + TK;
Chris@10: 		    R1[WS(rs, 1)] = FMA(KP1_662939224, TJ, KP1_111140466 * TM);
Chris@10: 		    R1[WS(rs, 7)] = FNMS(KP1_961570560, TN, KP390180644 * TO);
Chris@10: 		    R1[WS(rs, 5)] = FNMS(KP1_111140466, TJ, KP1_662939224 * TM);
Chris@10: 		    R1[WS(rs, 3)] = FMA(KP390180644, TN, KP1_961570560 * TO);
Chris@10: 	       }
Chris@10: 	       {
Chris@10: 		    E Tv, TF, TE, TG, Tu, Ty;
Chris@10: 		    Tu = KP707106781 * (To + Tt);
Chris@10: 		    Tv = Tj + Tu;
Chris@10: 		    TF = Tj - Tu;
Chris@10: 		    Ty = KP707106781 * (Tw - Tx);
Chris@10: 		    TE = Ty + TD;
Chris@10: 		    TG = Ty - TD;
Chris@10: 		    R1[0] = FNMS(KP390180644, TE, KP1_961570560 * Tv);
Chris@10: 		    R1[WS(rs, 6)] = FNMS(KP1_662939224, TF, KP1_111140466 * TG);
Chris@10: 		    R1[WS(rs, 4)] = -(FMA(KP390180644, Tv, KP1_961570560 * TE));
Chris@10: 		    R1[WS(rs, 2)] = FMA(KP1_111140466, TF, KP1_662939224 * TG);
Chris@10: 	       }
Chris@10: 	  }
Chris@10:      }
Chris@10: }
Chris@10: 
Chris@10: static const kr2c_desc desc = { 16, "r2cbIII_16", {54, 20, 12, 0}, &GENUS };
Chris@10: 
Chris@10: void X(codelet_r2cbIII_16) (planner *p) {
Chris@10:      X(kr2c_register) (p, r2cbIII_16, &desc);
Chris@10: }
Chris@10: 
Chris@10: #endif				/* HAVE_FMA */