cannam@95: /*
cannam@95:  * Copyright (c) 2003, 2007-11 Matteo Frigo
cannam@95:  * Copyright (c) 2003, 2007-11 Massachusetts Institute of Technology
cannam@95:  *
cannam@95:  * This program is free software; you can redistribute it and/or modify
cannam@95:  * it under the terms of the GNU General Public License as published by
cannam@95:  * the Free Software Foundation; either version 2 of the License, or
cannam@95:  * (at your option) any later version.
cannam@95:  *
cannam@95:  * This program is distributed in the hope that it will be useful,
cannam@95:  * but WITHOUT ANY WARRANTY; without even the implied warranty of
cannam@95:  * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
cannam@95:  * GNU General Public License for more details.
cannam@95:  *
cannam@95:  * You should have received a copy of the GNU General Public License
cannam@95:  * along with this program; if not, write to the Free Software
cannam@95:  * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA  02110-1301  USA
cannam@95:  *
cannam@95:  */
cannam@95: 
cannam@95: /* This file was automatically generated --- DO NOT EDIT */
cannam@95: /* Generated on Sun Nov 25 07:42:04 EST 2012 */
cannam@95: 
cannam@95: #include "codelet-rdft.h"
cannam@95: 
cannam@95: #ifdef HAVE_FMA
cannam@95: 
cannam@95: /* Generated by: ../../../genfft/gen_hc2cdft.native -fma -reorder-insns -schedule-for-pipeline -compact -variables 4 -pipeline-latency 4 -sign 1 -n 6 -dif -name hc2cbdft_6 -include hc2cb.h */
cannam@95: 
cannam@95: /*
cannam@95:  * This function contains 58 FP additions, 32 FP multiplications,
cannam@95:  * (or, 36 additions, 10 multiplications, 22 fused multiply/add),
cannam@95:  * 52 stack variables, 2 constants, and 24 memory accesses
cannam@95:  */
cannam@95: #include "hc2cb.h"
cannam@95: 
cannam@95: static void hc2cbdft_6(R *Rp, R *Ip, R *Rm, R *Im, const R *W, stride rs, INT mb, INT me, INT ms)
cannam@95: {
cannam@95:      DK(KP866025403, +0.866025403784438646763723170752936183471402627);
cannam@95:      DK(KP500000000, +0.500000000000000000000000000000000000000000000);
cannam@95:      {
cannam@95: 	  INT m;
cannam@95: 	  for (m = mb, W = W + ((mb - 1) * 10); m < me; m = m + 1, Rp = Rp + ms, Ip = Ip + ms, Rm = Rm - ms, Im = Im - ms, W = W + 10, MAKE_VOLATILE_STRIDE(24, rs)) {
cannam@95: 	       E T18, T1b, T16, T1e, T1a, T1f, T19, T1g, T1c;
cannam@95: 	       {
cannam@95: 		    E Tw, T4, TV, Tj, TP, TH, Tr, TY, T5, T6, Ta, Ty;
cannam@95: 		    {
cannam@95: 			 E Tg, TF, Tf, TD, Tp, Th;
cannam@95: 			 {
cannam@95: 			      E Td, Te, Tn, To;
cannam@95: 			      Td = Ip[WS(rs, 1)];
cannam@95: 			      Te = Im[WS(rs, 1)];
cannam@95: 			      Tn = Ip[0];
cannam@95: 			      To = Im[WS(rs, 2)];
cannam@95: 			      Tg = Ip[WS(rs, 2)];
cannam@95: 			      TF = Te + Td;
cannam@95: 			      Tf = Td - Te;
cannam@95: 			      TD = Tn + To;
cannam@95: 			      Tp = Tn - To;
cannam@95: 			      Th = Im[0];
cannam@95: 			 }
cannam@95: 			 {
cannam@95: 			      E T2, T3, T8, T9;
cannam@95: 			      T2 = Rp[0];
cannam@95: 			      T3 = Rm[WS(rs, 2)];
cannam@95: 			      {
cannam@95: 				   E Tq, TE, Ti, TG;
cannam@95: 				   T8 = Rm[WS(rs, 1)];
cannam@95: 				   TE = Tg + Th;
cannam@95: 				   Ti = Tg - Th;
cannam@95: 				   Tw = T2 - T3;
cannam@95: 				   T4 = T2 + T3;
cannam@95: 				   TG = TE - TF;
cannam@95: 				   TV = TF + TE;
cannam@95: 				   Tq = Tf + Ti;
cannam@95: 				   Tj = Tf - Ti;
cannam@95: 				   TP = FNMS(KP500000000, TG, TD);
cannam@95: 				   TH = TD + TG;
cannam@95: 				   T9 = Rp[WS(rs, 1)];
cannam@95: 				   Tr = FNMS(KP500000000, Tq, Tp);
cannam@95: 				   TY = Tp + Tq;
cannam@95: 			      }
cannam@95: 			      T5 = Rp[WS(rs, 2)];
cannam@95: 			      T6 = Rm[0];
cannam@95: 			      Ta = T8 + T9;
cannam@95: 			      Ty = T8 - T9;
cannam@95: 			 }
cannam@95: 		    }
cannam@95: 		    {
cannam@95: 			 E TO, TT, Ts, TA, TR, Tc, TN, TW, TS, Tx, T7;
cannam@95: 			 Tx = T5 - T6;
cannam@95: 			 T7 = T5 + T6;
cannam@95: 			 TO = W[0];
cannam@95: 			 TT = W[1];
cannam@95: 			 {
cannam@95: 			      E Tz, TQ, Tb, TU;
cannam@95: 			      Tz = Tx + Ty;
cannam@95: 			      TQ = Tx - Ty;
cannam@95: 			      Tb = T7 + Ta;
cannam@95: 			      Ts = T7 - Ta;
cannam@95: 			      TU = FNMS(KP500000000, Tz, Tw);
cannam@95: 			      TA = Tw + Tz;
cannam@95: 			      TR = FMA(KP866025403, TQ, TP);
cannam@95: 			      T18 = FNMS(KP866025403, TQ, TP);
cannam@95: 			      Tc = FNMS(KP500000000, Tb, T4);
cannam@95: 			      TN = T4 + Tb;
cannam@95: 			      T1b = FMA(KP866025403, TV, TU);
cannam@95: 			      TW = FNMS(KP866025403, TV, TU);
cannam@95: 			      TS = TO * TR;
cannam@95: 			 }
cannam@95: 			 {
cannam@95: 			      E T15, Tt, T12, T1, Tm, TI, TM, Tl, TJ;
cannam@95: 			      {
cannam@95: 				   E Tv, TC, TB, TL, Tk, TZ, TX, T10;
cannam@95: 				   T15 = FMA(KP866025403, Ts, Tr);
cannam@95: 				   Tt = FNMS(KP866025403, Ts, Tr);
cannam@95: 				   TZ = TO * TW;
cannam@95: 				   TX = FMA(TT, TW, TS);
cannam@95: 				   Tv = W[4];
cannam@95: 				   TC = W[5];
cannam@95: 				   T10 = FNMS(TT, TR, TZ);
cannam@95: 				   Rm[0] = TN + TX;
cannam@95: 				   Rp[0] = TN - TX;
cannam@95: 				   TB = Tv * TA;
cannam@95: 				   Im[0] = T10 - TY;
cannam@95: 				   Ip[0] = TY + T10;
cannam@95: 				   TL = TC * TA;
cannam@95: 				   Tk = FNMS(KP866025403, Tj, Tc);
cannam@95: 				   T12 = FMA(KP866025403, Tj, Tc);
cannam@95: 				   T1 = W[3];
cannam@95: 				   Tm = W[2];
cannam@95: 				   TI = FNMS(TC, TH, TB);
cannam@95: 				   TM = FMA(Tv, TH, TL);
cannam@95: 				   Tl = T1 * Tk;
cannam@95: 				   TJ = Tm * Tk;
cannam@95: 			      }
cannam@95: 			      {
cannam@95: 				   E T11, T14, T13, T1d, T17, Tu, TK;
cannam@95: 				   Tu = FMA(Tm, Tt, Tl);
cannam@95: 				   TK = FNMS(T1, Tt, TJ);
cannam@95: 				   T11 = W[6];
cannam@95: 				   T14 = W[7];
cannam@95: 				   Im[WS(rs, 1)] = TI - Tu;
cannam@95: 				   Ip[WS(rs, 1)] = Tu + TI;
cannam@95: 				   Rm[WS(rs, 1)] = TK + TM;
cannam@95: 				   Rp[WS(rs, 1)] = TK - TM;
cannam@95: 				   T13 = T11 * T12;
cannam@95: 				   T1d = T14 * T12;
cannam@95: 				   T17 = W[8];
cannam@95: 				   T16 = FNMS(T14, T15, T13);
cannam@95: 				   T1e = FMA(T11, T15, T1d);
cannam@95: 				   T1a = W[9];
cannam@95: 				   T1f = T17 * T1b;
cannam@95: 				   T19 = T17 * T18;
cannam@95: 			      }
cannam@95: 			 }
cannam@95: 		    }
cannam@95: 	       }
cannam@95: 	       T1g = FNMS(T1a, T18, T1f);
cannam@95: 	       T1c = FMA(T1a, T1b, T19);
cannam@95: 	       Im[WS(rs, 2)] = T1g - T1e;
cannam@95: 	       Ip[WS(rs, 2)] = T1e + T1g;
cannam@95: 	       Rm[WS(rs, 2)] = T16 + T1c;
cannam@95: 	       Rp[WS(rs, 2)] = T16 - T1c;
cannam@95: 	  }
cannam@95:      }
cannam@95: }
cannam@95: 
cannam@95: static const tw_instr twinstr[] = {
cannam@95:      {TW_FULL, 1, 6},
cannam@95:      {TW_NEXT, 1, 0}
cannam@95: };
cannam@95: 
cannam@95: static const hc2c_desc desc = { 6, "hc2cbdft_6", twinstr, &GENUS, {36, 10, 22, 0} };
cannam@95: 
cannam@95: void X(codelet_hc2cbdft_6) (planner *p) {
cannam@95:      X(khc2c_register) (p, hc2cbdft_6, &desc, HC2C_VIA_DFT);
cannam@95: }
cannam@95: #else				/* HAVE_FMA */
cannam@95: 
cannam@95: /* Generated by: ../../../genfft/gen_hc2cdft.native -compact -variables 4 -pipeline-latency 4 -sign 1 -n 6 -dif -name hc2cbdft_6 -include hc2cb.h */
cannam@95: 
cannam@95: /*
cannam@95:  * This function contains 58 FP additions, 28 FP multiplications,
cannam@95:  * (or, 44 additions, 14 multiplications, 14 fused multiply/add),
cannam@95:  * 29 stack variables, 2 constants, and 24 memory accesses
cannam@95:  */
cannam@95: #include "hc2cb.h"
cannam@95: 
cannam@95: static void hc2cbdft_6(R *Rp, R *Ip, R *Rm, R *Im, const R *W, stride rs, INT mb, INT me, INT ms)
cannam@95: {
cannam@95:      DK(KP500000000, +0.500000000000000000000000000000000000000000000);
cannam@95:      DK(KP866025403, +0.866025403784438646763723170752936183471402627);
cannam@95:      {
cannam@95: 	  INT m;
cannam@95: 	  for (m = mb, W = W + ((mb - 1) * 10); m < me; m = m + 1, Rp = Rp + ms, Ip = Ip + ms, Rm = Rm - ms, Im = Im - ms, W = W + 10, MAKE_VOLATILE_STRIDE(24, rs)) {
cannam@95: 	       E T4, Tv, Tr, TL, Tb, Tc, Ty, TP, To, TB, Tj, TQ, Tp, Tq, TE;
cannam@95: 	       E TM;
cannam@95: 	       {
cannam@95: 		    E Ta, Tx, T7, Tw, T2, T3;
cannam@95: 		    T2 = Rp[0];
cannam@95: 		    T3 = Rm[WS(rs, 2)];
cannam@95: 		    T4 = T2 + T3;
cannam@95: 		    Tv = T2 - T3;
cannam@95: 		    {
cannam@95: 			 E T8, T9, T5, T6;
cannam@95: 			 T8 = Rm[WS(rs, 1)];
cannam@95: 			 T9 = Rp[WS(rs, 1)];
cannam@95: 			 Ta = T8 + T9;
cannam@95: 			 Tx = T8 - T9;
cannam@95: 			 T5 = Rp[WS(rs, 2)];
cannam@95: 			 T6 = Rm[0];
cannam@95: 			 T7 = T5 + T6;
cannam@95: 			 Tw = T5 - T6;
cannam@95: 		    }
cannam@95: 		    Tr = KP866025403 * (T7 - Ta);
cannam@95: 		    TL = KP866025403 * (Tw - Tx);
cannam@95: 		    Tb = T7 + Ta;
cannam@95: 		    Tc = FNMS(KP500000000, Tb, T4);
cannam@95: 		    Ty = Tw + Tx;
cannam@95: 		    TP = FNMS(KP500000000, Ty, Tv);
cannam@95: 	       }
cannam@95: 	       {
cannam@95: 		    E Tf, TC, Ti, TD, Td, Te;
cannam@95: 		    Td = Ip[WS(rs, 1)];
cannam@95: 		    Te = Im[WS(rs, 1)];
cannam@95: 		    Tf = Td - Te;
cannam@95: 		    TC = Te + Td;
cannam@95: 		    {
cannam@95: 			 E Tm, Tn, Tg, Th;
cannam@95: 			 Tm = Ip[0];
cannam@95: 			 Tn = Im[WS(rs, 2)];
cannam@95: 			 To = Tm - Tn;
cannam@95: 			 TB = Tm + Tn;
cannam@95: 			 Tg = Ip[WS(rs, 2)];
cannam@95: 			 Th = Im[0];
cannam@95: 			 Ti = Tg - Th;
cannam@95: 			 TD = Tg + Th;
cannam@95: 		    }
cannam@95: 		    Tj = KP866025403 * (Tf - Ti);
cannam@95: 		    TQ = KP866025403 * (TC + TD);
cannam@95: 		    Tp = Tf + Ti;
cannam@95: 		    Tq = FNMS(KP500000000, Tp, To);
cannam@95: 		    TE = TC - TD;
cannam@95: 		    TM = FMA(KP500000000, TE, TB);
cannam@95: 	       }
cannam@95: 	       {
cannam@95: 		    E TJ, TT, TS, TU;
cannam@95: 		    TJ = T4 + Tb;
cannam@95: 		    TT = To + Tp;
cannam@95: 		    {
cannam@95: 			 E TN, TR, TK, TO;
cannam@95: 			 TN = TL + TM;
cannam@95: 			 TR = TP - TQ;
cannam@95: 			 TK = W[0];
cannam@95: 			 TO = W[1];
cannam@95: 			 TS = FMA(TK, TN, TO * TR);
cannam@95: 			 TU = FNMS(TO, TN, TK * TR);
cannam@95: 		    }
cannam@95: 		    Rp[0] = TJ - TS;
cannam@95: 		    Ip[0] = TT + TU;
cannam@95: 		    Rm[0] = TJ + TS;
cannam@95: 		    Im[0] = TU - TT;
cannam@95: 	       }
cannam@95: 	       {
cannam@95: 		    E TZ, T15, T14, T16;
cannam@95: 		    {
cannam@95: 			 E TW, TY, TV, TX;
cannam@95: 			 TW = Tc + Tj;
cannam@95: 			 TY = Tr + Tq;
cannam@95: 			 TV = W[6];
cannam@95: 			 TX = W[7];
cannam@95: 			 TZ = FNMS(TX, TY, TV * TW);
cannam@95: 			 T15 = FMA(TX, TW, TV * TY);
cannam@95: 		    }
cannam@95: 		    {
cannam@95: 			 E T11, T13, T10, T12;
cannam@95: 			 T11 = TM - TL;
cannam@95: 			 T13 = TP + TQ;
cannam@95: 			 T10 = W[8];
cannam@95: 			 T12 = W[9];
cannam@95: 			 T14 = FMA(T10, T11, T12 * T13);
cannam@95: 			 T16 = FNMS(T12, T11, T10 * T13);
cannam@95: 		    }
cannam@95: 		    Rp[WS(rs, 2)] = TZ - T14;
cannam@95: 		    Ip[WS(rs, 2)] = T15 + T16;
cannam@95: 		    Rm[WS(rs, 2)] = TZ + T14;
cannam@95: 		    Im[WS(rs, 2)] = T16 - T15;
cannam@95: 	       }
cannam@95: 	       {
cannam@95: 		    E Tt, TH, TG, TI;
cannam@95: 		    {
cannam@95: 			 E Tk, Ts, T1, Tl;
cannam@95: 			 Tk = Tc - Tj;
cannam@95: 			 Ts = Tq - Tr;
cannam@95: 			 T1 = W[3];
cannam@95: 			 Tl = W[2];
cannam@95: 			 Tt = FMA(T1, Tk, Tl * Ts);
cannam@95: 			 TH = FNMS(T1, Ts, Tl * Tk);
cannam@95: 		    }
cannam@95: 		    {
cannam@95: 			 E Tz, TF, Tu, TA;
cannam@95: 			 Tz = Tv + Ty;
cannam@95: 			 TF = TB - TE;
cannam@95: 			 Tu = W[4];
cannam@95: 			 TA = W[5];
cannam@95: 			 TG = FNMS(TA, TF, Tu * Tz);
cannam@95: 			 TI = FMA(TA, Tz, Tu * TF);
cannam@95: 		    }
cannam@95: 		    Ip[WS(rs, 1)] = Tt + TG;
cannam@95: 		    Rp[WS(rs, 1)] = TH - TI;
cannam@95: 		    Im[WS(rs, 1)] = TG - Tt;
cannam@95: 		    Rm[WS(rs, 1)] = TH + TI;
cannam@95: 	       }
cannam@95: 	  }
cannam@95:      }
cannam@95: }
cannam@95: 
cannam@95: static const tw_instr twinstr[] = {
cannam@95:      {TW_FULL, 1, 6},
cannam@95:      {TW_NEXT, 1, 0}
cannam@95: };
cannam@95: 
cannam@95: static const hc2c_desc desc = { 6, "hc2cbdft_6", twinstr, &GENUS, {44, 14, 14, 0} };
cannam@95: 
cannam@95: void X(codelet_hc2cbdft_6) (planner *p) {
cannam@95:      X(khc2c_register) (p, hc2cbdft_6, &desc, HC2C_VIA_DFT);
cannam@95: }
cannam@95: #endif				/* HAVE_FMA */