Chris@42: /*
Chris@42:  * Copyright (c) 2003, 2007-14 Matteo Frigo
Chris@42:  * Copyright (c) 2003, 2007-14 Massachusetts Institute of Technology
Chris@42:  *
Chris@42:  * This program is free software; you can redistribute it and/or modify
Chris@42:  * it under the terms of the GNU General Public License as published by
Chris@42:  * the Free Software Foundation; either version 2 of the License, or
Chris@42:  * (at your option) any later version.
Chris@42:  *
Chris@42:  * This program is distributed in the hope that it will be useful,
Chris@42:  * but WITHOUT ANY WARRANTY; without even the implied warranty of
Chris@42:  * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
Chris@42:  * GNU General Public License for more details.
Chris@42:  *
Chris@42:  * You should have received a copy of the GNU General Public License
Chris@42:  * along with this program; if not, write to the Free Software
Chris@42:  * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA  02110-1301  USA
Chris@42:  *
Chris@42:  */
Chris@42: 
Chris@42: /* This file was automatically generated --- DO NOT EDIT */
Chris@42: /* Generated on Sat Jul 30 16:49:41 EDT 2016 */
Chris@42: 
Chris@42: #include "codelet-rdft.h"
Chris@42: 
Chris@42: #ifdef HAVE_FMA
Chris@42: 
Chris@42: /* Generated by: ../../../genfft/gen_hc2hc.native -fma -reorder-insns -schedule-for-pipeline -compact -variables 4 -pipeline-latency 4 -sign 1 -n 6 -dif -name hb_6 -include hb.h */
Chris@42: 
Chris@42: /*
Chris@42:  * This function contains 46 FP additions, 32 FP multiplications,
Chris@42:  * (or, 24 additions, 10 multiplications, 22 fused multiply/add),
Chris@42:  * 45 stack variables, 2 constants, and 24 memory accesses
Chris@42:  */
Chris@42: #include "hb.h"
Chris@42: 
Chris@42: static void hb_6(R *cr, R *ci, const R *W, stride rs, INT mb, INT me, INT ms)
Chris@42: {
Chris@42:      DK(KP866025403, +0.866025403784438646763723170752936183471402627);
Chris@42:      DK(KP500000000, +0.500000000000000000000000000000000000000000000);
Chris@42:      {
Chris@42: 	  INT m;
Chris@42: 	  for (m = mb, W = W + ((mb - 1) * 10); m < me; m = m + 1, cr = cr + ms, ci = ci - ms, W = W + 10, MAKE_VOLATILE_STRIDE(12, rs)) {
Chris@42: 	       E TK, TR, TB, TM, TL, TS;
Chris@42: 	       {
Chris@42: 		    E Td, TN, TO, TJ, Tn, Tk, TC, T3, Tr, T4, T5, T7, T8;
Chris@42: 		    {
Chris@42: 			 E TH, Tg, Tj, TI, Th, Ti, T1, T2;
Chris@42: 			 {
Chris@42: 			      E Tb, Tc, Te, Tf;
Chris@42: 			      Tb = ci[WS(rs, 5)];
Chris@42: 			      Tc = cr[WS(rs, 3)];
Chris@42: 			      Te = ci[WS(rs, 3)];
Chris@42: 			      Tf = cr[WS(rs, 5)];
Chris@42: 			      Th = ci[WS(rs, 4)];
Chris@42: 			      Td = Tb - Tc;
Chris@42: 			      TN = Tb + Tc;
Chris@42: 			      Ti = cr[WS(rs, 4)];
Chris@42: 			      TH = Te + Tf;
Chris@42: 			      Tg = Te - Tf;
Chris@42: 			 }
Chris@42: 			 Tj = Th - Ti;
Chris@42: 			 TI = Th + Ti;
Chris@42: 			 T1 = cr[0];
Chris@42: 			 T2 = ci[WS(rs, 2)];
Chris@42: 			 TO = TH - TI;
Chris@42: 			 TJ = TH + TI;
Chris@42: 			 Tn = Tj - Tg;
Chris@42: 			 Tk = Tg + Tj;
Chris@42: 			 TC = T1 - T2;
Chris@42: 			 T3 = T1 + T2;
Chris@42: 			 Tr = FNMS(KP500000000, Tk, Td);
Chris@42: 			 T4 = cr[WS(rs, 2)];
Chris@42: 			 T5 = ci[0];
Chris@42: 			 T7 = ci[WS(rs, 1)];
Chris@42: 			 T8 = cr[WS(rs, 1)];
Chris@42: 		    }
Chris@42: 		    {
Chris@42: 			 E Tl, Tq, TQ, Ts, Ta, T10, TG;
Chris@42: 			 ci[0] = Td + Tk;
Chris@42: 			 {
Chris@42: 			      E T6, TD, T9, TE, TF;
Chris@42: 			      T6 = T4 + T5;
Chris@42: 			      TD = T4 - T5;
Chris@42: 			      T9 = T7 + T8;
Chris@42: 			      TE = T7 - T8;
Chris@42: 			      Tl = W[2];
Chris@42: 			      Tq = W[3];
Chris@42: 			      TQ = TD - TE;
Chris@42: 			      TF = TD + TE;
Chris@42: 			      Ts = T6 - T9;
Chris@42: 			      Ta = T6 + T9;
Chris@42: 			      T10 = TC + TF;
Chris@42: 			      TG = FNMS(KP500000000, TF, TC);
Chris@42: 			 }
Chris@42: 			 {
Chris@42: 			      E T13, TP, Tz, TZ, Tw, T14, Tv, Ty;
Chris@42: 			      {
Chris@42: 				   E Tt, T12, T11, Tp, Tm, To, Tu;
Chris@42: 				   T13 = TN + TO;
Chris@42: 				   TP = FNMS(KP500000000, TO, TN);
Chris@42: 				   cr[0] = T3 + Ta;
Chris@42: 				   Tm = FNMS(KP500000000, Ta, T3);
Chris@42: 				   Tz = FMA(KP866025403, Ts, Tr);
Chris@42: 				   Tt = FNMS(KP866025403, Ts, Tr);
Chris@42: 				   TZ = W[4];
Chris@42: 				   To = FNMS(KP866025403, Tn, Tm);
Chris@42: 				   Tw = FMA(KP866025403, Tn, Tm);
Chris@42: 				   Tu = Tl * Tt;
Chris@42: 				   T12 = W[5];
Chris@42: 				   T11 = TZ * T10;
Chris@42: 				   Tp = Tl * To;
Chris@42: 				   ci[WS(rs, 2)] = FMA(Tq, To, Tu);
Chris@42: 				   T14 = T12 * T10;
Chris@42: 				   cr[WS(rs, 3)] = FNMS(T12, T13, T11);
Chris@42: 				   cr[WS(rs, 2)] = FNMS(Tq, Tt, Tp);
Chris@42: 			      }
Chris@42: 			      ci[WS(rs, 3)] = FMA(TZ, T13, T14);
Chris@42: 			      Tv = W[6];
Chris@42: 			      Ty = W[7];
Chris@42: 			      {
Chris@42: 				   E TX, TT, TW, TV, TY, TU, TA, Tx;
Chris@42: 				   TK = FNMS(KP866025403, TJ, TG);
Chris@42: 				   TU = FMA(KP866025403, TJ, TG);
Chris@42: 				   TA = Tv * Tz;
Chris@42: 				   Tx = Tv * Tw;
Chris@42: 				   TX = FNMS(KP866025403, TQ, TP);
Chris@42: 				   TR = FMA(KP866025403, TQ, TP);
Chris@42: 				   ci[WS(rs, 4)] = FMA(Ty, Tw, TA);
Chris@42: 				   cr[WS(rs, 4)] = FNMS(Ty, Tz, Tx);
Chris@42: 				   TT = W[8];
Chris@42: 				   TW = W[9];
Chris@42: 				   TB = W[0];
Chris@42: 				   TV = TT * TU;
Chris@42: 				   TY = TW * TU;
Chris@42: 				   TM = W[1];
Chris@42: 				   TL = TB * TK;
Chris@42: 				   cr[WS(rs, 5)] = FNMS(TW, TX, TV);
Chris@42: 				   ci[WS(rs, 5)] = FMA(TT, TX, TY);
Chris@42: 			      }
Chris@42: 			 }
Chris@42: 		    }
Chris@42: 	       }
Chris@42: 	       cr[WS(rs, 1)] = FNMS(TM, TR, TL);
Chris@42: 	       TS = TM * TK;
Chris@42: 	       ci[WS(rs, 1)] = FMA(TB, TR, TS);
Chris@42: 	  }
Chris@42:      }
Chris@42: }
Chris@42: 
Chris@42: static const tw_instr twinstr[] = {
Chris@42:      {TW_FULL, 1, 6},
Chris@42:      {TW_NEXT, 1, 0}
Chris@42: };
Chris@42: 
Chris@42: static const hc2hc_desc desc = { 6, "hb_6", twinstr, &GENUS, {24, 10, 22, 0} };
Chris@42: 
Chris@42: void X(codelet_hb_6) (planner *p) {
Chris@42:      X(khc2hc_register) (p, hb_6, &desc);
Chris@42: }
Chris@42: #else				/* HAVE_FMA */
Chris@42: 
Chris@42: /* Generated by: ../../../genfft/gen_hc2hc.native -compact -variables 4 -pipeline-latency 4 -sign 1 -n 6 -dif -name hb_6 -include hb.h */
Chris@42: 
Chris@42: /*
Chris@42:  * This function contains 46 FP additions, 28 FP multiplications,
Chris@42:  * (or, 32 additions, 14 multiplications, 14 fused multiply/add),
Chris@42:  * 27 stack variables, 2 constants, and 24 memory accesses
Chris@42:  */
Chris@42: #include "hb.h"
Chris@42: 
Chris@42: static void hb_6(R *cr, R *ci, const R *W, stride rs, INT mb, INT me, INT ms)
Chris@42: {
Chris@42:      DK(KP500000000, +0.500000000000000000000000000000000000000000000);
Chris@42:      DK(KP866025403, +0.866025403784438646763723170752936183471402627);
Chris@42:      {
Chris@42: 	  INT m;
Chris@42: 	  for (m = mb, W = W + ((mb - 1) * 10); m < me; m = m + 1, cr = cr + ms, ci = ci - ms, W = W + 10, MAKE_VOLATILE_STRIDE(12, rs)) {
Chris@42: 	       E T3, Ty, Ta, TO, Tr, TB, Td, TE, Tk, TL, Tn, TH;
Chris@42: 	       {
Chris@42: 		    E T1, T2, Tb, Tc;
Chris@42: 		    T1 = cr[0];
Chris@42: 		    T2 = ci[WS(rs, 2)];
Chris@42: 		    T3 = T1 + T2;
Chris@42: 		    Ty = T1 - T2;
Chris@42: 		    {
Chris@42: 			 E T6, Tz, T9, TA;
Chris@42: 			 {
Chris@42: 			      E T4, T5, T7, T8;
Chris@42: 			      T4 = cr[WS(rs, 2)];
Chris@42: 			      T5 = ci[0];
Chris@42: 			      T6 = T4 + T5;
Chris@42: 			      Tz = T4 - T5;
Chris@42: 			      T7 = ci[WS(rs, 1)];
Chris@42: 			      T8 = cr[WS(rs, 1)];
Chris@42: 			      T9 = T7 + T8;
Chris@42: 			      TA = T7 - T8;
Chris@42: 			 }
Chris@42: 			 Ta = T6 + T9;
Chris@42: 			 TO = KP866025403 * (Tz - TA);
Chris@42: 			 Tr = KP866025403 * (T6 - T9);
Chris@42: 			 TB = Tz + TA;
Chris@42: 		    }
Chris@42: 		    Tb = ci[WS(rs, 5)];
Chris@42: 		    Tc = cr[WS(rs, 3)];
Chris@42: 		    Td = Tb - Tc;
Chris@42: 		    TE = Tb + Tc;
Chris@42: 		    {
Chris@42: 			 E Tg, TG, Tj, TF;
Chris@42: 			 {
Chris@42: 			      E Te, Tf, Th, Ti;
Chris@42: 			      Te = ci[WS(rs, 3)];
Chris@42: 			      Tf = cr[WS(rs, 5)];
Chris@42: 			      Tg = Te - Tf;
Chris@42: 			      TG = Te + Tf;
Chris@42: 			      Th = ci[WS(rs, 4)];
Chris@42: 			      Ti = cr[WS(rs, 4)];
Chris@42: 			      Tj = Th - Ti;
Chris@42: 			      TF = Th + Ti;
Chris@42: 			 }
Chris@42: 			 Tk = Tg + Tj;
Chris@42: 			 TL = KP866025403 * (TG + TF);
Chris@42: 			 Tn = KP866025403 * (Tj - Tg);
Chris@42: 			 TH = TF - TG;
Chris@42: 		    }
Chris@42: 	       }
Chris@42: 	       cr[0] = T3 + Ta;
Chris@42: 	       ci[0] = Td + Tk;
Chris@42: 	       {
Chris@42: 		    E TC, TI, Tx, TD;
Chris@42: 		    TC = Ty + TB;
Chris@42: 		    TI = TE - TH;
Chris@42: 		    Tx = W[4];
Chris@42: 		    TD = W[5];
Chris@42: 		    cr[WS(rs, 3)] = FNMS(TD, TI, Tx * TC);
Chris@42: 		    ci[WS(rs, 3)] = FMA(TD, TC, Tx * TI);
Chris@42: 	       }
Chris@42: 	       {
Chris@42: 		    E To, Tu, Ts, Tw, Tm, Tq;
Chris@42: 		    Tm = FNMS(KP500000000, Ta, T3);
Chris@42: 		    To = Tm - Tn;
Chris@42: 		    Tu = Tm + Tn;
Chris@42: 		    Tq = FNMS(KP500000000, Tk, Td);
Chris@42: 		    Ts = Tq - Tr;
Chris@42: 		    Tw = Tr + Tq;
Chris@42: 		    {
Chris@42: 			 E Tl, Tp, Tt, Tv;
Chris@42: 			 Tl = W[2];
Chris@42: 			 Tp = W[3];
Chris@42: 			 cr[WS(rs, 2)] = FNMS(Tp, Ts, Tl * To);
Chris@42: 			 ci[WS(rs, 2)] = FMA(Tl, Ts, Tp * To);
Chris@42: 			 Tt = W[6];
Chris@42: 			 Tv = W[7];
Chris@42: 			 cr[WS(rs, 4)] = FNMS(Tv, Tw, Tt * Tu);
Chris@42: 			 ci[WS(rs, 4)] = FMA(Tt, Tw, Tv * Tu);
Chris@42: 		    }
Chris@42: 	       }
Chris@42: 	       {
Chris@42: 		    E TM, TS, TQ, TU, TK, TP;
Chris@42: 		    TK = FNMS(KP500000000, TB, Ty);
Chris@42: 		    TM = TK - TL;
Chris@42: 		    TS = TK + TL;
Chris@42: 		    TP = FMA(KP500000000, TH, TE);
Chris@42: 		    TQ = TO + TP;
Chris@42: 		    TU = TP - TO;
Chris@42: 		    {
Chris@42: 			 E TJ, TN, TR, TT;
Chris@42: 			 TJ = W[0];
Chris@42: 			 TN = W[1];
Chris@42: 			 cr[WS(rs, 1)] = FNMS(TN, TQ, TJ * TM);
Chris@42: 			 ci[WS(rs, 1)] = FMA(TN, TM, TJ * TQ);
Chris@42: 			 TR = W[8];
Chris@42: 			 TT = W[9];
Chris@42: 			 cr[WS(rs, 5)] = FNMS(TT, TU, TR * TS);
Chris@42: 			 ci[WS(rs, 5)] = FMA(TT, TS, TR * TU);
Chris@42: 		    }
Chris@42: 	       }
Chris@42: 	  }
Chris@42:      }
Chris@42: }
Chris@42: 
Chris@42: static const tw_instr twinstr[] = {
Chris@42:      {TW_FULL, 1, 6},
Chris@42:      {TW_NEXT, 1, 0}
Chris@42: };
Chris@42: 
Chris@42: static const hc2hc_desc desc = { 6, "hb_6", twinstr, &GENUS, {32, 14, 14, 0} };
Chris@42: 
Chris@42: void X(codelet_hb_6) (planner *p) {
Chris@42:      X(khc2hc_register) (p, hb_6, &desc);
Chris@42: }
Chris@42: #endif				/* HAVE_FMA */