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 07:39:50 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_hc2hc.native -fma -reorder-insns -schedule-for-pipeline -compact -variables 4 -pipeline-latency 4 -n 7 -dit -name hf_7 -include hf.h */
Chris@10: 
Chris@10: /*
Chris@10:  * This function contains 72 FP additions, 66 FP multiplications,
Chris@10:  * (or, 18 additions, 12 multiplications, 54 fused multiply/add),
Chris@10:  * 62 stack variables, 6 constants, and 28 memory accesses
Chris@10:  */
Chris@10: #include "hf.h"
Chris@10: 
Chris@10: static void hf_7(R *cr, R *ci, const R *W, stride rs, INT mb, INT me, INT ms)
Chris@10: {
Chris@10:      DK(KP974927912, +0.974927912181823607018131682993931217232785801);
Chris@10:      DK(KP801937735, +0.801937735804838252472204639014890102331838324);
Chris@10:      DK(KP900968867, +0.900968867902419126236102319507445051165919162);
Chris@10:      DK(KP692021471, +0.692021471630095869627814897002069140197260599);
Chris@10:      DK(KP554958132, +0.554958132087371191422194871006410481067288862);
Chris@10:      DK(KP356895867, +0.356895867892209443894399510021300583399127187);
Chris@10:      {
Chris@10: 	  INT m;
Chris@10: 	  for (m = mb, W = W + ((mb - 1) * 12); m < me; m = m + 1, cr = cr + ms, ci = ci - ms, W = W + 12, MAKE_VOLATILE_STRIDE(14, rs)) {
Chris@10: 	       E T1, TR, T18, T10, T12, T16, T11, T13;
Chris@10: 	       {
Chris@10: 		    E T19, T1a, T1i, Te, Tt, Tw, T1b, TM, T1h, Tr, Tu, TS, Tz, TC, Ty;
Chris@10: 		    E Tv, TB;
Chris@10: 		    T1 = cr[0];
Chris@10: 		    T19 = ci[0];
Chris@10: 		    {
Chris@10: 			 E T9, Tc, TP, Ta, Tb, TO, T7;
Chris@10: 			 {
Chris@10: 			      E T3, T6, T8, TN, T4, T2, T5;
Chris@10: 			      T3 = cr[WS(rs, 1)];
Chris@10: 			      T6 = ci[WS(rs, 1)];
Chris@10: 			      T2 = W[0];
Chris@10: 			      T9 = cr[WS(rs, 6)];
Chris@10: 			      Tc = ci[WS(rs, 6)];
Chris@10: 			      T8 = W[10];
Chris@10: 			      TN = T2 * T6;
Chris@10: 			      T4 = T2 * T3;
Chris@10: 			      T5 = W[1];
Chris@10: 			      TP = T8 * Tc;
Chris@10: 			      Ta = T8 * T9;
Chris@10: 			      Tb = W[11];
Chris@10: 			      TO = FNMS(T5, T3, TN);
Chris@10: 			      T7 = FMA(T5, T6, T4);
Chris@10: 			 }
Chris@10: 			 {
Chris@10: 			      E Tg, Tj, Th, TI, Tm, Tp, Tl, Ti, To, TQ, Td, Tf;
Chris@10: 			      Tg = cr[WS(rs, 2)];
Chris@10: 			      TQ = FNMS(Tb, T9, TP);
Chris@10: 			      Td = FMA(Tb, Tc, Ta);
Chris@10: 			      Tj = ci[WS(rs, 2)];
Chris@10: 			      Tf = W[2];
Chris@10: 			      T1a = TO + TQ;
Chris@10: 			      TR = TO - TQ;
Chris@10: 			      T1i = Td - T7;
Chris@10: 			      Te = T7 + Td;
Chris@10: 			      Th = Tf * Tg;
Chris@10: 			      TI = Tf * Tj;
Chris@10: 			      Tm = cr[WS(rs, 5)];
Chris@10: 			      Tp = ci[WS(rs, 5)];
Chris@10: 			      Tl = W[8];
Chris@10: 			      Ti = W[3];
Chris@10: 			      To = W[9];
Chris@10: 			      {
Chris@10: 				   E TJ, Tk, TL, Tq, TK, Tn, Ts;
Chris@10: 				   Tt = cr[WS(rs, 3)];
Chris@10: 				   TK = Tl * Tp;
Chris@10: 				   Tn = Tl * Tm;
Chris@10: 				   TJ = FNMS(Ti, Tg, TI);
Chris@10: 				   Tk = FMA(Ti, Tj, Th);
Chris@10: 				   TL = FNMS(To, Tm, TK);
Chris@10: 				   Tq = FMA(To, Tp, Tn);
Chris@10: 				   Tw = ci[WS(rs, 3)];
Chris@10: 				   Ts = W[4];
Chris@10: 				   T1b = TJ + TL;
Chris@10: 				   TM = TJ - TL;
Chris@10: 				   T1h = Tq - Tk;
Chris@10: 				   Tr = Tk + Tq;
Chris@10: 				   Tu = Ts * Tt;
Chris@10: 				   TS = Ts * Tw;
Chris@10: 			      }
Chris@10: 			      Tz = cr[WS(rs, 4)];
Chris@10: 			      TC = ci[WS(rs, 4)];
Chris@10: 			      Ty = W[6];
Chris@10: 			      Tv = W[5];
Chris@10: 			      TB = W[7];
Chris@10: 			 }
Chris@10: 		    }
Chris@10: 		    {
Chris@10: 			 E TF, TT, Tx, TV, TD, T1q, TU, TA;
Chris@10: 			 TF = FNMS(KP356895867, Tr, Te);
Chris@10: 			 TU = Ty * TC;
Chris@10: 			 TA = Ty * Tz;
Chris@10: 			 TT = FNMS(Tv, Tt, TS);
Chris@10: 			 Tx = FMA(Tv, Tw, Tu);
Chris@10: 			 TV = FNMS(TB, Tz, TU);
Chris@10: 			 TD = FMA(TB, TC, TA);
Chris@10: 			 T1q = FNMS(KP356895867, T1b, T1a);
Chris@10: 			 {
Chris@10: 			      E TW, TE, T1k, T1f;
Chris@10: 			      {
Chris@10: 				   E T1e, T1s, TY, T1p, T1u, TH, T1n, T1j, T1c, T1g;
Chris@10: 				   T1j = FNMS(KP554958132, T1i, T1h);
Chris@10: 				   T1c = TT + TV;
Chris@10: 				   TW = TT - TV;
Chris@10: 				   T1g = TD - Tx;
Chris@10: 				   TE = Tx + TD;
Chris@10: 				   {
Chris@10: 					E T1d, T1l, T1r, TX;
Chris@10: 					T1d = FNMS(KP356895867, T1c, T1b);
Chris@10: 					T1l = FNMS(KP356895867, T1a, T1c);
Chris@10: 					T1r = FNMS(KP692021471, T1q, T1c);
Chris@10: 					ci[WS(rs, 6)] = T1a + T1b + T1c + T19;
Chris@10: 					TX = FMA(KP554958132, TW, TR);
Chris@10: 					{
Chris@10: 					     E T1o, T1t, TG, T1m;
Chris@10: 					     T1o = FMA(KP554958132, T1h, T1g);
Chris@10: 					     T1t = FMA(KP554958132, T1g, T1i);
Chris@10: 					     TG = FNMS(KP692021471, TF, TE);
Chris@10: 					     cr[0] = T1 + Te + Tr + TE;
Chris@10: 					     T1e = FNMS(KP692021471, T1d, T1a);
Chris@10: 					     T1m = FNMS(KP692021471, T1l, T1b);
Chris@10: 					     T1s = FNMS(KP900968867, T1r, T19);
Chris@10: 					     TY = FMA(KP801937735, TX, TM);
Chris@10: 					     T1p = FNMS(KP801937735, T1o, T1i);
Chris@10: 					     T1u = FMA(KP801937735, T1t, T1h);
Chris@10: 					     TH = FNMS(KP900968867, TG, T1);
Chris@10: 					     T1n = FNMS(KP900968867, T1m, T19);
Chris@10: 					     T1k = FNMS(KP801937735, T1j, T1g);
Chris@10: 					}
Chris@10: 				   }
Chris@10: 				   ci[WS(rs, 5)] = FMA(KP974927912, T1u, T1s);
Chris@10: 				   cr[WS(rs, 6)] = FMS(KP974927912, T1u, T1s);
Chris@10: 				   cr[WS(rs, 1)] = FMA(KP974927912, TY, TH);
Chris@10: 				   ci[0] = FNMS(KP974927912, TY, TH);
Chris@10: 				   ci[WS(rs, 4)] = FMA(KP974927912, T1p, T1n);
Chris@10: 				   cr[WS(rs, 5)] = FMS(KP974927912, T1p, T1n);
Chris@10: 				   T1f = FNMS(KP900968867, T1e, T19);
Chris@10: 			      }
Chris@10: 			      {
Chris@10: 				   E T14, T17, T15, TZ;
Chris@10: 				   T14 = FNMS(KP356895867, TE, Tr);
Chris@10: 				   T17 = FNMS(KP554958132, TR, TM);
Chris@10: 				   TZ = FNMS(KP356895867, Te, TE);
Chris@10: 				   ci[WS(rs, 3)] = FMA(KP974927912, T1k, T1f);
Chris@10: 				   cr[WS(rs, 4)] = FMS(KP974927912, T1k, T1f);
Chris@10: 				   T15 = FNMS(KP692021471, T14, Te);
Chris@10: 				   T18 = FNMS(KP801937735, T17, TW);
Chris@10: 				   T10 = FNMS(KP692021471, TZ, Tr);
Chris@10: 				   T12 = FMA(KP554958132, TM, TW);
Chris@10: 				   T16 = FNMS(KP900968867, T15, T1);
Chris@10: 			      }
Chris@10: 			 }
Chris@10: 		    }
Chris@10: 	       }
Chris@10: 	       T11 = FNMS(KP900968867, T10, T1);
Chris@10: 	       T13 = FNMS(KP801937735, T12, TR);
Chris@10: 	       cr[WS(rs, 3)] = FMA(KP974927912, T18, T16);
Chris@10: 	       ci[WS(rs, 2)] = FNMS(KP974927912, T18, T16);
Chris@10: 	       cr[WS(rs, 2)] = FMA(KP974927912, T13, T11);
Chris@10: 	       ci[WS(rs, 1)] = FNMS(KP974927912, T13, T11);
Chris@10: 	  }
Chris@10:      }
Chris@10: }
Chris@10: 
Chris@10: static const tw_instr twinstr[] = {
Chris@10:      {TW_FULL, 1, 7},
Chris@10:      {TW_NEXT, 1, 0}
Chris@10: };
Chris@10: 
Chris@10: static const hc2hc_desc desc = { 7, "hf_7", twinstr, &GENUS, {18, 12, 54, 0} };
Chris@10: 
Chris@10: void X(codelet_hf_7) (planner *p) {
Chris@10:      X(khc2hc_register) (p, hf_7, &desc);
Chris@10: }
Chris@10: #else				/* HAVE_FMA */
Chris@10: 
Chris@10: /* Generated by: ../../../genfft/gen_hc2hc.native -compact -variables 4 -pipeline-latency 4 -n 7 -dit -name hf_7 -include hf.h */
Chris@10: 
Chris@10: /*
Chris@10:  * This function contains 72 FP additions, 60 FP multiplications,
Chris@10:  * (or, 36 additions, 24 multiplications, 36 fused multiply/add),
Chris@10:  * 29 stack variables, 6 constants, and 28 memory accesses
Chris@10:  */
Chris@10: #include "hf.h"
Chris@10: 
Chris@10: static void hf_7(R *cr, R *ci, const R *W, stride rs, INT mb, INT me, INT ms)
Chris@10: {
Chris@10:      DK(KP222520933, +0.222520933956314404288902564496794759466355569);
Chris@10:      DK(KP900968867, +0.900968867902419126236102319507445051165919162);
Chris@10:      DK(KP623489801, +0.623489801858733530525004884004239810632274731);
Chris@10:      DK(KP433883739, +0.433883739117558120475768332848358754609990728);
Chris@10:      DK(KP974927912, +0.974927912181823607018131682993931217232785801);
Chris@10:      DK(KP781831482, +0.781831482468029808708444526674057750232334519);
Chris@10:      {
Chris@10: 	  INT m;
Chris@10: 	  for (m = mb, W = W + ((mb - 1) * 12); m < me; m = m + 1, cr = cr + ms, ci = ci - ms, W = W + 12, MAKE_VOLATILE_STRIDE(14, rs)) {
Chris@10: 	       E T1, TT, Tc, TV, TC, TO, Tn, TS, TI, TP, Ty, TU, TF, TQ;
Chris@10: 	       T1 = cr[0];
Chris@10: 	       TT = ci[0];
Chris@10: 	       {
Chris@10: 		    E T6, TA, Tb, TB;
Chris@10: 		    {
Chris@10: 			 E T3, T5, T2, T4;
Chris@10: 			 T3 = cr[WS(rs, 1)];
Chris@10: 			 T5 = ci[WS(rs, 1)];
Chris@10: 			 T2 = W[0];
Chris@10: 			 T4 = W[1];
Chris@10: 			 T6 = FMA(T2, T3, T4 * T5);
Chris@10: 			 TA = FNMS(T4, T3, T2 * T5);
Chris@10: 		    }
Chris@10: 		    {
Chris@10: 			 E T8, Ta, T7, T9;
Chris@10: 			 T8 = cr[WS(rs, 6)];
Chris@10: 			 Ta = ci[WS(rs, 6)];
Chris@10: 			 T7 = W[10];
Chris@10: 			 T9 = W[11];
Chris@10: 			 Tb = FMA(T7, T8, T9 * Ta);
Chris@10: 			 TB = FNMS(T9, T8, T7 * Ta);
Chris@10: 		    }
Chris@10: 		    Tc = T6 + Tb;
Chris@10: 		    TV = TA + TB;
Chris@10: 		    TC = TA - TB;
Chris@10: 		    TO = Tb - T6;
Chris@10: 	       }
Chris@10: 	       {
Chris@10: 		    E Th, TG, Tm, TH;
Chris@10: 		    {
Chris@10: 			 E Te, Tg, Td, Tf;
Chris@10: 			 Te = cr[WS(rs, 2)];
Chris@10: 			 Tg = ci[WS(rs, 2)];
Chris@10: 			 Td = W[2];
Chris@10: 			 Tf = W[3];
Chris@10: 			 Th = FMA(Td, Te, Tf * Tg);
Chris@10: 			 TG = FNMS(Tf, Te, Td * Tg);
Chris@10: 		    }
Chris@10: 		    {
Chris@10: 			 E Tj, Tl, Ti, Tk;
Chris@10: 			 Tj = cr[WS(rs, 5)];
Chris@10: 			 Tl = ci[WS(rs, 5)];
Chris@10: 			 Ti = W[8];
Chris@10: 			 Tk = W[9];
Chris@10: 			 Tm = FMA(Ti, Tj, Tk * Tl);
Chris@10: 			 TH = FNMS(Tk, Tj, Ti * Tl);
Chris@10: 		    }
Chris@10: 		    Tn = Th + Tm;
Chris@10: 		    TS = TG + TH;
Chris@10: 		    TI = TG - TH;
Chris@10: 		    TP = Th - Tm;
Chris@10: 	       }
Chris@10: 	       {
Chris@10: 		    E Ts, TD, Tx, TE;
Chris@10: 		    {
Chris@10: 			 E Tp, Tr, To, Tq;
Chris@10: 			 Tp = cr[WS(rs, 3)];
Chris@10: 			 Tr = ci[WS(rs, 3)];
Chris@10: 			 To = W[4];
Chris@10: 			 Tq = W[5];
Chris@10: 			 Ts = FMA(To, Tp, Tq * Tr);
Chris@10: 			 TD = FNMS(Tq, Tp, To * Tr);
Chris@10: 		    }
Chris@10: 		    {
Chris@10: 			 E Tu, Tw, Tt, Tv;
Chris@10: 			 Tu = cr[WS(rs, 4)];
Chris@10: 			 Tw = ci[WS(rs, 4)];
Chris@10: 			 Tt = W[6];
Chris@10: 			 Tv = W[7];
Chris@10: 			 Tx = FMA(Tt, Tu, Tv * Tw);
Chris@10: 			 TE = FNMS(Tv, Tu, Tt * Tw);
Chris@10: 		    }
Chris@10: 		    Ty = Ts + Tx;
Chris@10: 		    TU = TD + TE;
Chris@10: 		    TF = TD - TE;
Chris@10: 		    TQ = Tx - Ts;
Chris@10: 	       }
Chris@10: 	       {
Chris@10: 		    E TL, TK, TZ, T10;
Chris@10: 		    cr[0] = T1 + Tc + Tn + Ty;
Chris@10: 		    TL = FMA(KP781831482, TC, KP974927912 * TI) + (KP433883739 * TF);
Chris@10: 		    TK = FMA(KP623489801, Tc, T1) + FNMA(KP900968867, Ty, KP222520933 * Tn);
Chris@10: 		    ci[0] = TK - TL;
Chris@10: 		    cr[WS(rs, 1)] = TK + TL;
Chris@10: 		    ci[WS(rs, 6)] = TV + TS + TU + TT;
Chris@10: 		    TZ = FMA(KP781831482, TO, KP433883739 * TQ) - (KP974927912 * TP);
Chris@10: 		    T10 = FMA(KP623489801, TV, TT) + FNMA(KP900968867, TU, KP222520933 * TS);
Chris@10: 		    cr[WS(rs, 6)] = TZ - T10;
Chris@10: 		    ci[WS(rs, 5)] = TZ + T10;
Chris@10: 	       }
Chris@10: 	       {
Chris@10: 		    E TX, TY, TR, TW;
Chris@10: 		    TX = FMA(KP974927912, TO, KP433883739 * TP) - (KP781831482 * TQ);
Chris@10: 		    TY = FMA(KP623489801, TU, TT) + FNMA(KP900968867, TS, KP222520933 * TV);
Chris@10: 		    cr[WS(rs, 5)] = TX - TY;
Chris@10: 		    ci[WS(rs, 4)] = TX + TY;
Chris@10: 		    TR = FMA(KP433883739, TO, KP781831482 * TP) + (KP974927912 * TQ);
Chris@10: 		    TW = FMA(KP623489801, TS, TT) + FNMA(KP222520933, TU, KP900968867 * TV);
Chris@10: 		    cr[WS(rs, 4)] = TR - TW;
Chris@10: 		    ci[WS(rs, 3)] = TR + TW;
Chris@10: 	       }
Chris@10: 	       {
Chris@10: 		    E TN, TM, TJ, Tz;
Chris@10: 		    TN = FMA(KP433883739, TC, KP974927912 * TF) - (KP781831482 * TI);
Chris@10: 		    TM = FMA(KP623489801, Tn, T1) + FNMA(KP222520933, Ty, KP900968867 * Tc);
Chris@10: 		    ci[WS(rs, 2)] = TM - TN;
Chris@10: 		    cr[WS(rs, 3)] = TM + TN;
Chris@10: 		    TJ = FNMS(KP781831482, TF, KP974927912 * TC) - (KP433883739 * TI);
Chris@10: 		    Tz = FMA(KP623489801, Ty, T1) + FNMA(KP900968867, Tn, KP222520933 * Tc);
Chris@10: 		    ci[WS(rs, 1)] = Tz - TJ;
Chris@10: 		    cr[WS(rs, 2)] = Tz + TJ;
Chris@10: 	       }
Chris@10: 	  }
Chris@10:      }
Chris@10: }
Chris@10: 
Chris@10: static const tw_instr twinstr[] = {
Chris@10:      {TW_FULL, 1, 7},
Chris@10:      {TW_NEXT, 1, 0}
Chris@10: };
Chris@10: 
Chris@10: static const hc2hc_desc desc = { 7, "hf_7", twinstr, &GENUS, {36, 24, 36, 0} };
Chris@10: 
Chris@10: void X(codelet_hf_7) (planner *p) {
Chris@10:      X(khc2hc_register) (p, hf_7, &desc);
Chris@10: }
Chris@10: #endif				/* HAVE_FMA */