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 16:46:18 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 -n 9 -dit -name hf_9 -include hf.h */
Chris@42: 
Chris@42: /*
Chris@42:  * This function contains 96 FP additions, 88 FP multiplications,
Chris@42:  * (or, 24 additions, 16 multiplications, 72 fused multiply/add),
Chris@42:  * 69 stack variables, 10 constants, and 36 memory accesses
Chris@42:  */
Chris@42: #include "hf.h"
Chris@42: 
Chris@42: static void hf_9(R *cr, R *ci, const R *W, stride rs, INT mb, INT me, INT ms)
Chris@42: {
Chris@42:      DK(KP777861913, +0.777861913430206160028177977318626690410586096);
Chris@42:      DK(KP852868531, +0.852868531952443209628250963940074071936020296);
Chris@42:      DK(KP839099631, +0.839099631177280011763127298123181364687434283);
Chris@42:      DK(KP492403876, +0.492403876506104029683371512294761506835321626);
Chris@42:      DK(KP984807753, +0.984807753012208059366743024589523013670643252);
Chris@42:      DK(KP954188894, +0.954188894138671133499268364187245676532219158);
Chris@42:      DK(KP363970234, +0.363970234266202361351047882776834043890471784);
Chris@42:      DK(KP176326980, +0.176326980708464973471090386868618986121633062);
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) * 16); m < me; m = m + 1, cr = cr + ms, ci = ci - ms, W = W + 16, MAKE_VOLATILE_STRIDE(18, rs)) {
Chris@42: 	       E T20, T1Z;
Chris@42: 	       {
Chris@42: 		    E T1, T1P, T1Q, T10, T1S, Te, TB, T1d, T1a, T19, T1M, TE, T1c, Tz, T1n;
Chris@42: 		    E TC, TH, TK, T1k, TR, TG, TJ, TD;
Chris@42: 		    T1 = cr[0];
Chris@42: 		    T1P = ci[0];
Chris@42: 		    {
Chris@42: 			 E T9, Tc, TY, Ta, Tb, TX, T7;
Chris@42: 			 {
Chris@42: 			      E T3, T6, T8, TW, T4, T2, T5;
Chris@42: 			      T3 = cr[WS(rs, 3)];
Chris@42: 			      T6 = ci[WS(rs, 3)];
Chris@42: 			      T2 = W[4];
Chris@42: 			      T9 = cr[WS(rs, 6)];
Chris@42: 			      Tc = ci[WS(rs, 6)];
Chris@42: 			      T8 = W[10];
Chris@42: 			      TW = T2 * T6;
Chris@42: 			      T4 = T2 * T3;
Chris@42: 			      T5 = W[5];
Chris@42: 			      TY = T8 * Tc;
Chris@42: 			      Ta = T8 * T9;
Chris@42: 			      Tb = W[11];
Chris@42: 			      TX = FNMS(T5, T3, TW);
Chris@42: 			      T7 = FMA(T5, T6, T4);
Chris@42: 			 }
Chris@42: 			 {
Chris@42: 			      E Th, Tk, Ti, T12, Tn, Tq, Tp, T17, Tx, T14, To, Tj, TZ, Td, Tg;
Chris@42: 			      E TA, Tl, Ty;
Chris@42: 			      Th = cr[WS(rs, 1)];
Chris@42: 			      TZ = FNMS(Tb, T9, TY);
Chris@42: 			      Td = FMA(Tb, Tc, Ta);
Chris@42: 			      Tk = ci[WS(rs, 1)];
Chris@42: 			      Tg = W[0];
Chris@42: 			      T1Q = TX + TZ;
Chris@42: 			      T10 = TX - TZ;
Chris@42: 			      T1S = Td - T7;
Chris@42: 			      Te = T7 + Td;
Chris@42: 			      Ti = Tg * Th;
Chris@42: 			      T12 = Tg * Tk;
Chris@42: 			      {
Chris@42: 				   E Tt, Tw, Ts, Tv, T16, Tu, Tm;
Chris@42: 				   Tt = cr[WS(rs, 7)];
Chris@42: 				   Tw = ci[WS(rs, 7)];
Chris@42: 				   Ts = W[12];
Chris@42: 				   Tv = W[13];
Chris@42: 				   Tn = cr[WS(rs, 4)];
Chris@42: 				   Tq = ci[WS(rs, 4)];
Chris@42: 				   T16 = Ts * Tw;
Chris@42: 				   Tu = Ts * Tt;
Chris@42: 				   Tm = W[6];
Chris@42: 				   Tp = W[7];
Chris@42: 				   T17 = FNMS(Tv, Tt, T16);
Chris@42: 				   Tx = FMA(Tv, Tw, Tu);
Chris@42: 				   T14 = Tm * Tq;
Chris@42: 				   To = Tm * Tn;
Chris@42: 			      }
Chris@42: 			      Tj = W[1];
Chris@42: 			      TB = cr[WS(rs, 2)];
Chris@42: 			      {
Chris@42: 				   E T15, Tr, T13, T18;
Chris@42: 				   T15 = FNMS(Tp, Tn, T14);
Chris@42: 				   Tr = FMA(Tp, Tq, To);
Chris@42: 				   T13 = FNMS(Tj, Th, T12);
Chris@42: 				   Tl = FMA(Tj, Tk, Ti);
Chris@42: 				   T18 = T15 + T17;
Chris@42: 				   T1d = T15 - T17;
Chris@42: 				   Ty = Tr + Tx;
Chris@42: 				   T1a = Tr - Tx;
Chris@42: 				   T19 = FNMS(KP500000000, T18, T13);
Chris@42: 				   T1M = T13 + T18;
Chris@42: 				   TE = ci[WS(rs, 2)];
Chris@42: 			      }
Chris@42: 			      T1c = FNMS(KP500000000, Ty, Tl);
Chris@42: 			      Tz = Tl + Ty;
Chris@42: 			      TA = W[2];
Chris@42: 			      {
Chris@42: 				   E TN, TQ, TP, T1j, TO, TM;
Chris@42: 				   TN = cr[WS(rs, 8)];
Chris@42: 				   TQ = ci[WS(rs, 8)];
Chris@42: 				   TM = W[14];
Chris@42: 				   T1n = TA * TE;
Chris@42: 				   TC = TA * TB;
Chris@42: 				   TP = W[15];
Chris@42: 				   T1j = TM * TQ;
Chris@42: 				   TO = TM * TN;
Chris@42: 				   TH = cr[WS(rs, 5)];
Chris@42: 				   TK = ci[WS(rs, 5)];
Chris@42: 				   T1k = FNMS(TP, TN, T1j);
Chris@42: 				   TR = FMA(TP, TQ, TO);
Chris@42: 				   TG = W[8];
Chris@42: 				   TJ = W[9];
Chris@42: 			      }
Chris@42: 			      TD = W[3];
Chris@42: 			 }
Chris@42: 		    }
Chris@42: 		    {
Chris@42: 			 E TV, Tf, T21, T1R, T1l, T1r, T1q, T1N, TT, T1g;
Chris@42: 			 {
Chris@42: 			      E T1o, TF, T1i, TL, T1h, TI, TS, T1p;
Chris@42: 			      TV = FNMS(KP500000000, Te, T1);
Chris@42: 			      Tf = T1 + Te;
Chris@42: 			      T1h = TG * TK;
Chris@42: 			      TI = TG * TH;
Chris@42: 			      T1o = FNMS(TD, TB, T1n);
Chris@42: 			      TF = FMA(TD, TE, TC);
Chris@42: 			      T1i = FNMS(TJ, TH, T1h);
Chris@42: 			      TL = FMA(TJ, TK, TI);
Chris@42: 			      T21 = T1Q + T1P;
Chris@42: 			      T1R = FNMS(KP500000000, T1Q, T1P);
Chris@42: 			      T1p = T1i + T1k;
Chris@42: 			      T1l = T1i - T1k;
Chris@42: 			      TS = TL + TR;
Chris@42: 			      T1r = TR - TL;
Chris@42: 			      T1q = FNMS(KP500000000, T1p, T1o);
Chris@42: 			      T1N = T1o + T1p;
Chris@42: 			      TT = TF + TS;
Chris@42: 			      T1g = FNMS(KP500000000, TS, TF);
Chris@42: 			 }
Chris@42: 			 {
Chris@42: 			      E T11, T1z, T1E, T1D, T1X, T1T, T1I, T1C, T1Y, T1y, T1u, T24, TU;
Chris@42: 			      T24 = TT - Tz;
Chris@42: 			      TU = Tz + TT;
Chris@42: 			      {
Chris@42: 				   E T22, T1O, T1L, T23;
Chris@42: 				   T22 = T1M + T1N;
Chris@42: 				   T1O = T1M - T1N;
Chris@42: 				   T11 = FNMS(KP866025403, T10, TV);
Chris@42: 				   T1z = FMA(KP866025403, T10, TV);
Chris@42: 				   T1L = FNMS(KP500000000, TU, Tf);
Chris@42: 				   cr[0] = Tf + TU;
Chris@42: 				   T23 = FNMS(KP500000000, T22, T21);
Chris@42: 				   ci[WS(rs, 8)] = T22 + T21;
Chris@42: 				   cr[WS(rs, 3)] = FMA(KP866025403, T1O, T1L);
Chris@42: 				   ci[WS(rs, 2)] = FNMS(KP866025403, T1O, T1L);
Chris@42: 				   ci[WS(rs, 5)] = FMA(KP866025403, T24, T23);
Chris@42: 				   cr[WS(rs, 6)] = FMS(KP866025403, T24, T23);
Chris@42: 			      }
Chris@42: 			      {
Chris@42: 				   E T1B, T1m, T1w, T1f, T1s, T1A, T1b, T1e, T1x, T1t;
Chris@42: 				   T1E = FNMS(KP866025403, T1a, T19);
Chris@42: 				   T1b = FMA(KP866025403, T1a, T19);
Chris@42: 				   T1e = FNMS(KP866025403, T1d, T1c);
Chris@42: 				   T1D = FMA(KP866025403, T1d, T1c);
Chris@42: 				   T1B = FMA(KP866025403, T1l, T1g);
Chris@42: 				   T1m = FNMS(KP866025403, T1l, T1g);
Chris@42: 				   T1X = FNMS(KP866025403, T1S, T1R);
Chris@42: 				   T1T = FMA(KP866025403, T1S, T1R);
Chris@42: 				   T1w = FNMS(KP176326980, T1b, T1e);
Chris@42: 				   T1f = FMA(KP176326980, T1e, T1b);
Chris@42: 				   T1s = FNMS(KP866025403, T1r, T1q);
Chris@42: 				   T1A = FMA(KP866025403, T1r, T1q);
Chris@42: 				   T1x = FMA(KP363970234, T1m, T1s);
Chris@42: 				   T1t = FNMS(KP363970234, T1s, T1m);
Chris@42: 				   T1I = FNMS(KP176326980, T1A, T1B);
Chris@42: 				   T1C = FMA(KP176326980, T1B, T1A);
Chris@42: 				   T1Y = FMA(KP954188894, T1x, T1w);
Chris@42: 				   T1y = FNMS(KP954188894, T1x, T1w);
Chris@42: 				   T20 = FMA(KP954188894, T1t, T1f);
Chris@42: 				   T1u = FNMS(KP954188894, T1t, T1f);
Chris@42: 			      }
Chris@42: 			      {
Chris@42: 				   E T1F, T1J, T1v, T1U, T1K;
Chris@42: 				   ci[WS(rs, 6)] = FNMS(KP984807753, T1Y, T1X);
Chris@42: 				   T1v = FNMS(KP492403876, T1u, T11);
Chris@42: 				   cr[WS(rs, 2)] = FMA(KP984807753, T1u, T11);
Chris@42: 				   T1F = FMA(KP839099631, T1E, T1D);
Chris@42: 				   T1J = FNMS(KP839099631, T1D, T1E);
Chris@42: 				   ci[WS(rs, 3)] = FNMS(KP852868531, T1y, T1v);
Chris@42: 				   ci[0] = FMA(KP852868531, T1y, T1v);
Chris@42: 				   T1U = FNMS(KP777861913, T1J, T1I);
Chris@42: 				   T1K = FMA(KP777861913, T1J, T1I);
Chris@42: 				   {
Chris@42: 					E T1G, T1W, T1V, T1H;
Chris@42: 					T1G = FMA(KP777861913, T1F, T1C);
Chris@42: 					T1W = FNMS(KP777861913, T1F, T1C);
Chris@42: 					T1Z = FMA(KP492403876, T1Y, T1X);
Chris@42: 					T1V = FMA(KP492403876, T1U, T1T);
Chris@42: 					ci[WS(rs, 7)] = FNMS(KP984807753, T1U, T1T);
Chris@42: 					T1H = FNMS(KP492403876, T1G, T1z);
Chris@42: 					cr[WS(rs, 1)] = FMA(KP984807753, T1G, T1z);
Chris@42: 					ci[WS(rs, 4)] = FMA(KP852868531, T1W, T1V);
Chris@42: 					cr[WS(rs, 7)] = FMS(KP852868531, T1W, T1V);
Chris@42: 					cr[WS(rs, 4)] = FMA(KP852868531, T1K, T1H);
Chris@42: 					ci[WS(rs, 1)] = FNMS(KP852868531, T1K, T1H);
Chris@42: 				   }
Chris@42: 			      }
Chris@42: 			 }
Chris@42: 		    }
Chris@42: 	       }
Chris@42: 	       cr[WS(rs, 8)] = -(FMA(KP852868531, T20, T1Z));
Chris@42: 	       cr[WS(rs, 5)] = FMS(KP852868531, T20, T1Z);
Chris@42: 	  }
Chris@42:      }
Chris@42: }
Chris@42: 
Chris@42: static const tw_instr twinstr[] = {
Chris@42:      {TW_FULL, 1, 9},
Chris@42:      {TW_NEXT, 1, 0}
Chris@42: };
Chris@42: 
Chris@42: static const hc2hc_desc desc = { 9, "hf_9", twinstr, &GENUS, {24, 16, 72, 0} };
Chris@42: 
Chris@42: void X(codelet_hf_9) (planner *p) {
Chris@42:      X(khc2hc_register) (p, hf_9, &desc);
Chris@42: }
Chris@42: #else				/* HAVE_FMA */
Chris@42: 
Chris@42: /* Generated by: ../../../genfft/gen_hc2hc.native -compact -variables 4 -pipeline-latency 4 -n 9 -dit -name hf_9 -include hf.h */
Chris@42: 
Chris@42: /*
Chris@42:  * This function contains 96 FP additions, 72 FP multiplications,
Chris@42:  * (or, 60 additions, 36 multiplications, 36 fused multiply/add),
Chris@42:  * 41 stack variables, 8 constants, and 36 memory accesses
Chris@42:  */
Chris@42: #include "hf.h"
Chris@42: 
Chris@42: static void hf_9(R *cr, R *ci, const R *W, stride rs, INT mb, INT me, INT ms)
Chris@42: {
Chris@42:      DK(KP642787609, +0.642787609686539326322643409907263432907559884);
Chris@42:      DK(KP766044443, +0.766044443118978035202392650555416673935832457);
Chris@42:      DK(KP939692620, +0.939692620785908384054109277324731469936208134);
Chris@42:      DK(KP342020143, +0.342020143325668733044099614682259580763083368);
Chris@42:      DK(KP984807753, +0.984807753012208059366743024589523013670643252);
Chris@42:      DK(KP173648177, +0.173648177666930348851716626769314796000375677);
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) * 16); m < me; m = m + 1, cr = cr + ms, ci = ci - ms, W = W + 16, MAKE_VOLATILE_STRIDE(18, rs)) {
Chris@42: 	       E T1, T1B, TQ, T1A, Tc, TN, T1C, T1D, TL, T1x, T19, T1o, T1c, T1n, Tu;
Chris@42: 	       E T1w, TW, T1k, T11, T1l;
Chris@42: 	       {
Chris@42: 		    E T6, TO, Tb, TP;
Chris@42: 		    T1 = cr[0];
Chris@42: 		    T1B = ci[0];
Chris@42: 		    {
Chris@42: 			 E T3, T5, T2, T4;
Chris@42: 			 T3 = cr[WS(rs, 3)];
Chris@42: 			 T5 = ci[WS(rs, 3)];
Chris@42: 			 T2 = W[4];
Chris@42: 			 T4 = W[5];
Chris@42: 			 T6 = FMA(T2, T3, T4 * T5);
Chris@42: 			 TO = FNMS(T4, T3, T2 * T5);
Chris@42: 		    }
Chris@42: 		    {
Chris@42: 			 E T8, Ta, T7, T9;
Chris@42: 			 T8 = cr[WS(rs, 6)];
Chris@42: 			 Ta = ci[WS(rs, 6)];
Chris@42: 			 T7 = W[10];
Chris@42: 			 T9 = W[11];
Chris@42: 			 Tb = FMA(T7, T8, T9 * Ta);
Chris@42: 			 TP = FNMS(T9, T8, T7 * Ta);
Chris@42: 		    }
Chris@42: 		    TQ = KP866025403 * (TO - TP);
Chris@42: 		    T1A = KP866025403 * (Tb - T6);
Chris@42: 		    Tc = T6 + Tb;
Chris@42: 		    TN = FNMS(KP500000000, Tc, T1);
Chris@42: 		    T1C = TO + TP;
Chris@42: 		    T1D = FNMS(KP500000000, T1C, T1B);
Chris@42: 	       }
Chris@42: 	       {
Chris@42: 		    E Tz, T13, TE, T14, TJ, T15, TK, T16;
Chris@42: 		    {
Chris@42: 			 E Tw, Ty, Tv, Tx;
Chris@42: 			 Tw = cr[WS(rs, 2)];
Chris@42: 			 Ty = ci[WS(rs, 2)];
Chris@42: 			 Tv = W[2];
Chris@42: 			 Tx = W[3];
Chris@42: 			 Tz = FMA(Tv, Tw, Tx * Ty);
Chris@42: 			 T13 = FNMS(Tx, Tw, Tv * Ty);
Chris@42: 		    }
Chris@42: 		    {
Chris@42: 			 E TB, TD, TA, TC;
Chris@42: 			 TB = cr[WS(rs, 5)];
Chris@42: 			 TD = ci[WS(rs, 5)];
Chris@42: 			 TA = W[8];
Chris@42: 			 TC = W[9];
Chris@42: 			 TE = FMA(TA, TB, TC * TD);
Chris@42: 			 T14 = FNMS(TC, TB, TA * TD);
Chris@42: 		    }
Chris@42: 		    {
Chris@42: 			 E TG, TI, TF, TH;
Chris@42: 			 TG = cr[WS(rs, 8)];
Chris@42: 			 TI = ci[WS(rs, 8)];
Chris@42: 			 TF = W[14];
Chris@42: 			 TH = W[15];
Chris@42: 			 TJ = FMA(TF, TG, TH * TI);
Chris@42: 			 T15 = FNMS(TH, TG, TF * TI);
Chris@42: 		    }
Chris@42: 		    TK = TE + TJ;
Chris@42: 		    T16 = T14 + T15;
Chris@42: 		    TL = Tz + TK;
Chris@42: 		    T1x = T13 + T16;
Chris@42: 		    {
Chris@42: 			 E T17, T18, T1a, T1b;
Chris@42: 			 T17 = FNMS(KP500000000, T16, T13);
Chris@42: 			 T18 = KP866025403 * (TJ - TE);
Chris@42: 			 T19 = T17 - T18;
Chris@42: 			 T1o = T18 + T17;
Chris@42: 			 T1a = FNMS(KP500000000, TK, Tz);
Chris@42: 			 T1b = KP866025403 * (T14 - T15);
Chris@42: 			 T1c = T1a - T1b;
Chris@42: 			 T1n = T1a + T1b;
Chris@42: 		    }
Chris@42: 	       }
Chris@42: 	       {
Chris@42: 		    E Ti, TX, Tn, TT, Ts, TU, Tt, TY;
Chris@42: 		    {
Chris@42: 			 E Tf, Th, Te, Tg;
Chris@42: 			 Tf = cr[WS(rs, 1)];
Chris@42: 			 Th = ci[WS(rs, 1)];
Chris@42: 			 Te = W[0];
Chris@42: 			 Tg = W[1];
Chris@42: 			 Ti = FMA(Te, Tf, Tg * Th);
Chris@42: 			 TX = FNMS(Tg, Tf, Te * Th);
Chris@42: 		    }
Chris@42: 		    {
Chris@42: 			 E Tk, Tm, Tj, Tl;
Chris@42: 			 Tk = cr[WS(rs, 4)];
Chris@42: 			 Tm = ci[WS(rs, 4)];
Chris@42: 			 Tj = W[6];
Chris@42: 			 Tl = W[7];
Chris@42: 			 Tn = FMA(Tj, Tk, Tl * Tm);
Chris@42: 			 TT = FNMS(Tl, Tk, Tj * Tm);
Chris@42: 		    }
Chris@42: 		    {
Chris@42: 			 E Tp, Tr, To, Tq;
Chris@42: 			 Tp = cr[WS(rs, 7)];
Chris@42: 			 Tr = ci[WS(rs, 7)];
Chris@42: 			 To = W[12];
Chris@42: 			 Tq = W[13];
Chris@42: 			 Ts = FMA(To, Tp, Tq * Tr);
Chris@42: 			 TU = FNMS(Tq, Tp, To * Tr);
Chris@42: 		    }
Chris@42: 		    Tt = Tn + Ts;
Chris@42: 		    TY = TT + TU;
Chris@42: 		    Tu = Ti + Tt;
Chris@42: 		    T1w = TX + TY;
Chris@42: 		    {
Chris@42: 			 E TS, TV, TZ, T10;
Chris@42: 			 TS = FNMS(KP500000000, Tt, Ti);
Chris@42: 			 TV = KP866025403 * (TT - TU);
Chris@42: 			 TW = TS - TV;
Chris@42: 			 T1k = TS + TV;
Chris@42: 			 TZ = FNMS(KP500000000, TY, TX);
Chris@42: 			 T10 = KP866025403 * (Ts - Tn);
Chris@42: 			 T11 = TZ - T10;
Chris@42: 			 T1l = T10 + TZ;
Chris@42: 		    }
Chris@42: 	       }
Chris@42: 	       {
Chris@42: 		    E T1y, Td, TM, T1v;
Chris@42: 		    T1y = KP866025403 * (T1w - T1x);
Chris@42: 		    Td = T1 + Tc;
Chris@42: 		    TM = Tu + TL;
Chris@42: 		    T1v = FNMS(KP500000000, TM, Td);
Chris@42: 		    cr[0] = Td + TM;
Chris@42: 		    cr[WS(rs, 3)] = T1v + T1y;
Chris@42: 		    ci[WS(rs, 2)] = T1v - T1y;
Chris@42: 	       }
Chris@42: 	       {
Chris@42: 		    E TR, T1I, T1e, T1K, T1i, T1H, T1f, T1J;
Chris@42: 		    TR = TN - TQ;
Chris@42: 		    T1I = T1D - T1A;
Chris@42: 		    {
Chris@42: 			 E T12, T1d, T1g, T1h;
Chris@42: 			 T12 = FMA(KP173648177, TW, KP984807753 * T11);
Chris@42: 			 T1d = FNMS(KP939692620, T1c, KP342020143 * T19);
Chris@42: 			 T1e = T12 + T1d;
Chris@42: 			 T1K = KP866025403 * (T1d - T12);
Chris@42: 			 T1g = FNMS(KP984807753, TW, KP173648177 * T11);
Chris@42: 			 T1h = FMA(KP342020143, T1c, KP939692620 * T19);
Chris@42: 			 T1i = KP866025403 * (T1g + T1h);
Chris@42: 			 T1H = T1g - T1h;
Chris@42: 		    }
Chris@42: 		    cr[WS(rs, 2)] = TR + T1e;
Chris@42: 		    ci[WS(rs, 6)] = T1H + T1I;
Chris@42: 		    T1f = FNMS(KP500000000, T1e, TR);
Chris@42: 		    ci[0] = T1f - T1i;
Chris@42: 		    ci[WS(rs, 3)] = T1f + T1i;
Chris@42: 		    T1J = FMS(KP500000000, T1H, T1I);
Chris@42: 		    cr[WS(rs, 5)] = T1J - T1K;
Chris@42: 		    cr[WS(rs, 8)] = T1K + T1J;
Chris@42: 	       }
Chris@42: 	       {
Chris@42: 		    E T1L, T1M, T1N, T1O;
Chris@42: 		    T1L = KP866025403 * (TL - Tu);
Chris@42: 		    T1M = T1C + T1B;
Chris@42: 		    T1N = T1w + T1x;
Chris@42: 		    T1O = FNMS(KP500000000, T1N, T1M);
Chris@42: 		    cr[WS(rs, 6)] = T1L - T1O;
Chris@42: 		    ci[WS(rs, 8)] = T1N + T1M;
Chris@42: 		    ci[WS(rs, 5)] = T1L + T1O;
Chris@42: 	       }
Chris@42: 	       {
Chris@42: 		    E T1j, T1E, T1q, T1z, T1u, T1F, T1r, T1G;
Chris@42: 		    T1j = TN + TQ;
Chris@42: 		    T1E = T1A + T1D;
Chris@42: 		    {
Chris@42: 			 E T1m, T1p, T1s, T1t;
Chris@42: 			 T1m = FMA(KP766044443, T1k, KP642787609 * T1l);
Chris@42: 			 T1p = FMA(KP173648177, T1n, KP984807753 * T1o);
Chris@42: 			 T1q = T1m + T1p;
Chris@42: 			 T1z = KP866025403 * (T1p - T1m);
Chris@42: 			 T1s = FNMS(KP642787609, T1k, KP766044443 * T1l);
Chris@42: 			 T1t = FNMS(KP984807753, T1n, KP173648177 * T1o);
Chris@42: 			 T1u = KP866025403 * (T1s - T1t);
Chris@42: 			 T1F = T1s + T1t;
Chris@42: 		    }
Chris@42: 		    cr[WS(rs, 1)] = T1j + T1q;
Chris@42: 		    T1r = FNMS(KP500000000, T1q, T1j);
Chris@42: 		    ci[WS(rs, 1)] = T1r - T1u;
Chris@42: 		    cr[WS(rs, 4)] = T1r + T1u;
Chris@42: 		    ci[WS(rs, 7)] = T1F + T1E;
Chris@42: 		    T1G = FNMS(KP500000000, T1F, T1E);
Chris@42: 		    cr[WS(rs, 7)] = T1z - T1G;
Chris@42: 		    ci[WS(rs, 4)] = T1z + T1G;
Chris@42: 	       }
Chris@42: 	  }
Chris@42:      }
Chris@42: }
Chris@42: 
Chris@42: static const tw_instr twinstr[] = {
Chris@42:      {TW_FULL, 1, 9},
Chris@42:      {TW_NEXT, 1, 0}
Chris@42: };
Chris@42: 
Chris@42: static const hc2hc_desc desc = { 9, "hf_9", twinstr, &GENUS, {60, 36, 36, 0} };
Chris@42: 
Chris@42: void X(codelet_hf_9) (planner *p) {
Chris@42:      X(khc2hc_register) (p, hf_9, &desc);
Chris@42: }
Chris@42: #endif				/* HAVE_FMA */