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 07:41:13 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 -sign 1 -n 10 -dif -name hb_10 -include hb.h */
Chris@10: 
Chris@10: /*
Chris@10:  * This function contains 102 FP additions, 72 FP multiplications,
Chris@10:  * (or, 48 additions, 18 multiplications, 54 fused multiply/add),
Chris@10:  * 71 stack variables, 4 constants, and 40 memory accesses
Chris@10:  */
Chris@10: #include "hb.h"
Chris@10: 
Chris@10: static void hb_10(R *cr, R *ci, const R *W, stride rs, INT mb, INT me, INT ms)
Chris@10: {
Chris@10:      DK(KP951056516, +0.951056516295153572116439333379382143405698634);
Chris@10:      DK(KP559016994, +0.559016994374947424102293417182819058860154590);
Chris@10:      DK(KP250000000, +0.250000000000000000000000000000000000000000000);
Chris@10:      DK(KP618033988, +0.618033988749894848204586834365638117720309180);
Chris@10:      {
Chris@10: 	  INT m;
Chris@10: 	  for (m = mb, W = W + ((mb - 1) * 18); m < me; m = m + 1, cr = cr + ms, ci = ci - ms, W = W + 18, MAKE_VOLATILE_STRIDE(20, rs)) {
Chris@10: 	       E T21, T1Y, T1X;
Chris@10: 	       {
Chris@10: 		    E T1B, TH, T1g, T3, T1V, T1x, T1G, T1E, TM, TK, T11, TB, T7, T1m, T1J;
Chris@10: 		    E TO, Th, T1h, T6, T8, TF, TG, T1i, T9;
Chris@10: 		    TF = ci[WS(rs, 9)];
Chris@10: 		    TG = cr[WS(rs, 5)];
Chris@10: 		    {
Chris@10: 			 E T1u, Tp, Tu, T1s, Tz, T1v, Ts, Tv;
Chris@10: 			 {
Chris@10: 			      E Tx, Ty, Tn, To, Tq, Tr;
Chris@10: 			      Tn = ci[WS(rs, 5)];
Chris@10: 			      To = cr[WS(rs, 9)];
Chris@10: 			      Tx = ci[WS(rs, 6)];
Chris@10: 			      T1B = TF + TG;
Chris@10: 			      TH = TF - TG;
Chris@10: 			      T1u = Tn + To;
Chris@10: 			      Tp = Tn - To;
Chris@10: 			      Ty = cr[WS(rs, 8)];
Chris@10: 			      Tq = ci[WS(rs, 8)];
Chris@10: 			      Tr = cr[WS(rs, 6)];
Chris@10: 			      Tu = ci[WS(rs, 7)];
Chris@10: 			      T1s = Tx + Ty;
Chris@10: 			      Tz = Tx - Ty;
Chris@10: 			      T1v = Tq + Tr;
Chris@10: 			      Ts = Tq - Tr;
Chris@10: 			      Tv = cr[WS(rs, 7)];
Chris@10: 			 }
Chris@10: 			 {
Chris@10: 			      E T1, T1w, T1D, TJ, Tt, T1r, Tw, T2;
Chris@10: 			      T1 = cr[0];
Chris@10: 			      T1w = T1u + T1v;
Chris@10: 			      T1D = T1u - T1v;
Chris@10: 			      TJ = Tp + Ts;
Chris@10: 			      Tt = Tp - Ts;
Chris@10: 			      T1r = Tu + Tv;
Chris@10: 			      Tw = Tu - Tv;
Chris@10: 			      T2 = ci[WS(rs, 4)];
Chris@10: 			      {
Chris@10: 				   E Tb, Tc, Te, Tf;
Chris@10: 				   Tb = cr[WS(rs, 4)];
Chris@10: 				   {
Chris@10: 					E T1t, T1C, TI, TA;
Chris@10: 					T1t = T1r + T1s;
Chris@10: 					T1C = T1r - T1s;
Chris@10: 					TI = Tw + Tz;
Chris@10: 					TA = Tw - Tz;
Chris@10: 					T1g = T1 - T2;
Chris@10: 					T3 = T1 + T2;
Chris@10: 					T1V = FNMS(KP618033988, T1t, T1w);
Chris@10: 					T1x = FMA(KP618033988, T1w, T1t);
Chris@10: 					T1G = T1C - T1D;
Chris@10: 					T1E = T1C + T1D;
Chris@10: 					TM = TI - TJ;
Chris@10: 					TK = TI + TJ;
Chris@10: 					T11 = FMA(KP618033988, Tt, TA);
Chris@10: 					TB = FNMS(KP618033988, TA, Tt);
Chris@10: 					Tc = ci[0];
Chris@10: 				   }
Chris@10: 				   Te = ci[WS(rs, 3)];
Chris@10: 				   Tf = cr[WS(rs, 1)];
Chris@10: 				   {
Chris@10: 					E T4, T1k, Td, T1l, Tg, T5;
Chris@10: 					T4 = cr[WS(rs, 2)];
Chris@10: 					T1k = Tb - Tc;
Chris@10: 					Td = Tb + Tc;
Chris@10: 					T1l = Te - Tf;
Chris@10: 					Tg = Te + Tf;
Chris@10: 					T5 = ci[WS(rs, 2)];
Chris@10: 					T7 = ci[WS(rs, 1)];
Chris@10: 					T1m = T1k + T1l;
Chris@10: 					T1J = T1k - T1l;
Chris@10: 					TO = Td - Tg;
Chris@10: 					Th = Td + Tg;
Chris@10: 					T1h = T4 - T5;
Chris@10: 					T6 = T4 + T5;
Chris@10: 					T8 = cr[WS(rs, 3)];
Chris@10: 				   }
Chris@10: 			      }
Chris@10: 			 }
Chris@10: 		    }
Chris@10: 		    ci[0] = TH + TK;
Chris@10: 		    T1i = T7 - T8;
Chris@10: 		    T9 = T7 + T8;
Chris@10: 		    {
Chris@10: 			 E T2d, T1F, T29, T1I, TP, T2c, T1p, Tl, T1o, Tk, T2b, T2e, T17, T14, T13;
Chris@10: 			 T2d = T1B + T1E;
Chris@10: 			 T1F = FNMS(KP250000000, T1E, T1B);
Chris@10: 			 {
Chris@10: 			      E T1j, Ta, T1n, Ti, T2a;
Chris@10: 			      T29 = W[8];
Chris@10: 			      T1I = T1h - T1i;
Chris@10: 			      T1j = T1h + T1i;
Chris@10: 			      TP = T6 - T9;
Chris@10: 			      Ta = T6 + T9;
Chris@10: 			      T2c = W[9];
Chris@10: 			      T1p = T1j - T1m;
Chris@10: 			      T1n = T1j + T1m;
Chris@10: 			      Tl = Ta - Th;
Chris@10: 			      Ti = Ta + Th;
Chris@10: 			      T1o = FNMS(KP250000000, T1n, T1g);
Chris@10: 			      T2a = T1g + T1n;
Chris@10: 			      cr[0] = T3 + Ti;
Chris@10: 			      Tk = FNMS(KP250000000, Ti, T3);
Chris@10: 			      T2b = T29 * T2a;
Chris@10: 			      T2e = T2c * T2a;
Chris@10: 			 }
Chris@10: 			 {
Chris@10: 			      E T16, TQ, T10, Tm, TL;
Chris@10: 			      T16 = FMA(KP618033988, TO, TP);
Chris@10: 			      TQ = FNMS(KP618033988, TP, TO);
Chris@10: 			      cr[WS(rs, 5)] = FNMS(T2c, T2d, T2b);
Chris@10: 			      ci[WS(rs, 5)] = FMA(T29, T2d, T2e);
Chris@10: 			      T10 = FMA(KP559016994, Tl, Tk);
Chris@10: 			      Tm = FNMS(KP559016994, Tl, Tk);
Chris@10: 			      TL = FNMS(KP250000000, TK, TH);
Chris@10: 			      {
Chris@10: 				   E TE, TU, T12, TR, TX, T1d, T1c, T19, TD, T1e, T1b, TW, TT;
Chris@10: 				   {
Chris@10: 					E TC, T15, T1a, TS, Tj, TN;
Chris@10: 					TE = W[3];
Chris@10: 					TC = FMA(KP951056516, TB, Tm);
Chris@10: 					TU = FNMS(KP951056516, TB, Tm);
Chris@10: 					TN = FNMS(KP559016994, TM, TL);
Chris@10: 					T15 = FMA(KP559016994, TM, TL);
Chris@10: 					T12 = FMA(KP951056516, T11, T10);
Chris@10: 					T1a = FNMS(KP951056516, T11, T10);
Chris@10: 					TS = TE * TC;
Chris@10: 					TR = FNMS(KP951056516, TQ, TN);
Chris@10: 					TX = FMA(KP951056516, TQ, TN);
Chris@10: 					Tj = W[2];
Chris@10: 					T1d = FMA(KP951056516, T16, T15);
Chris@10: 					T17 = FNMS(KP951056516, T16, T15);
Chris@10: 					T1c = W[11];
Chris@10: 					T19 = W[10];
Chris@10: 					ci[WS(rs, 2)] = FMA(Tj, TR, TS);
Chris@10: 					TD = Tj * TC;
Chris@10: 					T1e = T1c * T1a;
Chris@10: 					T1b = T19 * T1a;
Chris@10: 				   }
Chris@10: 				   cr[WS(rs, 2)] = FNMS(TE, TR, TD);
Chris@10: 				   ci[WS(rs, 6)] = FMA(T19, T1d, T1e);
Chris@10: 				   cr[WS(rs, 6)] = FNMS(T1c, T1d, T1b);
Chris@10: 				   TW = W[15];
Chris@10: 				   TT = W[14];
Chris@10: 				   {
Chris@10: 					E TZ, T18, TY, TV;
Chris@10: 					T14 = W[7];
Chris@10: 					TY = TW * TU;
Chris@10: 					TV = TT * TU;
Chris@10: 					TZ = W[6];
Chris@10: 					T18 = T14 * T12;
Chris@10: 					ci[WS(rs, 8)] = FMA(TT, TX, TY);
Chris@10: 					cr[WS(rs, 8)] = FNMS(TW, TX, TV);
Chris@10: 					T13 = TZ * T12;
Chris@10: 					ci[WS(rs, 4)] = FMA(TZ, T17, T18);
Chris@10: 				   }
Chris@10: 			      }
Chris@10: 			 }
Chris@10: 			 {
Chris@10: 			      E T20, T1K, T1q, T1U;
Chris@10: 			      T20 = FNMS(KP618033988, T1I, T1J);
Chris@10: 			      T1K = FMA(KP618033988, T1J, T1I);
Chris@10: 			      cr[WS(rs, 4)] = FNMS(T14, T17, T13);
Chris@10: 			      T1q = FMA(KP559016994, T1p, T1o);
Chris@10: 			      T1U = FNMS(KP559016994, T1p, T1o);
Chris@10: 			      {
Chris@10: 				   E T1A, T1O, T1W, T1R, T1L, T27, T26, T23, T1z, T28, T25, T1Q, T1N;
Chris@10: 				   {
Chris@10: 					E T1y, T1Z, T24, T1M, T1f, T1H;
Chris@10: 					T1A = W[1];
Chris@10: 					T1O = FMA(KP951056516, T1x, T1q);
Chris@10: 					T1y = FNMS(KP951056516, T1x, T1q);
Chris@10: 					T1Z = FNMS(KP559016994, T1G, T1F);
Chris@10: 					T1H = FMA(KP559016994, T1G, T1F);
Chris@10: 					T24 = FMA(KP951056516, T1V, T1U);
Chris@10: 					T1W = FNMS(KP951056516, T1V, T1U);
Chris@10: 					T1M = T1A * T1y;
Chris@10: 					T1R = FNMS(KP951056516, T1K, T1H);
Chris@10: 					T1L = FMA(KP951056516, T1K, T1H);
Chris@10: 					T1f = W[0];
Chris@10: 					T21 = FMA(KP951056516, T20, T1Z);
Chris@10: 					T27 = FNMS(KP951056516, T20, T1Z);
Chris@10: 					T26 = W[13];
Chris@10: 					T23 = W[12];
Chris@10: 					ci[WS(rs, 1)] = FMA(T1f, T1L, T1M);
Chris@10: 					T1z = T1f * T1y;
Chris@10: 					T28 = T26 * T24;
Chris@10: 					T25 = T23 * T24;
Chris@10: 				   }
Chris@10: 				   cr[WS(rs, 1)] = FNMS(T1A, T1L, T1z);
Chris@10: 				   ci[WS(rs, 7)] = FMA(T23, T27, T28);
Chris@10: 				   cr[WS(rs, 7)] = FNMS(T26, T27, T25);
Chris@10: 				   T1Q = W[17];
Chris@10: 				   T1N = W[16];
Chris@10: 				   {
Chris@10: 					E T1T, T22, T1S, T1P;
Chris@10: 					T1Y = W[5];
Chris@10: 					T1S = T1Q * T1O;
Chris@10: 					T1P = T1N * T1O;
Chris@10: 					T1T = W[4];
Chris@10: 					T22 = T1Y * T1W;
Chris@10: 					ci[WS(rs, 9)] = FMA(T1N, T1R, T1S);
Chris@10: 					cr[WS(rs, 9)] = FNMS(T1Q, T1R, T1P);
Chris@10: 					T1X = T1T * T1W;
Chris@10: 					ci[WS(rs, 3)] = FMA(T1T, T21, T22);
Chris@10: 				   }
Chris@10: 			      }
Chris@10: 			 }
Chris@10: 		    }
Chris@10: 	       }
Chris@10: 	       cr[WS(rs, 3)] = FNMS(T1Y, T21, T1X);
Chris@10: 	  }
Chris@10:      }
Chris@10: }
Chris@10: 
Chris@10: static const tw_instr twinstr[] = {
Chris@10:      {TW_FULL, 1, 10},
Chris@10:      {TW_NEXT, 1, 0}
Chris@10: };
Chris@10: 
Chris@10: static const hc2hc_desc desc = { 10, "hb_10", twinstr, &GENUS, {48, 18, 54, 0} };
Chris@10: 
Chris@10: void X(codelet_hb_10) (planner *p) {
Chris@10:      X(khc2hc_register) (p, hb_10, &desc);
Chris@10: }
Chris@10: #else				/* HAVE_FMA */
Chris@10: 
Chris@10: /* Generated by: ../../../genfft/gen_hc2hc.native -compact -variables 4 -pipeline-latency 4 -sign 1 -n 10 -dif -name hb_10 -include hb.h */
Chris@10: 
Chris@10: /*
Chris@10:  * This function contains 102 FP additions, 60 FP multiplications,
Chris@10:  * (or, 72 additions, 30 multiplications, 30 fused multiply/add),
Chris@10:  * 41 stack variables, 4 constants, and 40 memory accesses
Chris@10:  */
Chris@10: #include "hb.h"
Chris@10: 
Chris@10: static void hb_10(R *cr, R *ci, const R *W, stride rs, INT mb, INT me, INT ms)
Chris@10: {
Chris@10:      DK(KP250000000, +0.250000000000000000000000000000000000000000000);
Chris@10:      DK(KP951056516, +0.951056516295153572116439333379382143405698634);
Chris@10:      DK(KP587785252, +0.587785252292473129168705954639072768597652438);
Chris@10:      DK(KP559016994, +0.559016994374947424102293417182819058860154590);
Chris@10:      {
Chris@10: 	  INT m;
Chris@10: 	  for (m = mb, W = W + ((mb - 1) * 18); m < me; m = m + 1, cr = cr + ms, ci = ci - ms, W = W + 18, MAKE_VOLATILE_STRIDE(20, rs)) {
Chris@10: 	       E T3, T18, TE, TF, T1B, T1A, T1f, T1t, Ti, Tl, TJ, T1i, Tt, TA, T1w;
Chris@10: 	       E T1v, T1p, T1E, TM, TO;
Chris@10: 	       {
Chris@10: 		    E T1, T2, TH, TI;
Chris@10: 		    T1 = cr[0];
Chris@10: 		    T2 = ci[WS(rs, 4)];
Chris@10: 		    T3 = T1 + T2;
Chris@10: 		    T18 = T1 - T2;
Chris@10: 		    {
Chris@10: 			 E T6, T19, Tg, T1d, T9, T1a, Td, T1c;
Chris@10: 			 {
Chris@10: 			      E T4, T5, Te, Tf;
Chris@10: 			      T4 = cr[WS(rs, 2)];
Chris@10: 			      T5 = ci[WS(rs, 2)];
Chris@10: 			      T6 = T4 + T5;
Chris@10: 			      T19 = T4 - T5;
Chris@10: 			      Te = ci[WS(rs, 3)];
Chris@10: 			      Tf = cr[WS(rs, 1)];
Chris@10: 			      Tg = Te + Tf;
Chris@10: 			      T1d = Te - Tf;
Chris@10: 			 }
Chris@10: 			 {
Chris@10: 			      E T7, T8, Tb, Tc;
Chris@10: 			      T7 = ci[WS(rs, 1)];
Chris@10: 			      T8 = cr[WS(rs, 3)];
Chris@10: 			      T9 = T7 + T8;
Chris@10: 			      T1a = T7 - T8;
Chris@10: 			      Tb = cr[WS(rs, 4)];
Chris@10: 			      Tc = ci[0];
Chris@10: 			      Td = Tb + Tc;
Chris@10: 			      T1c = Tb - Tc;
Chris@10: 			 }
Chris@10: 			 TE = T6 - T9;
Chris@10: 			 TF = Td - Tg;
Chris@10: 			 T1B = T1c - T1d;
Chris@10: 			 T1A = T19 - T1a;
Chris@10: 			 {
Chris@10: 			      E T1b, T1e, Ta, Th;
Chris@10: 			      T1b = T19 + T1a;
Chris@10: 			      T1e = T1c + T1d;
Chris@10: 			      T1f = T1b + T1e;
Chris@10: 			      T1t = KP559016994 * (T1b - T1e);
Chris@10: 			      Ta = T6 + T9;
Chris@10: 			      Th = Td + Tg;
Chris@10: 			      Ti = Ta + Th;
Chris@10: 			      Tl = KP559016994 * (Ta - Th);
Chris@10: 			 }
Chris@10: 		    }
Chris@10: 		    TH = ci[WS(rs, 9)];
Chris@10: 		    TI = cr[WS(rs, 5)];
Chris@10: 		    TJ = TH - TI;
Chris@10: 		    T1i = TH + TI;
Chris@10: 		    {
Chris@10: 			 E Tp, T1j, Tz, T1n, Ts, T1k, Tw, T1m;
Chris@10: 			 {
Chris@10: 			      E Tn, To, Tx, Ty;
Chris@10: 			      Tn = ci[WS(rs, 7)];
Chris@10: 			      To = cr[WS(rs, 7)];
Chris@10: 			      Tp = Tn - To;
Chris@10: 			      T1j = Tn + To;
Chris@10: 			      Tx = ci[WS(rs, 8)];
Chris@10: 			      Ty = cr[WS(rs, 6)];
Chris@10: 			      Tz = Tx - Ty;
Chris@10: 			      T1n = Tx + Ty;
Chris@10: 			 }
Chris@10: 			 {
Chris@10: 			      E Tq, Tr, Tu, Tv;
Chris@10: 			      Tq = ci[WS(rs, 6)];
Chris@10: 			      Tr = cr[WS(rs, 8)];
Chris@10: 			      Ts = Tq - Tr;
Chris@10: 			      T1k = Tq + Tr;
Chris@10: 			      Tu = ci[WS(rs, 5)];
Chris@10: 			      Tv = cr[WS(rs, 9)];
Chris@10: 			      Tw = Tu - Tv;
Chris@10: 			      T1m = Tu + Tv;
Chris@10: 			 }
Chris@10: 			 Tt = Tp - Ts;
Chris@10: 			 TA = Tw - Tz;
Chris@10: 			 T1w = T1m + T1n;
Chris@10: 			 T1v = T1j + T1k;
Chris@10: 			 {
Chris@10: 			      E T1l, T1o, TK, TL;
Chris@10: 			      T1l = T1j - T1k;
Chris@10: 			      T1o = T1m - T1n;
Chris@10: 			      T1p = T1l + T1o;
Chris@10: 			      T1E = KP559016994 * (T1l - T1o);
Chris@10: 			      TK = Tp + Ts;
Chris@10: 			      TL = Tw + Tz;
Chris@10: 			      TM = TK + TL;
Chris@10: 			      TO = KP559016994 * (TK - TL);
Chris@10: 			 }
Chris@10: 		    }
Chris@10: 	       }
Chris@10: 	       cr[0] = T3 + Ti;
Chris@10: 	       ci[0] = TJ + TM;
Chris@10: 	       {
Chris@10: 		    E T1g, T1q, T17, T1h;
Chris@10: 		    T1g = T18 + T1f;
Chris@10: 		    T1q = T1i + T1p;
Chris@10: 		    T17 = W[8];
Chris@10: 		    T1h = W[9];
Chris@10: 		    cr[WS(rs, 5)] = FNMS(T1h, T1q, T17 * T1g);
Chris@10: 		    ci[WS(rs, 5)] = FMA(T1h, T1g, T17 * T1q);
Chris@10: 	       }
Chris@10: 	       {
Chris@10: 		    E TB, TG, T11, TX, TP, T10, Tm, TW, TN, Tk;
Chris@10: 		    TB = FNMS(KP951056516, TA, KP587785252 * Tt);
Chris@10: 		    TG = FNMS(KP951056516, TF, KP587785252 * TE);
Chris@10: 		    T11 = FMA(KP951056516, TE, KP587785252 * TF);
Chris@10: 		    TX = FMA(KP951056516, Tt, KP587785252 * TA);
Chris@10: 		    TN = FNMS(KP250000000, TM, TJ);
Chris@10: 		    TP = TN - TO;
Chris@10: 		    T10 = TO + TN;
Chris@10: 		    Tk = FNMS(KP250000000, Ti, T3);
Chris@10: 		    Tm = Tk - Tl;
Chris@10: 		    TW = Tl + Tk;
Chris@10: 		    {
Chris@10: 			 E TC, TQ, Tj, TD;
Chris@10: 			 TC = Tm - TB;
Chris@10: 			 TQ = TG + TP;
Chris@10: 			 Tj = W[2];
Chris@10: 			 TD = W[3];
Chris@10: 			 cr[WS(rs, 2)] = FNMS(TD, TQ, Tj * TC);
Chris@10: 			 ci[WS(rs, 2)] = FMA(TD, TC, Tj * TQ);
Chris@10: 		    }
Chris@10: 		    {
Chris@10: 			 E T14, T16, T13, T15;
Chris@10: 			 T14 = TW - TX;
Chris@10: 			 T16 = T11 + T10;
Chris@10: 			 T13 = W[10];
Chris@10: 			 T15 = W[11];
Chris@10: 			 cr[WS(rs, 6)] = FNMS(T15, T16, T13 * T14);
Chris@10: 			 ci[WS(rs, 6)] = FMA(T15, T14, T13 * T16);
Chris@10: 		    }
Chris@10: 		    {
Chris@10: 			 E TS, TU, TR, TT;
Chris@10: 			 TS = Tm + TB;
Chris@10: 			 TU = TP - TG;
Chris@10: 			 TR = W[14];
Chris@10: 			 TT = W[15];
Chris@10: 			 cr[WS(rs, 8)] = FNMS(TT, TU, TR * TS);
Chris@10: 			 ci[WS(rs, 8)] = FMA(TT, TS, TR * TU);
Chris@10: 		    }
Chris@10: 		    {
Chris@10: 			 E TY, T12, TV, TZ;
Chris@10: 			 TY = TW + TX;
Chris@10: 			 T12 = T10 - T11;
Chris@10: 			 TV = W[6];
Chris@10: 			 TZ = W[7];
Chris@10: 			 cr[WS(rs, 4)] = FNMS(TZ, T12, TV * TY);
Chris@10: 			 ci[WS(rs, 4)] = FMA(TZ, TY, TV * T12);
Chris@10: 		    }
Chris@10: 	       }
Chris@10: 	       {
Chris@10: 		    E T1x, T1C, T1Q, T1N, T1F, T1R, T1u, T1M, T1D, T1s;
Chris@10: 		    T1x = FNMS(KP951056516, T1w, KP587785252 * T1v);
Chris@10: 		    T1C = FNMS(KP951056516, T1B, KP587785252 * T1A);
Chris@10: 		    T1Q = FMA(KP951056516, T1A, KP587785252 * T1B);
Chris@10: 		    T1N = FMA(KP951056516, T1v, KP587785252 * T1w);
Chris@10: 		    T1D = FNMS(KP250000000, T1p, T1i);
Chris@10: 		    T1F = T1D - T1E;
Chris@10: 		    T1R = T1E + T1D;
Chris@10: 		    T1s = FNMS(KP250000000, T1f, T18);
Chris@10: 		    T1u = T1s - T1t;
Chris@10: 		    T1M = T1t + T1s;
Chris@10: 		    {
Chris@10: 			 E T1y, T1G, T1r, T1z;
Chris@10: 			 T1y = T1u - T1x;
Chris@10: 			 T1G = T1C + T1F;
Chris@10: 			 T1r = W[12];
Chris@10: 			 T1z = W[13];
Chris@10: 			 cr[WS(rs, 7)] = FNMS(T1z, T1G, T1r * T1y);
Chris@10: 			 ci[WS(rs, 7)] = FMA(T1r, T1G, T1z * T1y);
Chris@10: 		    }
Chris@10: 		    {
Chris@10: 			 E T1U, T1W, T1T, T1V;
Chris@10: 			 T1U = T1M + T1N;
Chris@10: 			 T1W = T1R - T1Q;
Chris@10: 			 T1T = W[16];
Chris@10: 			 T1V = W[17];
Chris@10: 			 cr[WS(rs, 9)] = FNMS(T1V, T1W, T1T * T1U);
Chris@10: 			 ci[WS(rs, 9)] = FMA(T1T, T1W, T1V * T1U);
Chris@10: 		    }
Chris@10: 		    {
Chris@10: 			 E T1I, T1K, T1H, T1J;
Chris@10: 			 T1I = T1u + T1x;
Chris@10: 			 T1K = T1F - T1C;
Chris@10: 			 T1H = W[4];
Chris@10: 			 T1J = W[5];
Chris@10: 			 cr[WS(rs, 3)] = FNMS(T1J, T1K, T1H * T1I);
Chris@10: 			 ci[WS(rs, 3)] = FMA(T1H, T1K, T1J * T1I);
Chris@10: 		    }
Chris@10: 		    {
Chris@10: 			 E T1O, T1S, T1L, T1P;
Chris@10: 			 T1O = T1M - T1N;
Chris@10: 			 T1S = T1Q + T1R;
Chris@10: 			 T1L = W[0];
Chris@10: 			 T1P = W[1];
Chris@10: 			 cr[WS(rs, 1)] = FNMS(T1P, T1S, T1L * T1O);
Chris@10: 			 ci[WS(rs, 1)] = FMA(T1L, T1S, T1P * T1O);
Chris@10: 		    }
Chris@10: 	       }
Chris@10: 	  }
Chris@10:      }
Chris@10: }
Chris@10: 
Chris@10: static const tw_instr twinstr[] = {
Chris@10:      {TW_FULL, 1, 10},
Chris@10:      {TW_NEXT, 1, 0}
Chris@10: };
Chris@10: 
Chris@10: static const hc2hc_desc desc = { 10, "hb_10", twinstr, &GENUS, {72, 30, 30, 0} };
Chris@10: 
Chris@10: void X(codelet_hb_10) (planner *p) {
Chris@10:      X(khc2hc_register) (p, hb_10, &desc);
Chris@10: }
Chris@10: #endif				/* HAVE_FMA */