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 07:35:49 EST 2012 */
cannam@95: 
cannam@95: #include "codelet-dft.h"
cannam@95: 
cannam@95: #ifdef HAVE_FMA
cannam@95: 
cannam@95: /* Generated by: ../../../genfft/gen_twiddle.native -fma -reorder-insns -schedule-for-pipeline -compact -variables 4 -pipeline-latency 4 -n 9 -name t1_9 -include t.h */
cannam@95: 
cannam@95: /*
cannam@95:  * This function contains 96 FP additions, 88 FP multiplications,
cannam@95:  * (or, 24 additions, 16 multiplications, 72 fused multiply/add),
cannam@95:  * 72 stack variables, 10 constants, and 36 memory accesses
cannam@95:  */
cannam@95: #include "t.h"
cannam@95: 
cannam@95: static void t1_9(R *ri, R *ii, const R *W, stride rs, INT mb, INT me, INT ms)
cannam@95: {
cannam@95:      DK(KP954188894, +0.954188894138671133499268364187245676532219158);
cannam@95:      DK(KP852868531, +0.852868531952443209628250963940074071936020296);
cannam@95:      DK(KP363970234, +0.363970234266202361351047882776834043890471784);
cannam@95:      DK(KP492403876, +0.492403876506104029683371512294761506835321626);
cannam@95:      DK(KP984807753, +0.984807753012208059366743024589523013670643252);
cannam@95:      DK(KP777861913, +0.777861913430206160028177977318626690410586096);
cannam@95:      DK(KP839099631, +0.839099631177280011763127298123181364687434283);
cannam@95:      DK(KP176326980, +0.176326980708464973471090386868618986121633062);
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 * 16); m < me; m = m + 1, ri = ri + ms, ii = ii + ms, W = W + 16, MAKE_VOLATILE_STRIDE(18, rs)) {
cannam@95: 	       E T1K, T24, T1H, T23;
cannam@95: 	       {
cannam@95: 		    E T1, T1R, T1Q, T10, T1W, Te, TB, T1l, T1r, T1q, T1M, TE, T1g, Tz, T12;
cannam@95: 		    E TC, TH, TK, T17, TR, TG, TJ, TD;
cannam@95: 		    T1 = ri[0];
cannam@95: 		    T1R = ii[0];
cannam@95: 		    {
cannam@95: 			 E T9, Tc, TY, Ta, Tb, TX, T7;
cannam@95: 			 {
cannam@95: 			      E T3, T6, T8, TW, T4, T2, T5;
cannam@95: 			      T3 = ri[WS(rs, 3)];
cannam@95: 			      T6 = ii[WS(rs, 3)];
cannam@95: 			      T2 = W[4];
cannam@95: 			      T9 = ri[WS(rs, 6)];
cannam@95: 			      Tc = ii[WS(rs, 6)];
cannam@95: 			      T8 = W[10];
cannam@95: 			      TW = T2 * T6;
cannam@95: 			      T4 = T2 * T3;
cannam@95: 			      T5 = W[5];
cannam@95: 			      TY = T8 * Tc;
cannam@95: 			      Ta = T8 * T9;
cannam@95: 			      Tb = W[11];
cannam@95: 			      TX = FNMS(T5, T3, TW);
cannam@95: 			      T7 = FMA(T5, T6, T4);
cannam@95: 			 }
cannam@95: 			 {
cannam@95: 			      E Th, Tk, Ti, T1n, Tn, Tq, Tp, T1i, Tx, T1j, To, Tj, TZ, Td, Tg;
cannam@95: 			      E TA, Tl, Ty;
cannam@95: 			      Th = ri[WS(rs, 1)];
cannam@95: 			      TZ = FNMS(Tb, T9, TY);
cannam@95: 			      Td = FMA(Tb, Tc, Ta);
cannam@95: 			      Tk = ii[WS(rs, 1)];
cannam@95: 			      Tg = W[0];
cannam@95: 			      T1Q = TX + TZ;
cannam@95: 			      T10 = TX - TZ;
cannam@95: 			      T1W = Td - T7;
cannam@95: 			      Te = T7 + Td;
cannam@95: 			      Ti = Tg * Th;
cannam@95: 			      T1n = Tg * Tk;
cannam@95: 			      {
cannam@95: 				   E Tt, Tw, Ts, Tv, T1h, Tu, Tm;
cannam@95: 				   Tt = ri[WS(rs, 7)];
cannam@95: 				   Tw = ii[WS(rs, 7)];
cannam@95: 				   Ts = W[12];
cannam@95: 				   Tv = W[13];
cannam@95: 				   Tn = ri[WS(rs, 4)];
cannam@95: 				   Tq = ii[WS(rs, 4)];
cannam@95: 				   T1h = Ts * Tw;
cannam@95: 				   Tu = Ts * Tt;
cannam@95: 				   Tm = W[6];
cannam@95: 				   Tp = W[7];
cannam@95: 				   T1i = FNMS(Tv, Tt, T1h);
cannam@95: 				   Tx = FMA(Tv, Tw, Tu);
cannam@95: 				   T1j = Tm * Tq;
cannam@95: 				   To = Tm * Tn;
cannam@95: 			      }
cannam@95: 			      Tj = W[1];
cannam@95: 			      TB = ri[WS(rs, 2)];
cannam@95: 			      {
cannam@95: 				   E T1k, Tr, T1o, T1p;
cannam@95: 				   T1k = FNMS(Tp, Tn, T1j);
cannam@95: 				   Tr = FMA(Tp, Tq, To);
cannam@95: 				   T1o = FNMS(Tj, Th, T1n);
cannam@95: 				   Tl = FMA(Tj, Tk, Ti);
cannam@95: 				   T1p = T1k + T1i;
cannam@95: 				   T1l = T1i - T1k;
cannam@95: 				   Ty = Tr + Tx;
cannam@95: 				   T1r = Tr - Tx;
cannam@95: 				   T1q = FNMS(KP500000000, T1p, T1o);
cannam@95: 				   T1M = T1o + T1p;
cannam@95: 				   TE = ii[WS(rs, 2)];
cannam@95: 			      }
cannam@95: 			      T1g = FNMS(KP500000000, Ty, Tl);
cannam@95: 			      Tz = Tl + Ty;
cannam@95: 			      TA = W[2];
cannam@95: 			      {
cannam@95: 				   E TN, TQ, TP, T16, TO, TM;
cannam@95: 				   TN = ri[WS(rs, 8)];
cannam@95: 				   TQ = ii[WS(rs, 8)];
cannam@95: 				   TM = W[14];
cannam@95: 				   T12 = TA * TE;
cannam@95: 				   TC = TA * TB;
cannam@95: 				   TP = W[15];
cannam@95: 				   T16 = TM * TQ;
cannam@95: 				   TO = TM * TN;
cannam@95: 				   TH = ri[WS(rs, 5)];
cannam@95: 				   TK = ii[WS(rs, 5)];
cannam@95: 				   T17 = FNMS(TP, TN, T16);
cannam@95: 				   TR = FMA(TP, TQ, TO);
cannam@95: 				   TG = W[8];
cannam@95: 				   TJ = W[9];
cannam@95: 			      }
cannam@95: 			      TD = W[3];
cannam@95: 			 }
cannam@95: 		    }
cannam@95: 		    {
cannam@95: 			 E TV, Tf, T1S, T1V, T1d, T1a, T19, T1N, TT, T1c;
cannam@95: 			 {
cannam@95: 			      E T13, TF, T15, TL, T14, TI, TS, T18;
cannam@95: 			      TV = FNMS(KP500000000, Te, T1);
cannam@95: 			      Tf = T1 + Te;
cannam@95: 			      T14 = TG * TK;
cannam@95: 			      TI = TG * TH;
cannam@95: 			      T13 = FNMS(TD, TB, T12);
cannam@95: 			      TF = FMA(TD, TE, TC);
cannam@95: 			      T15 = FNMS(TJ, TH, T14);
cannam@95: 			      TL = FMA(TJ, TK, TI);
cannam@95: 			      T1S = T1Q + T1R;
cannam@95: 			      T1V = FNMS(KP500000000, T1Q, T1R);
cannam@95: 			      T18 = T15 + T17;
cannam@95: 			      T1d = T15 - T17;
cannam@95: 			      TS = TL + TR;
cannam@95: 			      T1a = TR - TL;
cannam@95: 			      T19 = FNMS(KP500000000, T18, T13);
cannam@95: 			      T1N = T13 + T18;
cannam@95: 			      TT = TF + TS;
cannam@95: 			      T1c = FNMS(KP500000000, TS, TF);
cannam@95: 			 }
cannam@95: 			 {
cannam@95: 			      E T11, T1z, T1E, T1D, T21, T1X, T1I, T1C, T1Y, T1y, T20, T1u, T1U, TU;
cannam@95: 			      T1U = TT - Tz;
cannam@95: 			      TU = Tz + TT;
cannam@95: 			      {
cannam@95: 				   E T1P, T1O, T1L, T1T;
cannam@95: 				   T1P = T1M + T1N;
cannam@95: 				   T1O = T1M - T1N;
cannam@95: 				   T11 = FMA(KP866025403, T10, TV);
cannam@95: 				   T1z = FNMS(KP866025403, T10, TV);
cannam@95: 				   T1L = FNMS(KP500000000, TU, Tf);
cannam@95: 				   ri[0] = Tf + TU;
cannam@95: 				   T1T = FNMS(KP500000000, T1P, T1S);
cannam@95: 				   ii[0] = T1P + T1S;
cannam@95: 				   ri[WS(rs, 3)] = FMA(KP866025403, T1O, T1L);
cannam@95: 				   ri[WS(rs, 6)] = FNMS(KP866025403, T1O, T1L);
cannam@95: 				   ii[WS(rs, 6)] = FNMS(KP866025403, T1U, T1T);
cannam@95: 				   ii[WS(rs, 3)] = FMA(KP866025403, T1U, T1T);
cannam@95: 			      }
cannam@95: 			      {
cannam@95: 				   E T1B, T1m, T1w, T1f, T1s, T1A, T1b, T1e, T1x, T1t;
cannam@95: 				   T1E = FNMS(KP866025403, T1a, T19);
cannam@95: 				   T1b = FMA(KP866025403, T1a, T19);
cannam@95: 				   T1e = FMA(KP866025403, T1d, T1c);
cannam@95: 				   T1D = FNMS(KP866025403, T1d, T1c);
cannam@95: 				   T1B = FMA(KP866025403, T1l, T1g);
cannam@95: 				   T1m = FNMS(KP866025403, T1l, T1g);
cannam@95: 				   T21 = FNMS(KP866025403, T1W, T1V);
cannam@95: 				   T1X = FMA(KP866025403, T1W, T1V);
cannam@95: 				   T1w = FNMS(KP176326980, T1b, T1e);
cannam@95: 				   T1f = FMA(KP176326980, T1e, T1b);
cannam@95: 				   T1s = FNMS(KP866025403, T1r, T1q);
cannam@95: 				   T1A = FMA(KP866025403, T1r, T1q);
cannam@95: 				   T1x = FNMS(KP839099631, T1m, T1s);
cannam@95: 				   T1t = FMA(KP839099631, T1s, T1m);
cannam@95: 				   T1I = FNMS(KP176326980, T1A, T1B);
cannam@95: 				   T1C = FMA(KP176326980, T1B, T1A);
cannam@95: 				   T1Y = FNMS(KP777861913, T1x, T1w);
cannam@95: 				   T1y = FMA(KP777861913, T1x, T1w);
cannam@95: 				   T20 = FNMS(KP777861913, T1t, T1f);
cannam@95: 				   T1u = FMA(KP777861913, T1t, T1f);
cannam@95: 			      }
cannam@95: 			      {
cannam@95: 				   E T22, T1G, T1Z, T1F, T1J, T1v;
cannam@95: 				   ii[WS(rs, 1)] = FNMS(KP984807753, T1Y, T1X);
cannam@95: 				   T1v = FNMS(KP492403876, T1u, T11);
cannam@95: 				   ri[WS(rs, 1)] = FMA(KP984807753, T1u, T11);
cannam@95: 				   T1F = FNMS(KP363970234, T1E, T1D);
cannam@95: 				   T1J = FMA(KP363970234, T1D, T1E);
cannam@95: 				   ri[WS(rs, 4)] = FMA(KP852868531, T1y, T1v);
cannam@95: 				   ri[WS(rs, 7)] = FNMS(KP852868531, T1y, T1v);
cannam@95: 				   T1K = FNMS(KP954188894, T1J, T1I);
cannam@95: 				   T22 = FMA(KP954188894, T1J, T1I);
cannam@95: 				   T1G = FNMS(KP954188894, T1F, T1C);
cannam@95: 				   T24 = FMA(KP954188894, T1F, T1C);
cannam@95: 				   T1Z = FMA(KP492403876, T1Y, T1X);
cannam@95: 				   ii[WS(rs, 2)] = FNMS(KP984807753, T22, T21);
cannam@95: 				   ri[WS(rs, 2)] = FMA(KP984807753, T1G, T1z);
cannam@95: 				   T1H = FNMS(KP492403876, T1G, T1z);
cannam@95: 				   ii[WS(rs, 7)] = FNMS(KP852868531, T20, T1Z);
cannam@95: 				   ii[WS(rs, 4)] = FMA(KP852868531, T20, T1Z);
cannam@95: 				   T23 = FMA(KP492403876, T22, T21);
cannam@95: 			      }
cannam@95: 			 }
cannam@95: 		    }
cannam@95: 	       }
cannam@95: 	       ri[WS(rs, 8)] = FMA(KP852868531, T1K, T1H);
cannam@95: 	       ri[WS(rs, 5)] = FNMS(KP852868531, T1K, T1H);
cannam@95: 	       ii[WS(rs, 8)] = FMA(KP852868531, T24, T23);
cannam@95: 	       ii[WS(rs, 5)] = FNMS(KP852868531, T24, T23);
cannam@95: 	  }
cannam@95:      }
cannam@95: }
cannam@95: 
cannam@95: static const tw_instr twinstr[] = {
cannam@95:      {TW_FULL, 0, 9},
cannam@95:      {TW_NEXT, 1, 0}
cannam@95: };
cannam@95: 
cannam@95: static const ct_desc desc = { 9, "t1_9", twinstr, &GENUS, {24, 16, 72, 0}, 0, 0, 0 };
cannam@95: 
cannam@95: void X(codelet_t1_9) (planner *p) {
cannam@95:      X(kdft_dit_register) (p, t1_9, &desc);
cannam@95: }
cannam@95: #else				/* HAVE_FMA */
cannam@95: 
cannam@95: /* Generated by: ../../../genfft/gen_twiddle.native -compact -variables 4 -pipeline-latency 4 -n 9 -name t1_9 -include t.h */
cannam@95: 
cannam@95: /*
cannam@95:  * This function contains 96 FP additions, 72 FP multiplications,
cannam@95:  * (or, 60 additions, 36 multiplications, 36 fused multiply/add),
cannam@95:  * 41 stack variables, 8 constants, and 36 memory accesses
cannam@95:  */
cannam@95: #include "t.h"
cannam@95: 
cannam@95: static void t1_9(R *ri, R *ii, const R *W, stride rs, INT mb, INT me, INT ms)
cannam@95: {
cannam@95:      DK(KP939692620, +0.939692620785908384054109277324731469936208134);
cannam@95:      DK(KP342020143, +0.342020143325668733044099614682259580763083368);
cannam@95:      DK(KP984807753, +0.984807753012208059366743024589523013670643252);
cannam@95:      DK(KP173648177, +0.173648177666930348851716626769314796000375677);
cannam@95:      DK(KP642787609, +0.642787609686539326322643409907263432907559884);
cannam@95:      DK(KP766044443, +0.766044443118978035202392650555416673935832457);
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 * 16); m < me; m = m + 1, ri = ri + ms, ii = ii + ms, W = W + 16, MAKE_VOLATILE_STRIDE(18, rs)) {
cannam@95: 	       E T1, T1B, TQ, T1G, Tc, TN, T1A, T1H, TL, T1x, T17, T1o, T1c, T1n, Tu;
cannam@95: 	       E T1w, TW, T1k, T11, T1l;
cannam@95: 	       {
cannam@95: 		    E T6, TO, Tb, TP;
cannam@95: 		    T1 = ri[0];
cannam@95: 		    T1B = ii[0];
cannam@95: 		    {
cannam@95: 			 E T3, T5, T2, T4;
cannam@95: 			 T3 = ri[WS(rs, 3)];
cannam@95: 			 T5 = ii[WS(rs, 3)];
cannam@95: 			 T2 = W[4];
cannam@95: 			 T4 = W[5];
cannam@95: 			 T6 = FMA(T2, T3, T4 * T5);
cannam@95: 			 TO = FNMS(T4, T3, T2 * T5);
cannam@95: 		    }
cannam@95: 		    {
cannam@95: 			 E T8, Ta, T7, T9;
cannam@95: 			 T8 = ri[WS(rs, 6)];
cannam@95: 			 Ta = ii[WS(rs, 6)];
cannam@95: 			 T7 = W[10];
cannam@95: 			 T9 = W[11];
cannam@95: 			 Tb = FMA(T7, T8, T9 * Ta);
cannam@95: 			 TP = FNMS(T9, T8, T7 * Ta);
cannam@95: 		    }
cannam@95: 		    TQ = KP866025403 * (TO - TP);
cannam@95: 		    T1G = KP866025403 * (Tb - T6);
cannam@95: 		    Tc = T6 + Tb;
cannam@95: 		    TN = FNMS(KP500000000, Tc, T1);
cannam@95: 		    T1A = TO + TP;
cannam@95: 		    T1H = FNMS(KP500000000, T1A, T1B);
cannam@95: 	       }
cannam@95: 	       {
cannam@95: 		    E Tz, T19, TE, T14, TJ, T15, TK, T1a;
cannam@95: 		    {
cannam@95: 			 E Tw, Ty, Tv, Tx;
cannam@95: 			 Tw = ri[WS(rs, 2)];
cannam@95: 			 Ty = ii[WS(rs, 2)];
cannam@95: 			 Tv = W[2];
cannam@95: 			 Tx = W[3];
cannam@95: 			 Tz = FMA(Tv, Tw, Tx * Ty);
cannam@95: 			 T19 = FNMS(Tx, Tw, Tv * Ty);
cannam@95: 		    }
cannam@95: 		    {
cannam@95: 			 E TB, TD, TA, TC;
cannam@95: 			 TB = ri[WS(rs, 5)];
cannam@95: 			 TD = ii[WS(rs, 5)];
cannam@95: 			 TA = W[8];
cannam@95: 			 TC = W[9];
cannam@95: 			 TE = FMA(TA, TB, TC * TD);
cannam@95: 			 T14 = FNMS(TC, TB, TA * TD);
cannam@95: 		    }
cannam@95: 		    {
cannam@95: 			 E TG, TI, TF, TH;
cannam@95: 			 TG = ri[WS(rs, 8)];
cannam@95: 			 TI = ii[WS(rs, 8)];
cannam@95: 			 TF = W[14];
cannam@95: 			 TH = W[15];
cannam@95: 			 TJ = FMA(TF, TG, TH * TI);
cannam@95: 			 T15 = FNMS(TH, TG, TF * TI);
cannam@95: 		    }
cannam@95: 		    TK = TE + TJ;
cannam@95: 		    T1a = T14 + T15;
cannam@95: 		    TL = Tz + TK;
cannam@95: 		    T1x = T19 + T1a;
cannam@95: 		    {
cannam@95: 			 E T13, T16, T18, T1b;
cannam@95: 			 T13 = FNMS(KP500000000, TK, Tz);
cannam@95: 			 T16 = KP866025403 * (T14 - T15);
cannam@95: 			 T17 = T13 + T16;
cannam@95: 			 T1o = T13 - T16;
cannam@95: 			 T18 = KP866025403 * (TJ - TE);
cannam@95: 			 T1b = FNMS(KP500000000, T1a, T19);
cannam@95: 			 T1c = T18 + T1b;
cannam@95: 			 T1n = T1b - T18;
cannam@95: 		    }
cannam@95: 	       }
cannam@95: 	       {
cannam@95: 		    E Ti, TY, Tn, TT, Ts, TU, Tt, TZ;
cannam@95: 		    {
cannam@95: 			 E Tf, Th, Te, Tg;
cannam@95: 			 Tf = ri[WS(rs, 1)];
cannam@95: 			 Th = ii[WS(rs, 1)];
cannam@95: 			 Te = W[0];
cannam@95: 			 Tg = W[1];
cannam@95: 			 Ti = FMA(Te, Tf, Tg * Th);
cannam@95: 			 TY = FNMS(Tg, Tf, Te * Th);
cannam@95: 		    }
cannam@95: 		    {
cannam@95: 			 E Tk, Tm, Tj, Tl;
cannam@95: 			 Tk = ri[WS(rs, 4)];
cannam@95: 			 Tm = ii[WS(rs, 4)];
cannam@95: 			 Tj = W[6];
cannam@95: 			 Tl = W[7];
cannam@95: 			 Tn = FMA(Tj, Tk, Tl * Tm);
cannam@95: 			 TT = FNMS(Tl, Tk, Tj * Tm);
cannam@95: 		    }
cannam@95: 		    {
cannam@95: 			 E Tp, Tr, To, Tq;
cannam@95: 			 Tp = ri[WS(rs, 7)];
cannam@95: 			 Tr = ii[WS(rs, 7)];
cannam@95: 			 To = W[12];
cannam@95: 			 Tq = W[13];
cannam@95: 			 Ts = FMA(To, Tp, Tq * Tr);
cannam@95: 			 TU = FNMS(Tq, Tp, To * Tr);
cannam@95: 		    }
cannam@95: 		    Tt = Tn + Ts;
cannam@95: 		    TZ = TT + TU;
cannam@95: 		    Tu = Ti + Tt;
cannam@95: 		    T1w = TY + TZ;
cannam@95: 		    {
cannam@95: 			 E TS, TV, TX, T10;
cannam@95: 			 TS = FNMS(KP500000000, Tt, Ti);
cannam@95: 			 TV = KP866025403 * (TT - TU);
cannam@95: 			 TW = TS + TV;
cannam@95: 			 T1k = TS - TV;
cannam@95: 			 TX = KP866025403 * (Ts - Tn);
cannam@95: 			 T10 = FNMS(KP500000000, TZ, TY);
cannam@95: 			 T11 = TX + T10;
cannam@95: 			 T1l = T10 - TX;
cannam@95: 		    }
cannam@95: 	       }
cannam@95: 	       {
cannam@95: 		    E T1y, Td, TM, T1v;
cannam@95: 		    T1y = KP866025403 * (T1w - T1x);
cannam@95: 		    Td = T1 + Tc;
cannam@95: 		    TM = Tu + TL;
cannam@95: 		    T1v = FNMS(KP500000000, TM, Td);
cannam@95: 		    ri[0] = Td + TM;
cannam@95: 		    ri[WS(rs, 3)] = T1v + T1y;
cannam@95: 		    ri[WS(rs, 6)] = T1v - T1y;
cannam@95: 	       }
cannam@95: 	       {
cannam@95: 		    E T1D, T1z, T1C, T1E;
cannam@95: 		    T1D = KP866025403 * (TL - Tu);
cannam@95: 		    T1z = T1w + T1x;
cannam@95: 		    T1C = T1A + T1B;
cannam@95: 		    T1E = FNMS(KP500000000, T1z, T1C);
cannam@95: 		    ii[0] = T1z + T1C;
cannam@95: 		    ii[WS(rs, 6)] = T1E - T1D;
cannam@95: 		    ii[WS(rs, 3)] = T1D + T1E;
cannam@95: 	       }
cannam@95: 	       {
cannam@95: 		    E TR, T1I, T1e, T1J, T1i, T1F, T1f, T1K;
cannam@95: 		    TR = TN + TQ;
cannam@95: 		    T1I = T1G + T1H;
cannam@95: 		    {
cannam@95: 			 E T12, T1d, T1g, T1h;
cannam@95: 			 T12 = FMA(KP766044443, TW, KP642787609 * T11);
cannam@95: 			 T1d = FMA(KP173648177, T17, KP984807753 * T1c);
cannam@95: 			 T1e = T12 + T1d;
cannam@95: 			 T1J = KP866025403 * (T1d - T12);
cannam@95: 			 T1g = FNMS(KP642787609, TW, KP766044443 * T11);
cannam@95: 			 T1h = FNMS(KP984807753, T17, KP173648177 * T1c);
cannam@95: 			 T1i = KP866025403 * (T1g - T1h);
cannam@95: 			 T1F = T1g + T1h;
cannam@95: 		    }
cannam@95: 		    ri[WS(rs, 1)] = TR + T1e;
cannam@95: 		    ii[WS(rs, 1)] = T1F + T1I;
cannam@95: 		    T1f = FNMS(KP500000000, T1e, TR);
cannam@95: 		    ri[WS(rs, 7)] = T1f - T1i;
cannam@95: 		    ri[WS(rs, 4)] = T1f + T1i;
cannam@95: 		    T1K = FNMS(KP500000000, T1F, T1I);
cannam@95: 		    ii[WS(rs, 4)] = T1J + T1K;
cannam@95: 		    ii[WS(rs, 7)] = T1K - T1J;
cannam@95: 	       }
cannam@95: 	       {
cannam@95: 		    E T1j, T1M, T1q, T1N, T1u, T1L, T1r, T1O;
cannam@95: 		    T1j = TN - TQ;
cannam@95: 		    T1M = T1H - T1G;
cannam@95: 		    {
cannam@95: 			 E T1m, T1p, T1s, T1t;
cannam@95: 			 T1m = FMA(KP173648177, T1k, KP984807753 * T1l);
cannam@95: 			 T1p = FNMS(KP939692620, T1o, KP342020143 * T1n);
cannam@95: 			 T1q = T1m + T1p;
cannam@95: 			 T1N = KP866025403 * (T1p - T1m);
cannam@95: 			 T1s = FNMS(KP984807753, T1k, KP173648177 * T1l);
cannam@95: 			 T1t = FMA(KP342020143, T1o, KP939692620 * T1n);
cannam@95: 			 T1u = KP866025403 * (T1s + T1t);
cannam@95: 			 T1L = T1s - T1t;
cannam@95: 		    }
cannam@95: 		    ri[WS(rs, 2)] = T1j + T1q;
cannam@95: 		    ii[WS(rs, 2)] = T1L + T1M;
cannam@95: 		    T1r = FNMS(KP500000000, T1q, T1j);
cannam@95: 		    ri[WS(rs, 8)] = T1r - T1u;
cannam@95: 		    ri[WS(rs, 5)] = T1r + T1u;
cannam@95: 		    T1O = FNMS(KP500000000, T1L, T1M);
cannam@95: 		    ii[WS(rs, 5)] = T1N + T1O;
cannam@95: 		    ii[WS(rs, 8)] = T1O - T1N;
cannam@95: 	       }
cannam@95: 	  }
cannam@95:      }
cannam@95: }
cannam@95: 
cannam@95: static const tw_instr twinstr[] = {
cannam@95:      {TW_FULL, 0, 9},
cannam@95:      {TW_NEXT, 1, 0}
cannam@95: };
cannam@95: 
cannam@95: static const ct_desc desc = { 9, "t1_9", twinstr, &GENUS, {60, 36, 36, 0}, 0, 0, 0 };
cannam@95: 
cannam@95: void X(codelet_t1_9) (planner *p) {
cannam@95:      X(kdft_dit_register) (p, t1_9, &desc);
cannam@95: }
cannam@95: #endif				/* HAVE_FMA */