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 07:41:53 EST 2012 */
cannam@95: 
cannam@95: #include "codelet-rdft.h"
cannam@95: 
cannam@95: #ifdef HAVE_FMA
cannam@95: 
cannam@95: /* Generated by: ../../../genfft/gen_hc2c.native -fma -reorder-insns -schedule-for-pipeline -compact -variables 4 -pipeline-latency 4 -sign 1 -n 12 -dif -name hc2cb_12 -include hc2cb.h */
cannam@95: 
cannam@95: /*
cannam@95:  * This function contains 118 FP additions, 68 FP multiplications,
cannam@95:  * (or, 72 additions, 22 multiplications, 46 fused multiply/add),
cannam@95:  * 64 stack variables, 2 constants, and 48 memory accesses
cannam@95:  */
cannam@95: #include "hc2cb.h"
cannam@95: 
cannam@95: static void hc2cb_12(R *Rp, R *Ip, R *Rm, R *Im, const R *W, stride rs, INT mb, INT me, INT ms)
cannam@95: {
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 - 1) * 22); m < me; m = m + 1, Rp = Rp + ms, Ip = Ip + ms, Rm = Rm - ms, Im = Im - ms, W = W + 22, MAKE_VOLATILE_STRIDE(48, rs)) {
cannam@95: 	       E T1U, T1X, T1W, T1Y, T1V;
cannam@95: 	       {
cannam@95: 		    E T18, T20, T21, T1b, T2a, T1s, T29, T1p, TO, T11, To, Tb, Tg, T23, T1f;
cannam@95: 		    E Tl, Ty, Tt, T1i, T24, T1z, T2d, T1w, T2c;
cannam@95: 		    {
cannam@95: 			 E T5, Ta, TN, TI;
cannam@95: 			 {
cannam@95: 			      E T1, TE, T6, TM, T7, T1o, T4, T17, TH, T8, TJ, TK;
cannam@95: 			      T1 = Rp[0];
cannam@95: 			      TE = Ip[0];
cannam@95: 			      T6 = Rm[WS(rs, 5)];
cannam@95: 			      TM = Im[WS(rs, 5)];
cannam@95: 			      {
cannam@95: 				   E T2, T3, TF, TG;
cannam@95: 				   T2 = Rp[WS(rs, 4)];
cannam@95: 				   T3 = Rm[WS(rs, 3)];
cannam@95: 				   TF = Ip[WS(rs, 4)];
cannam@95: 				   TG = Im[WS(rs, 3)];
cannam@95: 				   T7 = Rm[WS(rs, 1)];
cannam@95: 				   T1o = T2 - T3;
cannam@95: 				   T4 = T2 + T3;
cannam@95: 				   T17 = TF + TG;
cannam@95: 				   TH = TF - TG;
cannam@95: 				   T8 = Rp[WS(rs, 2)];
cannam@95: 				   TJ = Ip[WS(rs, 2)];
cannam@95: 				   TK = Im[WS(rs, 1)];
cannam@95: 			      }
cannam@95: 			      {
cannam@95: 				   E T1r, T1a, T19, T1q, T9, TL, T16, T1n;
cannam@95: 				   T5 = T1 + T4;
cannam@95: 				   T16 = FNMS(KP500000000, T4, T1);
cannam@95: 				   T1r = T7 - T8;
cannam@95: 				   T9 = T7 + T8;
cannam@95: 				   T1a = TJ + TK;
cannam@95: 				   TL = TJ - TK;
cannam@95: 				   T18 = FNMS(KP866025403, T17, T16);
cannam@95: 				   T20 = FMA(KP866025403, T17, T16);
cannam@95: 				   T19 = FNMS(KP500000000, T9, T6);
cannam@95: 				   Ta = T6 + T9;
cannam@95: 				   TN = TL - TM;
cannam@95: 				   T1q = FMA(KP500000000, TL, TM);
cannam@95: 				   T1n = FNMS(KP500000000, TH, TE);
cannam@95: 				   TI = TE + TH;
cannam@95: 				   T21 = FNMS(KP866025403, T1a, T19);
cannam@95: 				   T1b = FMA(KP866025403, T1a, T19);
cannam@95: 				   T2a = FMA(KP866025403, T1r, T1q);
cannam@95: 				   T1s = FNMS(KP866025403, T1r, T1q);
cannam@95: 				   T29 = FNMS(KP866025403, T1o, T1n);
cannam@95: 				   T1p = FMA(KP866025403, T1o, T1n);
cannam@95: 			      }
cannam@95: 			 }
cannam@95: 			 {
cannam@95: 			      E Tc, Tp, Th, Tx, Ti, Tf, T1v, Ts, T1e, Tj, Tu, Tv;
cannam@95: 			      Tc = Rp[WS(rs, 3)];
cannam@95: 			      TO = TI - TN;
cannam@95: 			      T11 = TI + TN;
cannam@95: 			      Tp = Ip[WS(rs, 3)];
cannam@95: 			      To = T5 - Ta;
cannam@95: 			      Tb = T5 + Ta;
cannam@95: 			      Th = Rm[WS(rs, 2)];
cannam@95: 			      Tx = Im[WS(rs, 2)];
cannam@95: 			      {
cannam@95: 				   E Td, Te, Tq, Tr;
cannam@95: 				   Td = Rm[WS(rs, 4)];
cannam@95: 				   Te = Rm[0];
cannam@95: 				   Tq = Im[WS(rs, 4)];
cannam@95: 				   Tr = Im[0];
cannam@95: 				   Ti = Rp[WS(rs, 1)];
cannam@95: 				   Tf = Td + Te;
cannam@95: 				   T1v = Td - Te;
cannam@95: 				   Ts = Tq + Tr;
cannam@95: 				   T1e = Tq - Tr;
cannam@95: 				   Tj = Rp[WS(rs, 5)];
cannam@95: 				   Tu = Ip[WS(rs, 1)];
cannam@95: 				   Tv = Ip[WS(rs, 5)];
cannam@95: 			      }
cannam@95: 			      {
cannam@95: 				   E T1y, T1h, T1g, T1x, Tk, Tw, T1d, T1u;
cannam@95: 				   T1d = FNMS(KP500000000, Tf, Tc);
cannam@95: 				   Tg = Tc + Tf;
cannam@95: 				   Tk = Ti + Tj;
cannam@95: 				   T1y = Ti - Tj;
cannam@95: 				   Tw = Tu + Tv;
cannam@95: 				   T1h = Tv - Tu;
cannam@95: 				   T23 = FNMS(KP866025403, T1e, T1d);
cannam@95: 				   T1f = FMA(KP866025403, T1e, T1d);
cannam@95: 				   Tl = Th + Tk;
cannam@95: 				   T1g = FNMS(KP500000000, Tk, Th);
cannam@95: 				   T1x = FMA(KP500000000, Tw, Tx);
cannam@95: 				   Ty = Tw - Tx;
cannam@95: 				   Tt = Tp - Ts;
cannam@95: 				   T1u = FMA(KP500000000, Ts, Tp);
cannam@95: 				   T1i = FMA(KP866025403, T1h, T1g);
cannam@95: 				   T24 = FNMS(KP866025403, T1h, T1g);
cannam@95: 				   T1z = FNMS(KP866025403, T1y, T1x);
cannam@95: 				   T2d = FMA(KP866025403, T1y, T1x);
cannam@95: 				   T1w = FMA(KP866025403, T1v, T1u);
cannam@95: 				   T2c = FNMS(KP866025403, T1v, T1u);
cannam@95: 			      }
cannam@95: 			 }
cannam@95: 		    }
cannam@95: 		    {
cannam@95: 			 E TY, T13, TX, T10;
cannam@95: 			 {
cannam@95: 			      E Tn, T12, TC, Tm, TD, TS, TA, Tz;
cannam@95: 			      Tn = W[16];
cannam@95: 			      T12 = Tt + Ty;
cannam@95: 			      Tz = Tt - Ty;
cannam@95: 			      TC = W[17];
cannam@95: 			      Tm = Tg + Tl;
cannam@95: 			      TD = Tg - Tl;
cannam@95: 			      TS = To + Tz;
cannam@95: 			      TA = To - Tz;
cannam@95: 			      {
cannam@95: 				   E TV, TU, TW, TT;
cannam@95: 				   {
cannam@95: 					E TQ, TR, TP, TB;
cannam@95: 					TV = TO - TD;
cannam@95: 					TP = TD + TO;
cannam@95: 					Rp[0] = Tb + Tm;
cannam@95: 					TB = Tn * TA;
cannam@95: 					TQ = Tn * TP;
cannam@95: 					TR = W[4];
cannam@95: 					Ip[WS(rs, 4)] = FNMS(TC, TP, TB);
cannam@95: 					TU = W[5];
cannam@95: 					Im[WS(rs, 4)] = FMA(TC, TA, TQ);
cannam@95: 					TW = TR * TV;
cannam@95: 					TT = TR * TS;
cannam@95: 				   }
cannam@95: 				   Im[WS(rs, 1)] = FMA(TU, TS, TW);
cannam@95: 				   Ip[WS(rs, 1)] = FNMS(TU, TV, TT);
cannam@95: 				   TY = Tb - Tm;
cannam@95: 				   T13 = T11 - T12;
cannam@95: 				   TX = W[10];
cannam@95: 				   T10 = W[11];
cannam@95: 				   Rm[0] = T11 + T12;
cannam@95: 			      }
cannam@95: 			 }
cannam@95: 			 {
cannam@95: 			      E T1K, T1Q, T1P, T1L, T2o, T2u, T2t, T2p;
cannam@95: 			      {
cannam@95: 				   E T1E, T1D, T1H, T1F, T1G, T1t, T1k, T1A;
cannam@95: 				   {
cannam@95: 					E T1c, TZ, T14, T1j;
cannam@95: 					T1K = T18 - T1b;
cannam@95: 					T1c = T18 + T1b;
cannam@95: 					TZ = TX * TY;
cannam@95: 					T14 = T10 * TY;
cannam@95: 					T1j = T1f + T1i;
cannam@95: 					T1Q = T1f - T1i;
cannam@95: 					T1P = T1p + T1s;
cannam@95: 					T1t = T1p - T1s;
cannam@95: 					Rp[WS(rs, 3)] = FNMS(T10, T13, TZ);
cannam@95: 					Rm[WS(rs, 3)] = FMA(TX, T13, T14);
cannam@95: 					T1E = T1c + T1j;
cannam@95: 					T1k = T1c - T1j;
cannam@95: 					T1A = T1w - T1z;
cannam@95: 					T1L = T1w + T1z;
cannam@95: 				   }
cannam@95: 				   {
cannam@95: 					E T15, T1m, T1B, T1l, T1C;
cannam@95: 					T15 = W[18];
cannam@95: 					T1m = W[19];
cannam@95: 					T1D = W[6];
cannam@95: 					T1H = T1t + T1A;
cannam@95: 					T1B = T1t - T1A;
cannam@95: 					T1l = T15 * T1k;
cannam@95: 					T1C = T1m * T1k;
cannam@95: 					T1F = T1D * T1E;
cannam@95: 					T1G = W[7];
cannam@95: 					Rp[WS(rs, 5)] = FNMS(T1m, T1B, T1l);
cannam@95: 					Rm[WS(rs, 5)] = FMA(T15, T1B, T1C);
cannam@95: 				   }
cannam@95: 				   {
cannam@95: 					E T26, T2i, T2l, T2f, T1Z, T28;
cannam@95: 					{
cannam@95: 					     E T22, T1I, T25, T2b, T2e;
cannam@95: 					     T22 = T20 + T21;
cannam@95: 					     T2o = T20 - T21;
cannam@95: 					     Rp[WS(rs, 2)] = FNMS(T1G, T1H, T1F);
cannam@95: 					     T1I = T1G * T1E;
cannam@95: 					     T2u = T23 - T24;
cannam@95: 					     T25 = T23 + T24;
cannam@95: 					     T2b = T29 - T2a;
cannam@95: 					     T2t = T29 + T2a;
cannam@95: 					     T2p = T2c + T2d;
cannam@95: 					     T2e = T2c - T2d;
cannam@95: 					     Rm[WS(rs, 2)] = FMA(T1D, T1H, T1I);
cannam@95: 					     T26 = T22 - T25;
cannam@95: 					     T2i = T22 + T25;
cannam@95: 					     T2l = T2b + T2e;
cannam@95: 					     T2f = T2b - T2e;
cannam@95: 					}
cannam@95: 					T1Z = W[2];
cannam@95: 					T28 = W[3];
cannam@95: 					{
cannam@95: 					     E T2h, T2k, T27, T2g, T2j, T2m;
cannam@95: 					     T2h = W[14];
cannam@95: 					     T2k = W[15];
cannam@95: 					     T27 = T1Z * T26;
cannam@95: 					     T2g = T28 * T26;
cannam@95: 					     T2j = T2h * T2i;
cannam@95: 					     T2m = T2k * T2i;
cannam@95: 					     Rp[WS(rs, 1)] = FNMS(T28, T2f, T27);
cannam@95: 					     Rm[WS(rs, 1)] = FMA(T1Z, T2f, T2g);
cannam@95: 					     Rp[WS(rs, 4)] = FNMS(T2k, T2l, T2j);
cannam@95: 					     Rm[WS(rs, 4)] = FMA(T2h, T2l, T2m);
cannam@95: 					}
cannam@95: 				   }
cannam@95: 			      }
cannam@95: 			      {
cannam@95: 				   E T2y, T2B, T2A, T2C, T2z;
cannam@95: 				   {
cannam@95: 					E T2n, T2q, T2v, T2s, T2r, T2x, T2w;
cannam@95: 					T2n = W[8];
cannam@95: 					T2y = T2o + T2p;
cannam@95: 					T2q = T2o - T2p;
cannam@95: 					T2B = T2t - T2u;
cannam@95: 					T2v = T2t + T2u;
cannam@95: 					T2s = W[9];
cannam@95: 					T2r = T2n * T2q;
cannam@95: 					T2x = W[20];
cannam@95: 					T2w = T2n * T2v;
cannam@95: 					T2A = W[21];
cannam@95: 					Ip[WS(rs, 2)] = FNMS(T2s, T2v, T2r);
cannam@95: 					T2C = T2x * T2B;
cannam@95: 					T2z = T2x * T2y;
cannam@95: 					Im[WS(rs, 2)] = FMA(T2s, T2q, T2w);
cannam@95: 				   }
cannam@95: 				   Im[WS(rs, 5)] = FMA(T2A, T2y, T2C);
cannam@95: 				   Ip[WS(rs, 5)] = FNMS(T2A, T2B, T2z);
cannam@95: 				   {
cannam@95: 					E T1J, T1M, T1R, T1O, T1N, T1T, T1S;
cannam@95: 					T1J = W[0];
cannam@95: 					T1U = T1K + T1L;
cannam@95: 					T1M = T1K - T1L;
cannam@95: 					T1X = T1P - T1Q;
cannam@95: 					T1R = T1P + T1Q;
cannam@95: 					T1O = W[1];
cannam@95: 					T1N = T1J * T1M;
cannam@95: 					T1T = W[12];
cannam@95: 					T1S = T1J * T1R;
cannam@95: 					T1W = W[13];
cannam@95: 					Ip[0] = FNMS(T1O, T1R, T1N);
cannam@95: 					T1Y = T1T * T1X;
cannam@95: 					T1V = T1T * T1U;
cannam@95: 					Im[0] = FMA(T1O, T1M, T1S);
cannam@95: 				   }
cannam@95: 			      }
cannam@95: 			 }
cannam@95: 		    }
cannam@95: 	       }
cannam@95: 	       Im[WS(rs, 3)] = FMA(T1W, T1U, T1Y);
cannam@95: 	       Ip[WS(rs, 3)] = FNMS(T1W, T1X, T1V);
cannam@95: 	  }
cannam@95:      }
cannam@95: }
cannam@95: 
cannam@95: static const tw_instr twinstr[] = {
cannam@95:      {TW_FULL, 1, 12},
cannam@95:      {TW_NEXT, 1, 0}
cannam@95: };
cannam@95: 
cannam@95: static const hc2c_desc desc = { 12, "hc2cb_12", twinstr, &GENUS, {72, 22, 46, 0} };
cannam@95: 
cannam@95: void X(codelet_hc2cb_12) (planner *p) {
cannam@95:      X(khc2c_register) (p, hc2cb_12, &desc, HC2C_VIA_RDFT);
cannam@95: }
cannam@95: #else				/* HAVE_FMA */
cannam@95: 
cannam@95: /* Generated by: ../../../genfft/gen_hc2c.native -compact -variables 4 -pipeline-latency 4 -sign 1 -n 12 -dif -name hc2cb_12 -include hc2cb.h */
cannam@95: 
cannam@95: /*
cannam@95:  * This function contains 118 FP additions, 60 FP multiplications,
cannam@95:  * (or, 88 additions, 30 multiplications, 30 fused multiply/add),
cannam@95:  * 39 stack variables, 2 constants, and 48 memory accesses
cannam@95:  */
cannam@95: #include "hc2cb.h"
cannam@95: 
cannam@95: static void hc2cb_12(R *Rp, R *Ip, R *Rm, R *Im, const R *W, stride rs, INT mb, INT me, INT ms)
cannam@95: {
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 - 1) * 22); m < me; m = m + 1, Rp = Rp + ms, Ip = Ip + ms, Rm = Rm - ms, Im = Im - ms, W = W + 22, MAKE_VOLATILE_STRIDE(48, rs)) {
cannam@95: 	       E T5, TH, T12, T1M, T1i, T1U, Tl, Ty, T1c, T1Y, T1s, T1Q, Ta, TM, T15;
cannam@95: 	       E T1N, T1l, T1V, Tg, Tt, T19, T1X, T1p, T1P;
cannam@95: 	       {
cannam@95: 		    E T1, TD, T4, T1g, TG, T11, T10, T1h;
cannam@95: 		    T1 = Rp[0];
cannam@95: 		    TD = Ip[0];
cannam@95: 		    {
cannam@95: 			 E T2, T3, TE, TF;
cannam@95: 			 T2 = Rp[WS(rs, 4)];
cannam@95: 			 T3 = Rm[WS(rs, 3)];
cannam@95: 			 T4 = T2 + T3;
cannam@95: 			 T1g = KP866025403 * (T2 - T3);
cannam@95: 			 TE = Ip[WS(rs, 4)];
cannam@95: 			 TF = Im[WS(rs, 3)];
cannam@95: 			 TG = TE - TF;
cannam@95: 			 T11 = KP866025403 * (TE + TF);
cannam@95: 		    }
cannam@95: 		    T5 = T1 + T4;
cannam@95: 		    TH = TD + TG;
cannam@95: 		    T10 = FNMS(KP500000000, T4, T1);
cannam@95: 		    T12 = T10 - T11;
cannam@95: 		    T1M = T10 + T11;
cannam@95: 		    T1h = FNMS(KP500000000, TG, TD);
cannam@95: 		    T1i = T1g + T1h;
cannam@95: 		    T1U = T1h - T1g;
cannam@95: 	       }
cannam@95: 	       {
cannam@95: 		    E Th, Tx, Tk, T1a, Tw, T1r, T1b, T1q;
cannam@95: 		    Th = Rm[WS(rs, 2)];
cannam@95: 		    Tx = Im[WS(rs, 2)];
cannam@95: 		    {
cannam@95: 			 E Ti, Tj, Tu, Tv;
cannam@95: 			 Ti = Rp[WS(rs, 1)];
cannam@95: 			 Tj = Rp[WS(rs, 5)];
cannam@95: 			 Tk = Ti + Tj;
cannam@95: 			 T1a = KP866025403 * (Ti - Tj);
cannam@95: 			 Tu = Ip[WS(rs, 1)];
cannam@95: 			 Tv = Ip[WS(rs, 5)];
cannam@95: 			 Tw = Tu + Tv;
cannam@95: 			 T1r = KP866025403 * (Tv - Tu);
cannam@95: 		    }
cannam@95: 		    Tl = Th + Tk;
cannam@95: 		    Ty = Tw - Tx;
cannam@95: 		    T1b = FMA(KP500000000, Tw, Tx);
cannam@95: 		    T1c = T1a - T1b;
cannam@95: 		    T1Y = T1a + T1b;
cannam@95: 		    T1q = FNMS(KP500000000, Tk, Th);
cannam@95: 		    T1s = T1q + T1r;
cannam@95: 		    T1Q = T1q - T1r;
cannam@95: 	       }
cannam@95: 	       {
cannam@95: 		    E T6, TL, T9, T1j, TK, T14, T13, T1k;
cannam@95: 		    T6 = Rm[WS(rs, 5)];
cannam@95: 		    TL = Im[WS(rs, 5)];
cannam@95: 		    {
cannam@95: 			 E T7, T8, TI, TJ;
cannam@95: 			 T7 = Rm[WS(rs, 1)];
cannam@95: 			 T8 = Rp[WS(rs, 2)];
cannam@95: 			 T9 = T7 + T8;
cannam@95: 			 T1j = KP866025403 * (T7 - T8);
cannam@95: 			 TI = Ip[WS(rs, 2)];
cannam@95: 			 TJ = Im[WS(rs, 1)];
cannam@95: 			 TK = TI - TJ;
cannam@95: 			 T14 = KP866025403 * (TI + TJ);
cannam@95: 		    }
cannam@95: 		    Ta = T6 + T9;
cannam@95: 		    TM = TK - TL;
cannam@95: 		    T13 = FNMS(KP500000000, T9, T6);
cannam@95: 		    T15 = T13 + T14;
cannam@95: 		    T1N = T13 - T14;
cannam@95: 		    T1k = FMA(KP500000000, TK, TL);
cannam@95: 		    T1l = T1j - T1k;
cannam@95: 		    T1V = T1j + T1k;
cannam@95: 	       }
cannam@95: 	       {
cannam@95: 		    E Tc, Tp, Tf, T17, Ts, T1o, T18, T1n;
cannam@95: 		    Tc = Rp[WS(rs, 3)];
cannam@95: 		    Tp = Ip[WS(rs, 3)];
cannam@95: 		    {
cannam@95: 			 E Td, Te, Tq, Tr;
cannam@95: 			 Td = Rm[WS(rs, 4)];
cannam@95: 			 Te = Rm[0];
cannam@95: 			 Tf = Td + Te;
cannam@95: 			 T17 = KP866025403 * (Td - Te);
cannam@95: 			 Tq = Im[WS(rs, 4)];
cannam@95: 			 Tr = Im[0];
cannam@95: 			 Ts = Tq + Tr;
cannam@95: 			 T1o = KP866025403 * (Tq - Tr);
cannam@95: 		    }
cannam@95: 		    Tg = Tc + Tf;
cannam@95: 		    Tt = Tp - Ts;
cannam@95: 		    T18 = FMA(KP500000000, Ts, Tp);
cannam@95: 		    T19 = T17 + T18;
cannam@95: 		    T1X = T18 - T17;
cannam@95: 		    T1n = FNMS(KP500000000, Tf, Tc);
cannam@95: 		    T1p = T1n + T1o;
cannam@95: 		    T1P = T1n - T1o;
cannam@95: 	       }
cannam@95: 	       {
cannam@95: 		    E Tb, Tm, TU, TW, TX, TY, TT, TV;
cannam@95: 		    Tb = T5 + Ta;
cannam@95: 		    Tm = Tg + Tl;
cannam@95: 		    TU = Tb - Tm;
cannam@95: 		    TW = TH + TM;
cannam@95: 		    TX = Tt + Ty;
cannam@95: 		    TY = TW - TX;
cannam@95: 		    Rp[0] = Tb + Tm;
cannam@95: 		    Rm[0] = TW + TX;
cannam@95: 		    TT = W[10];
cannam@95: 		    TV = W[11];
cannam@95: 		    Rp[WS(rs, 3)] = FNMS(TV, TY, TT * TU);
cannam@95: 		    Rm[WS(rs, 3)] = FMA(TV, TU, TT * TY);
cannam@95: 	       }
cannam@95: 	       {
cannam@95: 		    E TA, TQ, TO, TS;
cannam@95: 		    {
cannam@95: 			 E To, Tz, TC, TN;
cannam@95: 			 To = T5 - Ta;
cannam@95: 			 Tz = Tt - Ty;
cannam@95: 			 TA = To - Tz;
cannam@95: 			 TQ = To + Tz;
cannam@95: 			 TC = Tg - Tl;
cannam@95: 			 TN = TH - TM;
cannam@95: 			 TO = TC + TN;
cannam@95: 			 TS = TN - TC;
cannam@95: 		    }
cannam@95: 		    {
cannam@95: 			 E Tn, TB, TP, TR;
cannam@95: 			 Tn = W[16];
cannam@95: 			 TB = W[17];
cannam@95: 			 Ip[WS(rs, 4)] = FNMS(TB, TO, Tn * TA);
cannam@95: 			 Im[WS(rs, 4)] = FMA(Tn, TO, TB * TA);
cannam@95: 			 TP = W[4];
cannam@95: 			 TR = W[5];
cannam@95: 			 Ip[WS(rs, 1)] = FNMS(TR, TS, TP * TQ);
cannam@95: 			 Im[WS(rs, 1)] = FMA(TP, TS, TR * TQ);
cannam@95: 		    }
cannam@95: 	       }
cannam@95: 	       {
cannam@95: 		    E T28, T2e, T2c, T2g;
cannam@95: 		    {
cannam@95: 			 E T26, T27, T2a, T2b;
cannam@95: 			 T26 = T1M - T1N;
cannam@95: 			 T27 = T1X + T1Y;
cannam@95: 			 T28 = T26 - T27;
cannam@95: 			 T2e = T26 + T27;
cannam@95: 			 T2a = T1U + T1V;
cannam@95: 			 T2b = T1P - T1Q;
cannam@95: 			 T2c = T2a + T2b;
cannam@95: 			 T2g = T2a - T2b;
cannam@95: 		    }
cannam@95: 		    {
cannam@95: 			 E T25, T29, T2d, T2f;
cannam@95: 			 T25 = W[8];
cannam@95: 			 T29 = W[9];
cannam@95: 			 Ip[WS(rs, 2)] = FNMS(T29, T2c, T25 * T28);
cannam@95: 			 Im[WS(rs, 2)] = FMA(T25, T2c, T29 * T28);
cannam@95: 			 T2d = W[20];
cannam@95: 			 T2f = W[21];
cannam@95: 			 Ip[WS(rs, 5)] = FNMS(T2f, T2g, T2d * T2e);
cannam@95: 			 Im[WS(rs, 5)] = FMA(T2d, T2g, T2f * T2e);
cannam@95: 		    }
cannam@95: 	       }
cannam@95: 	       {
cannam@95: 		    E T1S, T22, T20, T24;
cannam@95: 		    {
cannam@95: 			 E T1O, T1R, T1W, T1Z;
cannam@95: 			 T1O = T1M + T1N;
cannam@95: 			 T1R = T1P + T1Q;
cannam@95: 			 T1S = T1O - T1R;
cannam@95: 			 T22 = T1O + T1R;
cannam@95: 			 T1W = T1U - T1V;
cannam@95: 			 T1Z = T1X - T1Y;
cannam@95: 			 T20 = T1W - T1Z;
cannam@95: 			 T24 = T1W + T1Z;
cannam@95: 		    }
cannam@95: 		    {
cannam@95: 			 E T1L, T1T, T21, T23;
cannam@95: 			 T1L = W[2];
cannam@95: 			 T1T = W[3];
cannam@95: 			 Rp[WS(rs, 1)] = FNMS(T1T, T20, T1L * T1S);
cannam@95: 			 Rm[WS(rs, 1)] = FMA(T1T, T1S, T1L * T20);
cannam@95: 			 T21 = W[14];
cannam@95: 			 T23 = W[15];
cannam@95: 			 Rp[WS(rs, 4)] = FNMS(T23, T24, T21 * T22);
cannam@95: 			 Rm[WS(rs, 4)] = FMA(T23, T22, T21 * T24);
cannam@95: 		    }
cannam@95: 	       }
cannam@95: 	       {
cannam@95: 		    E T1C, T1I, T1G, T1K;
cannam@95: 		    {
cannam@95: 			 E T1A, T1B, T1E, T1F;
cannam@95: 			 T1A = T12 + T15;
cannam@95: 			 T1B = T1p + T1s;
cannam@95: 			 T1C = T1A - T1B;
cannam@95: 			 T1I = T1A + T1B;
cannam@95: 			 T1E = T1i + T1l;
cannam@95: 			 T1F = T19 + T1c;
cannam@95: 			 T1G = T1E - T1F;
cannam@95: 			 T1K = T1E + T1F;
cannam@95: 		    }
cannam@95: 		    {
cannam@95: 			 E T1z, T1D, T1H, T1J;
cannam@95: 			 T1z = W[18];
cannam@95: 			 T1D = W[19];
cannam@95: 			 Rp[WS(rs, 5)] = FNMS(T1D, T1G, T1z * T1C);
cannam@95: 			 Rm[WS(rs, 5)] = FMA(T1D, T1C, T1z * T1G);
cannam@95: 			 T1H = W[6];
cannam@95: 			 T1J = W[7];
cannam@95: 			 Rp[WS(rs, 2)] = FNMS(T1J, T1K, T1H * T1I);
cannam@95: 			 Rm[WS(rs, 2)] = FMA(T1J, T1I, T1H * T1K);
cannam@95: 		    }
cannam@95: 	       }
cannam@95: 	       {
cannam@95: 		    E T1e, T1w, T1u, T1y;
cannam@95: 		    {
cannam@95: 			 E T16, T1d, T1m, T1t;
cannam@95: 			 T16 = T12 - T15;
cannam@95: 			 T1d = T19 - T1c;
cannam@95: 			 T1e = T16 - T1d;
cannam@95: 			 T1w = T16 + T1d;
cannam@95: 			 T1m = T1i - T1l;
cannam@95: 			 T1t = T1p - T1s;
cannam@95: 			 T1u = T1m + T1t;
cannam@95: 			 T1y = T1m - T1t;
cannam@95: 		    }
cannam@95: 		    {
cannam@95: 			 E TZ, T1f, T1v, T1x;
cannam@95: 			 TZ = W[0];
cannam@95: 			 T1f = W[1];
cannam@95: 			 Ip[0] = FNMS(T1f, T1u, TZ * T1e);
cannam@95: 			 Im[0] = FMA(TZ, T1u, T1f * T1e);
cannam@95: 			 T1v = W[12];
cannam@95: 			 T1x = W[13];
cannam@95: 			 Ip[WS(rs, 3)] = FNMS(T1x, T1y, T1v * T1w);
cannam@95: 			 Im[WS(rs, 3)] = FMA(T1v, T1y, T1x * T1w);
cannam@95: 		    }
cannam@95: 	       }
cannam@95: 	  }
cannam@95:      }
cannam@95: }
cannam@95: 
cannam@95: static const tw_instr twinstr[] = {
cannam@95:      {TW_FULL, 1, 12},
cannam@95:      {TW_NEXT, 1, 0}
cannam@95: };
cannam@95: 
cannam@95: static const hc2c_desc desc = { 12, "hc2cb_12", twinstr, &GENUS, {88, 30, 30, 0} };
cannam@95: 
cannam@95: void X(codelet_hc2cb_12) (planner *p) {
cannam@95:      X(khc2c_register) (p, hc2cb_12, &desc, HC2C_VIA_RDFT);
cannam@95: }
cannam@95: #endif				/* HAVE_FMA */