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 07:41:57 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_hc2c.native -fma -reorder-insns -schedule-for-pipeline -compact -variables 4 -pipeline-latency 4 -sign 1 -twiddle-log3 -precompute-twiddles -n 16 -dif -name hc2cb2_16 -include hc2cb.h */
Chris@10: 
Chris@10: /*
Chris@10:  * This function contains 196 FP additions, 134 FP multiplications,
Chris@10:  * (or, 104 additions, 42 multiplications, 92 fused multiply/add),
Chris@10:  * 112 stack variables, 3 constants, and 64 memory accesses
Chris@10:  */
Chris@10: #include "hc2cb.h"
Chris@10: 
Chris@10: static void hc2cb2_16(R *Rp, R *Ip, R *Rm, R *Im, const R *W, stride rs, INT mb, INT me, INT ms)
Chris@10: {
Chris@10:      DK(KP923879532, +0.923879532511286756128183189396788286822416626);
Chris@10:      DK(KP707106781, +0.707106781186547524400844362104849039284835938);
Chris@10:      DK(KP414213562, +0.414213562373095048801688724209698078569671875);
Chris@10:      {
Chris@10: 	  INT m;
Chris@10: 	  for (m = mb, W = W + ((mb - 1) * 8); m < me; m = m + 1, Rp = Rp + ms, Ip = Ip + ms, Rm = Rm - ms, Im = Im - ms, W = W + 8, MAKE_VOLATILE_STRIDE(64, rs)) {
Chris@10: 	       E Tv, TB, TF, Ty, T1J, T1O, T1N, T1K;
Chris@10: 	       {
Chris@10: 		    E Tw, T2z, T2C, Tx, T3f, T3l, T2F, T3r, Tz;
Chris@10: 		    Tv = W[0];
Chris@10: 		    Tw = W[2];
Chris@10: 		    T2z = W[6];
Chris@10: 		    T2C = W[7];
Chris@10: 		    TB = W[4];
Chris@10: 		    Tx = Tv * Tw;
Chris@10: 		    T3f = Tv * T2z;
Chris@10: 		    T3l = Tv * T2C;
Chris@10: 		    T2F = Tv * TB;
Chris@10: 		    T3r = Tw * TB;
Chris@10: 		    TF = W[5];
Chris@10: 		    Ty = W[1];
Chris@10: 		    Tz = W[3];
Chris@10: 		    {
Chris@10: 			 E T2G, T3z, T3m, T3g, T3L, T3s, T1V, TA, T3w, T3Q, T30, T3C, TE, T1X, T1D;
Chris@10: 			 E TG, T1G, T1o, T2p, T1Y, T2u, T2c, T1Z, TL, T1t, T2d, T3n, T35, T3R, T3F;
Chris@10: 			 E T1w, T20, T3M, Tf, T3h, T2L, T2e, TW, T2Q, T36, T3I, T3N, T2V, T37, T1d;
Chris@10: 			 E Tu, T3S, T18, T1z, T1i, T24, T2g, T27, T2h;
Chris@10: 			 {
Chris@10: 			      E T2K, TQ, TV, T2H;
Chris@10: 			      {
Chris@10: 				   E TH, T3, T32, T1s, T1p, T6, T33, TK, TM, Ta, TS, T2J, TP, TR, Td;
Chris@10: 				   E TT, TI, TJ;
Chris@10: 				   {
Chris@10: 					E T1q, T1r, T4, T5;
Chris@10: 					{
Chris@10: 					     E T1, T1n, TC, T2b, T1W, T2, T3v, T2Z, TD;
Chris@10: 					     T1 = Rp[0];
Chris@10: 					     T3v = Tw * TF;
Chris@10: 					     T2Z = Tv * TF;
Chris@10: 					     T2G = FNMS(Ty, TF, T2F);
Chris@10: 					     T3z = FMA(Ty, TF, T2F);
Chris@10: 					     T3m = FNMS(Ty, T2z, T3l);
Chris@10: 					     T3g = FMA(Ty, T2C, T3f);
Chris@10: 					     T3L = FNMS(Tz, TF, T3r);
Chris@10: 					     T3s = FMA(Tz, TF, T3r);
Chris@10: 					     T1V = FMA(Ty, Tz, Tx);
Chris@10: 					     TA = FNMS(Ty, Tz, Tx);
Chris@10: 					     TD = Tv * Tz;
Chris@10: 					     T3w = FNMS(Tz, TB, T3v);
Chris@10: 					     T3Q = FMA(Tz, TB, T3v);
Chris@10: 					     T30 = FMA(Ty, TB, T2Z);
Chris@10: 					     T3C = FNMS(Ty, TB, T2Z);
Chris@10: 					     T1n = TA * TF;
Chris@10: 					     TC = TA * TB;
Chris@10: 					     T2b = T1V * TF;
Chris@10: 					     T1W = T1V * TB;
Chris@10: 					     TE = FMA(Ty, Tw, TD);
Chris@10: 					     T1X = FNMS(Ty, Tw, TD);
Chris@10: 					     T2 = Rm[WS(rs, 7)];
Chris@10: 					     T1q = Ip[0];
Chris@10: 					     T1D = FMA(TE, TF, TC);
Chris@10: 					     TG = FNMS(TE, TF, TC);
Chris@10: 					     T1G = FNMS(TE, TB, T1n);
Chris@10: 					     T1o = FMA(TE, TB, T1n);
Chris@10: 					     T2p = FMA(T1X, TF, T1W);
Chris@10: 					     T1Y = FNMS(T1X, TF, T1W);
Chris@10: 					     T2u = FNMS(T1X, TB, T2b);
Chris@10: 					     T2c = FMA(T1X, TB, T2b);
Chris@10: 					     TH = T1 - T2;
Chris@10: 					     T3 = T1 + T2;
Chris@10: 					     T1r = Im[WS(rs, 7)];
Chris@10: 					}
Chris@10: 					T4 = Rp[WS(rs, 4)];
Chris@10: 					T5 = Rm[WS(rs, 3)];
Chris@10: 					TI = Ip[WS(rs, 4)];
Chris@10: 					T32 = T1q - T1r;
Chris@10: 					T1s = T1q + T1r;
Chris@10: 					T1p = T4 - T5;
Chris@10: 					T6 = T4 + T5;
Chris@10: 					TJ = Im[WS(rs, 3)];
Chris@10: 				   }
Chris@10: 				   {
Chris@10: 					E TN, TO, T8, T9, Tb, Tc;
Chris@10: 					T8 = Rp[WS(rs, 2)];
Chris@10: 					T9 = Rm[WS(rs, 5)];
Chris@10: 					TN = Ip[WS(rs, 2)];
Chris@10: 					T33 = TI - TJ;
Chris@10: 					TK = TI + TJ;
Chris@10: 					TM = T8 - T9;
Chris@10: 					Ta = T8 + T9;
Chris@10: 					TO = Im[WS(rs, 5)];
Chris@10: 					Tb = Rm[WS(rs, 1)];
Chris@10: 					Tc = Rp[WS(rs, 6)];
Chris@10: 					TS = Ip[WS(rs, 6)];
Chris@10: 					T2J = TN - TO;
Chris@10: 					TP = TN + TO;
Chris@10: 					TR = Tb - Tc;
Chris@10: 					Td = Tb + Tc;
Chris@10: 					TT = Im[WS(rs, 1)];
Chris@10: 				   }
Chris@10: 				   {
Chris@10: 					E T2I, TU, Te, T31, T34, T3D;
Chris@10: 					T1Z = TH + TK;
Chris@10: 					TL = TH - TK;
Chris@10: 					T1t = T1p + T1s;
Chris@10: 					T2d = T1s - T1p;
Chris@10: 					T2I = TS - TT;
Chris@10: 					TU = TS + TT;
Chris@10: 					Te = Ta + Td;
Chris@10: 					T31 = Ta - Td;
Chris@10: 					T34 = T32 - T33;
Chris@10: 					T3D = T32 + T33;
Chris@10: 					{
Chris@10: 					     E T1u, T1v, T3E, T7;
Chris@10: 					     T3E = T2J + T2I;
Chris@10: 					     T2K = T2I - T2J;
Chris@10: 					     TQ = TM - TP;
Chris@10: 					     T1u = TM + TP;
Chris@10: 					     T3n = T34 - T31;
Chris@10: 					     T35 = T31 + T34;
Chris@10: 					     TV = TR - TU;
Chris@10: 					     T1v = TR + TU;
Chris@10: 					     T3R = T3D - T3E;
Chris@10: 					     T3F = T3D + T3E;
Chris@10: 					     T2H = T3 - T6;
Chris@10: 					     T7 = T3 + T6;
Chris@10: 					     T1w = T1u - T1v;
Chris@10: 					     T20 = T1u + T1v;
Chris@10: 					     T3M = T7 - Te;
Chris@10: 					     Tf = T7 + Te;
Chris@10: 					}
Chris@10: 				   }
Chris@10: 			      }
Chris@10: 			      {
Chris@10: 				   E T1e, Ti, T2N, T1c, T19, Tl, T2O, T1h, Tq, T13, Tp, T2S, T11, Tr, T14;
Chris@10: 				   E T15;
Chris@10: 				   {
Chris@10: 					E Tj, Tk, T1f, T1g;
Chris@10: 					{
Chris@10: 					     E Tg, Th, T1a, T1b;
Chris@10: 					     Tg = Rp[WS(rs, 1)];
Chris@10: 					     T3h = T2H - T2K;
Chris@10: 					     T2L = T2H + T2K;
Chris@10: 					     T2e = TQ - TV;
Chris@10: 					     TW = TQ + TV;
Chris@10: 					     Th = Rm[WS(rs, 6)];
Chris@10: 					     T1a = Ip[WS(rs, 1)];
Chris@10: 					     T1b = Im[WS(rs, 6)];
Chris@10: 					     Tj = Rp[WS(rs, 5)];
Chris@10: 					     T1e = Tg - Th;
Chris@10: 					     Ti = Tg + Th;
Chris@10: 					     T2N = T1a - T1b;
Chris@10: 					     T1c = T1a + T1b;
Chris@10: 					     Tk = Rm[WS(rs, 2)];
Chris@10: 					     T1f = Ip[WS(rs, 5)];
Chris@10: 					     T1g = Im[WS(rs, 2)];
Chris@10: 					}
Chris@10: 					{
Chris@10: 					     E Tn, To, TZ, T10;
Chris@10: 					     Tn = Rm[0];
Chris@10: 					     T19 = Tj - Tk;
Chris@10: 					     Tl = Tj + Tk;
Chris@10: 					     T2O = T1f - T1g;
Chris@10: 					     T1h = T1f + T1g;
Chris@10: 					     To = Rp[WS(rs, 7)];
Chris@10: 					     TZ = Ip[WS(rs, 7)];
Chris@10: 					     T10 = Im[0];
Chris@10: 					     Tq = Rp[WS(rs, 3)];
Chris@10: 					     T13 = Tn - To;
Chris@10: 					     Tp = Tn + To;
Chris@10: 					     T2S = TZ - T10;
Chris@10: 					     T11 = TZ + T10;
Chris@10: 					     Tr = Rm[WS(rs, 4)];
Chris@10: 					     T14 = Ip[WS(rs, 3)];
Chris@10: 					     T15 = Im[WS(rs, 4)];
Chris@10: 					}
Chris@10: 				   }
Chris@10: 				   {
Chris@10: 					E TY, T16, Tm, Tt;
Chris@10: 					{
Chris@10: 					     E T2P, T3G, Ts, T2M, T3H, T2U, T2T, T2R;
Chris@10: 					     T2P = T2N - T2O;
Chris@10: 					     T3G = T2N + T2O;
Chris@10: 					     TY = Tq - Tr;
Chris@10: 					     Ts = Tq + Tr;
Chris@10: 					     T2T = T14 - T15;
Chris@10: 					     T16 = T14 + T15;
Chris@10: 					     T2M = Ti - Tl;
Chris@10: 					     Tm = Ti + Tl;
Chris@10: 					     T3H = T2S + T2T;
Chris@10: 					     T2U = T2S - T2T;
Chris@10: 					     Tt = Tp + Ts;
Chris@10: 					     T2R = Tp - Ts;
Chris@10: 					     T2Q = T2M - T2P;
Chris@10: 					     T36 = T2M + T2P;
Chris@10: 					     T3I = T3G + T3H;
Chris@10: 					     T3N = T3H - T3G;
Chris@10: 					     T2V = T2R + T2U;
Chris@10: 					     T37 = T2U - T2R;
Chris@10: 					}
Chris@10: 					{
Chris@10: 					     E T25, T26, T22, T23, T12, T17;
Chris@10: 					     T12 = TY - T11;
Chris@10: 					     T25 = TY + T11;
Chris@10: 					     T26 = T13 + T16;
Chris@10: 					     T17 = T13 - T16;
Chris@10: 					     T22 = T1c - T19;
Chris@10: 					     T1d = T19 + T1c;
Chris@10: 					     Tu = Tm + Tt;
Chris@10: 					     T3S = Tm - Tt;
Chris@10: 					     T18 = FNMS(KP414213562, T17, T12);
Chris@10: 					     T1z = FMA(KP414213562, T12, T17);
Chris@10: 					     T1i = T1e - T1h;
Chris@10: 					     T23 = T1e + T1h;
Chris@10: 					     T24 = FNMS(KP414213562, T23, T22);
Chris@10: 					     T2g = FMA(KP414213562, T22, T23);
Chris@10: 					     T27 = FNMS(KP414213562, T26, T25);
Chris@10: 					     T2h = FMA(KP414213562, T25, T26);
Chris@10: 					}
Chris@10: 				   }
Chris@10: 			      }
Chris@10: 			 }
Chris@10: 			 {
Chris@10: 			      E T1j, T1y, T3V, T3X, T3W, T38, T3i, T3o, T2W, T3K, T3B, T3A;
Chris@10: 			      Rp[0] = Tf + Tu;
Chris@10: 			      T3A = Tf - Tu;
Chris@10: 			      T1j = FMA(KP414213562, T1i, T1d);
Chris@10: 			      T1y = FNMS(KP414213562, T1d, T1i);
Chris@10: 			      T3K = T3C * T3A;
Chris@10: 			      T3B = T3z * T3A;
Chris@10: 			      {
Chris@10: 				   E T3O, T3T, T3J, T3P, T3U;
Chris@10: 				   T3O = T3M - T3N;
Chris@10: 				   T3V = T3M + T3N;
Chris@10: 				   T3X = T3S + T3R;
Chris@10: 				   T3T = T3R - T3S;
Chris@10: 				   Rm[0] = T3F + T3I;
Chris@10: 				   T3J = T3F - T3I;
Chris@10: 				   T3P = T3L * T3O;
Chris@10: 				   T3U = T3L * T3T;
Chris@10: 				   T3W = TA * T3V;
Chris@10: 				   Rp[WS(rs, 4)] = FNMS(T3C, T3J, T3B);
Chris@10: 				   Rm[WS(rs, 4)] = FMA(T3z, T3J, T3K);
Chris@10: 				   Rp[WS(rs, 6)] = FNMS(T3Q, T3T, T3P);
Chris@10: 				   Rm[WS(rs, 6)] = FMA(T3Q, T3O, T3U);
Chris@10: 				   T38 = T36 + T37;
Chris@10: 				   T3i = T37 - T36;
Chris@10: 				   T3o = T2Q - T2V;
Chris@10: 				   T2W = T2Q + T2V;
Chris@10: 			      }
Chris@10: 			      {
Chris@10: 				   E T2q, T21, T28, T2w, T2v, T2f, T2i, T2r;
Chris@10: 				   {
Chris@10: 					E T2Y, T3a, T3c, T3d, T39, T3e, T3b, T2X, T3Y;
Chris@10: 					Rp[WS(rs, 2)] = FNMS(TE, T3X, T3W);
Chris@10: 					T3Y = TA * T3X;
Chris@10: 					{
Chris@10: 					     E T3t, T3j, T3x, T3p;
Chris@10: 					     T3t = FMA(KP707106781, T3i, T3h);
Chris@10: 					     T3j = FNMS(KP707106781, T3i, T3h);
Chris@10: 					     T3x = FMA(KP707106781, T3o, T3n);
Chris@10: 					     T3p = FNMS(KP707106781, T3o, T3n);
Chris@10: 					     Rm[WS(rs, 2)] = FMA(TE, T3V, T3Y);
Chris@10: 					     {
Chris@10: 						  E T3u, T3k, T3y, T3q;
Chris@10: 						  T3u = T3s * T3t;
Chris@10: 						  T3k = T3g * T3j;
Chris@10: 						  T3y = T3s * T3x;
Chris@10: 						  T3q = T3g * T3p;
Chris@10: 						  Rp[WS(rs, 3)] = FNMS(T3w, T3x, T3u);
Chris@10: 						  Rp[WS(rs, 7)] = FNMS(T3m, T3p, T3k);
Chris@10: 						  Rm[WS(rs, 3)] = FMA(T3w, T3t, T3y);
Chris@10: 						  Rm[WS(rs, 7)] = FMA(T3m, T3j, T3q);
Chris@10: 						  T3b = FMA(KP707106781, T2W, T2L);
Chris@10: 						  T2X = FNMS(KP707106781, T2W, T2L);
Chris@10: 					     }
Chris@10: 					}
Chris@10: 					T2Y = T2G * T2X;
Chris@10: 					T3a = T30 * T2X;
Chris@10: 					T3c = T1V * T3b;
Chris@10: 					T3d = FMA(KP707106781, T38, T35);
Chris@10: 					T39 = FNMS(KP707106781, T38, T35);
Chris@10: 					T3e = T1X * T3b;
Chris@10: 					T2q = FMA(KP707106781, T20, T1Z);
Chris@10: 					T21 = FNMS(KP707106781, T20, T1Z);
Chris@10: 					Rp[WS(rs, 1)] = FNMS(T1X, T3d, T3c);
Chris@10: 					Rm[WS(rs, 5)] = FMA(T2G, T39, T3a);
Chris@10: 					Rp[WS(rs, 5)] = FNMS(T30, T39, T2Y);
Chris@10: 					Rm[WS(rs, 1)] = FMA(T1V, T3d, T3e);
Chris@10: 					T28 = T24 + T27;
Chris@10: 					T2w = T27 - T24;
Chris@10: 					T2v = FNMS(KP707106781, T2e, T2d);
Chris@10: 					T2f = FMA(KP707106781, T2e, T2d);
Chris@10: 					T2i = T2g - T2h;
Chris@10: 					T2r = T2g + T2h;
Chris@10: 				   }
Chris@10: 				   {
Chris@10: 					E TX, T1k, T1x, T1A;
Chris@10: 					T1J = FMA(KP707106781, TW, TL);
Chris@10: 					TX = FNMS(KP707106781, TW, TL);
Chris@10: 					{
Chris@10: 					     E T2l, T29, T2n, T2j;
Chris@10: 					     T2l = FNMS(KP923879532, T28, T21);
Chris@10: 					     T29 = FMA(KP923879532, T28, T21);
Chris@10: 					     T2n = FMA(KP923879532, T2i, T2f);
Chris@10: 					     T2j = FNMS(KP923879532, T2i, T2f);
Chris@10: 					     {
Chris@10: 						  E T2o, T2m, T2k, T2a;
Chris@10: 						  T2o = Tz * T2l;
Chris@10: 						  T2m = Tw * T2l;
Chris@10: 						  T2k = T2c * T29;
Chris@10: 						  T2a = T1Y * T29;
Chris@10: 						  Im[WS(rs, 1)] = FMA(Tw, T2n, T2o);
Chris@10: 						  Ip[WS(rs, 1)] = FNMS(Tz, T2n, T2m);
Chris@10: 						  Im[WS(rs, 5)] = FMA(T1Y, T2j, T2k);
Chris@10: 						  Ip[WS(rs, 5)] = FNMS(T2c, T2j, T2a);
Chris@10: 						  T1k = T18 - T1j;
Chris@10: 						  T1O = T1j + T18;
Chris@10: 					     }
Chris@10: 					}
Chris@10: 					T1N = FMA(KP707106781, T1w, T1t);
Chris@10: 					T1x = FNMS(KP707106781, T1w, T1t);
Chris@10: 					T1A = T1y - T1z;
Chris@10: 					T1K = T1y + T1z;
Chris@10: 					{
Chris@10: 					     E T1E, T1l, T1H, T1B;
Chris@10: 					     T1E = FMA(KP923879532, T1k, TX);
Chris@10: 					     T1l = FNMS(KP923879532, T1k, TX);
Chris@10: 					     T1H = FMA(KP923879532, T1A, T1x);
Chris@10: 					     T1B = FNMS(KP923879532, T1A, T1x);
Chris@10: 					     {
Chris@10: 						  E T1I, T1F, T1C, T1m;
Chris@10: 						  T1I = T1G * T1E;
Chris@10: 						  T1F = T1D * T1E;
Chris@10: 						  T1C = T1o * T1l;
Chris@10: 						  T1m = TG * T1l;
Chris@10: 						  Im[WS(rs, 2)] = FMA(T1D, T1H, T1I);
Chris@10: 						  Ip[WS(rs, 2)] = FNMS(T1G, T1H, T1F);
Chris@10: 						  Im[WS(rs, 6)] = FMA(TG, T1B, T1C);
Chris@10: 						  Ip[WS(rs, 6)] = FNMS(T1o, T1B, T1m);
Chris@10: 					     }
Chris@10: 					}
Chris@10: 					{
Chris@10: 					     E T2A, T2s, T2D, T2x;
Chris@10: 					     T2A = FMA(KP923879532, T2r, T2q);
Chris@10: 					     T2s = FNMS(KP923879532, T2r, T2q);
Chris@10: 					     T2D = FNMS(KP923879532, T2w, T2v);
Chris@10: 					     T2x = FMA(KP923879532, T2w, T2v);
Chris@10: 					     {
Chris@10: 						  E T2B, T2t, T2E, T2y;
Chris@10: 						  T2B = T2z * T2A;
Chris@10: 						  T2t = T2p * T2s;
Chris@10: 						  T2E = T2z * T2D;
Chris@10: 						  T2y = T2p * T2x;
Chris@10: 						  Ip[WS(rs, 7)] = FNMS(T2C, T2D, T2B);
Chris@10: 						  Ip[WS(rs, 3)] = FNMS(T2u, T2x, T2t);
Chris@10: 						  Im[WS(rs, 7)] = FMA(T2C, T2A, T2E);
Chris@10: 						  Im[WS(rs, 3)] = FMA(T2u, T2s, T2y);
Chris@10: 					     }
Chris@10: 					}
Chris@10: 				   }
Chris@10: 			      }
Chris@10: 			 }
Chris@10: 		    }
Chris@10: 	       }
Chris@10: 	       {
Chris@10: 		    E T1L, T1R, T1P, T1T;
Chris@10: 		    T1L = FNMS(KP923879532, T1K, T1J);
Chris@10: 		    T1R = FMA(KP923879532, T1K, T1J);
Chris@10: 		    T1P = FNMS(KP923879532, T1O, T1N);
Chris@10: 		    T1T = FMA(KP923879532, T1O, T1N);
Chris@10: 		    {
Chris@10: 			 E T1S, T1M, T1U, T1Q;
Chris@10: 			 T1S = Tv * T1R;
Chris@10: 			 T1M = TB * T1L;
Chris@10: 			 T1U = Tv * T1T;
Chris@10: 			 T1Q = TB * T1P;
Chris@10: 			 Ip[0] = FNMS(Ty, T1T, T1S);
Chris@10: 			 Ip[WS(rs, 4)] = FNMS(TF, T1P, T1M);
Chris@10: 			 Im[0] = FMA(Ty, T1R, T1U);
Chris@10: 			 Im[WS(rs, 4)] = FMA(TF, T1L, T1Q);
Chris@10: 		    }
Chris@10: 	       }
Chris@10: 	  }
Chris@10:      }
Chris@10: }
Chris@10: 
Chris@10: static const tw_instr twinstr[] = {
Chris@10:      {TW_CEXP, 1, 1},
Chris@10:      {TW_CEXP, 1, 3},
Chris@10:      {TW_CEXP, 1, 9},
Chris@10:      {TW_CEXP, 1, 15},
Chris@10:      {TW_NEXT, 1, 0}
Chris@10: };
Chris@10: 
Chris@10: static const hc2c_desc desc = { 16, "hc2cb2_16", twinstr, &GENUS, {104, 42, 92, 0} };
Chris@10: 
Chris@10: void X(codelet_hc2cb2_16) (planner *p) {
Chris@10:      X(khc2c_register) (p, hc2cb2_16, &desc, HC2C_VIA_RDFT);
Chris@10: }
Chris@10: #else				/* HAVE_FMA */
Chris@10: 
Chris@10: /* Generated by: ../../../genfft/gen_hc2c.native -compact -variables 4 -pipeline-latency 4 -sign 1 -twiddle-log3 -precompute-twiddles -n 16 -dif -name hc2cb2_16 -include hc2cb.h */
Chris@10: 
Chris@10: /*
Chris@10:  * This function contains 196 FP additions, 108 FP multiplications,
Chris@10:  * (or, 156 additions, 68 multiplications, 40 fused multiply/add),
Chris@10:  * 80 stack variables, 3 constants, and 64 memory accesses
Chris@10:  */
Chris@10: #include "hc2cb.h"
Chris@10: 
Chris@10: static void hc2cb2_16(R *Rp, R *Ip, R *Rm, R *Im, const R *W, stride rs, INT mb, INT me, INT ms)
Chris@10: {
Chris@10:      DK(KP382683432, +0.382683432365089771728459984030398866761344562);
Chris@10:      DK(KP923879532, +0.923879532511286756128183189396788286822416626);
Chris@10:      DK(KP707106781, +0.707106781186547524400844362104849039284835938);
Chris@10:      {
Chris@10: 	  INT m;
Chris@10: 	  for (m = mb, W = W + ((mb - 1) * 8); m < me; m = m + 1, Rp = Rp + ms, Ip = Ip + ms, Rm = Rm - ms, Im = Im - ms, W = W + 8, MAKE_VOLATILE_STRIDE(64, rs)) {
Chris@10: 	       E Tv, Ty, T1l, T1n, T1p, T1t, T27, T25, Tz, Tw, TB, T21, T1P, T1H, T1X;
Chris@10: 	       E T17, T1L, T1N, T1v, T1w, T1x, T1B, T2F, T2T, T2b, T2R, T3j, T3x, T35, T3t;
Chris@10: 	       {
Chris@10: 		    E TA, T1J, T15, T1G, Tx, T1K, T16, T1F;
Chris@10: 		    {
Chris@10: 			 E T1m, T1s, T1o, T1r;
Chris@10: 			 Tv = W[0];
Chris@10: 			 Ty = W[1];
Chris@10: 			 T1l = W[2];
Chris@10: 			 T1n = W[3];
Chris@10: 			 T1m = Tv * T1l;
Chris@10: 			 T1s = Ty * T1l;
Chris@10: 			 T1o = Ty * T1n;
Chris@10: 			 T1r = Tv * T1n;
Chris@10: 			 T1p = T1m + T1o;
Chris@10: 			 T1t = T1r - T1s;
Chris@10: 			 T27 = T1r + T1s;
Chris@10: 			 T25 = T1m - T1o;
Chris@10: 			 Tz = W[5];
Chris@10: 			 TA = Ty * Tz;
Chris@10: 			 T1J = T1l * Tz;
Chris@10: 			 T15 = Tv * Tz;
Chris@10: 			 T1G = T1n * Tz;
Chris@10: 			 Tw = W[4];
Chris@10: 			 Tx = Tv * Tw;
Chris@10: 			 T1K = T1n * Tw;
Chris@10: 			 T16 = Ty * Tw;
Chris@10: 			 T1F = T1l * Tw;
Chris@10: 		    }
Chris@10: 		    TB = Tx - TA;
Chris@10: 		    T21 = T1J + T1K;
Chris@10: 		    T1P = T15 - T16;
Chris@10: 		    T1H = T1F + T1G;
Chris@10: 		    T1X = T1F - T1G;
Chris@10: 		    T17 = T15 + T16;
Chris@10: 		    T1L = T1J - T1K;
Chris@10: 		    T1N = Tx + TA;
Chris@10: 		    T1v = W[6];
Chris@10: 		    T1w = W[7];
Chris@10: 		    T1x = FMA(Tv, T1v, Ty * T1w);
Chris@10: 		    T1B = FNMS(Ty, T1v, Tv * T1w);
Chris@10: 		    {
Chris@10: 			 E T2D, T2E, T29, T2a;
Chris@10: 			 T2D = T25 * Tz;
Chris@10: 			 T2E = T27 * Tw;
Chris@10: 			 T2F = T2D + T2E;
Chris@10: 			 T2T = T2D - T2E;
Chris@10: 			 T29 = T25 * Tw;
Chris@10: 			 T2a = T27 * Tz;
Chris@10: 			 T2b = T29 - T2a;
Chris@10: 			 T2R = T29 + T2a;
Chris@10: 		    }
Chris@10: 		    {
Chris@10: 			 E T3h, T3i, T33, T34;
Chris@10: 			 T3h = T1p * Tz;
Chris@10: 			 T3i = T1t * Tw;
Chris@10: 			 T3j = T3h + T3i;
Chris@10: 			 T3x = T3h - T3i;
Chris@10: 			 T33 = T1p * Tw;
Chris@10: 			 T34 = T1t * Tz;
Chris@10: 			 T35 = T33 - T34;
Chris@10: 			 T3t = T33 + T34;
Chris@10: 		    }
Chris@10: 	       }
Chris@10: 	       {
Chris@10: 		    E T7, T36, T3k, TC, T1f, T2e, T2I, T1Q, Te, TJ, T1R, T18, T2L, T37, T2l;
Chris@10: 		    E T3l, Tm, T1T, TT, T1h, T2A, T2N, T3b, T3n, Tt, T1U, T12, T1i, T2t, T2O;
Chris@10: 		    E T3e, T3o;
Chris@10: 		    {
Chris@10: 			 E T3, T2c, T1b, T2H, T6, T2G, T1e, T2d;
Chris@10: 			 {
Chris@10: 			      E T1, T2, T19, T1a;
Chris@10: 			      T1 = Rp[0];
Chris@10: 			      T2 = Rm[WS(rs, 7)];
Chris@10: 			      T3 = T1 + T2;
Chris@10: 			      T2c = T1 - T2;
Chris@10: 			      T19 = Ip[0];
Chris@10: 			      T1a = Im[WS(rs, 7)];
Chris@10: 			      T1b = T19 - T1a;
Chris@10: 			      T2H = T19 + T1a;
Chris@10: 			 }
Chris@10: 			 {
Chris@10: 			      E T4, T5, T1c, T1d;
Chris@10: 			      T4 = Rp[WS(rs, 4)];
Chris@10: 			      T5 = Rm[WS(rs, 3)];
Chris@10: 			      T6 = T4 + T5;
Chris@10: 			      T2G = T4 - T5;
Chris@10: 			      T1c = Ip[WS(rs, 4)];
Chris@10: 			      T1d = Im[WS(rs, 3)];
Chris@10: 			      T1e = T1c - T1d;
Chris@10: 			      T2d = T1c + T1d;
Chris@10: 			 }
Chris@10: 			 T7 = T3 + T6;
Chris@10: 			 T36 = T2c + T2d;
Chris@10: 			 T3k = T2H - T2G;
Chris@10: 			 TC = T3 - T6;
Chris@10: 			 T1f = T1b - T1e;
Chris@10: 			 T2e = T2c - T2d;
Chris@10: 			 T2I = T2G + T2H;
Chris@10: 			 T1Q = T1b + T1e;
Chris@10: 		    }
Chris@10: 		    {
Chris@10: 			 E Ta, T2f, TI, T2g, Td, T2i, TF, T2j;
Chris@10: 			 {
Chris@10: 			      E T8, T9, TG, TH;
Chris@10: 			      T8 = Rp[WS(rs, 2)];
Chris@10: 			      T9 = Rm[WS(rs, 5)];
Chris@10: 			      Ta = T8 + T9;
Chris@10: 			      T2f = T8 - T9;
Chris@10: 			      TG = Ip[WS(rs, 2)];
Chris@10: 			      TH = Im[WS(rs, 5)];
Chris@10: 			      TI = TG - TH;
Chris@10: 			      T2g = TG + TH;
Chris@10: 			 }
Chris@10: 			 {
Chris@10: 			      E Tb, Tc, TD, TE;
Chris@10: 			      Tb = Rm[WS(rs, 1)];
Chris@10: 			      Tc = Rp[WS(rs, 6)];
Chris@10: 			      Td = Tb + Tc;
Chris@10: 			      T2i = Tb - Tc;
Chris@10: 			      TD = Ip[WS(rs, 6)];
Chris@10: 			      TE = Im[WS(rs, 1)];
Chris@10: 			      TF = TD - TE;
Chris@10: 			      T2j = TD + TE;
Chris@10: 			 }
Chris@10: 			 Te = Ta + Td;
Chris@10: 			 TJ = TF - TI;
Chris@10: 			 T1R = TI + TF;
Chris@10: 			 T18 = Ta - Td;
Chris@10: 			 {
Chris@10: 			      E T2J, T2K, T2h, T2k;
Chris@10: 			      T2J = T2f + T2g;
Chris@10: 			      T2K = T2i + T2j;
Chris@10: 			      T2L = KP707106781 * (T2J - T2K);
Chris@10: 			      T37 = KP707106781 * (T2J + T2K);
Chris@10: 			      T2h = T2f - T2g;
Chris@10: 			      T2k = T2i - T2j;
Chris@10: 			      T2l = KP707106781 * (T2h + T2k);
Chris@10: 			      T3l = KP707106781 * (T2h - T2k);
Chris@10: 			 }
Chris@10: 		    }
Chris@10: 		    {
Chris@10: 			 E Ti, T2x, TO, T2v, Tl, T2u, TR, T2y, TL, TS;
Chris@10: 			 {
Chris@10: 			      E Tg, Th, TM, TN;
Chris@10: 			      Tg = Rp[WS(rs, 1)];
Chris@10: 			      Th = Rm[WS(rs, 6)];
Chris@10: 			      Ti = Tg + Th;
Chris@10: 			      T2x = Tg - Th;
Chris@10: 			      TM = Ip[WS(rs, 1)];
Chris@10: 			      TN = Im[WS(rs, 6)];
Chris@10: 			      TO = TM - TN;
Chris@10: 			      T2v = TM + TN;
Chris@10: 			 }
Chris@10: 			 {
Chris@10: 			      E Tj, Tk, TP, TQ;
Chris@10: 			      Tj = Rp[WS(rs, 5)];
Chris@10: 			      Tk = Rm[WS(rs, 2)];
Chris@10: 			      Tl = Tj + Tk;
Chris@10: 			      T2u = Tj - Tk;
Chris@10: 			      TP = Ip[WS(rs, 5)];
Chris@10: 			      TQ = Im[WS(rs, 2)];
Chris@10: 			      TR = TP - TQ;
Chris@10: 			      T2y = TP + TQ;
Chris@10: 			 }
Chris@10: 			 Tm = Ti + Tl;
Chris@10: 			 T1T = TO + TR;
Chris@10: 			 TL = Ti - Tl;
Chris@10: 			 TS = TO - TR;
Chris@10: 			 TT = TL - TS;
Chris@10: 			 T1h = TL + TS;
Chris@10: 			 {
Chris@10: 			      E T2w, T2z, T39, T3a;
Chris@10: 			      T2w = T2u + T2v;
Chris@10: 			      T2z = T2x - T2y;
Chris@10: 			      T2A = FMA(KP923879532, T2w, KP382683432 * T2z);
Chris@10: 			      T2N = FNMS(KP382683432, T2w, KP923879532 * T2z);
Chris@10: 			      T39 = T2x + T2y;
Chris@10: 			      T3a = T2v - T2u;
Chris@10: 			      T3b = FNMS(KP923879532, T3a, KP382683432 * T39);
Chris@10: 			      T3n = FMA(KP382683432, T3a, KP923879532 * T39);
Chris@10: 			 }
Chris@10: 		    }
Chris@10: 		    {
Chris@10: 			 E Tp, T2q, TX, T2o, Ts, T2n, T10, T2r, TU, T11;
Chris@10: 			 {
Chris@10: 			      E Tn, To, TV, TW;
Chris@10: 			      Tn = Rm[0];
Chris@10: 			      To = Rp[WS(rs, 7)];
Chris@10: 			      Tp = Tn + To;
Chris@10: 			      T2q = Tn - To;
Chris@10: 			      TV = Ip[WS(rs, 7)];
Chris@10: 			      TW = Im[0];
Chris@10: 			      TX = TV - TW;
Chris@10: 			      T2o = TV + TW;
Chris@10: 			 }
Chris@10: 			 {
Chris@10: 			      E Tq, Tr, TY, TZ;
Chris@10: 			      Tq = Rp[WS(rs, 3)];
Chris@10: 			      Tr = Rm[WS(rs, 4)];
Chris@10: 			      Ts = Tq + Tr;
Chris@10: 			      T2n = Tq - Tr;
Chris@10: 			      TY = Ip[WS(rs, 3)];
Chris@10: 			      TZ = Im[WS(rs, 4)];
Chris@10: 			      T10 = TY - TZ;
Chris@10: 			      T2r = TY + TZ;
Chris@10: 			 }
Chris@10: 			 Tt = Tp + Ts;
Chris@10: 			 T1U = TX + T10;
Chris@10: 			 TU = Tp - Ts;
Chris@10: 			 T11 = TX - T10;
Chris@10: 			 T12 = TU + T11;
Chris@10: 			 T1i = T11 - TU;
Chris@10: 			 {
Chris@10: 			      E T2p, T2s, T3c, T3d;
Chris@10: 			      T2p = T2n - T2o;
Chris@10: 			      T2s = T2q - T2r;
Chris@10: 			      T2t = FNMS(KP382683432, T2s, KP923879532 * T2p);
Chris@10: 			      T2O = FMA(KP382683432, T2p, KP923879532 * T2s);
Chris@10: 			      T3c = T2q + T2r;
Chris@10: 			      T3d = T2n + T2o;
Chris@10: 			      T3e = FNMS(KP923879532, T3d, KP382683432 * T3c);
Chris@10: 			      T3o = FMA(KP382683432, T3d, KP923879532 * T3c);
Chris@10: 			 }
Chris@10: 		    }
Chris@10: 		    {
Chris@10: 			 E Tf, Tu, T1O, T1S, T1V, T1W;
Chris@10: 			 Tf = T7 + Te;
Chris@10: 			 Tu = Tm + Tt;
Chris@10: 			 T1O = Tf - Tu;
Chris@10: 			 T1S = T1Q + T1R;
Chris@10: 			 T1V = T1T + T1U;
Chris@10: 			 T1W = T1S - T1V;
Chris@10: 			 Rp[0] = Tf + Tu;
Chris@10: 			 Rm[0] = T1S + T1V;
Chris@10: 			 Rp[WS(rs, 4)] = FNMS(T1P, T1W, T1N * T1O);
Chris@10: 			 Rm[WS(rs, 4)] = FMA(T1P, T1O, T1N * T1W);
Chris@10: 		    }
Chris@10: 		    {
Chris@10: 			 E T3g, T3r, T3q, T3s;
Chris@10: 			 {
Chris@10: 			      E T38, T3f, T3m, T3p;
Chris@10: 			      T38 = T36 - T37;
Chris@10: 			      T3f = T3b + T3e;
Chris@10: 			      T3g = T38 - T3f;
Chris@10: 			      T3r = T38 + T3f;
Chris@10: 			      T3m = T3k + T3l;
Chris@10: 			      T3p = T3n - T3o;
Chris@10: 			      T3q = T3m - T3p;
Chris@10: 			      T3s = T3m + T3p;
Chris@10: 			 }
Chris@10: 			 Ip[WS(rs, 5)] = FNMS(T3j, T3q, T35 * T3g);
Chris@10: 			 Im[WS(rs, 5)] = FMA(T3j, T3g, T35 * T3q);
Chris@10: 			 Ip[WS(rs, 1)] = FNMS(T1n, T3s, T1l * T3r);
Chris@10: 			 Im[WS(rs, 1)] = FMA(T1n, T3r, T1l * T3s);
Chris@10: 		    }
Chris@10: 		    {
Chris@10: 			 E T3w, T3B, T3A, T3C;
Chris@10: 			 {
Chris@10: 			      E T3u, T3v, T3y, T3z;
Chris@10: 			      T3u = T36 + T37;
Chris@10: 			      T3v = T3n + T3o;
Chris@10: 			      T3w = T3u - T3v;
Chris@10: 			      T3B = T3u + T3v;
Chris@10: 			      T3y = T3k - T3l;
Chris@10: 			      T3z = T3b - T3e;
Chris@10: 			      T3A = T3y + T3z;
Chris@10: 			      T3C = T3y - T3z;
Chris@10: 			 }
Chris@10: 			 Ip[WS(rs, 3)] = FNMS(T3x, T3A, T3t * T3w);
Chris@10: 			 Im[WS(rs, 3)] = FMA(T3t, T3A, T3x * T3w);
Chris@10: 			 Ip[WS(rs, 7)] = FNMS(T1w, T3C, T1v * T3B);
Chris@10: 			 Im[WS(rs, 7)] = FMA(T1v, T3C, T1w * T3B);
Chris@10: 		    }
Chris@10: 		    {
Chris@10: 			 E T14, T1q, T1k, T1u;
Chris@10: 			 {
Chris@10: 			      E TK, T13, T1g, T1j;
Chris@10: 			      TK = TC + TJ;
Chris@10: 			      T13 = KP707106781 * (TT + T12);
Chris@10: 			      T14 = TK - T13;
Chris@10: 			      T1q = TK + T13;
Chris@10: 			      T1g = T18 + T1f;
Chris@10: 			      T1j = KP707106781 * (T1h + T1i);
Chris@10: 			      T1k = T1g - T1j;
Chris@10: 			      T1u = T1g + T1j;
Chris@10: 			 }
Chris@10: 			 Rp[WS(rs, 5)] = FNMS(T17, T1k, TB * T14);
Chris@10: 			 Rm[WS(rs, 5)] = FMA(T17, T14, TB * T1k);
Chris@10: 			 Rp[WS(rs, 1)] = FNMS(T1t, T1u, T1p * T1q);
Chris@10: 			 Rm[WS(rs, 1)] = FMA(T1t, T1q, T1p * T1u);
Chris@10: 		    }
Chris@10: 		    {
Chris@10: 			 E T1A, T1I, T1E, T1M;
Chris@10: 			 {
Chris@10: 			      E T1y, T1z, T1C, T1D;
Chris@10: 			      T1y = TC - TJ;
Chris@10: 			      T1z = KP707106781 * (T1i - T1h);
Chris@10: 			      T1A = T1y - T1z;
Chris@10: 			      T1I = T1y + T1z;
Chris@10: 			      T1C = T1f - T18;
Chris@10: 			      T1D = KP707106781 * (TT - T12);
Chris@10: 			      T1E = T1C - T1D;
Chris@10: 			      T1M = T1C + T1D;
Chris@10: 			 }
Chris@10: 			 Rp[WS(rs, 7)] = FNMS(T1B, T1E, T1x * T1A);
Chris@10: 			 Rm[WS(rs, 7)] = FMA(T1x, T1E, T1B * T1A);
Chris@10: 			 Rp[WS(rs, 3)] = FNMS(T1L, T1M, T1H * T1I);
Chris@10: 			 Rm[WS(rs, 3)] = FMA(T1H, T1M, T1L * T1I);
Chris@10: 		    }
Chris@10: 		    {
Chris@10: 			 E T2C, T2S, T2Q, T2U;
Chris@10: 			 {
Chris@10: 			      E T2m, T2B, T2M, T2P;
Chris@10: 			      T2m = T2e - T2l;
Chris@10: 			      T2B = T2t - T2A;
Chris@10: 			      T2C = T2m - T2B;
Chris@10: 			      T2S = T2m + T2B;
Chris@10: 			      T2M = T2I - T2L;
Chris@10: 			      T2P = T2N - T2O;
Chris@10: 			      T2Q = T2M - T2P;
Chris@10: 			      T2U = T2M + T2P;
Chris@10: 			 }
Chris@10: 			 Ip[WS(rs, 6)] = FNMS(T2F, T2Q, T2b * T2C);
Chris@10: 			 Im[WS(rs, 6)] = FMA(T2F, T2C, T2b * T2Q);
Chris@10: 			 Ip[WS(rs, 2)] = FNMS(T2T, T2U, T2R * T2S);
Chris@10: 			 Im[WS(rs, 2)] = FMA(T2T, T2S, T2R * T2U);
Chris@10: 		    }
Chris@10: 		    {
Chris@10: 			 E T2X, T31, T30, T32;
Chris@10: 			 {
Chris@10: 			      E T2V, T2W, T2Y, T2Z;
Chris@10: 			      T2V = T2e + T2l;
Chris@10: 			      T2W = T2N + T2O;
Chris@10: 			      T2X = T2V - T2W;
Chris@10: 			      T31 = T2V + T2W;
Chris@10: 			      T2Y = T2I + T2L;
Chris@10: 			      T2Z = T2A + T2t;
Chris@10: 			      T30 = T2Y - T2Z;
Chris@10: 			      T32 = T2Y + T2Z;
Chris@10: 			 }
Chris@10: 			 Ip[WS(rs, 4)] = FNMS(Tz, T30, Tw * T2X);
Chris@10: 			 Im[WS(rs, 4)] = FMA(Tw, T30, Tz * T2X);
Chris@10: 			 Ip[0] = FNMS(Ty, T32, Tv * T31);
Chris@10: 			 Im[0] = FMA(Tv, T32, Ty * T31);
Chris@10: 		    }
Chris@10: 		    {
Chris@10: 			 E T20, T26, T24, T28;
Chris@10: 			 {
Chris@10: 			      E T1Y, T1Z, T22, T23;
Chris@10: 			      T1Y = T7 - Te;
Chris@10: 			      T1Z = T1U - T1T;
Chris@10: 			      T20 = T1Y - T1Z;
Chris@10: 			      T26 = T1Y + T1Z;
Chris@10: 			      T22 = T1Q - T1R;
Chris@10: 			      T23 = Tm - Tt;
Chris@10: 			      T24 = T22 - T23;
Chris@10: 			      T28 = T23 + T22;
Chris@10: 			 }
Chris@10: 			 Rp[WS(rs, 6)] = FNMS(T21, T24, T1X * T20);
Chris@10: 			 Rm[WS(rs, 6)] = FMA(T1X, T24, T21 * T20);
Chris@10: 			 Rp[WS(rs, 2)] = FNMS(T27, T28, T25 * T26);
Chris@10: 			 Rm[WS(rs, 2)] = FMA(T25, T28, T27 * T26);
Chris@10: 		    }
Chris@10: 	       }
Chris@10: 	  }
Chris@10:      }
Chris@10: }
Chris@10: 
Chris@10: static const tw_instr twinstr[] = {
Chris@10:      {TW_CEXP, 1, 1},
Chris@10:      {TW_CEXP, 1, 3},
Chris@10:      {TW_CEXP, 1, 9},
Chris@10:      {TW_CEXP, 1, 15},
Chris@10:      {TW_NEXT, 1, 0}
Chris@10: };
Chris@10: 
Chris@10: static const hc2c_desc desc = { 16, "hc2cb2_16", twinstr, &GENUS, {156, 68, 40, 0} };
Chris@10: 
Chris@10: void X(codelet_hc2cb2_16) (planner *p) {
Chris@10:      X(khc2c_register) (p, hc2cb2_16, &desc, HC2C_VIA_RDFT);
Chris@10: }
Chris@10: #endif				/* HAVE_FMA */