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 8 -dif -name hc2cb2_8 -include hc2cb.h */
Chris@10: 
Chris@10: /*
Chris@10:  * This function contains 74 FP additions, 50 FP multiplications,
Chris@10:  * (or, 44 additions, 20 multiplications, 30 fused multiply/add),
Chris@10:  * 64 stack variables, 1 constants, and 32 memory accesses
Chris@10:  */
Chris@10: #include "hc2cb.h"
Chris@10: 
Chris@10: static void hc2cb2_8(R *Rp, R *Ip, R *Rm, R *Im, const R *W, stride rs, INT mb, INT me, INT ms)
Chris@10: {
Chris@10:      DK(KP707106781, +0.707106781186547524400844362104849039284835938);
Chris@10:      {
Chris@10: 	  INT m;
Chris@10: 	  for (m = mb, W = W + ((mb - 1) * 6); m < me; m = m + 1, Rp = Rp + ms, Ip = Ip + ms, Rm = Rm - ms, Im = Im - ms, W = W + 6, MAKE_VOLATILE_STRIDE(32, rs)) {
Chris@10: 	       E Tf, Ti, TK, Tq, TH, TT, TX, TW, TY, TU, TI;
Chris@10: 	       {
Chris@10: 		    E Tg, Tl, Tp, Th, T1n, T1t, Tj;
Chris@10: 		    Tf = W[0];
Chris@10: 		    Tg = W[2];
Chris@10: 		    Tl = W[4];
Chris@10: 		    Tp = W[5];
Chris@10: 		    Ti = W[1];
Chris@10: 		    Th = Tf * Tg;
Chris@10: 		    T1n = Tf * Tl;
Chris@10: 		    T1t = Tf * Tp;
Chris@10: 		    Tj = W[3];
Chris@10: 		    {
Chris@10: 			 E T1o, T1u, Tk, T1b, To, T1e, T13, TP, T1p, T7, T1h, T1v, TZ, Tv, T1i;
Chris@10: 			 E TB, TA, TQ, Te, T1w, TE, T1j;
Chris@10: 			 {
Chris@10: 			      E Tr, T3, Ts, T1f, TO, TL, T6, Tt;
Chris@10: 			      {
Chris@10: 				   E TM, TN, T4, T5;
Chris@10: 				   {
Chris@10: 					E T1, Tn, T2, TJ, Tm;
Chris@10: 					T1 = Rp[0];
Chris@10: 					T1o = FMA(Ti, Tp, T1n);
Chris@10: 					T1u = FNMS(Ti, Tl, T1t);
Chris@10: 					Tk = FMA(Ti, Tj, Th);
Chris@10: 					T1b = FNMS(Ti, Tj, Th);
Chris@10: 					Tn = Tf * Tj;
Chris@10: 					T2 = Rm[WS(rs, 3)];
Chris@10: 					TM = Ip[0];
Chris@10: 					TJ = Tk * Tp;
Chris@10: 					Tm = Tk * Tl;
Chris@10: 					To = FNMS(Ti, Tg, Tn);
Chris@10: 					T1e = FMA(Ti, Tg, Tn);
Chris@10: 					Tr = T1 - T2;
Chris@10: 					T3 = T1 + T2;
Chris@10: 					TK = FNMS(To, Tl, TJ);
Chris@10: 					Tq = FMA(To, Tp, Tm);
Chris@10: 					TN = Im[WS(rs, 3)];
Chris@10: 				   }
Chris@10: 				   T4 = Rp[WS(rs, 2)];
Chris@10: 				   T5 = Rm[WS(rs, 1)];
Chris@10: 				   Ts = Ip[WS(rs, 2)];
Chris@10: 				   T1f = TM - TN;
Chris@10: 				   TO = TM + TN;
Chris@10: 				   TL = T4 - T5;
Chris@10: 				   T6 = T4 + T5;
Chris@10: 				   Tt = Im[WS(rs, 1)];
Chris@10: 			      }
Chris@10: 			      {
Chris@10: 				   E Tw, Ta, TC, Tz, Td, TD;
Chris@10: 				   {
Chris@10: 					E Tx, Ty, Tb, Tc;
Chris@10: 					{
Chris@10: 					     E T8, T1g, Tu, T9;
Chris@10: 					     T8 = Rp[WS(rs, 1)];
Chris@10: 					     T13 = TO - TL;
Chris@10: 					     TP = TL + TO;
Chris@10: 					     T1p = T3 - T6;
Chris@10: 					     T7 = T3 + T6;
Chris@10: 					     T1g = Ts - Tt;
Chris@10: 					     Tu = Ts + Tt;
Chris@10: 					     T9 = Rm[WS(rs, 2)];
Chris@10: 					     Tx = Ip[WS(rs, 1)];
Chris@10: 					     T1h = T1f + T1g;
Chris@10: 					     T1v = T1f - T1g;
Chris@10: 					     TZ = Tr + Tu;
Chris@10: 					     Tv = Tr - Tu;
Chris@10: 					     Tw = T8 - T9;
Chris@10: 					     Ta = T8 + T9;
Chris@10: 					     Ty = Im[WS(rs, 2)];
Chris@10: 					}
Chris@10: 					Tb = Rm[0];
Chris@10: 					Tc = Rp[WS(rs, 3)];
Chris@10: 					TC = Ip[WS(rs, 3)];
Chris@10: 					T1i = Tx - Ty;
Chris@10: 					Tz = Tx + Ty;
Chris@10: 					TB = Tb - Tc;
Chris@10: 					Td = Tb + Tc;
Chris@10: 					TD = Im[0];
Chris@10: 				   }
Chris@10: 				   TA = Tw - Tz;
Chris@10: 				   TQ = Tw + Tz;
Chris@10: 				   Te = Ta + Td;
Chris@10: 				   T1w = Ta - Td;
Chris@10: 				   TE = TC + TD;
Chris@10: 				   T1j = TC - TD;
Chris@10: 			      }
Chris@10: 			 }
Chris@10: 			 {
Chris@10: 			      E T1x, T1k, T1r, TG, TS, T19, T15, T17, T11, T16, T12;
Chris@10: 			      {
Chris@10: 				   E T1B, T1z, T10, T1A, T1C;
Chris@10: 				   T1x = T1v - T1w;
Chris@10: 				   T1B = T1w + T1v;
Chris@10: 				   Rp[0] = T7 + Te;
Chris@10: 				   {
Chris@10: 					E T1q, TR, TF, T14;
Chris@10: 					T1k = T1i + T1j;
Chris@10: 					T1q = T1j - T1i;
Chris@10: 					TR = TB + TE;
Chris@10: 					TF = TB - TE;
Chris@10: 					T1r = T1p - T1q;
Chris@10: 					T1z = T1p + T1q;
Chris@10: 					Rm[0] = T1h + T1k;
Chris@10: 					TG = TA + TF;
Chris@10: 					T14 = TA - TF;
Chris@10: 					TS = TQ - TR;
Chris@10: 					T10 = TQ + TR;
Chris@10: 					T1A = Tk * T1z;
Chris@10: 					T19 = FNMS(KP707106781, T14, T13);
Chris@10: 					T15 = FMA(KP707106781, T14, T13);
Chris@10: 					T1C = Tk * T1B;
Chris@10: 				   }
Chris@10: 				   T17 = FMA(KP707106781, T10, TZ);
Chris@10: 				   T11 = FNMS(KP707106781, T10, TZ);
Chris@10: 				   Rp[WS(rs, 1)] = FNMS(To, T1B, T1A);
Chris@10: 				   T16 = Tg * T15;
Chris@10: 				   Rm[WS(rs, 1)] = FMA(To, T1z, T1C);
Chris@10: 			      }
Chris@10: 			      T12 = Tg * T11;
Chris@10: 			      {
Chris@10: 				   E T1l, T1a, T1c, T18;
Chris@10: 				   Im[WS(rs, 1)] = FMA(Tj, T11, T16);
Chris@10: 				   Ip[WS(rs, 1)] = FNMS(Tj, T15, T12);
Chris@10: 				   T18 = Tl * T17;
Chris@10: 				   T1l = T1h - T1k;
Chris@10: 				   T1a = Tl * T19;
Chris@10: 				   T1c = T7 - Te;
Chris@10: 				   Ip[WS(rs, 3)] = FNMS(Tp, T19, T18);
Chris@10: 				   {
Chris@10: 					E T1s, T1m, T1d, T1y, TV;
Chris@10: 					Im[WS(rs, 3)] = FMA(Tp, T17, T1a);
Chris@10: 					T1m = T1e * T1c;
Chris@10: 					T1d = T1b * T1c;
Chris@10: 					T1s = T1o * T1r;
Chris@10: 					Rm[WS(rs, 2)] = FMA(T1b, T1l, T1m);
Chris@10: 					Rp[WS(rs, 2)] = FNMS(T1e, T1l, T1d);
Chris@10: 					Rp[WS(rs, 3)] = FNMS(T1u, T1x, T1s);
Chris@10: 					T1y = T1o * T1x;
Chris@10: 					TV = FMA(KP707106781, TG, Tv);
Chris@10: 					TH = FNMS(KP707106781, TG, Tv);
Chris@10: 					TT = FNMS(KP707106781, TS, TP);
Chris@10: 					TX = FMA(KP707106781, TS, TP);
Chris@10: 					Rm[WS(rs, 3)] = FMA(T1u, T1r, T1y);
Chris@10: 					TW = Tf * TV;
Chris@10: 					TY = Ti * TV;
Chris@10: 				   }
Chris@10: 			      }
Chris@10: 			 }
Chris@10: 		    }
Chris@10: 	       }
Chris@10: 	       Ip[0] = FNMS(Ti, TX, TW);
Chris@10: 	       Im[0] = FMA(Tf, TX, TY);
Chris@10: 	       TU = TK * TH;
Chris@10: 	       TI = Tq * TH;
Chris@10: 	       Im[WS(rs, 2)] = FMA(Tq, TT, TU);
Chris@10: 	       Ip[WS(rs, 2)] = FNMS(TK, TT, TI);
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, 7},
Chris@10:      {TW_NEXT, 1, 0}
Chris@10: };
Chris@10: 
Chris@10: static const hc2c_desc desc = { 8, "hc2cb2_8", twinstr, &GENUS, {44, 20, 30, 0} };
Chris@10: 
Chris@10: void X(codelet_hc2cb2_8) (planner *p) {
Chris@10:      X(khc2c_register) (p, hc2cb2_8, &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 8 -dif -name hc2cb2_8 -include hc2cb.h */
Chris@10: 
Chris@10: /*
Chris@10:  * This function contains 74 FP additions, 44 FP multiplications,
Chris@10:  * (or, 56 additions, 26 multiplications, 18 fused multiply/add),
Chris@10:  * 46 stack variables, 1 constants, and 32 memory accesses
Chris@10:  */
Chris@10: #include "hc2cb.h"
Chris@10: 
Chris@10: static void hc2cb2_8(R *Rp, R *Ip, R *Rm, R *Im, const R *W, stride rs, INT mb, INT me, INT ms)
Chris@10: {
Chris@10:      DK(KP707106781, +0.707106781186547524400844362104849039284835938);
Chris@10:      {
Chris@10: 	  INT m;
Chris@10: 	  for (m = mb, W = W + ((mb - 1) * 6); m < me; m = m + 1, Rp = Rp + ms, Ip = Ip + ms, Rm = Rm - ms, Im = Im - ms, W = W + 6, MAKE_VOLATILE_STRIDE(32, rs)) {
Chris@10: 	       E Tf, Ti, Tg, Tj, Tl, Tp, TP, TR, TF, TG, TH, T15, TL, TT;
Chris@10: 	       {
Chris@10: 		    E Th, To, Tk, Tn;
Chris@10: 		    Tf = W[0];
Chris@10: 		    Ti = W[1];
Chris@10: 		    Tg = W[2];
Chris@10: 		    Tj = W[3];
Chris@10: 		    Th = Tf * Tg;
Chris@10: 		    To = Ti * Tg;
Chris@10: 		    Tk = Ti * Tj;
Chris@10: 		    Tn = Tf * Tj;
Chris@10: 		    Tl = Th - Tk;
Chris@10: 		    Tp = Tn + To;
Chris@10: 		    TP = Th + Tk;
Chris@10: 		    TR = Tn - To;
Chris@10: 		    TF = W[4];
Chris@10: 		    TG = W[5];
Chris@10: 		    TH = FMA(Tf, TF, Ti * TG);
Chris@10: 		    T15 = FNMS(TR, TF, TP * TG);
Chris@10: 		    TL = FNMS(Ti, TF, Tf * TG);
Chris@10: 		    TT = FMA(TP, TF, TR * TG);
Chris@10: 	       }
Chris@10: 	       {
Chris@10: 		    E T7, T1f, T1i, Tw, TI, TW, T18, TM, Te, T19, T1a, TD, TJ, TZ, T12;
Chris@10: 		    E TN, Tm, TE;
Chris@10: 		    {
Chris@10: 			 E T3, TU, Ts, T17, T6, T16, Tv, TV;
Chris@10: 			 {
Chris@10: 			      E T1, T2, Tq, Tr;
Chris@10: 			      T1 = Rp[0];
Chris@10: 			      T2 = Rm[WS(rs, 3)];
Chris@10: 			      T3 = T1 + T2;
Chris@10: 			      TU = T1 - T2;
Chris@10: 			      Tq = Ip[0];
Chris@10: 			      Tr = Im[WS(rs, 3)];
Chris@10: 			      Ts = Tq - Tr;
Chris@10: 			      T17 = Tq + Tr;
Chris@10: 			 }
Chris@10: 			 {
Chris@10: 			      E T4, T5, Tt, Tu;
Chris@10: 			      T4 = Rp[WS(rs, 2)];
Chris@10: 			      T5 = Rm[WS(rs, 1)];
Chris@10: 			      T6 = T4 + T5;
Chris@10: 			      T16 = T4 - T5;
Chris@10: 			      Tt = Ip[WS(rs, 2)];
Chris@10: 			      Tu = Im[WS(rs, 1)];
Chris@10: 			      Tv = Tt - Tu;
Chris@10: 			      TV = Tt + Tu;
Chris@10: 			 }
Chris@10: 			 T7 = T3 + T6;
Chris@10: 			 T1f = TU + TV;
Chris@10: 			 T1i = T17 - T16;
Chris@10: 			 Tw = Ts + Tv;
Chris@10: 			 TI = T3 - T6;
Chris@10: 			 TW = TU - TV;
Chris@10: 			 T18 = T16 + T17;
Chris@10: 			 TM = Ts - Tv;
Chris@10: 		    }
Chris@10: 		    {
Chris@10: 			 E Ta, TX, Tz, TY, Td, T10, TC, T11;
Chris@10: 			 {
Chris@10: 			      E T8, T9, Tx, Ty;
Chris@10: 			      T8 = Rp[WS(rs, 1)];
Chris@10: 			      T9 = Rm[WS(rs, 2)];
Chris@10: 			      Ta = T8 + T9;
Chris@10: 			      TX = T8 - T9;
Chris@10: 			      Tx = Ip[WS(rs, 1)];
Chris@10: 			      Ty = Im[WS(rs, 2)];
Chris@10: 			      Tz = Tx - Ty;
Chris@10: 			      TY = Tx + Ty;
Chris@10: 			 }
Chris@10: 			 {
Chris@10: 			      E Tb, Tc, TA, TB;
Chris@10: 			      Tb = Rm[0];
Chris@10: 			      Tc = Rp[WS(rs, 3)];
Chris@10: 			      Td = Tb + Tc;
Chris@10: 			      T10 = Tb - Tc;
Chris@10: 			      TA = Ip[WS(rs, 3)];
Chris@10: 			      TB = Im[0];
Chris@10: 			      TC = TA - TB;
Chris@10: 			      T11 = TA + TB;
Chris@10: 			 }
Chris@10: 			 Te = Ta + Td;
Chris@10: 			 T19 = TX + TY;
Chris@10: 			 T1a = T10 + T11;
Chris@10: 			 TD = Tz + TC;
Chris@10: 			 TJ = TC - Tz;
Chris@10: 			 TZ = TX - TY;
Chris@10: 			 T12 = T10 - T11;
Chris@10: 			 TN = Ta - Td;
Chris@10: 		    }
Chris@10: 		    Rp[0] = T7 + Te;
Chris@10: 		    Rm[0] = Tw + TD;
Chris@10: 		    Tm = T7 - Te;
Chris@10: 		    TE = Tw - TD;
Chris@10: 		    Rp[WS(rs, 2)] = FNMS(Tp, TE, Tl * Tm);
Chris@10: 		    Rm[WS(rs, 2)] = FMA(Tp, Tm, Tl * TE);
Chris@10: 		    {
Chris@10: 			 E TQ, TS, TK, TO;
Chris@10: 			 TQ = TI + TJ;
Chris@10: 			 TS = TN + TM;
Chris@10: 			 Rp[WS(rs, 1)] = FNMS(TR, TS, TP * TQ);
Chris@10: 			 Rm[WS(rs, 1)] = FMA(TP, TS, TR * TQ);
Chris@10: 			 TK = TI - TJ;
Chris@10: 			 TO = TM - TN;
Chris@10: 			 Rp[WS(rs, 3)] = FNMS(TL, TO, TH * TK);
Chris@10: 			 Rm[WS(rs, 3)] = FMA(TH, TO, TL * TK);
Chris@10: 		    }
Chris@10: 		    {
Chris@10: 			 E T1h, T1l, T1k, T1m, T1g, T1j;
Chris@10: 			 T1g = KP707106781 * (T19 + T1a);
Chris@10: 			 T1h = T1f - T1g;
Chris@10: 			 T1l = T1f + T1g;
Chris@10: 			 T1j = KP707106781 * (TZ - T12);
Chris@10: 			 T1k = T1i + T1j;
Chris@10: 			 T1m = T1i - T1j;
Chris@10: 			 Ip[WS(rs, 1)] = FNMS(Tj, T1k, Tg * T1h);
Chris@10: 			 Im[WS(rs, 1)] = FMA(Tg, T1k, Tj * T1h);
Chris@10: 			 Ip[WS(rs, 3)] = FNMS(TG, T1m, TF * T1l);
Chris@10: 			 Im[WS(rs, 3)] = FMA(TF, T1m, TG * T1l);
Chris@10: 		    }
Chris@10: 		    {
Chris@10: 			 E T14, T1d, T1c, T1e, T13, T1b;
Chris@10: 			 T13 = KP707106781 * (TZ + T12);
Chris@10: 			 T14 = TW - T13;
Chris@10: 			 T1d = TW + T13;
Chris@10: 			 T1b = KP707106781 * (T19 - T1a);
Chris@10: 			 T1c = T18 - T1b;
Chris@10: 			 T1e = T18 + T1b;
Chris@10: 			 Ip[WS(rs, 2)] = FNMS(T15, T1c, TT * T14);
Chris@10: 			 Im[WS(rs, 2)] = FMA(T15, T14, TT * T1c);
Chris@10: 			 Ip[0] = FNMS(Ti, T1e, Tf * T1d);
Chris@10: 			 Im[0] = FMA(Ti, T1d, Tf * T1e);
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, 7},
Chris@10:      {TW_NEXT, 1, 0}
Chris@10: };
Chris@10: 
Chris@10: static const hc2c_desc desc = { 8, "hc2cb2_8", twinstr, &GENUS, {56, 26, 18, 0} };
Chris@10: 
Chris@10: void X(codelet_hc2cb2_8) (planner *p) {
Chris@10:      X(khc2c_register) (p, hc2cb2_8, &desc, HC2C_VIA_RDFT);
Chris@10: }
Chris@10: #endif				/* HAVE_FMA */