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 07:39:51 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 -n 8 -dit -name hf_8 -include hf.h */
Chris@10: 
Chris@10: /*
Chris@10:  * This function contains 66 FP additions, 36 FP multiplications,
Chris@10:  * (or, 44 additions, 14 multiplications, 22 fused multiply/add),
Chris@10:  * 61 stack variables, 1 constants, and 32 memory accesses
Chris@10:  */
Chris@10: #include "hf.h"
Chris@10: 
Chris@10: static void hf_8(R *cr, R *ci, 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) * 14); m < me; m = m + 1, cr = cr + ms, ci = ci - ms, W = W + 14, MAKE_VOLATILE_STRIDE(16, rs)) {
Chris@10: 	       E T1f, T1g, T1e, Tm, T1q, T1o, T1p, TN, T1h, T1i;
Chris@10: 	       {
Chris@10: 		    E T1, T1m, T1l, T7, TS, Tk, TQ, Te, To, Tr, T17, TM, T12, Tu, TW;
Chris@10: 		    E Tp, Tx, Tt, Tq, Tw;
Chris@10: 		    {
Chris@10: 			 E T3, T6, T2, T5;
Chris@10: 			 T1 = cr[0];
Chris@10: 			 T1m = ci[0];
Chris@10: 			 T3 = cr[WS(rs, 4)];
Chris@10: 			 T6 = ci[WS(rs, 4)];
Chris@10: 			 T2 = W[6];
Chris@10: 			 T5 = W[7];
Chris@10: 			 {
Chris@10: 			      E Ta, Td, T9, Tc;
Chris@10: 			      {
Chris@10: 				   E Tg, Tj, Ti, TR, Th, T1k, T4, Tf;
Chris@10: 				   Tg = cr[WS(rs, 6)];
Chris@10: 				   Tj = ci[WS(rs, 6)];
Chris@10: 				   T1k = T2 * T6;
Chris@10: 				   T4 = T2 * T3;
Chris@10: 				   Tf = W[10];
Chris@10: 				   Ti = W[11];
Chris@10: 				   T1l = FNMS(T5, T3, T1k);
Chris@10: 				   T7 = FMA(T5, T6, T4);
Chris@10: 				   TR = Tf * Tj;
Chris@10: 				   Th = Tf * Tg;
Chris@10: 				   Ta = cr[WS(rs, 2)];
Chris@10: 				   Td = ci[WS(rs, 2)];
Chris@10: 				   TS = FNMS(Ti, Tg, TR);
Chris@10: 				   Tk = FMA(Ti, Tj, Th);
Chris@10: 				   T9 = W[2];
Chris@10: 				   Tc = W[3];
Chris@10: 			      }
Chris@10: 			      {
Chris@10: 				   E TB, TE, TH, T13, TC, TK, TG, TD, TJ, TP, Tb, TA, Tn;
Chris@10: 				   TB = cr[WS(rs, 7)];
Chris@10: 				   TE = ci[WS(rs, 7)];
Chris@10: 				   TP = T9 * Td;
Chris@10: 				   Tb = T9 * Ta;
Chris@10: 				   TA = W[12];
Chris@10: 				   TH = cr[WS(rs, 3)];
Chris@10: 				   TQ = FNMS(Tc, Ta, TP);
Chris@10: 				   Te = FMA(Tc, Td, Tb);
Chris@10: 				   T13 = TA * TE;
Chris@10: 				   TC = TA * TB;
Chris@10: 				   TK = ci[WS(rs, 3)];
Chris@10: 				   TG = W[4];
Chris@10: 				   TD = W[13];
Chris@10: 				   TJ = W[5];
Chris@10: 				   {
Chris@10: 					E T14, TF, T16, TL, T15, TI;
Chris@10: 					To = cr[WS(rs, 1)];
Chris@10: 					T15 = TG * TK;
Chris@10: 					TI = TG * TH;
Chris@10: 					T14 = FNMS(TD, TB, T13);
Chris@10: 					TF = FMA(TD, TE, TC);
Chris@10: 					T16 = FNMS(TJ, TH, T15);
Chris@10: 					TL = FMA(TJ, TK, TI);
Chris@10: 					Tr = ci[WS(rs, 1)];
Chris@10: 					Tn = W[0];
Chris@10: 					T17 = T14 - T16;
Chris@10: 					T1f = T14 + T16;
Chris@10: 					TM = TF + TL;
Chris@10: 					T12 = TF - TL;
Chris@10: 				   }
Chris@10: 				   Tu = cr[WS(rs, 5)];
Chris@10: 				   TW = Tn * Tr;
Chris@10: 				   Tp = Tn * To;
Chris@10: 				   Tx = ci[WS(rs, 5)];
Chris@10: 				   Tt = W[8];
Chris@10: 				   Tq = W[1];
Chris@10: 				   Tw = W[9];
Chris@10: 			      }
Chris@10: 			 }
Chris@10: 		    }
Chris@10: 		    {
Chris@10: 			 E T8, T1j, Tl, Tz, T1a, TU, T1n, T1b, T1c, T1v, T1t, T1u, T19, T1w, T1d;
Chris@10: 			 {
Chris@10: 			      E T1r, T10, TV, T1s, T11, T18;
Chris@10: 			      {
Chris@10: 				   E TO, TX, Ts, TZ, Ty, TT, TY, Tv;
Chris@10: 				   T8 = T1 + T7;
Chris@10: 				   TO = T1 - T7;
Chris@10: 				   TY = Tt * Tx;
Chris@10: 				   Tv = Tt * Tu;
Chris@10: 				   TX = FNMS(Tq, To, TW);
Chris@10: 				   Ts = FMA(Tq, Tr, Tp);
Chris@10: 				   TZ = FNMS(Tw, Tu, TY);
Chris@10: 				   Ty = FMA(Tw, Tx, Tv);
Chris@10: 				   TT = TQ - TS;
Chris@10: 				   T1j = TQ + TS;
Chris@10: 				   Tl = Te + Tk;
Chris@10: 				   T1r = Te - Tk;
Chris@10: 				   T10 = TX - TZ;
Chris@10: 				   T1g = TX + TZ;
Chris@10: 				   Tz = Ts + Ty;
Chris@10: 				   TV = Ts - Ty;
Chris@10: 				   T1a = TO - TT;
Chris@10: 				   TU = TO + TT;
Chris@10: 				   T1s = T1m - T1l;
Chris@10: 				   T1n = T1l + T1m;
Chris@10: 			      }
Chris@10: 			      T1b = TV - T10;
Chris@10: 			      T11 = TV + T10;
Chris@10: 			      T18 = T12 - T17;
Chris@10: 			      T1c = T12 + T17;
Chris@10: 			      T1v = T1s - T1r;
Chris@10: 			      T1t = T1r + T1s;
Chris@10: 			      T1u = T18 - T11;
Chris@10: 			      T19 = T11 + T18;
Chris@10: 			 }
Chris@10: 			 ci[WS(rs, 4)] = FMA(KP707106781, T1u, T1t);
Chris@10: 			 cr[WS(rs, 7)] = FMS(KP707106781, T1u, T1t);
Chris@10: 			 cr[WS(rs, 1)] = FMA(KP707106781, T19, TU);
Chris@10: 			 ci[WS(rs, 2)] = FNMS(KP707106781, T19, TU);
Chris@10: 			 T1w = T1c - T1b;
Chris@10: 			 T1d = T1b + T1c;
Chris@10: 			 ci[WS(rs, 6)] = FMA(KP707106781, T1w, T1v);
Chris@10: 			 cr[WS(rs, 5)] = FMS(KP707106781, T1w, T1v);
Chris@10: 			 ci[0] = FMA(KP707106781, T1d, T1a);
Chris@10: 			 cr[WS(rs, 3)] = FNMS(KP707106781, T1d, T1a);
Chris@10: 			 T1e = T8 - Tl;
Chris@10: 			 Tm = T8 + Tl;
Chris@10: 			 T1q = T1n - T1j;
Chris@10: 			 T1o = T1j + T1n;
Chris@10: 			 T1p = TM - Tz;
Chris@10: 			 TN = Tz + TM;
Chris@10: 		    }
Chris@10: 	       }
Chris@10: 	       ci[WS(rs, 5)] = T1p + T1q;
Chris@10: 	       cr[WS(rs, 6)] = T1p - T1q;
Chris@10: 	       cr[0] = Tm + TN;
Chris@10: 	       ci[WS(rs, 3)] = Tm - TN;
Chris@10: 	       T1h = T1f - T1g;
Chris@10: 	       T1i = T1g + T1f;
Chris@10: 	       ci[WS(rs, 7)] = T1i + T1o;
Chris@10: 	       cr[WS(rs, 4)] = T1i - T1o;
Chris@10: 	       ci[WS(rs, 1)] = T1e + T1h;
Chris@10: 	       cr[WS(rs, 2)] = T1e - T1h;
Chris@10: 	  }
Chris@10:      }
Chris@10: }
Chris@10: 
Chris@10: static const tw_instr twinstr[] = {
Chris@10:      {TW_FULL, 1, 8},
Chris@10:      {TW_NEXT, 1, 0}
Chris@10: };
Chris@10: 
Chris@10: static const hc2hc_desc desc = { 8, "hf_8", twinstr, &GENUS, {44, 14, 22, 0} };
Chris@10: 
Chris@10: void X(codelet_hf_8) (planner *p) {
Chris@10:      X(khc2hc_register) (p, hf_8, &desc);
Chris@10: }
Chris@10: #else				/* HAVE_FMA */
Chris@10: 
Chris@10: /* Generated by: ../../../genfft/gen_hc2hc.native -compact -variables 4 -pipeline-latency 4 -n 8 -dit -name hf_8 -include hf.h */
Chris@10: 
Chris@10: /*
Chris@10:  * This function contains 66 FP additions, 32 FP multiplications,
Chris@10:  * (or, 52 additions, 18 multiplications, 14 fused multiply/add),
Chris@10:  * 28 stack variables, 1 constants, and 32 memory accesses
Chris@10:  */
Chris@10: #include "hf.h"
Chris@10: 
Chris@10: static void hf_8(R *cr, R *ci, 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) * 14); m < me; m = m + 1, cr = cr + ms, ci = ci - ms, W = W + 14, MAKE_VOLATILE_STRIDE(16, rs)) {
Chris@10: 	       E T7, T1f, TH, T19, TF, T12, TR, TU, Ti, T1e, TK, T16, Tu, T13, TM;
Chris@10: 	       E TP;
Chris@10: 	       {
Chris@10: 		    E T1, T18, T6, T17;
Chris@10: 		    T1 = cr[0];
Chris@10: 		    T18 = ci[0];
Chris@10: 		    {
Chris@10: 			 E T3, T5, T2, T4;
Chris@10: 			 T3 = cr[WS(rs, 4)];
Chris@10: 			 T5 = ci[WS(rs, 4)];
Chris@10: 			 T2 = W[6];
Chris@10: 			 T4 = W[7];
Chris@10: 			 T6 = FMA(T2, T3, T4 * T5);
Chris@10: 			 T17 = FNMS(T4, T3, T2 * T5);
Chris@10: 		    }
Chris@10: 		    T7 = T1 + T6;
Chris@10: 		    T1f = T18 - T17;
Chris@10: 		    TH = T1 - T6;
Chris@10: 		    T19 = T17 + T18;
Chris@10: 	       }
Chris@10: 	       {
Chris@10: 		    E Tz, TS, TE, TT;
Chris@10: 		    {
Chris@10: 			 E Tw, Ty, Tv, Tx;
Chris@10: 			 Tw = cr[WS(rs, 7)];
Chris@10: 			 Ty = ci[WS(rs, 7)];
Chris@10: 			 Tv = W[12];
Chris@10: 			 Tx = W[13];
Chris@10: 			 Tz = FMA(Tv, Tw, Tx * Ty);
Chris@10: 			 TS = FNMS(Tx, Tw, Tv * Ty);
Chris@10: 		    }
Chris@10: 		    {
Chris@10: 			 E TB, TD, TA, TC;
Chris@10: 			 TB = cr[WS(rs, 3)];
Chris@10: 			 TD = ci[WS(rs, 3)];
Chris@10: 			 TA = W[4];
Chris@10: 			 TC = W[5];
Chris@10: 			 TE = FMA(TA, TB, TC * TD);
Chris@10: 			 TT = FNMS(TC, TB, TA * TD);
Chris@10: 		    }
Chris@10: 		    TF = Tz + TE;
Chris@10: 		    T12 = TS + TT;
Chris@10: 		    TR = Tz - TE;
Chris@10: 		    TU = TS - TT;
Chris@10: 	       }
Chris@10: 	       {
Chris@10: 		    E Tc, TI, Th, TJ;
Chris@10: 		    {
Chris@10: 			 E T9, Tb, T8, Ta;
Chris@10: 			 T9 = cr[WS(rs, 2)];
Chris@10: 			 Tb = ci[WS(rs, 2)];
Chris@10: 			 T8 = W[2];
Chris@10: 			 Ta = W[3];
Chris@10: 			 Tc = FMA(T8, T9, Ta * Tb);
Chris@10: 			 TI = FNMS(Ta, T9, T8 * Tb);
Chris@10: 		    }
Chris@10: 		    {
Chris@10: 			 E Te, Tg, Td, Tf;
Chris@10: 			 Te = cr[WS(rs, 6)];
Chris@10: 			 Tg = ci[WS(rs, 6)];
Chris@10: 			 Td = W[10];
Chris@10: 			 Tf = W[11];
Chris@10: 			 Th = FMA(Td, Te, Tf * Tg);
Chris@10: 			 TJ = FNMS(Tf, Te, Td * Tg);
Chris@10: 		    }
Chris@10: 		    Ti = Tc + Th;
Chris@10: 		    T1e = Tc - Th;
Chris@10: 		    TK = TI - TJ;
Chris@10: 		    T16 = TI + TJ;
Chris@10: 	       }
Chris@10: 	       {
Chris@10: 		    E To, TN, Tt, TO;
Chris@10: 		    {
Chris@10: 			 E Tl, Tn, Tk, Tm;
Chris@10: 			 Tl = cr[WS(rs, 1)];
Chris@10: 			 Tn = ci[WS(rs, 1)];
Chris@10: 			 Tk = W[0];
Chris@10: 			 Tm = W[1];
Chris@10: 			 To = FMA(Tk, Tl, Tm * Tn);
Chris@10: 			 TN = FNMS(Tm, Tl, Tk * Tn);
Chris@10: 		    }
Chris@10: 		    {
Chris@10: 			 E Tq, Ts, Tp, Tr;
Chris@10: 			 Tq = cr[WS(rs, 5)];
Chris@10: 			 Ts = ci[WS(rs, 5)];
Chris@10: 			 Tp = W[8];
Chris@10: 			 Tr = W[9];
Chris@10: 			 Tt = FMA(Tp, Tq, Tr * Ts);
Chris@10: 			 TO = FNMS(Tr, Tq, Tp * Ts);
Chris@10: 		    }
Chris@10: 		    Tu = To + Tt;
Chris@10: 		    T13 = TN + TO;
Chris@10: 		    TM = To - Tt;
Chris@10: 		    TP = TN - TO;
Chris@10: 	       }
Chris@10: 	       {
Chris@10: 		    E Tj, TG, T1b, T1c;
Chris@10: 		    Tj = T7 + Ti;
Chris@10: 		    TG = Tu + TF;
Chris@10: 		    ci[WS(rs, 3)] = Tj - TG;
Chris@10: 		    cr[0] = Tj + TG;
Chris@10: 		    T1b = TF - Tu;
Chris@10: 		    T1c = T19 - T16;
Chris@10: 		    cr[WS(rs, 6)] = T1b - T1c;
Chris@10: 		    ci[WS(rs, 5)] = T1b + T1c;
Chris@10: 		    {
Chris@10: 			 E TX, T1i, T10, T1h, TY, TZ;
Chris@10: 			 TX = TH - TK;
Chris@10: 			 T1i = T1f - T1e;
Chris@10: 			 TY = TM - TP;
Chris@10: 			 TZ = TR + TU;
Chris@10: 			 T10 = KP707106781 * (TY + TZ);
Chris@10: 			 T1h = KP707106781 * (TZ - TY);
Chris@10: 			 cr[WS(rs, 3)] = TX - T10;
Chris@10: 			 ci[WS(rs, 6)] = T1h + T1i;
Chris@10: 			 ci[0] = TX + T10;
Chris@10: 			 cr[WS(rs, 5)] = T1h - T1i;
Chris@10: 		    }
Chris@10: 	       }
Chris@10: 	       {
Chris@10: 		    E T15, T1a, T11, T14;
Chris@10: 		    T15 = T13 + T12;
Chris@10: 		    T1a = T16 + T19;
Chris@10: 		    cr[WS(rs, 4)] = T15 - T1a;
Chris@10: 		    ci[WS(rs, 7)] = T15 + T1a;
Chris@10: 		    T11 = T7 - Ti;
Chris@10: 		    T14 = T12 - T13;
Chris@10: 		    cr[WS(rs, 2)] = T11 - T14;
Chris@10: 		    ci[WS(rs, 1)] = T11 + T14;
Chris@10: 		    {
Chris@10: 			 E TL, T1g, TW, T1d, TQ, TV;
Chris@10: 			 TL = TH + TK;
Chris@10: 			 T1g = T1e + T1f;
Chris@10: 			 TQ = TM + TP;
Chris@10: 			 TV = TR - TU;
Chris@10: 			 TW = KP707106781 * (TQ + TV);
Chris@10: 			 T1d = KP707106781 * (TV - TQ);
Chris@10: 			 ci[WS(rs, 2)] = TL - TW;
Chris@10: 			 ci[WS(rs, 4)] = T1d + T1g;
Chris@10: 			 cr[WS(rs, 1)] = TL + TW;
Chris@10: 			 cr[WS(rs, 7)] = T1d - T1g;
Chris@10: 		    }
Chris@10: 	       }
Chris@10: 	  }
Chris@10:      }
Chris@10: }
Chris@10: 
Chris@10: static const tw_instr twinstr[] = {
Chris@10:      {TW_FULL, 1, 8},
Chris@10:      {TW_NEXT, 1, 0}
Chris@10: };
Chris@10: 
Chris@10: static const hc2hc_desc desc = { 8, "hf_8", twinstr, &GENUS, {52, 18, 14, 0} };
Chris@10: 
Chris@10: void X(codelet_hf_8) (planner *p) {
Chris@10:      X(khc2hc_register) (p, hf_8, &desc);
Chris@10: }
Chris@10: #endif				/* HAVE_FMA */