Chris@82: /*
Chris@82:  * Copyright (c) 2003, 2007-14 Matteo Frigo
Chris@82:  * Copyright (c) 2003, 2007-14 Massachusetts Institute of Technology
Chris@82:  *
Chris@82:  * This program is free software; you can redistribute it and/or modify
Chris@82:  * it under the terms of the GNU General Public License as published by
Chris@82:  * the Free Software Foundation; either version 2 of the License, or
Chris@82:  * (at your option) any later version.
Chris@82:  *
Chris@82:  * This program is distributed in the hope that it will be useful,
Chris@82:  * but WITHOUT ANY WARRANTY; without even the implied warranty of
Chris@82:  * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
Chris@82:  * GNU General Public License for more details.
Chris@82:  *
Chris@82:  * You should have received a copy of the GNU General Public License
Chris@82:  * along with this program; if not, write to the Free Software
Chris@82:  * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA  02110-1301  USA
Chris@82:  *
Chris@82:  */
Chris@82: 
Chris@82: /* This file was automatically generated --- DO NOT EDIT */
Chris@82: /* Generated on Thu May 24 08:04:25 EDT 2018 */
Chris@82: 
Chris@82: #include "dft/codelet-dft.h"
Chris@82: 
Chris@82: #if defined(ARCH_PREFERS_FMA) || defined(ISA_EXTENSION_PREFERS_FMA)
Chris@82: 
Chris@82: /* Generated by: ../../../genfft/gen_twiddle.native -fma -compact -variables 4 -pipeline-latency 4 -twiddle-log3 -precompute-twiddles -n 5 -name t2_5 -include dft/scalar/t.h */
Chris@82: 
Chris@82: /*
Chris@82:  * This function contains 44 FP additions, 40 FP multiplications,
Chris@82:  * (or, 14 additions, 10 multiplications, 30 fused multiply/add),
Chris@82:  * 38 stack variables, 4 constants, and 20 memory accesses
Chris@82:  */
Chris@82: #include "dft/scalar/t.h"
Chris@82: 
Chris@82: static void t2_5(R *ri, R *ii, const R *W, stride rs, INT mb, INT me, INT ms)
Chris@82: {
Chris@82:      DK(KP951056516, +0.951056516295153572116439333379382143405698634);
Chris@82:      DK(KP559016994, +0.559016994374947424102293417182819058860154590);
Chris@82:      DK(KP618033988, +0.618033988749894848204586834365638117720309180);
Chris@82:      DK(KP250000000, +0.250000000000000000000000000000000000000000000);
Chris@82:      {
Chris@82: 	  INT m;
Chris@82: 	  for (m = mb, W = W + (mb * 4); m < me; m = m + 1, ri = ri + ms, ii = ii + ms, W = W + 4, MAKE_VOLATILE_STRIDE(10, rs)) {
Chris@82: 	       E T2, Ta, T8, T5, Tb, Tm, Tf, Tj, T9, Te;
Chris@82: 	       T2 = W[0];
Chris@82: 	       Ta = W[3];
Chris@82: 	       T8 = W[2];
Chris@82: 	       T9 = T2 * T8;
Chris@82: 	       Te = T2 * Ta;
Chris@82: 	       T5 = W[1];
Chris@82: 	       Tb = FNMS(T5, Ta, T9);
Chris@82: 	       Tm = FNMS(T5, T8, Te);
Chris@82: 	       Tf = FMA(T5, T8, Te);
Chris@82: 	       Tj = FMA(T5, Ta, T9);
Chris@82: 	       {
Chris@82: 		    E T1, TO, T7, Th, Ti, Tz, TB, TL, To, Ts, Tt, TE, TG, TM;
Chris@82: 		    T1 = ri[0];
Chris@82: 		    TO = ii[0];
Chris@82: 		    {
Chris@82: 			 E T3, T4, T6, Ty, Tc, Td, Tg, TA;
Chris@82: 			 T3 = ri[WS(rs, 1)];
Chris@82: 			 T4 = T2 * T3;
Chris@82: 			 T6 = ii[WS(rs, 1)];
Chris@82: 			 Ty = T2 * T6;
Chris@82: 			 Tc = ri[WS(rs, 4)];
Chris@82: 			 Td = Tb * Tc;
Chris@82: 			 Tg = ii[WS(rs, 4)];
Chris@82: 			 TA = Tb * Tg;
Chris@82: 			 T7 = FMA(T5, T6, T4);
Chris@82: 			 Th = FMA(Tf, Tg, Td);
Chris@82: 			 Ti = T7 + Th;
Chris@82: 			 Tz = FNMS(T5, T3, Ty);
Chris@82: 			 TB = FNMS(Tf, Tc, TA);
Chris@82: 			 TL = Tz + TB;
Chris@82: 		    }
Chris@82: 		    {
Chris@82: 			 E Tk, Tl, Tn, TD, Tp, Tq, Tr, TF;
Chris@82: 			 Tk = ri[WS(rs, 2)];
Chris@82: 			 Tl = Tj * Tk;
Chris@82: 			 Tn = ii[WS(rs, 2)];
Chris@82: 			 TD = Tj * Tn;
Chris@82: 			 Tp = ri[WS(rs, 3)];
Chris@82: 			 Tq = T8 * Tp;
Chris@82: 			 Tr = ii[WS(rs, 3)];
Chris@82: 			 TF = T8 * Tr;
Chris@82: 			 To = FMA(Tm, Tn, Tl);
Chris@82: 			 Ts = FMA(Ta, Tr, Tq);
Chris@82: 			 Tt = To + Ts;
Chris@82: 			 TE = FNMS(Tm, Tk, TD);
Chris@82: 			 TG = FNMS(Ta, Tp, TF);
Chris@82: 			 TM = TE + TG;
Chris@82: 		    }
Chris@82: 		    {
Chris@82: 			 E Tw, Tu, Tv, TI, TK, TC, TH, TJ, Tx;
Chris@82: 			 Tw = Ti - Tt;
Chris@82: 			 Tu = Ti + Tt;
Chris@82: 			 Tv = FNMS(KP250000000, Tu, T1);
Chris@82: 			 TC = Tz - TB;
Chris@82: 			 TH = TE - TG;
Chris@82: 			 TI = FMA(KP618033988, TH, TC);
Chris@82: 			 TK = FNMS(KP618033988, TC, TH);
Chris@82: 			 ri[0] = T1 + Tu;
Chris@82: 			 TJ = FNMS(KP559016994, Tw, Tv);
Chris@82: 			 ri[WS(rs, 2)] = FNMS(KP951056516, TK, TJ);
Chris@82: 			 ri[WS(rs, 3)] = FMA(KP951056516, TK, TJ);
Chris@82: 			 Tx = FMA(KP559016994, Tw, Tv);
Chris@82: 			 ri[WS(rs, 4)] = FNMS(KP951056516, TI, Tx);
Chris@82: 			 ri[WS(rs, 1)] = FMA(KP951056516, TI, Tx);
Chris@82: 		    }
Chris@82: 		    {
Chris@82: 			 E TQ, TN, TP, TU, TW, TS, TT, TV, TR;
Chris@82: 			 TQ = TL - TM;
Chris@82: 			 TN = TL + TM;
Chris@82: 			 TP = FNMS(KP250000000, TN, TO);
Chris@82: 			 TS = T7 - Th;
Chris@82: 			 TT = To - Ts;
Chris@82: 			 TU = FMA(KP618033988, TT, TS);
Chris@82: 			 TW = FNMS(KP618033988, TS, TT);
Chris@82: 			 ii[0] = TN + TO;
Chris@82: 			 TV = FNMS(KP559016994, TQ, TP);
Chris@82: 			 ii[WS(rs, 2)] = FMA(KP951056516, TW, TV);
Chris@82: 			 ii[WS(rs, 3)] = FNMS(KP951056516, TW, TV);
Chris@82: 			 TR = FMA(KP559016994, TQ, TP);
Chris@82: 			 ii[WS(rs, 1)] = FNMS(KP951056516, TU, TR);
Chris@82: 			 ii[WS(rs, 4)] = FMA(KP951056516, TU, TR);
Chris@82: 		    }
Chris@82: 	       }
Chris@82: 	  }
Chris@82:      }
Chris@82: }
Chris@82: 
Chris@82: static const tw_instr twinstr[] = {
Chris@82:      {TW_CEXP, 0, 1},
Chris@82:      {TW_CEXP, 0, 3},
Chris@82:      {TW_NEXT, 1, 0}
Chris@82: };
Chris@82: 
Chris@82: static const ct_desc desc = { 5, "t2_5", twinstr, &GENUS, {14, 10, 30, 0}, 0, 0, 0 };
Chris@82: 
Chris@82: void X(codelet_t2_5) (planner *p) {
Chris@82:      X(kdft_dit_register) (p, t2_5, &desc);
Chris@82: }
Chris@82: #else
Chris@82: 
Chris@82: /* Generated by: ../../../genfft/gen_twiddle.native -compact -variables 4 -pipeline-latency 4 -twiddle-log3 -precompute-twiddles -n 5 -name t2_5 -include dft/scalar/t.h */
Chris@82: 
Chris@82: /*
Chris@82:  * This function contains 44 FP additions, 32 FP multiplications,
Chris@82:  * (or, 30 additions, 18 multiplications, 14 fused multiply/add),
Chris@82:  * 37 stack variables, 4 constants, and 20 memory accesses
Chris@82:  */
Chris@82: #include "dft/scalar/t.h"
Chris@82: 
Chris@82: static void t2_5(R *ri, R *ii, const R *W, stride rs, INT mb, INT me, INT ms)
Chris@82: {
Chris@82:      DK(KP250000000, +0.250000000000000000000000000000000000000000000);
Chris@82:      DK(KP559016994, +0.559016994374947424102293417182819058860154590);
Chris@82:      DK(KP587785252, +0.587785252292473129168705954639072768597652438);
Chris@82:      DK(KP951056516, +0.951056516295153572116439333379382143405698634);
Chris@82:      {
Chris@82: 	  INT m;
Chris@82: 	  for (m = mb, W = W + (mb * 4); m < me; m = m + 1, ri = ri + ms, ii = ii + ms, W = W + 4, MAKE_VOLATILE_STRIDE(10, rs)) {
Chris@82: 	       E T2, T4, T7, T9, Tb, Tl, Tf, Tj;
Chris@82: 	       {
Chris@82: 		    E T8, Te, Ta, Td;
Chris@82: 		    T2 = W[0];
Chris@82: 		    T4 = W[1];
Chris@82: 		    T7 = W[2];
Chris@82: 		    T9 = W[3];
Chris@82: 		    T8 = T2 * T7;
Chris@82: 		    Te = T4 * T7;
Chris@82: 		    Ta = T4 * T9;
Chris@82: 		    Td = T2 * T9;
Chris@82: 		    Tb = T8 - Ta;
Chris@82: 		    Tl = Td - Te;
Chris@82: 		    Tf = Td + Te;
Chris@82: 		    Tj = T8 + Ta;
Chris@82: 	       }
Chris@82: 	       {
Chris@82: 		    E T1, TI, Ty, TB, TN, TM, TF, TG, TH, Ti, Tr, Ts;
Chris@82: 		    T1 = ri[0];
Chris@82: 		    TI = ii[0];
Chris@82: 		    {
Chris@82: 			 E T6, Tw, Tq, TA, Th, Tx, Tn, Tz;
Chris@82: 			 {
Chris@82: 			      E T3, T5, To, Tp;
Chris@82: 			      T3 = ri[WS(rs, 1)];
Chris@82: 			      T5 = ii[WS(rs, 1)];
Chris@82: 			      T6 = FMA(T2, T3, T4 * T5);
Chris@82: 			      Tw = FNMS(T4, T3, T2 * T5);
Chris@82: 			      To = ri[WS(rs, 3)];
Chris@82: 			      Tp = ii[WS(rs, 3)];
Chris@82: 			      Tq = FMA(T7, To, T9 * Tp);
Chris@82: 			      TA = FNMS(T9, To, T7 * Tp);
Chris@82: 			 }
Chris@82: 			 {
Chris@82: 			      E Tc, Tg, Tk, Tm;
Chris@82: 			      Tc = ri[WS(rs, 4)];
Chris@82: 			      Tg = ii[WS(rs, 4)];
Chris@82: 			      Th = FMA(Tb, Tc, Tf * Tg);
Chris@82: 			      Tx = FNMS(Tf, Tc, Tb * Tg);
Chris@82: 			      Tk = ri[WS(rs, 2)];
Chris@82: 			      Tm = ii[WS(rs, 2)];
Chris@82: 			      Tn = FMA(Tj, Tk, Tl * Tm);
Chris@82: 			      Tz = FNMS(Tl, Tk, Tj * Tm);
Chris@82: 			 }
Chris@82: 			 Ty = Tw - Tx;
Chris@82: 			 TB = Tz - TA;
Chris@82: 			 TN = Tn - Tq;
Chris@82: 			 TM = T6 - Th;
Chris@82: 			 TF = Tw + Tx;
Chris@82: 			 TG = Tz + TA;
Chris@82: 			 TH = TF + TG;
Chris@82: 			 Ti = T6 + Th;
Chris@82: 			 Tr = Tn + Tq;
Chris@82: 			 Ts = Ti + Tr;
Chris@82: 		    }
Chris@82: 		    ri[0] = T1 + Ts;
Chris@82: 		    ii[0] = TH + TI;
Chris@82: 		    {
Chris@82: 			 E TC, TE, Tv, TD, Tt, Tu;
Chris@82: 			 TC = FMA(KP951056516, Ty, KP587785252 * TB);
Chris@82: 			 TE = FNMS(KP587785252, Ty, KP951056516 * TB);
Chris@82: 			 Tt = KP559016994 * (Ti - Tr);
Chris@82: 			 Tu = FNMS(KP250000000, Ts, T1);
Chris@82: 			 Tv = Tt + Tu;
Chris@82: 			 TD = Tu - Tt;
Chris@82: 			 ri[WS(rs, 4)] = Tv - TC;
Chris@82: 			 ri[WS(rs, 3)] = TD + TE;
Chris@82: 			 ri[WS(rs, 1)] = Tv + TC;
Chris@82: 			 ri[WS(rs, 2)] = TD - TE;
Chris@82: 		    }
Chris@82: 		    {
Chris@82: 			 E TO, TP, TL, TQ, TJ, TK;
Chris@82: 			 TO = FMA(KP951056516, TM, KP587785252 * TN);
Chris@82: 			 TP = FNMS(KP587785252, TM, KP951056516 * TN);
Chris@82: 			 TJ = KP559016994 * (TF - TG);
Chris@82: 			 TK = FNMS(KP250000000, TH, TI);
Chris@82: 			 TL = TJ + TK;
Chris@82: 			 TQ = TK - TJ;
Chris@82: 			 ii[WS(rs, 1)] = TL - TO;
Chris@82: 			 ii[WS(rs, 3)] = TQ - TP;
Chris@82: 			 ii[WS(rs, 4)] = TO + TL;
Chris@82: 			 ii[WS(rs, 2)] = TP + TQ;
Chris@82: 		    }
Chris@82: 	       }
Chris@82: 	  }
Chris@82:      }
Chris@82: }
Chris@82: 
Chris@82: static const tw_instr twinstr[] = {
Chris@82:      {TW_CEXP, 0, 1},
Chris@82:      {TW_CEXP, 0, 3},
Chris@82:      {TW_NEXT, 1, 0}
Chris@82: };
Chris@82: 
Chris@82: static const ct_desc desc = { 5, "t2_5", twinstr, &GENUS, {30, 18, 14, 0}, 0, 0, 0 };
Chris@82: 
Chris@82: void X(codelet_t2_5) (planner *p) {
Chris@82:      X(kdft_dit_register) (p, t2_5, &desc);
Chris@82: }
Chris@82: #endif