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 08:07:37 EDT 2018 */
Chris@82: 
Chris@82: #include "rdft/codelet-rdft.h"
Chris@82: 
Chris@82: #if defined(ARCH_PREFERS_FMA) || defined(ISA_EXTENSION_PREFERS_FMA)
Chris@82: 
Chris@82: /* Generated by: ../../../genfft/gen_hc2hc.native -fma -compact -variables 4 -pipeline-latency 4 -sign 1 -twiddle-log3 -precompute-twiddles -n 4 -dif -name hb2_4 -include rdft/scalar/hb.h */
Chris@82: 
Chris@82: /*
Chris@82:  * This function contains 24 FP additions, 16 FP multiplications,
Chris@82:  * (or, 16 additions, 8 multiplications, 8 fused multiply/add),
Chris@82:  * 33 stack variables, 0 constants, and 16 memory accesses
Chris@82:  */
Chris@82: #include "rdft/scalar/hb.h"
Chris@82: 
Chris@82: static void hb2_4(R *cr, R *ci, const R *W, stride rs, INT mb, INT me, INT ms)
Chris@82: {
Chris@82:      {
Chris@82: 	  INT m;
Chris@82: 	  for (m = mb, W = W + ((mb - 1) * 4); m < me; m = m + 1, cr = cr + ms, ci = ci - ms, W = W + 4, MAKE_VOLATILE_STRIDE(8, rs)) {
Chris@82: 	       E T7, Tb, T8, Ta, Tc, Tg, T9, Tf;
Chris@82: 	       T7 = W[0];
Chris@82: 	       Tb = W[3];
Chris@82: 	       T8 = W[2];
Chris@82: 	       T9 = T7 * T8;
Chris@82: 	       Tf = T7 * Tb;
Chris@82: 	       Ta = W[1];
Chris@82: 	       Tc = FMA(Ta, Tb, T9);
Chris@82: 	       Tg = FNMS(Ta, T8, Tf);
Chris@82: 	       {
Chris@82: 		    E T3, T6, Td, Tj, Tz, Tx, Tr, Tm, Tv, Ts, Tw, TA;
Chris@82: 		    {
Chris@82: 			 E Th, Ti, Tu, Tk, Tl, Tq, Tp, Tt;
Chris@82: 			 Th = ci[WS(rs, 3)];
Chris@82: 			 Ti = cr[WS(rs, 2)];
Chris@82: 			 Tu = Th + Ti;
Chris@82: 			 Tk = ci[WS(rs, 2)];
Chris@82: 			 Tl = cr[WS(rs, 3)];
Chris@82: 			 Tq = Tk + Tl;
Chris@82: 			 {
Chris@82: 			      E T1, T2, T4, T5;
Chris@82: 			      T1 = cr[0];
Chris@82: 			      T2 = ci[WS(rs, 1)];
Chris@82: 			      T3 = T1 + T2;
Chris@82: 			      Tp = T1 - T2;
Chris@82: 			      T4 = cr[WS(rs, 1)];
Chris@82: 			      T5 = ci[0];
Chris@82: 			      T6 = T4 + T5;
Chris@82: 			      Tt = T4 - T5;
Chris@82: 			 }
Chris@82: 			 Td = T3 - T6;
Chris@82: 			 Tj = Th - Ti;
Chris@82: 			 Tz = Tu - Tt;
Chris@82: 			 Tx = Tp + Tq;
Chris@82: 			 Tr = Tp - Tq;
Chris@82: 			 Tm = Tk - Tl;
Chris@82: 			 Tv = Tt + Tu;
Chris@82: 		    }
Chris@82: 		    cr[0] = T3 + T6;
Chris@82: 		    ci[0] = Tj + Tm;
Chris@82: 		    Ts = T7 * Tr;
Chris@82: 		    cr[WS(rs, 1)] = FNMS(Ta, Tv, Ts);
Chris@82: 		    Tw = T7 * Tv;
Chris@82: 		    ci[WS(rs, 1)] = FMA(Ta, Tr, Tw);
Chris@82: 		    TA = T8 * Tz;
Chris@82: 		    ci[WS(rs, 3)] = FMA(Tb, Tx, TA);
Chris@82: 		    {
Chris@82: 			 E Ty, Te, To, Tn;
Chris@82: 			 Ty = T8 * Tx;
Chris@82: 			 cr[WS(rs, 3)] = FNMS(Tb, Tz, Ty);
Chris@82: 			 Te = Tc * Td;
Chris@82: 			 To = Tg * Td;
Chris@82: 			 Tn = Tj - Tm;
Chris@82: 			 cr[WS(rs, 2)] = FNMS(Tg, Tn, Te);
Chris@82: 			 ci[WS(rs, 2)] = FMA(Tc, Tn, To);
Chris@82: 		    }
Chris@82: 	       }
Chris@82: 	  }
Chris@82:      }
Chris@82: }
Chris@82: 
Chris@82: static const tw_instr twinstr[] = {
Chris@82:      {TW_CEXP, 1, 1},
Chris@82:      {TW_CEXP, 1, 3},
Chris@82:      {TW_NEXT, 1, 0}
Chris@82: };
Chris@82: 
Chris@82: static const hc2hc_desc desc = { 4, "hb2_4", twinstr, &GENUS, {16, 8, 8, 0} };
Chris@82: 
Chris@82: void X(codelet_hb2_4) (planner *p) {
Chris@82:      X(khc2hc_register) (p, hb2_4, &desc);
Chris@82: }
Chris@82: #else
Chris@82: 
Chris@82: /* Generated by: ../../../genfft/gen_hc2hc.native -compact -variables 4 -pipeline-latency 4 -sign 1 -twiddle-log3 -precompute-twiddles -n 4 -dif -name hb2_4 -include rdft/scalar/hb.h */
Chris@82: 
Chris@82: /*
Chris@82:  * This function contains 24 FP additions, 16 FP multiplications,
Chris@82:  * (or, 16 additions, 8 multiplications, 8 fused multiply/add),
Chris@82:  * 21 stack variables, 0 constants, and 16 memory accesses
Chris@82:  */
Chris@82: #include "rdft/scalar/hb.h"
Chris@82: 
Chris@82: static void hb2_4(R *cr, R *ci, const R *W, stride rs, INT mb, INT me, INT ms)
Chris@82: {
Chris@82:      {
Chris@82: 	  INT m;
Chris@82: 	  for (m = mb, W = W + ((mb - 1) * 4); m < me; m = m + 1, cr = cr + ms, ci = ci - ms, W = W + 4, MAKE_VOLATILE_STRIDE(8, rs)) {
Chris@82: 	       E T7, T9, T8, Ta, Tb, Td;
Chris@82: 	       T7 = W[0];
Chris@82: 	       T9 = W[1];
Chris@82: 	       T8 = W[2];
Chris@82: 	       Ta = W[3];
Chris@82: 	       Tb = FMA(T7, T8, T9 * Ta);
Chris@82: 	       Td = FNMS(T9, T8, T7 * Ta);
Chris@82: 	       {
Chris@82: 		    E T3, Tl, T6, To, Tg, Tp, Tj, Tm, Tc, Tk;
Chris@82: 		    {
Chris@82: 			 E T1, T2, T4, T5;
Chris@82: 			 T1 = cr[0];
Chris@82: 			 T2 = ci[WS(rs, 1)];
Chris@82: 			 T3 = T1 + T2;
Chris@82: 			 Tl = T1 - T2;
Chris@82: 			 T4 = cr[WS(rs, 1)];
Chris@82: 			 T5 = ci[0];
Chris@82: 			 T6 = T4 + T5;
Chris@82: 			 To = T4 - T5;
Chris@82: 		    }
Chris@82: 		    {
Chris@82: 			 E Te, Tf, Th, Ti;
Chris@82: 			 Te = ci[WS(rs, 3)];
Chris@82: 			 Tf = cr[WS(rs, 2)];
Chris@82: 			 Tg = Te - Tf;
Chris@82: 			 Tp = Te + Tf;
Chris@82: 			 Th = ci[WS(rs, 2)];
Chris@82: 			 Ti = cr[WS(rs, 3)];
Chris@82: 			 Tj = Th - Ti;
Chris@82: 			 Tm = Th + Ti;
Chris@82: 		    }
Chris@82: 		    cr[0] = T3 + T6;
Chris@82: 		    ci[0] = Tg + Tj;
Chris@82: 		    Tc = T3 - T6;
Chris@82: 		    Tk = Tg - Tj;
Chris@82: 		    cr[WS(rs, 2)] = FNMS(Td, Tk, Tb * Tc);
Chris@82: 		    ci[WS(rs, 2)] = FMA(Td, Tc, Tb * Tk);
Chris@82: 		    {
Chris@82: 			 E Tn, Tq, Tr, Ts;
Chris@82: 			 Tn = Tl - Tm;
Chris@82: 			 Tq = To + Tp;
Chris@82: 			 cr[WS(rs, 1)] = FNMS(T9, Tq, T7 * Tn);
Chris@82: 			 ci[WS(rs, 1)] = FMA(T7, Tq, T9 * Tn);
Chris@82: 			 Tr = Tl + Tm;
Chris@82: 			 Ts = Tp - To;
Chris@82: 			 cr[WS(rs, 3)] = FNMS(Ta, Ts, T8 * Tr);
Chris@82: 			 ci[WS(rs, 3)] = FMA(T8, Ts, Ta * Tr);
Chris@82: 		    }
Chris@82: 	       }
Chris@82: 	  }
Chris@82:      }
Chris@82: }
Chris@82: 
Chris@82: static const tw_instr twinstr[] = {
Chris@82:      {TW_CEXP, 1, 1},
Chris@82:      {TW_CEXP, 1, 3},
Chris@82:      {TW_NEXT, 1, 0}
Chris@82: };
Chris@82: 
Chris@82: static const hc2hc_desc desc = { 4, "hb2_4", twinstr, &GENUS, {16, 8, 8, 0} };
Chris@82: 
Chris@82: void X(codelet_hb2_4) (planner *p) {
Chris@82:      X(khc2hc_register) (p, hb2_4, &desc);
Chris@82: }
Chris@82: #endif