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 08:04:29 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_twidsq.native -fma -compact -variables 4 -pipeline-latency 4 -reload-twiddle -dif -n 2 -name q1_2 -include dft/scalar/q.h */
Chris@82: 
Chris@82: /*
Chris@82:  * This function contains 12 FP additions, 8 FP multiplications,
Chris@82:  * (or, 8 additions, 4 multiplications, 4 fused multiply/add),
Chris@82:  * 17 stack variables, 0 constants, and 16 memory accesses
Chris@82:  */
Chris@82: #include "dft/scalar/q.h"
Chris@82: 
Chris@82: static void q1_2(R *rio, R *iio, const R *W, stride rs, stride vs, INT mb, INT me, INT ms)
Chris@82: {
Chris@82:      {
Chris@82: 	  INT m;
Chris@82: 	  for (m = mb, W = W + (mb * 2); m < me; m = m + 1, rio = rio + ms, iio = iio + ms, W = W + 2, MAKE_VOLATILE_STRIDE(4, rs), MAKE_VOLATILE_STRIDE(0, vs)) {
Chris@82: 	       E T1, T2, T4, T7, T8, T9, Tb, Tc, Te, Th, Ti, Tj;
Chris@82: 	       T1 = rio[0];
Chris@82: 	       T2 = rio[WS(rs, 1)];
Chris@82: 	       T4 = T1 - T2;
Chris@82: 	       T7 = iio[0];
Chris@82: 	       T8 = iio[WS(rs, 1)];
Chris@82: 	       T9 = T7 - T8;
Chris@82: 	       Tb = rio[WS(vs, 1)];
Chris@82: 	       Tc = rio[WS(vs, 1) + WS(rs, 1)];
Chris@82: 	       Te = Tb - Tc;
Chris@82: 	       Th = iio[WS(vs, 1)];
Chris@82: 	       Ti = iio[WS(vs, 1) + WS(rs, 1)];
Chris@82: 	       Tj = Th - Ti;
Chris@82: 	       rio[0] = T1 + T2;
Chris@82: 	       iio[0] = T7 + T8;
Chris@82: 	       rio[WS(rs, 1)] = Tb + Tc;
Chris@82: 	       iio[WS(rs, 1)] = Th + Ti;
Chris@82: 	       {
Chris@82: 		    E Tf, Tk, Td, Tg;
Chris@82: 		    Td = W[0];
Chris@82: 		    Tf = Td * Te;
Chris@82: 		    Tk = Td * Tj;
Chris@82: 		    Tg = W[1];
Chris@82: 		    rio[WS(vs, 1) + WS(rs, 1)] = FMA(Tg, Tj, Tf);
Chris@82: 		    iio[WS(vs, 1) + WS(rs, 1)] = FNMS(Tg, Te, Tk);
Chris@82: 	       }
Chris@82: 	       {
Chris@82: 		    E T5, Ta, T3, T6;
Chris@82: 		    T3 = W[0];
Chris@82: 		    T5 = T3 * T4;
Chris@82: 		    Ta = T3 * T9;
Chris@82: 		    T6 = W[1];
Chris@82: 		    rio[WS(vs, 1)] = FMA(T6, T9, T5);
Chris@82: 		    iio[WS(vs, 1)] = FNMS(T6, T4, Ta);
Chris@82: 	       }
Chris@82: 	  }
Chris@82:      }
Chris@82: }
Chris@82: 
Chris@82: static const tw_instr twinstr[] = {
Chris@82:      {TW_FULL, 0, 2},
Chris@82:      {TW_NEXT, 1, 0}
Chris@82: };
Chris@82: 
Chris@82: static const ct_desc desc = { 2, "q1_2", twinstr, &GENUS, {8, 4, 4, 0}, 0, 0, 0 };
Chris@82: 
Chris@82: void X(codelet_q1_2) (planner *p) {
Chris@82:      X(kdft_difsq_register) (p, q1_2, &desc);
Chris@82: }
Chris@82: #else
Chris@82: 
Chris@82: /* Generated by: ../../../genfft/gen_twidsq.native -compact -variables 4 -pipeline-latency 4 -reload-twiddle -dif -n 2 -name q1_2 -include dft/scalar/q.h */
Chris@82: 
Chris@82: /*
Chris@82:  * This function contains 12 FP additions, 8 FP multiplications,
Chris@82:  * (or, 8 additions, 4 multiplications, 4 fused multiply/add),
Chris@82:  * 17 stack variables, 0 constants, and 16 memory accesses
Chris@82:  */
Chris@82: #include "dft/scalar/q.h"
Chris@82: 
Chris@82: static void q1_2(R *rio, R *iio, const R *W, stride rs, stride vs, INT mb, INT me, INT ms)
Chris@82: {
Chris@82:      {
Chris@82: 	  INT m;
Chris@82: 	  for (m = mb, W = W + (mb * 2); m < me; m = m + 1, rio = rio + ms, iio = iio + ms, W = W + 2, MAKE_VOLATILE_STRIDE(4, rs), MAKE_VOLATILE_STRIDE(0, vs)) {
Chris@82: 	       E T1, T2, T4, T6, T7, T8, T9, Ta, Tc, Te, Tf, Tg;
Chris@82: 	       T1 = rio[0];
Chris@82: 	       T2 = rio[WS(rs, 1)];
Chris@82: 	       T4 = T1 - T2;
Chris@82: 	       T6 = iio[0];
Chris@82: 	       T7 = iio[WS(rs, 1)];
Chris@82: 	       T8 = T6 - T7;
Chris@82: 	       T9 = rio[WS(vs, 1)];
Chris@82: 	       Ta = rio[WS(vs, 1) + WS(rs, 1)];
Chris@82: 	       Tc = T9 - Ta;
Chris@82: 	       Te = iio[WS(vs, 1)];
Chris@82: 	       Tf = iio[WS(vs, 1) + WS(rs, 1)];
Chris@82: 	       Tg = Te - Tf;
Chris@82: 	       rio[0] = T1 + T2;
Chris@82: 	       iio[0] = T6 + T7;
Chris@82: 	       rio[WS(rs, 1)] = T9 + Ta;
Chris@82: 	       iio[WS(rs, 1)] = Te + Tf;
Chris@82: 	       {
Chris@82: 		    E Tb, Td, T3, T5;
Chris@82: 		    Tb = W[0];
Chris@82: 		    Td = W[1];
Chris@82: 		    rio[WS(vs, 1) + WS(rs, 1)] = FMA(Tb, Tc, Td * Tg);
Chris@82: 		    iio[WS(vs, 1) + WS(rs, 1)] = FNMS(Td, Tc, Tb * Tg);
Chris@82: 		    T3 = W[0];
Chris@82: 		    T5 = W[1];
Chris@82: 		    rio[WS(vs, 1)] = FMA(T3, T4, T5 * T8);
Chris@82: 		    iio[WS(vs, 1)] = FNMS(T5, T4, T3 * T8);
Chris@82: 	       }
Chris@82: 	  }
Chris@82:      }
Chris@82: }
Chris@82: 
Chris@82: static const tw_instr twinstr[] = {
Chris@82:      {TW_FULL, 0, 2},
Chris@82:      {TW_NEXT, 1, 0}
Chris@82: };
Chris@82: 
Chris@82: static const ct_desc desc = { 2, "q1_2", twinstr, &GENUS, {8, 4, 4, 0}, 0, 0, 0 };
Chris@82: 
Chris@82: void X(codelet_q1_2) (planner *p) {
Chris@82:      X(kdft_difsq_register) (p, q1_2, &desc);
Chris@82: }
Chris@82: #endif