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 07:38:01 EST 2012 */
Chris@10: 
Chris@10: #include "codelet-dft.h"
Chris@10: 
Chris@10: #ifdef HAVE_FMA
Chris@10: 
Chris@10: /* Generated by: ../../../genfft/gen_twiddle_c.native -fma -reorder-insns -schedule-for-pipeline -simd -compact -variables 4 -pipeline-latency 8 -n 2 -name t1fv_2 -include t1f.h */
Chris@10: 
Chris@10: /*
Chris@10:  * This function contains 3 FP additions, 2 FP multiplications,
Chris@10:  * (or, 3 additions, 2 multiplications, 0 fused multiply/add),
Chris@10:  * 5 stack variables, 0 constants, and 4 memory accesses
Chris@10:  */
Chris@10: #include "t1f.h"
Chris@10: 
Chris@10: static void t1fv_2(R *ri, R *ii, const R *W, stride rs, INT mb, INT me, INT ms)
Chris@10: {
Chris@10:      {
Chris@10: 	  INT m;
Chris@10: 	  R *x;
Chris@10: 	  x = ri;
Chris@10: 	  for (m = mb, W = W + (mb * ((TWVL / VL) * 2)); m < me; m = m + VL, x = x + (VL * ms), W = W + (TWVL * 2), MAKE_VOLATILE_STRIDE(2, rs)) {
Chris@10: 	       V T1, T2, T3;
Chris@10: 	       T1 = LD(&(x[0]), ms, &(x[0]));
Chris@10: 	       T2 = LD(&(x[WS(rs, 1)]), ms, &(x[WS(rs, 1)]));
Chris@10: 	       T3 = BYTWJ(&(W[0]), T2);
Chris@10: 	       ST(&(x[0]), VADD(T1, T3), ms, &(x[0]));
Chris@10: 	       ST(&(x[WS(rs, 1)]), VSUB(T1, T3), ms, &(x[WS(rs, 1)]));
Chris@10: 	  }
Chris@10:      }
Chris@10:      VLEAVE();
Chris@10: }
Chris@10: 
Chris@10: static const tw_instr twinstr[] = {
Chris@10:      VTW(0, 1),
Chris@10:      {TW_NEXT, VL, 0}
Chris@10: };
Chris@10: 
Chris@10: static const ct_desc desc = { 2, XSIMD_STRING("t1fv_2"), twinstr, &GENUS, {3, 2, 0, 0}, 0, 0, 0 };
Chris@10: 
Chris@10: void XSIMD(codelet_t1fv_2) (planner *p) {
Chris@10:      X(kdft_dit_register) (p, t1fv_2, &desc);
Chris@10: }
Chris@10: #else				/* HAVE_FMA */
Chris@10: 
Chris@10: /* Generated by: ../../../genfft/gen_twiddle_c.native -simd -compact -variables 4 -pipeline-latency 8 -n 2 -name t1fv_2 -include t1f.h */
Chris@10: 
Chris@10: /*
Chris@10:  * This function contains 3 FP additions, 2 FP multiplications,
Chris@10:  * (or, 3 additions, 2 multiplications, 0 fused multiply/add),
Chris@10:  * 5 stack variables, 0 constants, and 4 memory accesses
Chris@10:  */
Chris@10: #include "t1f.h"
Chris@10: 
Chris@10: static void t1fv_2(R *ri, R *ii, const R *W, stride rs, INT mb, INT me, INT ms)
Chris@10: {
Chris@10:      {
Chris@10: 	  INT m;
Chris@10: 	  R *x;
Chris@10: 	  x = ri;
Chris@10: 	  for (m = mb, W = W + (mb * ((TWVL / VL) * 2)); m < me; m = m + VL, x = x + (VL * ms), W = W + (TWVL * 2), MAKE_VOLATILE_STRIDE(2, rs)) {
Chris@10: 	       V T1, T3, T2;
Chris@10: 	       T1 = LD(&(x[0]), ms, &(x[0]));
Chris@10: 	       T2 = LD(&(x[WS(rs, 1)]), ms, &(x[WS(rs, 1)]));
Chris@10: 	       T3 = BYTWJ(&(W[0]), T2);
Chris@10: 	       ST(&(x[WS(rs, 1)]), VSUB(T1, T3), ms, &(x[WS(rs, 1)]));
Chris@10: 	       ST(&(x[0]), VADD(T1, T3), ms, &(x[0]));
Chris@10: 	  }
Chris@10:      }
Chris@10:      VLEAVE();
Chris@10: }
Chris@10: 
Chris@10: static const tw_instr twinstr[] = {
Chris@10:      VTW(0, 1),
Chris@10:      {TW_NEXT, VL, 0}
Chris@10: };
Chris@10: 
Chris@10: static const ct_desc desc = { 2, XSIMD_STRING("t1fv_2"), twinstr, &GENUS, {3, 2, 0, 0}, 0, 0, 0 };
Chris@10: 
Chris@10: void XSIMD(codelet_t1fv_2) (planner *p) {
Chris@10:      X(kdft_dit_register) (p, t1fv_2, &desc);
Chris@10: }
Chris@10: #endif				/* HAVE_FMA */