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 07:37:29 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_notw_c.native -fma -reorder-insns -schedule-for-pipeline -simd -compact -variables 4 -pipeline-latency 8 -sign 1 -n 4 -name n2bv_4 -with-ostride 2 -include n2b.h -store-multiple 2 */
Chris@10: 
Chris@10: /*
Chris@10:  * This function contains 8 FP additions, 2 FP multiplications,
Chris@10:  * (or, 6 additions, 0 multiplications, 2 fused multiply/add),
Chris@10:  * 15 stack variables, 0 constants, and 10 memory accesses
Chris@10:  */
Chris@10: #include "n2b.h"
Chris@10: 
Chris@10: static void n2bv_4(const R *ri, const R *ii, R *ro, R *io, stride is, stride os, INT v, INT ivs, INT ovs)
Chris@10: {
Chris@10:      {
Chris@10: 	  INT i;
Chris@10: 	  const R *xi;
Chris@10: 	  R *xo;
Chris@10: 	  xi = ii;
Chris@10: 	  xo = io;
Chris@10: 	  for (i = v; i > 0; i = i - VL, xi = xi + (VL * ivs), xo = xo + (VL * ovs), MAKE_VOLATILE_STRIDE(8, is), MAKE_VOLATILE_STRIDE(8, os)) {
Chris@10: 	       V T1, T2, T4, T5;
Chris@10: 	       T1 = LD(&(xi[0]), ivs, &(xi[0]));
Chris@10: 	       T2 = LD(&(xi[WS(is, 2)]), ivs, &(xi[0]));
Chris@10: 	       T4 = LD(&(xi[WS(is, 1)]), ivs, &(xi[WS(is, 1)]));
Chris@10: 	       T5 = LD(&(xi[WS(is, 3)]), ivs, &(xi[WS(is, 1)]));
Chris@10: 	       {
Chris@10: 		    V T3, T7, T6, T8;
Chris@10: 		    T3 = VSUB(T1, T2);
Chris@10: 		    T7 = VADD(T1, T2);
Chris@10: 		    T6 = VSUB(T4, T5);
Chris@10: 		    T8 = VADD(T4, T5);
Chris@10: 		    {
Chris@10: 			 V T9, Ta, Tb, Tc;
Chris@10: 			 T9 = VSUB(T7, T8);
Chris@10: 			 STM2(&(xo[4]), T9, ovs, &(xo[0]));
Chris@10: 			 Ta = VADD(T7, T8);
Chris@10: 			 STM2(&(xo[0]), Ta, ovs, &(xo[0]));
Chris@10: 			 Tb = VFMAI(T6, T3);
Chris@10: 			 STM2(&(xo[2]), Tb, ovs, &(xo[2]));
Chris@10: 			 STN2(&(xo[0]), Ta, Tb, ovs);
Chris@10: 			 Tc = VFNMSI(T6, T3);
Chris@10: 			 STM2(&(xo[6]), Tc, ovs, &(xo[2]));
Chris@10: 			 STN2(&(xo[4]), T9, Tc, ovs);
Chris@10: 		    }
Chris@10: 	       }
Chris@10: 	  }
Chris@10:      }
Chris@10:      VLEAVE();
Chris@10: }
Chris@10: 
Chris@10: static const kdft_desc desc = { 4, XSIMD_STRING("n2bv_4"), {6, 0, 2, 0}, &GENUS, 0, 2, 0, 0 };
Chris@10: 
Chris@10: void XSIMD(codelet_n2bv_4) (planner *p) {
Chris@10:      X(kdft_register) (p, n2bv_4, &desc);
Chris@10: }
Chris@10: 
Chris@10: #else				/* HAVE_FMA */
Chris@10: 
Chris@10: /* Generated by: ../../../genfft/gen_notw_c.native -simd -compact -variables 4 -pipeline-latency 8 -sign 1 -n 4 -name n2bv_4 -with-ostride 2 -include n2b.h -store-multiple 2 */
Chris@10: 
Chris@10: /*
Chris@10:  * This function contains 8 FP additions, 0 FP multiplications,
Chris@10:  * (or, 8 additions, 0 multiplications, 0 fused multiply/add),
Chris@10:  * 11 stack variables, 0 constants, and 10 memory accesses
Chris@10:  */
Chris@10: #include "n2b.h"
Chris@10: 
Chris@10: static void n2bv_4(const R *ri, const R *ii, R *ro, R *io, stride is, stride os, INT v, INT ivs, INT ovs)
Chris@10: {
Chris@10:      {
Chris@10: 	  INT i;
Chris@10: 	  const R *xi;
Chris@10: 	  R *xo;
Chris@10: 	  xi = ii;
Chris@10: 	  xo = io;
Chris@10: 	  for (i = v; i > 0; i = i - VL, xi = xi + (VL * ivs), xo = xo + (VL * ovs), MAKE_VOLATILE_STRIDE(8, is), MAKE_VOLATILE_STRIDE(8, os)) {
Chris@10: 	       V T3, T7, T6, T8;
Chris@10: 	       {
Chris@10: 		    V T1, T2, T4, T5;
Chris@10: 		    T1 = LD(&(xi[0]), ivs, &(xi[0]));
Chris@10: 		    T2 = LD(&(xi[WS(is, 2)]), ivs, &(xi[0]));
Chris@10: 		    T3 = VSUB(T1, T2);
Chris@10: 		    T7 = VADD(T1, T2);
Chris@10: 		    T4 = LD(&(xi[WS(is, 1)]), ivs, &(xi[WS(is, 1)]));
Chris@10: 		    T5 = LD(&(xi[WS(is, 3)]), ivs, &(xi[WS(is, 1)]));
Chris@10: 		    T6 = VBYI(VSUB(T4, T5));
Chris@10: 		    T8 = VADD(T4, T5);
Chris@10: 	       }
Chris@10: 	       {
Chris@10: 		    V T9, Ta, Tb, Tc;
Chris@10: 		    T9 = VSUB(T3, T6);
Chris@10: 		    STM2(&(xo[6]), T9, ovs, &(xo[2]));
Chris@10: 		    Ta = VADD(T7, T8);
Chris@10: 		    STM2(&(xo[0]), Ta, ovs, &(xo[0]));
Chris@10: 		    Tb = VADD(T3, T6);
Chris@10: 		    STM2(&(xo[2]), Tb, ovs, &(xo[2]));
Chris@10: 		    STN2(&(xo[0]), Ta, Tb, ovs);
Chris@10: 		    Tc = VSUB(T7, T8);
Chris@10: 		    STM2(&(xo[4]), Tc, ovs, &(xo[0]));
Chris@10: 		    STN2(&(xo[4]), Tc, T9, ovs);
Chris@10: 	       }
Chris@10: 	  }
Chris@10:      }
Chris@10:      VLEAVE();
Chris@10: }
Chris@10: 
Chris@10: static const kdft_desc desc = { 4, XSIMD_STRING("n2bv_4"), {8, 0, 0, 0}, &GENUS, 0, 2, 0, 0 };
Chris@10: 
Chris@10: void XSIMD(codelet_n2bv_4) (planner *p) {
Chris@10:      X(kdft_register) (p, n2bv_4, &desc);
Chris@10: }
Chris@10: 
Chris@10: #endif				/* HAVE_FMA */