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 07:38:49 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 -twiddle-log3 -precompute-twiddles -no-generate-bytw -n 4 -name t3fv_4 -include t3f.h */
Chris@10: 
Chris@10: /*
Chris@10:  * This function contains 12 FP additions, 10 FP multiplications,
Chris@10:  * (or, 10 additions, 8 multiplications, 2 fused multiply/add),
Chris@10:  * 16 stack variables, 0 constants, and 8 memory accesses
Chris@10:  */
Chris@10: #include "t3f.h"
Chris@10: 
Chris@10: static void t3fv_4(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) * 4)); m < me; m = m + VL, x = x + (VL * ms), W = W + (TWVL * 4), MAKE_VOLATILE_STRIDE(4, rs)) {
Chris@10: 	       V T2, T3, T1, Ta, T5, T8;
Chris@10: 	       T2 = LDW(&(W[0]));
Chris@10: 	       T3 = LDW(&(W[TWVL * 2]));
Chris@10: 	       T1 = LD(&(x[0]), ms, &(x[0]));
Chris@10: 	       Ta = LD(&(x[WS(rs, 3)]), ms, &(x[WS(rs, 1)]));
Chris@10: 	       T5 = LD(&(x[WS(rs, 2)]), ms, &(x[0]));
Chris@10: 	       T8 = LD(&(x[WS(rs, 1)]), ms, &(x[WS(rs, 1)]));
Chris@10: 	       {
Chris@10: 		    V T4, Tb, T9, T6;
Chris@10: 		    T4 = VZMULJ(T2, T3);
Chris@10: 		    Tb = VZMULJ(T3, Ta);
Chris@10: 		    T9 = VZMULJ(T2, T8);
Chris@10: 		    T6 = VZMULJ(T4, T5);
Chris@10: 		    {
Chris@10: 			 V Tc, Te, T7, Td;
Chris@10: 			 Tc = VSUB(T9, Tb);
Chris@10: 			 Te = VADD(T9, Tb);
Chris@10: 			 T7 = VSUB(T1, T6);
Chris@10: 			 Td = VADD(T1, T6);
Chris@10: 			 ST(&(x[0]), VADD(Td, Te), ms, &(x[0]));
Chris@10: 			 ST(&(x[WS(rs, 2)]), VSUB(Td, Te), ms, &(x[0]));
Chris@10: 			 ST(&(x[WS(rs, 3)]), VFMAI(Tc, T7), ms, &(x[WS(rs, 1)]));
Chris@10: 			 ST(&(x[WS(rs, 1)]), VFNMSI(Tc, T7), ms, &(x[WS(rs, 1)]));
Chris@10: 		    }
Chris@10: 	       }
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:      VTW(0, 3),
Chris@10:      {TW_NEXT, VL, 0}
Chris@10: };
Chris@10: 
Chris@10: static const ct_desc desc = { 4, XSIMD_STRING("t3fv_4"), twinstr, &GENUS, {10, 8, 2, 0}, 0, 0, 0 };
Chris@10: 
Chris@10: void XSIMD(codelet_t3fv_4) (planner *p) {
Chris@10:      X(kdft_dit_register) (p, t3fv_4, &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 -twiddle-log3 -precompute-twiddles -no-generate-bytw -n 4 -name t3fv_4 -include t3f.h */
Chris@10: 
Chris@10: /*
Chris@10:  * This function contains 12 FP additions, 8 FP multiplications,
Chris@10:  * (or, 12 additions, 8 multiplications, 0 fused multiply/add),
Chris@10:  * 16 stack variables, 0 constants, and 8 memory accesses
Chris@10:  */
Chris@10: #include "t3f.h"
Chris@10: 
Chris@10: static void t3fv_4(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) * 4)); m < me; m = m + VL, x = x + (VL * ms), W = W + (TWVL * 4), MAKE_VOLATILE_STRIDE(4, rs)) {
Chris@10: 	       V T2, T3, T4;
Chris@10: 	       T2 = LDW(&(W[0]));
Chris@10: 	       T3 = LDW(&(W[TWVL * 2]));
Chris@10: 	       T4 = VZMULJ(T2, T3);
Chris@10: 	       {
Chris@10: 		    V T1, Tb, T6, T9, Ta, T5, T8;
Chris@10: 		    T1 = LD(&(x[0]), ms, &(x[0]));
Chris@10: 		    Ta = LD(&(x[WS(rs, 3)]), ms, &(x[WS(rs, 1)]));
Chris@10: 		    Tb = VZMULJ(T3, Ta);
Chris@10: 		    T5 = LD(&(x[WS(rs, 2)]), ms, &(x[0]));
Chris@10: 		    T6 = VZMULJ(T4, T5);
Chris@10: 		    T8 = LD(&(x[WS(rs, 1)]), ms, &(x[WS(rs, 1)]));
Chris@10: 		    T9 = VZMULJ(T2, T8);
Chris@10: 		    {
Chris@10: 			 V T7, Tc, Td, Te;
Chris@10: 			 T7 = VSUB(T1, T6);
Chris@10: 			 Tc = VBYI(VSUB(T9, Tb));
Chris@10: 			 ST(&(x[WS(rs, 1)]), VSUB(T7, Tc), ms, &(x[WS(rs, 1)]));
Chris@10: 			 ST(&(x[WS(rs, 3)]), VADD(T7, Tc), ms, &(x[WS(rs, 1)]));
Chris@10: 			 Td = VADD(T1, T6);
Chris@10: 			 Te = VADD(T9, Tb);
Chris@10: 			 ST(&(x[WS(rs, 2)]), VSUB(Td, Te), ms, &(x[0]));
Chris@10: 			 ST(&(x[0]), VADD(Td, Te), ms, &(x[0]));
Chris@10: 		    }
Chris@10: 	       }
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:      VTW(0, 3),
Chris@10:      {TW_NEXT, VL, 0}
Chris@10: };
Chris@10: 
Chris@10: static const ct_desc desc = { 4, XSIMD_STRING("t3fv_4"), twinstr, &GENUS, {12, 8, 0, 0}, 0, 0, 0 };
Chris@10: 
Chris@10: void XSIMD(codelet_t3fv_4) (planner *p) {
Chris@10:      X(kdft_dit_register) (p, t3fv_4, &desc);
Chris@10: }
Chris@10: #endif				/* HAVE_FMA */