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 07:38:41 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 5 -name t2fv_5 -include t2f.h */
Chris@10: 
Chris@10: /*
Chris@10:  * This function contains 20 FP additions, 19 FP multiplications,
Chris@10:  * (or, 11 additions, 10 multiplications, 9 fused multiply/add),
Chris@10:  * 26 stack variables, 4 constants, and 10 memory accesses
Chris@10:  */
Chris@10: #include "t2f.h"
Chris@10: 
Chris@10: static void t2fv_5(R *ri, R *ii, const R *W, stride rs, INT mb, INT me, INT ms)
Chris@10: {
Chris@10:      DVK(KP559016994, +0.559016994374947424102293417182819058860154590);
Chris@10:      DVK(KP250000000, +0.250000000000000000000000000000000000000000000);
Chris@10:      DVK(KP618033988, +0.618033988749894848204586834365638117720309180);
Chris@10:      DVK(KP951056516, +0.951056516295153572116439333379382143405698634);
Chris@10:      {
Chris@10: 	  INT m;
Chris@10: 	  R *x;
Chris@10: 	  x = ri;
Chris@10: 	  for (m = mb, W = W + (mb * ((TWVL / VL) * 8)); m < me; m = m + VL, x = x + (VL * ms), W = W + (TWVL * 8), MAKE_VOLATILE_STRIDE(5, rs)) {
Chris@10: 	       V T1, T2, T9, T4, T7;
Chris@10: 	       T1 = LD(&(x[0]), ms, &(x[0]));
Chris@10: 	       T2 = LD(&(x[WS(rs, 1)]), ms, &(x[WS(rs, 1)]));
Chris@10: 	       T9 = LD(&(x[WS(rs, 3)]), ms, &(x[WS(rs, 1)]));
Chris@10: 	       T4 = LD(&(x[WS(rs, 4)]), ms, &(x[0]));
Chris@10: 	       T7 = LD(&(x[WS(rs, 2)]), ms, &(x[0]));
Chris@10: 	       {
Chris@10: 		    V T3, Ta, T5, T8;
Chris@10: 		    T3 = BYTWJ(&(W[0]), T2);
Chris@10: 		    Ta = BYTWJ(&(W[TWVL * 4]), T9);
Chris@10: 		    T5 = BYTWJ(&(W[TWVL * 6]), T4);
Chris@10: 		    T8 = BYTWJ(&(W[TWVL * 2]), T7);
Chris@10: 		    {
Chris@10: 			 V T6, Tg, Tb, Th;
Chris@10: 			 T6 = VADD(T3, T5);
Chris@10: 			 Tg = VSUB(T3, T5);
Chris@10: 			 Tb = VADD(T8, Ta);
Chris@10: 			 Th = VSUB(T8, Ta);
Chris@10: 			 {
Chris@10: 			      V Te, Tc, Tk, Ti, Td, Tj, Tf;
Chris@10: 			      Te = VSUB(T6, Tb);
Chris@10: 			      Tc = VADD(T6, Tb);
Chris@10: 			      Tk = VMUL(LDK(KP951056516), VFNMS(LDK(KP618033988), Tg, Th));
Chris@10: 			      Ti = VMUL(LDK(KP951056516), VFMA(LDK(KP618033988), Th, Tg));
Chris@10: 			      Td = VFNMS(LDK(KP250000000), Tc, T1);
Chris@10: 			      ST(&(x[0]), VADD(T1, Tc), ms, &(x[0]));
Chris@10: 			      Tj = VFNMS(LDK(KP559016994), Te, Td);
Chris@10: 			      Tf = VFMA(LDK(KP559016994), Te, Td);
Chris@10: 			      ST(&(x[WS(rs, 2)]), VFMAI(Tk, Tj), ms, &(x[0]));
Chris@10: 			      ST(&(x[WS(rs, 3)]), VFNMSI(Tk, Tj), ms, &(x[WS(rs, 1)]));
Chris@10: 			      ST(&(x[WS(rs, 4)]), VFMAI(Ti, Tf), ms, &(x[0]));
Chris@10: 			      ST(&(x[WS(rs, 1)]), VFNMSI(Ti, Tf), ms, &(x[WS(rs, 1)]));
Chris@10: 			 }
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, 2),
Chris@10:      VTW(0, 3),
Chris@10:      VTW(0, 4),
Chris@10:      {TW_NEXT, VL, 0}
Chris@10: };
Chris@10: 
Chris@10: static const ct_desc desc = { 5, XSIMD_STRING("t2fv_5"), twinstr, &GENUS, {11, 10, 9, 0}, 0, 0, 0 };
Chris@10: 
Chris@10: void XSIMD(codelet_t2fv_5) (planner *p) {
Chris@10:      X(kdft_dit_register) (p, t2fv_5, &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 5 -name t2fv_5 -include t2f.h */
Chris@10: 
Chris@10: /*
Chris@10:  * This function contains 20 FP additions, 14 FP multiplications,
Chris@10:  * (or, 17 additions, 11 multiplications, 3 fused multiply/add),
Chris@10:  * 20 stack variables, 4 constants, and 10 memory accesses
Chris@10:  */
Chris@10: #include "t2f.h"
Chris@10: 
Chris@10: static void t2fv_5(R *ri, R *ii, const R *W, stride rs, INT mb, INT me, INT ms)
Chris@10: {
Chris@10:      DVK(KP250000000, +0.250000000000000000000000000000000000000000000);
Chris@10:      DVK(KP559016994, +0.559016994374947424102293417182819058860154590);
Chris@10:      DVK(KP587785252, +0.587785252292473129168705954639072768597652438);
Chris@10:      DVK(KP951056516, +0.951056516295153572116439333379382143405698634);
Chris@10:      {
Chris@10: 	  INT m;
Chris@10: 	  R *x;
Chris@10: 	  x = ri;
Chris@10: 	  for (m = mb, W = W + (mb * ((TWVL / VL) * 8)); m < me; m = m + VL, x = x + (VL * ms), W = W + (TWVL * 8), MAKE_VOLATILE_STRIDE(5, rs)) {
Chris@10: 	       V Tc, Tg, Th, T5, Ta, Td;
Chris@10: 	       Tc = LD(&(x[0]), ms, &(x[0]));
Chris@10: 	       {
Chris@10: 		    V T2, T9, T4, T7;
Chris@10: 		    {
Chris@10: 			 V T1, T8, T3, T6;
Chris@10: 			 T1 = LD(&(x[WS(rs, 1)]), ms, &(x[WS(rs, 1)]));
Chris@10: 			 T2 = BYTWJ(&(W[0]), T1);
Chris@10: 			 T8 = LD(&(x[WS(rs, 3)]), ms, &(x[WS(rs, 1)]));
Chris@10: 			 T9 = BYTWJ(&(W[TWVL * 4]), T8);
Chris@10: 			 T3 = LD(&(x[WS(rs, 4)]), ms, &(x[0]));
Chris@10: 			 T4 = BYTWJ(&(W[TWVL * 6]), T3);
Chris@10: 			 T6 = LD(&(x[WS(rs, 2)]), ms, &(x[0]));
Chris@10: 			 T7 = BYTWJ(&(W[TWVL * 2]), T6);
Chris@10: 		    }
Chris@10: 		    Tg = VSUB(T2, T4);
Chris@10: 		    Th = VSUB(T7, T9);
Chris@10: 		    T5 = VADD(T2, T4);
Chris@10: 		    Ta = VADD(T7, T9);
Chris@10: 		    Td = VADD(T5, Ta);
Chris@10: 	       }
Chris@10: 	       ST(&(x[0]), VADD(Tc, Td), ms, &(x[0]));
Chris@10: 	       {
Chris@10: 		    V Ti, Tj, Tf, Tk, Tb, Te;
Chris@10: 		    Ti = VBYI(VFMA(LDK(KP951056516), Tg, VMUL(LDK(KP587785252), Th)));
Chris@10: 		    Tj = VBYI(VFNMS(LDK(KP587785252), Tg, VMUL(LDK(KP951056516), Th)));
Chris@10: 		    Tb = VMUL(LDK(KP559016994), VSUB(T5, Ta));
Chris@10: 		    Te = VFNMS(LDK(KP250000000), Td, Tc);
Chris@10: 		    Tf = VADD(Tb, Te);
Chris@10: 		    Tk = VSUB(Te, Tb);
Chris@10: 		    ST(&(x[WS(rs, 1)]), VSUB(Tf, Ti), ms, &(x[WS(rs, 1)]));
Chris@10: 		    ST(&(x[WS(rs, 3)]), VSUB(Tk, Tj), ms, &(x[WS(rs, 1)]));
Chris@10: 		    ST(&(x[WS(rs, 4)]), VADD(Ti, Tf), ms, &(x[0]));
Chris@10: 		    ST(&(x[WS(rs, 2)]), VADD(Tj, Tk), ms, &(x[0]));
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, 2),
Chris@10:      VTW(0, 3),
Chris@10:      VTW(0, 4),
Chris@10:      {TW_NEXT, VL, 0}
Chris@10: };
Chris@10: 
Chris@10: static const ct_desc desc = { 5, XSIMD_STRING("t2fv_5"), twinstr, &GENUS, {17, 11, 3, 0}, 0, 0, 0 };
Chris@10: 
Chris@10: void XSIMD(codelet_t2fv_5) (planner *p) {
Chris@10:      X(kdft_dit_register) (p, t2fv_5, &desc);
Chris@10: }
Chris@10: #endif				/* HAVE_FMA */