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 07:39:18 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 8 -name t3bv_8 -include t3b.h -sign 1 */
Chris@10: 
Chris@10: /*
Chris@10:  * This function contains 37 FP additions, 32 FP multiplications,
Chris@10:  * (or, 27 additions, 22 multiplications, 10 fused multiply/add),
Chris@10:  * 43 stack variables, 1 constants, and 16 memory accesses
Chris@10:  */
Chris@10: #include "t3b.h"
Chris@10: 
Chris@10: static void t3bv_8(R *ri, R *ii, const R *W, stride rs, INT mb, INT me, INT ms)
Chris@10: {
Chris@10:      DVK(KP707106781, +0.707106781186547524400844362104849039284835938);
Chris@10:      {
Chris@10: 	  INT m;
Chris@10: 	  R *x;
Chris@10: 	  x = ii;
Chris@10: 	  for (m = mb, W = W + (mb * ((TWVL / VL) * 6)); m < me; m = m + VL, x = x + (VL * ms), W = W + (TWVL * 6), MAKE_VOLATILE_STRIDE(8, rs)) {
Chris@10: 	       V T2, T3, Tb, T1, T5, Tn, Tq, T8, Td, T4, Ta, Tp, Tg, Ti, T9;
Chris@10: 	       T2 = LDW(&(W[0]));
Chris@10: 	       T3 = LDW(&(W[TWVL * 2]));
Chris@10: 	       Tb = LDW(&(W[TWVL * 4]));
Chris@10: 	       T1 = LD(&(x[0]), ms, &(x[0]));
Chris@10: 	       T5 = LD(&(x[WS(rs, 4)]), ms, &(x[0]));
Chris@10: 	       Tn = LD(&(x[WS(rs, 2)]), ms, &(x[0]));
Chris@10: 	       Tq = LD(&(x[WS(rs, 6)]), ms, &(x[0]));
Chris@10: 	       T8 = LD(&(x[WS(rs, 1)]), ms, &(x[WS(rs, 1)]));
Chris@10: 	       Td = LD(&(x[WS(rs, 5)]), ms, &(x[WS(rs, 1)]));
Chris@10: 	       T4 = VZMUL(T2, T3);
Chris@10: 	       Ta = VZMULJ(T2, T3);
Chris@10: 	       Tp = VZMULJ(T2, Tb);
Chris@10: 	       Tg = LD(&(x[WS(rs, 7)]), ms, &(x[WS(rs, 1)]));
Chris@10: 	       Ti = LD(&(x[WS(rs, 3)]), ms, &(x[WS(rs, 1)]));
Chris@10: 	       T9 = VZMUL(T2, T8);
Chris@10: 	       {
Chris@10: 		    V T6, To, Tc, Tr, Th, Tj;
Chris@10: 		    T6 = VZMUL(T4, T5);
Chris@10: 		    To = VZMUL(Ta, Tn);
Chris@10: 		    Tc = VZMULJ(Ta, Tb);
Chris@10: 		    Tr = VZMUL(Tp, Tq);
Chris@10: 		    Th = VZMUL(Tb, Tg);
Chris@10: 		    Tj = VZMUL(T3, Ti);
Chris@10: 		    {
Chris@10: 			 V Tx, T7, Te, Ts, Ty, Tk, TB;
Chris@10: 			 Tx = VADD(T1, T6);
Chris@10: 			 T7 = VSUB(T1, T6);
Chris@10: 			 Te = VZMUL(Tc, Td);
Chris@10: 			 Ts = VSUB(To, Tr);
Chris@10: 			 Ty = VADD(To, Tr);
Chris@10: 			 Tk = VSUB(Th, Tj);
Chris@10: 			 TB = VADD(Th, Tj);
Chris@10: 			 {
Chris@10: 			      V Tf, TA, Tz, TD;
Chris@10: 			      Tf = VSUB(T9, Te);
Chris@10: 			      TA = VADD(T9, Te);
Chris@10: 			      Tz = VSUB(Tx, Ty);
Chris@10: 			      TD = VADD(Tx, Ty);
Chris@10: 			      {
Chris@10: 				   V TC, TE, Tl, Tt;
Chris@10: 				   TC = VSUB(TA, TB);
Chris@10: 				   TE = VADD(TA, TB);
Chris@10: 				   Tl = VADD(Tf, Tk);
Chris@10: 				   Tt = VSUB(Tf, Tk);
Chris@10: 				   {
Chris@10: 					V Tu, Tw, Tm, Tv;
Chris@10: 					ST(&(x[0]), VADD(TD, TE), ms, &(x[0]));
Chris@10: 					ST(&(x[WS(rs, 4)]), VSUB(TD, TE), ms, &(x[0]));
Chris@10: 					ST(&(x[WS(rs, 2)]), VFMAI(TC, Tz), ms, &(x[0]));
Chris@10: 					ST(&(x[WS(rs, 6)]), VFNMSI(TC, Tz), ms, &(x[0]));
Chris@10: 					Tu = VFNMS(LDK(KP707106781), Tt, Ts);
Chris@10: 					Tw = VFMA(LDK(KP707106781), Tt, Ts);
Chris@10: 					Tm = VFNMS(LDK(KP707106781), Tl, T7);
Chris@10: 					Tv = VFMA(LDK(KP707106781), Tl, T7);
Chris@10: 					ST(&(x[WS(rs, 1)]), VFMAI(Tw, Tv), ms, &(x[WS(rs, 1)]));
Chris@10: 					ST(&(x[WS(rs, 7)]), VFNMSI(Tw, Tv), ms, &(x[WS(rs, 1)]));
Chris@10: 					ST(&(x[WS(rs, 5)]), VFMAI(Tu, Tm), ms, &(x[WS(rs, 1)]));
Chris@10: 					ST(&(x[WS(rs, 3)]), VFNMSI(Tu, Tm), ms, &(x[WS(rs, 1)]));
Chris@10: 				   }
Chris@10: 			      }
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, 3),
Chris@10:      VTW(0, 7),
Chris@10:      {TW_NEXT, VL, 0}
Chris@10: };
Chris@10: 
Chris@10: static const ct_desc desc = { 8, XSIMD_STRING("t3bv_8"), twinstr, &GENUS, {27, 22, 10, 0}, 0, 0, 0 };
Chris@10: 
Chris@10: void XSIMD(codelet_t3bv_8) (planner *p) {
Chris@10:      X(kdft_dit_register) (p, t3bv_8, &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 8 -name t3bv_8 -include t3b.h -sign 1 */
Chris@10: 
Chris@10: /*
Chris@10:  * This function contains 37 FP additions, 24 FP multiplications,
Chris@10:  * (or, 37 additions, 24 multiplications, 0 fused multiply/add),
Chris@10:  * 31 stack variables, 1 constants, and 16 memory accesses
Chris@10:  */
Chris@10: #include "t3b.h"
Chris@10: 
Chris@10: static void t3bv_8(R *ri, R *ii, const R *W, stride rs, INT mb, INT me, INT ms)
Chris@10: {
Chris@10:      DVK(KP707106781, +0.707106781186547524400844362104849039284835938);
Chris@10:      {
Chris@10: 	  INT m;
Chris@10: 	  R *x;
Chris@10: 	  x = ii;
Chris@10: 	  for (m = mb, W = W + (mb * ((TWVL / VL) * 6)); m < me; m = m + VL, x = x + (VL * ms), W = W + (TWVL * 6), MAKE_VOLATILE_STRIDE(8, rs)) {
Chris@10: 	       V T1, T4, T5, Tp, T6, T7, Tj;
Chris@10: 	       T1 = LDW(&(W[0]));
Chris@10: 	       T4 = LDW(&(W[TWVL * 2]));
Chris@10: 	       T5 = VZMULJ(T1, T4);
Chris@10: 	       Tp = VZMUL(T1, T4);
Chris@10: 	       T6 = LDW(&(W[TWVL * 4]));
Chris@10: 	       T7 = VZMULJ(T5, T6);
Chris@10: 	       Tj = VZMULJ(T1, T6);
Chris@10: 	       {
Chris@10: 		    V Ts, Tx, Tm, Ty, Ta, TA, Tf, TB, To, Tr, Tq;
Chris@10: 		    To = LD(&(x[0]), ms, &(x[0]));
Chris@10: 		    Tq = LD(&(x[WS(rs, 4)]), ms, &(x[0]));
Chris@10: 		    Tr = VZMUL(Tp, Tq);
Chris@10: 		    Ts = VSUB(To, Tr);
Chris@10: 		    Tx = VADD(To, Tr);
Chris@10: 		    {
Chris@10: 			 V Ti, Tl, Th, Tk;
Chris@10: 			 Th = LD(&(x[WS(rs, 2)]), ms, &(x[0]));
Chris@10: 			 Ti = VZMUL(T5, Th);
Chris@10: 			 Tk = LD(&(x[WS(rs, 6)]), ms, &(x[0]));
Chris@10: 			 Tl = VZMUL(Tj, Tk);
Chris@10: 			 Tm = VSUB(Ti, Tl);
Chris@10: 			 Ty = VADD(Ti, Tl);
Chris@10: 		    }
Chris@10: 		    {
Chris@10: 			 V T3, T9, T2, T8;
Chris@10: 			 T2 = LD(&(x[WS(rs, 1)]), ms, &(x[WS(rs, 1)]));
Chris@10: 			 T3 = VZMUL(T1, T2);
Chris@10: 			 T8 = LD(&(x[WS(rs, 5)]), ms, &(x[WS(rs, 1)]));
Chris@10: 			 T9 = VZMUL(T7, T8);
Chris@10: 			 Ta = VSUB(T3, T9);
Chris@10: 			 TA = VADD(T3, T9);
Chris@10: 		    }
Chris@10: 		    {
Chris@10: 			 V Tc, Te, Tb, Td;
Chris@10: 			 Tb = LD(&(x[WS(rs, 7)]), ms, &(x[WS(rs, 1)]));
Chris@10: 			 Tc = VZMUL(T6, Tb);
Chris@10: 			 Td = LD(&(x[WS(rs, 3)]), ms, &(x[WS(rs, 1)]));
Chris@10: 			 Te = VZMUL(T4, Td);
Chris@10: 			 Tf = VSUB(Tc, Te);
Chris@10: 			 TB = VADD(Tc, Te);
Chris@10: 		    }
Chris@10: 		    {
Chris@10: 			 V Tz, TC, TD, TE;
Chris@10: 			 Tz = VSUB(Tx, Ty);
Chris@10: 			 TC = VBYI(VSUB(TA, TB));
Chris@10: 			 ST(&(x[WS(rs, 6)]), VSUB(Tz, TC), ms, &(x[0]));
Chris@10: 			 ST(&(x[WS(rs, 2)]), VADD(Tz, TC), ms, &(x[0]));
Chris@10: 			 TD = VADD(Tx, Ty);
Chris@10: 			 TE = VADD(TA, TB);
Chris@10: 			 ST(&(x[WS(rs, 4)]), VSUB(TD, TE), ms, &(x[0]));
Chris@10: 			 ST(&(x[0]), VADD(TD, TE), ms, &(x[0]));
Chris@10: 			 {
Chris@10: 			      V Tn, Tv, Tu, Tw, Tg, Tt;
Chris@10: 			      Tg = VMUL(LDK(KP707106781), VSUB(Ta, Tf));
Chris@10: 			      Tn = VBYI(VSUB(Tg, Tm));
Chris@10: 			      Tv = VBYI(VADD(Tm, Tg));
Chris@10: 			      Tt = VMUL(LDK(KP707106781), VADD(Ta, Tf));
Chris@10: 			      Tu = VSUB(Ts, Tt);
Chris@10: 			      Tw = VADD(Ts, Tt);
Chris@10: 			      ST(&(x[WS(rs, 3)]), VADD(Tn, Tu), ms, &(x[WS(rs, 1)]));
Chris@10: 			      ST(&(x[WS(rs, 7)]), VSUB(Tw, Tv), ms, &(x[WS(rs, 1)]));
Chris@10: 			      ST(&(x[WS(rs, 5)]), VSUB(Tu, Tn), ms, &(x[WS(rs, 1)]));
Chris@10: 			      ST(&(x[WS(rs, 1)]), VADD(Tv, Tw), 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, 3),
Chris@10:      VTW(0, 7),
Chris@10:      {TW_NEXT, VL, 0}
Chris@10: };
Chris@10: 
Chris@10: static const ct_desc desc = { 8, XSIMD_STRING("t3bv_8"), twinstr, &GENUS, {37, 24, 0, 0}, 0, 0, 0 };
Chris@10: 
Chris@10: void XSIMD(codelet_t3bv_8) (planner *p) {
Chris@10:      X(kdft_dit_register) (p, t3bv_8, &desc);
Chris@10: }
Chris@10: #endif				/* HAVE_FMA */