cannam@167: /*
cannam@167:  * Copyright (c) 2003, 2007-14 Matteo Frigo
cannam@167:  * Copyright (c) 2003, 2007-14 Massachusetts Institute of Technology
cannam@167:  *
cannam@167:  * This program is free software; you can redistribute it and/or modify
cannam@167:  * it under the terms of the GNU General Public License as published by
cannam@167:  * the Free Software Foundation; either version 2 of the License, or
cannam@167:  * (at your option) any later version.
cannam@167:  *
cannam@167:  * This program is distributed in the hope that it will be useful,
cannam@167:  * but WITHOUT ANY WARRANTY; without even the implied warranty of
cannam@167:  * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
cannam@167:  * GNU General Public License for more details.
cannam@167:  *
cannam@167:  * You should have received a copy of the GNU General Public License
cannam@167:  * along with this program; if not, write to the Free Software
cannam@167:  * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA  02110-1301  USA
cannam@167:  *
cannam@167:  */
cannam@167: 
cannam@167: /* This file was automatically generated --- DO NOT EDIT */
cannam@167: /* Generated on Thu May 24 08:06:11 EDT 2018 */
cannam@167: 
cannam@167: #include "dft/codelet-dft.h"
cannam@167: 
cannam@167: #if defined(ARCH_PREFERS_FMA) || defined(ISA_EXTENSION_PREFERS_FMA)
cannam@167: 
cannam@167: /* Generated by: ../../../genfft/gen_twiddle.native -fma -simd -compact -variables 4 -pipeline-latency 8 -twiddle-log3 -precompute-twiddles -n 4 -name t2sv_4 -include dft/simd/ts.h */
cannam@167: 
cannam@167: /*
cannam@167:  * This function contains 24 FP additions, 16 FP multiplications,
cannam@167:  * (or, 16 additions, 8 multiplications, 8 fused multiply/add),
cannam@167:  * 21 stack variables, 0 constants, and 16 memory accesses
cannam@167:  */
cannam@167: #include "dft/simd/ts.h"
cannam@167: 
cannam@167: static void t2sv_4(R *ri, R *ii, const R *W, stride rs, INT mb, INT me, INT ms)
cannam@167: {
cannam@167:      {
cannam@167: 	  INT m;
cannam@167: 	  for (m = mb, W = W + (mb * 4); m < me; m = m + (2 * VL), ri = ri + ((2 * VL) * ms), ii = ii + ((2 * VL) * ms), W = W + ((2 * VL) * 4), MAKE_VOLATILE_STRIDE(8, rs)) {
cannam@167: 	       V T2, T6, T3, T5, T7, Tb, T4, Ta;
cannam@167: 	       T2 = LDW(&(W[0]));
cannam@167: 	       T6 = LDW(&(W[TWVL * 3]));
cannam@167: 	       T3 = LDW(&(W[TWVL * 2]));
cannam@167: 	       T4 = VMUL(T2, T3);
cannam@167: 	       Ta = VMUL(T2, T6);
cannam@167: 	       T5 = LDW(&(W[TWVL * 1]));
cannam@167: 	       T7 = VFMA(T5, T6, T4);
cannam@167: 	       Tb = VFNMS(T5, T3, Ta);
cannam@167: 	       {
cannam@167: 		    V T1, Tx, Td, Tw, Ti, Tq, Tm, Ts;
cannam@167: 		    T1 = LD(&(ri[0]), ms, &(ri[0]));
cannam@167: 		    Tx = LD(&(ii[0]), ms, &(ii[0]));
cannam@167: 		    {
cannam@167: 			 V T8, T9, Tc, Tv;
cannam@167: 			 T8 = LD(&(ri[WS(rs, 2)]), ms, &(ri[0]));
cannam@167: 			 T9 = VMUL(T7, T8);
cannam@167: 			 Tc = LD(&(ii[WS(rs, 2)]), ms, &(ii[0]));
cannam@167: 			 Tv = VMUL(T7, Tc);
cannam@167: 			 Td = VFMA(Tb, Tc, T9);
cannam@167: 			 Tw = VFNMS(Tb, T8, Tv);
cannam@167: 		    }
cannam@167: 		    {
cannam@167: 			 V Tf, Tg, Th, Tp;
cannam@167: 			 Tf = LD(&(ri[WS(rs, 1)]), ms, &(ri[WS(rs, 1)]));
cannam@167: 			 Tg = VMUL(T2, Tf);
cannam@167: 			 Th = LD(&(ii[WS(rs, 1)]), ms, &(ii[WS(rs, 1)]));
cannam@167: 			 Tp = VMUL(T2, Th);
cannam@167: 			 Ti = VFMA(T5, Th, Tg);
cannam@167: 			 Tq = VFNMS(T5, Tf, Tp);
cannam@167: 		    }
cannam@167: 		    {
cannam@167: 			 V Tj, Tk, Tl, Tr;
cannam@167: 			 Tj = LD(&(ri[WS(rs, 3)]), ms, &(ri[WS(rs, 1)]));
cannam@167: 			 Tk = VMUL(T3, Tj);
cannam@167: 			 Tl = LD(&(ii[WS(rs, 3)]), ms, &(ii[WS(rs, 1)]));
cannam@167: 			 Tr = VMUL(T3, Tl);
cannam@167: 			 Tm = VFMA(T6, Tl, Tk);
cannam@167: 			 Ts = VFNMS(T6, Tj, Tr);
cannam@167: 		    }
cannam@167: 		    {
cannam@167: 			 V Te, Tn, Tu, Ty;
cannam@167: 			 Te = VADD(T1, Td);
cannam@167: 			 Tn = VADD(Ti, Tm);
cannam@167: 			 ST(&(ri[WS(rs, 2)]), VSUB(Te, Tn), ms, &(ri[0]));
cannam@167: 			 ST(&(ri[0]), VADD(Te, Tn), ms, &(ri[0]));
cannam@167: 			 Tu = VADD(Tq, Ts);
cannam@167: 			 Ty = VADD(Tw, Tx);
cannam@167: 			 ST(&(ii[0]), VADD(Tu, Ty), ms, &(ii[0]));
cannam@167: 			 ST(&(ii[WS(rs, 2)]), VSUB(Ty, Tu), ms, &(ii[0]));
cannam@167: 		    }
cannam@167: 		    {
cannam@167: 			 V To, Tt, Tz, TA;
cannam@167: 			 To = VSUB(T1, Td);
cannam@167: 			 Tt = VSUB(Tq, Ts);
cannam@167: 			 ST(&(ri[WS(rs, 3)]), VSUB(To, Tt), ms, &(ri[WS(rs, 1)]));
cannam@167: 			 ST(&(ri[WS(rs, 1)]), VADD(To, Tt), ms, &(ri[WS(rs, 1)]));
cannam@167: 			 Tz = VSUB(Tx, Tw);
cannam@167: 			 TA = VSUB(Ti, Tm);
cannam@167: 			 ST(&(ii[WS(rs, 1)]), VSUB(Tz, TA), ms, &(ii[WS(rs, 1)]));
cannam@167: 			 ST(&(ii[WS(rs, 3)]), VADD(TA, Tz), ms, &(ii[WS(rs, 1)]));
cannam@167: 		    }
cannam@167: 	       }
cannam@167: 	  }
cannam@167:      }
cannam@167:      VLEAVE();
cannam@167: }
cannam@167: 
cannam@167: static const tw_instr twinstr[] = {
cannam@167:      VTW(0, 1),
cannam@167:      VTW(0, 3),
cannam@167:      {TW_NEXT, (2 * VL), 0}
cannam@167: };
cannam@167: 
cannam@167: static const ct_desc desc = { 4, XSIMD_STRING("t2sv_4"), twinstr, &GENUS, {16, 8, 8, 0}, 0, 0, 0 };
cannam@167: 
cannam@167: void XSIMD(codelet_t2sv_4) (planner *p) {
cannam@167:      X(kdft_dit_register) (p, t2sv_4, &desc);
cannam@167: }
cannam@167: #else
cannam@167: 
cannam@167: /* Generated by: ../../../genfft/gen_twiddle.native -simd -compact -variables 4 -pipeline-latency 8 -twiddle-log3 -precompute-twiddles -n 4 -name t2sv_4 -include dft/simd/ts.h */
cannam@167: 
cannam@167: /*
cannam@167:  * This function contains 24 FP additions, 16 FP multiplications,
cannam@167:  * (or, 16 additions, 8 multiplications, 8 fused multiply/add),
cannam@167:  * 21 stack variables, 0 constants, and 16 memory accesses
cannam@167:  */
cannam@167: #include "dft/simd/ts.h"
cannam@167: 
cannam@167: static void t2sv_4(R *ri, R *ii, const R *W, stride rs, INT mb, INT me, INT ms)
cannam@167: {
cannam@167:      {
cannam@167: 	  INT m;
cannam@167: 	  for (m = mb, W = W + (mb * 4); m < me; m = m + (2 * VL), ri = ri + ((2 * VL) * ms), ii = ii + ((2 * VL) * ms), W = W + ((2 * VL) * 4), MAKE_VOLATILE_STRIDE(8, rs)) {
cannam@167: 	       V T2, T4, T3, T5, T6, T8;
cannam@167: 	       T2 = LDW(&(W[0]));
cannam@167: 	       T4 = LDW(&(W[TWVL * 1]));
cannam@167: 	       T3 = LDW(&(W[TWVL * 2]));
cannam@167: 	       T5 = LDW(&(W[TWVL * 3]));
cannam@167: 	       T6 = VFMA(T2, T3, VMUL(T4, T5));
cannam@167: 	       T8 = VFNMS(T4, T3, VMUL(T2, T5));
cannam@167: 	       {
cannam@167: 		    V T1, Tp, Ta, To, Te, Tk, Th, Tl, T7, T9;
cannam@167: 		    T1 = LD(&(ri[0]), ms, &(ri[0]));
cannam@167: 		    Tp = LD(&(ii[0]), ms, &(ii[0]));
cannam@167: 		    T7 = LD(&(ri[WS(rs, 2)]), ms, &(ri[0]));
cannam@167: 		    T9 = LD(&(ii[WS(rs, 2)]), ms, &(ii[0]));
cannam@167: 		    Ta = VFMA(T6, T7, VMUL(T8, T9));
cannam@167: 		    To = VFNMS(T8, T7, VMUL(T6, T9));
cannam@167: 		    {
cannam@167: 			 V Tc, Td, Tf, Tg;
cannam@167: 			 Tc = LD(&(ri[WS(rs, 1)]), ms, &(ri[WS(rs, 1)]));
cannam@167: 			 Td = LD(&(ii[WS(rs, 1)]), ms, &(ii[WS(rs, 1)]));
cannam@167: 			 Te = VFMA(T2, Tc, VMUL(T4, Td));
cannam@167: 			 Tk = VFNMS(T4, Tc, VMUL(T2, Td));
cannam@167: 			 Tf = LD(&(ri[WS(rs, 3)]), ms, &(ri[WS(rs, 1)]));
cannam@167: 			 Tg = LD(&(ii[WS(rs, 3)]), ms, &(ii[WS(rs, 1)]));
cannam@167: 			 Th = VFMA(T3, Tf, VMUL(T5, Tg));
cannam@167: 			 Tl = VFNMS(T5, Tf, VMUL(T3, Tg));
cannam@167: 		    }
cannam@167: 		    {
cannam@167: 			 V Tb, Ti, Tn, Tq;
cannam@167: 			 Tb = VADD(T1, Ta);
cannam@167: 			 Ti = VADD(Te, Th);
cannam@167: 			 ST(&(ri[WS(rs, 2)]), VSUB(Tb, Ti), ms, &(ri[0]));
cannam@167: 			 ST(&(ri[0]), VADD(Tb, Ti), ms, &(ri[0]));
cannam@167: 			 Tn = VADD(Tk, Tl);
cannam@167: 			 Tq = VADD(To, Tp);
cannam@167: 			 ST(&(ii[0]), VADD(Tn, Tq), ms, &(ii[0]));
cannam@167: 			 ST(&(ii[WS(rs, 2)]), VSUB(Tq, Tn), ms, &(ii[0]));
cannam@167: 		    }
cannam@167: 		    {
cannam@167: 			 V Tj, Tm, Tr, Ts;
cannam@167: 			 Tj = VSUB(T1, Ta);
cannam@167: 			 Tm = VSUB(Tk, Tl);
cannam@167: 			 ST(&(ri[WS(rs, 3)]), VSUB(Tj, Tm), ms, &(ri[WS(rs, 1)]));
cannam@167: 			 ST(&(ri[WS(rs, 1)]), VADD(Tj, Tm), ms, &(ri[WS(rs, 1)]));
cannam@167: 			 Tr = VSUB(Tp, To);
cannam@167: 			 Ts = VSUB(Te, Th);
cannam@167: 			 ST(&(ii[WS(rs, 1)]), VSUB(Tr, Ts), ms, &(ii[WS(rs, 1)]));
cannam@167: 			 ST(&(ii[WS(rs, 3)]), VADD(Ts, Tr), ms, &(ii[WS(rs, 1)]));
cannam@167: 		    }
cannam@167: 	       }
cannam@167: 	  }
cannam@167:      }
cannam@167:      VLEAVE();
cannam@167: }
cannam@167: 
cannam@167: static const tw_instr twinstr[] = {
cannam@167:      VTW(0, 1),
cannam@167:      VTW(0, 3),
cannam@167:      {TW_NEXT, (2 * VL), 0}
cannam@167: };
cannam@167: 
cannam@167: static const ct_desc desc = { 4, XSIMD_STRING("t2sv_4"), twinstr, &GENUS, {16, 8, 8, 0}, 0, 0, 0 };
cannam@167: 
cannam@167: void XSIMD(codelet_t2sv_4) (planner *p) {
cannam@167:      X(kdft_dit_register) (p, t2sv_4, &desc);
cannam@167: }
cannam@167: #endif