Chris@82: /*
Chris@82:  * Copyright (c) 2003, 2007-14 Matteo Frigo
Chris@82:  * Copyright (c) 2003, 2007-14 Massachusetts Institute of Technology
Chris@82:  *
Chris@82:  * This program is free software; you can redistribute it and/or modify
Chris@82:  * it under the terms of the GNU General Public License as published by
Chris@82:  * the Free Software Foundation; either version 2 of the License, or
Chris@82:  * (at your option) any later version.
Chris@82:  *
Chris@82:  * This program is distributed in the hope that it will be useful,
Chris@82:  * but WITHOUT ANY WARRANTY; without even the implied warranty of
Chris@82:  * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
Chris@82:  * GNU General Public License for more details.
Chris@82:  *
Chris@82:  * You should have received a copy of the GNU General Public License
Chris@82:  * along with this program; if not, write to the Free Software
Chris@82:  * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA  02110-1301  USA
Chris@82:  *
Chris@82:  */
Chris@82: 
Chris@82: /* This file was automatically generated --- DO NOT EDIT */
Chris@82: /* Generated on Thu May 24 08:05:56 EDT 2018 */
Chris@82: 
Chris@82: #include "dft/codelet-dft.h"
Chris@82: 
Chris@82: #if defined(ARCH_PREFERS_FMA) || defined(ISA_EXTENSION_PREFERS_FMA)
Chris@82: 
Chris@82: /* Generated by: ../../../genfft/gen_twiddle_c.native -fma -simd -compact -variables 4 -pipeline-latency 8 -n 8 -name t1buv_8 -include dft/simd/t1bu.h -sign 1 */
Chris@82: 
Chris@82: /*
Chris@82:  * This function contains 33 FP additions, 24 FP multiplications,
Chris@82:  * (or, 23 additions, 14 multiplications, 10 fused multiply/add),
Chris@82:  * 24 stack variables, 1 constants, and 16 memory accesses
Chris@82:  */
Chris@82: #include "dft/simd/t1bu.h"
Chris@82: 
Chris@82: static void t1buv_8(R *ri, R *ii, const R *W, stride rs, INT mb, INT me, INT ms)
Chris@82: {
Chris@82:      DVK(KP707106781, +0.707106781186547524400844362104849039284835938);
Chris@82:      {
Chris@82: 	  INT m;
Chris@82: 	  R *x;
Chris@82: 	  x = ii;
Chris@82: 	  for (m = mb, W = W + (mb * ((TWVL / VL) * 14)); m < me; m = m + VL, x = x + (VL * ms), W = W + (TWVL * 14), MAKE_VOLATILE_STRIDE(8, rs)) {
Chris@82: 	       V T4, Tq, Tl, Tr, T9, Tt, Te, Tu, T1, T3, T2;
Chris@82: 	       T1 = LD(&(x[0]), ms, &(x[0]));
Chris@82: 	       T2 = LD(&(x[WS(rs, 4)]), ms, &(x[0]));
Chris@82: 	       T3 = BYTW(&(W[TWVL * 6]), T2);
Chris@82: 	       T4 = VSUB(T1, T3);
Chris@82: 	       Tq = VADD(T1, T3);
Chris@82: 	       {
Chris@82: 		    V Ti, Tk, Th, Tj;
Chris@82: 		    Th = LD(&(x[WS(rs, 2)]), ms, &(x[0]));
Chris@82: 		    Ti = BYTW(&(W[TWVL * 2]), Th);
Chris@82: 		    Tj = LD(&(x[WS(rs, 6)]), ms, &(x[0]));
Chris@82: 		    Tk = BYTW(&(W[TWVL * 10]), Tj);
Chris@82: 		    Tl = VSUB(Ti, Tk);
Chris@82: 		    Tr = VADD(Ti, Tk);
Chris@82: 	       }
Chris@82: 	       {
Chris@82: 		    V T6, T8, T5, T7;
Chris@82: 		    T5 = LD(&(x[WS(rs, 1)]), ms, &(x[WS(rs, 1)]));
Chris@82: 		    T6 = BYTW(&(W[0]), T5);
Chris@82: 		    T7 = LD(&(x[WS(rs, 5)]), ms, &(x[WS(rs, 1)]));
Chris@82: 		    T8 = BYTW(&(W[TWVL * 8]), T7);
Chris@82: 		    T9 = VSUB(T6, T8);
Chris@82: 		    Tt = VADD(T6, T8);
Chris@82: 	       }
Chris@82: 	       {
Chris@82: 		    V Tb, Td, Ta, Tc;
Chris@82: 		    Ta = LD(&(x[WS(rs, 7)]), ms, &(x[WS(rs, 1)]));
Chris@82: 		    Tb = BYTW(&(W[TWVL * 12]), Ta);
Chris@82: 		    Tc = LD(&(x[WS(rs, 3)]), ms, &(x[WS(rs, 1)]));
Chris@82: 		    Td = BYTW(&(W[TWVL * 4]), Tc);
Chris@82: 		    Te = VSUB(Tb, Td);
Chris@82: 		    Tu = VADD(Tb, Td);
Chris@82: 	       }
Chris@82: 	       {
Chris@82: 		    V Ts, Tv, Tw, Tx;
Chris@82: 		    Ts = VSUB(Tq, Tr);
Chris@82: 		    Tv = VSUB(Tt, Tu);
Chris@82: 		    ST(&(x[WS(rs, 6)]), VFNMSI(Tv, Ts), ms, &(x[0]));
Chris@82: 		    ST(&(x[WS(rs, 2)]), VFMAI(Tv, Ts), ms, &(x[0]));
Chris@82: 		    Tw = VADD(Tq, Tr);
Chris@82: 		    Tx = VADD(Tt, Tu);
Chris@82: 		    ST(&(x[WS(rs, 4)]), VSUB(Tw, Tx), ms, &(x[0]));
Chris@82: 		    ST(&(x[0]), VADD(Tw, Tx), ms, &(x[0]));
Chris@82: 		    {
Chris@82: 			 V Tg, To, Tn, Tp, Tf, Tm;
Chris@82: 			 Tf = VADD(T9, Te);
Chris@82: 			 Tg = VFNMS(LDK(KP707106781), Tf, T4);
Chris@82: 			 To = VFMA(LDK(KP707106781), Tf, T4);
Chris@82: 			 Tm = VSUB(T9, Te);
Chris@82: 			 Tn = VFNMS(LDK(KP707106781), Tm, Tl);
Chris@82: 			 Tp = VFMA(LDK(KP707106781), Tm, Tl);
Chris@82: 			 ST(&(x[WS(rs, 3)]), VFNMSI(Tn, Tg), ms, &(x[WS(rs, 1)]));
Chris@82: 			 ST(&(x[WS(rs, 7)]), VFNMSI(Tp, To), ms, &(x[WS(rs, 1)]));
Chris@82: 			 ST(&(x[WS(rs, 5)]), VFMAI(Tn, Tg), ms, &(x[WS(rs, 1)]));
Chris@82: 			 ST(&(x[WS(rs, 1)]), VFMAI(Tp, To), ms, &(x[WS(rs, 1)]));
Chris@82: 		    }
Chris@82: 	       }
Chris@82: 	  }
Chris@82:      }
Chris@82:      VLEAVE();
Chris@82: }
Chris@82: 
Chris@82: static const tw_instr twinstr[] = {
Chris@82:      VTW(0, 1),
Chris@82:      VTW(0, 2),
Chris@82:      VTW(0, 3),
Chris@82:      VTW(0, 4),
Chris@82:      VTW(0, 5),
Chris@82:      VTW(0, 6),
Chris@82:      VTW(0, 7),
Chris@82:      {TW_NEXT, VL, 0}
Chris@82: };
Chris@82: 
Chris@82: static const ct_desc desc = { 8, XSIMD_STRING("t1buv_8"), twinstr, &GENUS, {23, 14, 10, 0}, 0, 0, 0 };
Chris@82: 
Chris@82: void XSIMD(codelet_t1buv_8) (planner *p) {
Chris@82:      X(kdft_dit_register) (p, t1buv_8, &desc);
Chris@82: }
Chris@82: #else
Chris@82: 
Chris@82: /* Generated by: ../../../genfft/gen_twiddle_c.native -simd -compact -variables 4 -pipeline-latency 8 -n 8 -name t1buv_8 -include dft/simd/t1bu.h -sign 1 */
Chris@82: 
Chris@82: /*
Chris@82:  * This function contains 33 FP additions, 16 FP multiplications,
Chris@82:  * (or, 33 additions, 16 multiplications, 0 fused multiply/add),
Chris@82:  * 24 stack variables, 1 constants, and 16 memory accesses
Chris@82:  */
Chris@82: #include "dft/simd/t1bu.h"
Chris@82: 
Chris@82: static void t1buv_8(R *ri, R *ii, const R *W, stride rs, INT mb, INT me, INT ms)
Chris@82: {
Chris@82:      DVK(KP707106781, +0.707106781186547524400844362104849039284835938);
Chris@82:      {
Chris@82: 	  INT m;
Chris@82: 	  R *x;
Chris@82: 	  x = ii;
Chris@82: 	  for (m = mb, W = W + (mb * ((TWVL / VL) * 14)); m < me; m = m + VL, x = x + (VL * ms), W = W + (TWVL * 14), MAKE_VOLATILE_STRIDE(8, rs)) {
Chris@82: 	       V Tl, Tq, Tg, Tr, T5, Tt, Ta, Tu, Ti, Tk, Tj;
Chris@82: 	       Ti = LD(&(x[0]), ms, &(x[0]));
Chris@82: 	       Tj = LD(&(x[WS(rs, 4)]), ms, &(x[0]));
Chris@82: 	       Tk = BYTW(&(W[TWVL * 6]), Tj);
Chris@82: 	       Tl = VSUB(Ti, Tk);
Chris@82: 	       Tq = VADD(Ti, Tk);
Chris@82: 	       {
Chris@82: 		    V Td, Tf, Tc, Te;
Chris@82: 		    Tc = LD(&(x[WS(rs, 2)]), ms, &(x[0]));
Chris@82: 		    Td = BYTW(&(W[TWVL * 2]), Tc);
Chris@82: 		    Te = LD(&(x[WS(rs, 6)]), ms, &(x[0]));
Chris@82: 		    Tf = BYTW(&(W[TWVL * 10]), Te);
Chris@82: 		    Tg = VSUB(Td, Tf);
Chris@82: 		    Tr = VADD(Td, Tf);
Chris@82: 	       }
Chris@82: 	       {
Chris@82: 		    V T2, T4, T1, T3;
Chris@82: 		    T1 = LD(&(x[WS(rs, 1)]), ms, &(x[WS(rs, 1)]));
Chris@82: 		    T2 = BYTW(&(W[0]), T1);
Chris@82: 		    T3 = LD(&(x[WS(rs, 5)]), ms, &(x[WS(rs, 1)]));
Chris@82: 		    T4 = BYTW(&(W[TWVL * 8]), T3);
Chris@82: 		    T5 = VSUB(T2, T4);
Chris@82: 		    Tt = VADD(T2, T4);
Chris@82: 	       }
Chris@82: 	       {
Chris@82: 		    V T7, T9, T6, T8;
Chris@82: 		    T6 = LD(&(x[WS(rs, 7)]), ms, &(x[WS(rs, 1)]));
Chris@82: 		    T7 = BYTW(&(W[TWVL * 12]), T6);
Chris@82: 		    T8 = LD(&(x[WS(rs, 3)]), ms, &(x[WS(rs, 1)]));
Chris@82: 		    T9 = BYTW(&(W[TWVL * 4]), T8);
Chris@82: 		    Ta = VSUB(T7, T9);
Chris@82: 		    Tu = VADD(T7, T9);
Chris@82: 	       }
Chris@82: 	       {
Chris@82: 		    V Ts, Tv, Tw, Tx;
Chris@82: 		    Ts = VSUB(Tq, Tr);
Chris@82: 		    Tv = VBYI(VSUB(Tt, Tu));
Chris@82: 		    ST(&(x[WS(rs, 6)]), VSUB(Ts, Tv), ms, &(x[0]));
Chris@82: 		    ST(&(x[WS(rs, 2)]), VADD(Ts, Tv), ms, &(x[0]));
Chris@82: 		    Tw = VADD(Tq, Tr);
Chris@82: 		    Tx = VADD(Tt, Tu);
Chris@82: 		    ST(&(x[WS(rs, 4)]), VSUB(Tw, Tx), ms, &(x[0]));
Chris@82: 		    ST(&(x[0]), VADD(Tw, Tx), ms, &(x[0]));
Chris@82: 		    {
Chris@82: 			 V Th, To, Tn, Tp, Tb, Tm;
Chris@82: 			 Tb = VMUL(LDK(KP707106781), VSUB(T5, Ta));
Chris@82: 			 Th = VBYI(VSUB(Tb, Tg));
Chris@82: 			 To = VBYI(VADD(Tg, Tb));
Chris@82: 			 Tm = VMUL(LDK(KP707106781), VADD(T5, Ta));
Chris@82: 			 Tn = VSUB(Tl, Tm);
Chris@82: 			 Tp = VADD(Tl, Tm);
Chris@82: 			 ST(&(x[WS(rs, 3)]), VADD(Th, Tn), ms, &(x[WS(rs, 1)]));
Chris@82: 			 ST(&(x[WS(rs, 7)]), VSUB(Tp, To), ms, &(x[WS(rs, 1)]));
Chris@82: 			 ST(&(x[WS(rs, 5)]), VSUB(Tn, Th), ms, &(x[WS(rs, 1)]));
Chris@82: 			 ST(&(x[WS(rs, 1)]), VADD(To, Tp), ms, &(x[WS(rs, 1)]));
Chris@82: 		    }
Chris@82: 	       }
Chris@82: 	  }
Chris@82:      }
Chris@82:      VLEAVE();
Chris@82: }
Chris@82: 
Chris@82: static const tw_instr twinstr[] = {
Chris@82:      VTW(0, 1),
Chris@82:      VTW(0, 2),
Chris@82:      VTW(0, 3),
Chris@82:      VTW(0, 4),
Chris@82:      VTW(0, 5),
Chris@82:      VTW(0, 6),
Chris@82:      VTW(0, 7),
Chris@82:      {TW_NEXT, VL, 0}
Chris@82: };
Chris@82: 
Chris@82: static const ct_desc desc = { 8, XSIMD_STRING("t1buv_8"), twinstr, &GENUS, {33, 16, 0, 0}, 0, 0, 0 };
Chris@82: 
Chris@82: void XSIMD(codelet_t1buv_8) (planner *p) {
Chris@82:      X(kdft_dit_register) (p, t1buv_8, &desc);
Chris@82: }
Chris@82: #endif