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 08:05:58 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 12 -name t1bv_12 -include dft/simd/t1b.h -sign 1 */
Chris@82: 
Chris@82: /*
Chris@82:  * This function contains 59 FP additions, 42 FP multiplications,
Chris@82:  * (or, 41 additions, 24 multiplications, 18 fused multiply/add),
Chris@82:  * 28 stack variables, 2 constants, and 24 memory accesses
Chris@82:  */
Chris@82: #include "dft/simd/t1b.h"
Chris@82: 
Chris@82: static void t1bv_12(R *ri, R *ii, const R *W, stride rs, INT mb, INT me, INT ms)
Chris@82: {
Chris@82:      DVK(KP866025403, +0.866025403784438646763723170752936183471402627);
Chris@82:      DVK(KP500000000, +0.500000000000000000000000000000000000000000000);
Chris@82:      {
Chris@82: 	  INT m;
Chris@82: 	  R *x;
Chris@82: 	  x = ii;
Chris@82: 	  for (m = mb, W = W + (mb * ((TWVL / VL) * 22)); m < me; m = m + VL, x = x + (VL * ms), W = W + (TWVL * 22), MAKE_VOLATILE_STRIDE(12, rs)) {
Chris@82: 	       V T1, TK, T6, TA, Tq, TI, Tv, TE, T9, TL, Te, TB, Ti, TH, Tn;
Chris@82: 	       V TD;
Chris@82: 	       {
Chris@82: 		    V T5, T3, T4, T2;
Chris@82: 		    T1 = LD(&(x[0]), ms, &(x[0]));
Chris@82: 		    T4 = LD(&(x[WS(rs, 8)]), ms, &(x[0]));
Chris@82: 		    T5 = BYTW(&(W[TWVL * 14]), T4);
Chris@82: 		    T2 = LD(&(x[WS(rs, 4)]), ms, &(x[0]));
Chris@82: 		    T3 = BYTW(&(W[TWVL * 6]), T2);
Chris@82: 		    TK = VSUB(T3, T5);
Chris@82: 		    T6 = VADD(T3, T5);
Chris@82: 		    TA = VFNMS(LDK(KP500000000), T6, T1);
Chris@82: 	       }
Chris@82: 	       {
Chris@82: 		    V Tu, Ts, Tp, Tt, Tr;
Chris@82: 		    Tp = LD(&(x[WS(rs, 9)]), ms, &(x[WS(rs, 1)]));
Chris@82: 		    Tq = BYTW(&(W[TWVL * 16]), Tp);
Chris@82: 		    Tt = LD(&(x[WS(rs, 5)]), ms, &(x[WS(rs, 1)]));
Chris@82: 		    Tu = BYTW(&(W[TWVL * 8]), Tt);
Chris@82: 		    Tr = LD(&(x[WS(rs, 1)]), ms, &(x[WS(rs, 1)]));
Chris@82: 		    Ts = BYTW(&(W[0]), Tr);
Chris@82: 		    TI = VSUB(Tu, Ts);
Chris@82: 		    Tv = VADD(Ts, Tu);
Chris@82: 		    TE = VFNMS(LDK(KP500000000), Tv, Tq);
Chris@82: 	       }
Chris@82: 	       {
Chris@82: 		    V Td, Tb, T8, Tc, Ta;
Chris@82: 		    T8 = LD(&(x[WS(rs, 6)]), ms, &(x[0]));
Chris@82: 		    T9 = BYTW(&(W[TWVL * 10]), T8);
Chris@82: 		    Tc = LD(&(x[WS(rs, 2)]), ms, &(x[0]));
Chris@82: 		    Td = BYTW(&(W[TWVL * 2]), Tc);
Chris@82: 		    Ta = LD(&(x[WS(rs, 10)]), ms, &(x[0]));
Chris@82: 		    Tb = BYTW(&(W[TWVL * 18]), Ta);
Chris@82: 		    TL = VSUB(Tb, Td);
Chris@82: 		    Te = VADD(Tb, Td);
Chris@82: 		    TB = VFNMS(LDK(KP500000000), Te, T9);
Chris@82: 	       }
Chris@82: 	       {
Chris@82: 		    V Tm, Tk, Th, Tl, Tj;
Chris@82: 		    Th = LD(&(x[WS(rs, 3)]), ms, &(x[WS(rs, 1)]));
Chris@82: 		    Ti = BYTW(&(W[TWVL * 4]), Th);
Chris@82: 		    Tl = LD(&(x[WS(rs, 11)]), ms, &(x[WS(rs, 1)]));
Chris@82: 		    Tm = BYTW(&(W[TWVL * 20]), Tl);
Chris@82: 		    Tj = LD(&(x[WS(rs, 7)]), ms, &(x[WS(rs, 1)]));
Chris@82: 		    Tk = BYTW(&(W[TWVL * 12]), Tj);
Chris@82: 		    TH = VSUB(Tk, Tm);
Chris@82: 		    Tn = VADD(Tk, Tm);
Chris@82: 		    TD = VFNMS(LDK(KP500000000), Tn, Ti);
Chris@82: 	       }
Chris@82: 	       {
Chris@82: 		    V Tg, Ty, Tx, Tz;
Chris@82: 		    {
Chris@82: 			 V T7, Tf, To, Tw;
Chris@82: 			 T7 = VADD(T1, T6);
Chris@82: 			 Tf = VADD(T9, Te);
Chris@82: 			 Tg = VSUB(T7, Tf);
Chris@82: 			 Ty = VADD(T7, Tf);
Chris@82: 			 To = VADD(Ti, Tn);
Chris@82: 			 Tw = VADD(Tq, Tv);
Chris@82: 			 Tx = VSUB(To, Tw);
Chris@82: 			 Tz = VADD(To, Tw);
Chris@82: 		    }
Chris@82: 		    ST(&(x[WS(rs, 3)]), VFNMSI(Tx, Tg), ms, &(x[WS(rs, 1)]));
Chris@82: 		    ST(&(x[0]), VADD(Ty, Tz), ms, &(x[0]));
Chris@82: 		    ST(&(x[WS(rs, 9)]), VFMAI(Tx, Tg), ms, &(x[WS(rs, 1)]));
Chris@82: 		    ST(&(x[WS(rs, 6)]), VSUB(Ty, Tz), ms, &(x[0]));
Chris@82: 	       }
Chris@82: 	       {
Chris@82: 		    V TS, TW, TV, TX;
Chris@82: 		    {
Chris@82: 			 V TQ, TR, TT, TU;
Chris@82: 			 TQ = VSUB(TA, TB);
Chris@82: 			 TR = VADD(TH, TI);
Chris@82: 			 TS = VFNMS(LDK(KP866025403), TR, TQ);
Chris@82: 			 TW = VFMA(LDK(KP866025403), TR, TQ);
Chris@82: 			 TT = VSUB(TD, TE);
Chris@82: 			 TU = VSUB(TK, TL);
Chris@82: 			 TV = VFMA(LDK(KP866025403), TU, TT);
Chris@82: 			 TX = VFNMS(LDK(KP866025403), TU, TT);
Chris@82: 		    }
Chris@82: 		    ST(&(x[WS(rs, 1)]), VFMAI(TV, TS), ms, &(x[WS(rs, 1)]));
Chris@82: 		    ST(&(x[WS(rs, 7)]), VFNMSI(TX, TW), ms, &(x[WS(rs, 1)]));
Chris@82: 		    ST(&(x[WS(rs, 11)]), VFNMSI(TV, TS), ms, &(x[WS(rs, 1)]));
Chris@82: 		    ST(&(x[WS(rs, 5)]), VFMAI(TX, TW), ms, &(x[WS(rs, 1)]));
Chris@82: 	       }
Chris@82: 	       {
Chris@82: 		    V TG, TO, TN, TP;
Chris@82: 		    {
Chris@82: 			 V TC, TF, TJ, TM;
Chris@82: 			 TC = VADD(TA, TB);
Chris@82: 			 TF = VADD(TD, TE);
Chris@82: 			 TG = VSUB(TC, TF);
Chris@82: 			 TO = VADD(TC, TF);
Chris@82: 			 TJ = VSUB(TH, TI);
Chris@82: 			 TM = VADD(TK, TL);
Chris@82: 			 TN = VMUL(LDK(KP866025403), VSUB(TJ, TM));
Chris@82: 			 TP = VMUL(LDK(KP866025403), VADD(TM, TJ));
Chris@82: 		    }
Chris@82: 		    ST(&(x[WS(rs, 10)]), VFNMSI(TN, TG), ms, &(x[0]));
Chris@82: 		    ST(&(x[WS(rs, 4)]), VFMAI(TP, TO), ms, &(x[0]));
Chris@82: 		    ST(&(x[WS(rs, 2)]), VFMAI(TN, TG), ms, &(x[0]));
Chris@82: 		    ST(&(x[WS(rs, 8)]), VFNMSI(TP, TO), ms, &(x[0]));
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:      VTW(0, 8),
Chris@82:      VTW(0, 9),
Chris@82:      VTW(0, 10),
Chris@82:      VTW(0, 11),
Chris@82:      {TW_NEXT, VL, 0}
Chris@82: };
Chris@82: 
Chris@82: static const ct_desc desc = { 12, XSIMD_STRING("t1bv_12"), twinstr, &GENUS, {41, 24, 18, 0}, 0, 0, 0 };
Chris@82: 
Chris@82: void XSIMD(codelet_t1bv_12) (planner *p) {
Chris@82:      X(kdft_dit_register) (p, t1bv_12, &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 12 -name t1bv_12 -include dft/simd/t1b.h -sign 1 */
Chris@82: 
Chris@82: /*
Chris@82:  * This function contains 59 FP additions, 30 FP multiplications,
Chris@82:  * (or, 55 additions, 26 multiplications, 4 fused multiply/add),
Chris@82:  * 28 stack variables, 2 constants, and 24 memory accesses
Chris@82:  */
Chris@82: #include "dft/simd/t1b.h"
Chris@82: 
Chris@82: static void t1bv_12(R *ri, R *ii, const R *W, stride rs, INT mb, INT me, INT ms)
Chris@82: {
Chris@82:      DVK(KP866025403, +0.866025403784438646763723170752936183471402627);
Chris@82:      DVK(KP500000000, +0.500000000000000000000000000000000000000000000);
Chris@82:      {
Chris@82: 	  INT m;
Chris@82: 	  R *x;
Chris@82: 	  x = ii;
Chris@82: 	  for (m = mb, W = W + (mb * ((TWVL / VL) * 22)); m < me; m = m + VL, x = x + (VL * ms), W = W + (TWVL * 22), MAKE_VOLATILE_STRIDE(12, rs)) {
Chris@82: 	       V T1, Tt, T6, T7, TB, Tq, TC, TD, T9, Tu, Te, Tf, Tx, Tl, Ty;
Chris@82: 	       V Tz;
Chris@82: 	       {
Chris@82: 		    V T5, T3, T4, T2;
Chris@82: 		    T1 = LD(&(x[0]), ms, &(x[0]));
Chris@82: 		    T4 = LD(&(x[WS(rs, 8)]), ms, &(x[0]));
Chris@82: 		    T5 = BYTW(&(W[TWVL * 14]), T4);
Chris@82: 		    T2 = LD(&(x[WS(rs, 4)]), ms, &(x[0]));
Chris@82: 		    T3 = BYTW(&(W[TWVL * 6]), T2);
Chris@82: 		    Tt = VSUB(T3, T5);
Chris@82: 		    T6 = VADD(T3, T5);
Chris@82: 		    T7 = VFNMS(LDK(KP500000000), T6, T1);
Chris@82: 	       }
Chris@82: 	       {
Chris@82: 		    V Tn, Tp, Tm, TA, To;
Chris@82: 		    Tm = LD(&(x[WS(rs, 1)]), ms, &(x[WS(rs, 1)]));
Chris@82: 		    Tn = BYTW(&(W[0]), Tm);
Chris@82: 		    TA = LD(&(x[WS(rs, 9)]), ms, &(x[WS(rs, 1)]));
Chris@82: 		    TB = BYTW(&(W[TWVL * 16]), TA);
Chris@82: 		    To = LD(&(x[WS(rs, 5)]), ms, &(x[WS(rs, 1)]));
Chris@82: 		    Tp = BYTW(&(W[TWVL * 8]), To);
Chris@82: 		    Tq = VSUB(Tn, Tp);
Chris@82: 		    TC = VADD(Tn, Tp);
Chris@82: 		    TD = VFNMS(LDK(KP500000000), TC, TB);
Chris@82: 	       }
Chris@82: 	       {
Chris@82: 		    V Td, Tb, T8, Tc, Ta;
Chris@82: 		    T8 = LD(&(x[WS(rs, 6)]), ms, &(x[0]));
Chris@82: 		    T9 = BYTW(&(W[TWVL * 10]), T8);
Chris@82: 		    Tc = LD(&(x[WS(rs, 2)]), ms, &(x[0]));
Chris@82: 		    Td = BYTW(&(W[TWVL * 2]), Tc);
Chris@82: 		    Ta = LD(&(x[WS(rs, 10)]), ms, &(x[0]));
Chris@82: 		    Tb = BYTW(&(W[TWVL * 18]), Ta);
Chris@82: 		    Tu = VSUB(Tb, Td);
Chris@82: 		    Te = VADD(Tb, Td);
Chris@82: 		    Tf = VFNMS(LDK(KP500000000), Te, T9);
Chris@82: 	       }
Chris@82: 	       {
Chris@82: 		    V Ti, Tk, Th, Tw, Tj;
Chris@82: 		    Th = LD(&(x[WS(rs, 7)]), ms, &(x[WS(rs, 1)]));
Chris@82: 		    Ti = BYTW(&(W[TWVL * 12]), Th);
Chris@82: 		    Tw = LD(&(x[WS(rs, 3)]), ms, &(x[WS(rs, 1)]));
Chris@82: 		    Tx = BYTW(&(W[TWVL * 4]), Tw);
Chris@82: 		    Tj = LD(&(x[WS(rs, 11)]), ms, &(x[WS(rs, 1)]));
Chris@82: 		    Tk = BYTW(&(W[TWVL * 20]), Tj);
Chris@82: 		    Tl = VSUB(Ti, Tk);
Chris@82: 		    Ty = VADD(Ti, Tk);
Chris@82: 		    Tz = VFNMS(LDK(KP500000000), Ty, Tx);
Chris@82: 	       }
Chris@82: 	       {
Chris@82: 		    V Ts, TG, TF, TH;
Chris@82: 		    {
Chris@82: 			 V Tg, Tr, Tv, TE;
Chris@82: 			 Tg = VSUB(T7, Tf);
Chris@82: 			 Tr = VMUL(LDK(KP866025403), VSUB(Tl, Tq));
Chris@82: 			 Ts = VSUB(Tg, Tr);
Chris@82: 			 TG = VADD(Tg, Tr);
Chris@82: 			 Tv = VMUL(LDK(KP866025403), VSUB(Tt, Tu));
Chris@82: 			 TE = VSUB(Tz, TD);
Chris@82: 			 TF = VBYI(VADD(Tv, TE));
Chris@82: 			 TH = VBYI(VSUB(TE, Tv));
Chris@82: 		    }
Chris@82: 		    ST(&(x[WS(rs, 11)]), VSUB(Ts, TF), ms, &(x[WS(rs, 1)]));
Chris@82: 		    ST(&(x[WS(rs, 5)]), VADD(TG, TH), ms, &(x[WS(rs, 1)]));
Chris@82: 		    ST(&(x[WS(rs, 1)]), VADD(Ts, TF), ms, &(x[WS(rs, 1)]));
Chris@82: 		    ST(&(x[WS(rs, 7)]), VSUB(TG, TH), ms, &(x[WS(rs, 1)]));
Chris@82: 	       }
Chris@82: 	       {
Chris@82: 		    V TS, TW, TV, TX;
Chris@82: 		    {
Chris@82: 			 V TQ, TR, TT, TU;
Chris@82: 			 TQ = VADD(T1, T6);
Chris@82: 			 TR = VADD(T9, Te);
Chris@82: 			 TS = VSUB(TQ, TR);
Chris@82: 			 TW = VADD(TQ, TR);
Chris@82: 			 TT = VADD(Tx, Ty);
Chris@82: 			 TU = VADD(TB, TC);
Chris@82: 			 TV = VBYI(VSUB(TT, TU));
Chris@82: 			 TX = VADD(TT, TU);
Chris@82: 		    }
Chris@82: 		    ST(&(x[WS(rs, 3)]), VSUB(TS, TV), ms, &(x[WS(rs, 1)]));
Chris@82: 		    ST(&(x[0]), VADD(TW, TX), ms, &(x[0]));
Chris@82: 		    ST(&(x[WS(rs, 9)]), VADD(TS, TV), ms, &(x[WS(rs, 1)]));
Chris@82: 		    ST(&(x[WS(rs, 6)]), VSUB(TW, TX), ms, &(x[0]));
Chris@82: 	       }
Chris@82: 	       {
Chris@82: 		    V TK, TO, TN, TP;
Chris@82: 		    {
Chris@82: 			 V TI, TJ, TL, TM;
Chris@82: 			 TI = VADD(Tl, Tq);
Chris@82: 			 TJ = VADD(Tt, Tu);
Chris@82: 			 TK = VBYI(VMUL(LDK(KP866025403), VSUB(TI, TJ)));
Chris@82: 			 TO = VBYI(VMUL(LDK(KP866025403), VADD(TJ, TI)));
Chris@82: 			 TL = VADD(T7, Tf);
Chris@82: 			 TM = VADD(Tz, TD);
Chris@82: 			 TN = VSUB(TL, TM);
Chris@82: 			 TP = VADD(TL, TM);
Chris@82: 		    }
Chris@82: 		    ST(&(x[WS(rs, 2)]), VADD(TK, TN), ms, &(x[0]));
Chris@82: 		    ST(&(x[WS(rs, 8)]), VSUB(TP, TO), ms, &(x[0]));
Chris@82: 		    ST(&(x[WS(rs, 10)]), VSUB(TN, TK), ms, &(x[0]));
Chris@82: 		    ST(&(x[WS(rs, 4)]), VADD(TO, TP), ms, &(x[0]));
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:      VTW(0, 8),
Chris@82:      VTW(0, 9),
Chris@82:      VTW(0, 10),
Chris@82:      VTW(0, 11),
Chris@82:      {TW_NEXT, VL, 0}
Chris@82: };
Chris@82: 
Chris@82: static const ct_desc desc = { 12, XSIMD_STRING("t1bv_12"), twinstr, &GENUS, {55, 26, 4, 0}, 0, 0, 0 };
Chris@82: 
Chris@82: void XSIMD(codelet_t1bv_12) (planner *p) {
Chris@82:      X(kdft_dit_register) (p, t1bv_12, &desc);
Chris@82: }
Chris@82: #endif