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 07:39:04 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 16 -name t1bv_16 -include t1b.h -sign 1 */
Chris@10: 
Chris@10: /*
Chris@10:  * This function contains 87 FP additions, 64 FP multiplications,
Chris@10:  * (or, 53 additions, 30 multiplications, 34 fused multiply/add),
Chris@10:  * 61 stack variables, 3 constants, and 32 memory accesses
Chris@10:  */
Chris@10: #include "t1b.h"
Chris@10: 
Chris@10: static void t1bv_16(R *ri, R *ii, const R *W, stride rs, INT mb, INT me, INT ms)
Chris@10: {
Chris@10:      DVK(KP923879532, +0.923879532511286756128183189396788286822416626);
Chris@10:      DVK(KP414213562, +0.414213562373095048801688724209698078569671875);
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) * 30)); m < me; m = m + VL, x = x + (VL * ms), W = W + (TWVL * 30), MAKE_VOLATILE_STRIDE(16, rs)) {
Chris@10: 	       V TO, Ta, TJ, TP, T14, Tq, T1i, T10, T1b, T1l, T13, T1c, TR, Tl, T15;
Chris@10: 	       V Tv;
Chris@10: 	       {
Chris@10: 		    V Tc, TW, T4, T19, T9, TD, TI, Tj, TZ, T1a, Te, Th, Tn, Tr, Tu;
Chris@10: 		    V Tp;
Chris@10: 		    {
Chris@10: 			 V T1, T2, T5, T7;
Chris@10: 			 T1 = LD(&(x[0]), ms, &(x[0]));
Chris@10: 			 T2 = LD(&(x[WS(rs, 8)]), ms, &(x[0]));
Chris@10: 			 T5 = LD(&(x[WS(rs, 4)]), ms, &(x[0]));
Chris@10: 			 T7 = LD(&(x[WS(rs, 12)]), ms, &(x[0]));
Chris@10: 			 {
Chris@10: 			      V Tz, TG, TB, TE;
Chris@10: 			      Tz = LD(&(x[WS(rs, 2)]), ms, &(x[0]));
Chris@10: 			      TG = LD(&(x[WS(rs, 6)]), ms, &(x[0]));
Chris@10: 			      TB = LD(&(x[WS(rs, 10)]), ms, &(x[0]));
Chris@10: 			      TE = LD(&(x[WS(rs, 14)]), ms, &(x[0]));
Chris@10: 			      {
Chris@10: 				   V Ti, TX, TY, Td, Tg, Tm, Tt, To;
Chris@10: 				   {
Chris@10: 					V T3, T6, T8, TA, TH, TC, TF, Tb;
Chris@10: 					Tb = LD(&(x[WS(rs, 1)]), ms, &(x[WS(rs, 1)]));
Chris@10: 					T3 = BYTW(&(W[TWVL * 14]), T2);
Chris@10: 					T6 = BYTW(&(W[TWVL * 6]), T5);
Chris@10: 					T8 = BYTW(&(W[TWVL * 22]), T7);
Chris@10: 					TA = BYTW(&(W[TWVL * 2]), Tz);
Chris@10: 					TH = BYTW(&(W[TWVL * 10]), TG);
Chris@10: 					TC = BYTW(&(W[TWVL * 18]), TB);
Chris@10: 					TF = BYTW(&(W[TWVL * 26]), TE);
Chris@10: 					Tc = BYTW(&(W[0]), Tb);
Chris@10: 					TW = VSUB(T1, T3);
Chris@10: 					T4 = VADD(T1, T3);
Chris@10: 					T19 = VSUB(T6, T8);
Chris@10: 					T9 = VADD(T6, T8);
Chris@10: 					Ti = LD(&(x[WS(rs, 13)]), ms, &(x[WS(rs, 1)]));
Chris@10: 					TD = VADD(TA, TC);
Chris@10: 					TX = VSUB(TA, TC);
Chris@10: 					TI = VADD(TF, TH);
Chris@10: 					TY = VSUB(TF, TH);
Chris@10: 				   }
Chris@10: 				   Td = LD(&(x[WS(rs, 9)]), ms, &(x[WS(rs, 1)]));
Chris@10: 				   Tg = LD(&(x[WS(rs, 5)]), ms, &(x[WS(rs, 1)]));
Chris@10: 				   Tm = LD(&(x[WS(rs, 15)]), ms, &(x[WS(rs, 1)]));
Chris@10: 				   Tj = BYTW(&(W[TWVL * 24]), Ti);
Chris@10: 				   Tt = LD(&(x[WS(rs, 11)]), ms, &(x[WS(rs, 1)]));
Chris@10: 				   To = LD(&(x[WS(rs, 7)]), ms, &(x[WS(rs, 1)]));
Chris@10: 				   TZ = VADD(TX, TY);
Chris@10: 				   T1a = VSUB(TX, TY);
Chris@10: 				   Te = BYTW(&(W[TWVL * 16]), Td);
Chris@10: 				   Th = BYTW(&(W[TWVL * 8]), Tg);
Chris@10: 				   Tn = BYTW(&(W[TWVL * 28]), Tm);
Chris@10: 				   Tr = LD(&(x[WS(rs, 3)]), ms, &(x[WS(rs, 1)]));
Chris@10: 				   Tu = BYTW(&(W[TWVL * 20]), Tt);
Chris@10: 				   Tp = BYTW(&(W[TWVL * 12]), To);
Chris@10: 			      }
Chris@10: 			 }
Chris@10: 		    }
Chris@10: 		    {
Chris@10: 			 V Tf, T11, Tk, T12, Ts;
Chris@10: 			 TO = VADD(T4, T9);
Chris@10: 			 Ta = VSUB(T4, T9);
Chris@10: 			 TJ = VSUB(TD, TI);
Chris@10: 			 TP = VADD(TD, TI);
Chris@10: 			 Tf = VADD(Tc, Te);
Chris@10: 			 T11 = VSUB(Tc, Te);
Chris@10: 			 Tk = VADD(Th, Tj);
Chris@10: 			 T12 = VSUB(Th, Tj);
Chris@10: 			 Ts = BYTW(&(W[TWVL * 4]), Tr);
Chris@10: 			 T14 = VSUB(Tn, Tp);
Chris@10: 			 Tq = VADD(Tn, Tp);
Chris@10: 			 T1i = VFNMS(LDK(KP707106781), TZ, TW);
Chris@10: 			 T10 = VFMA(LDK(KP707106781), TZ, TW);
Chris@10: 			 T1b = VFMA(LDK(KP707106781), T1a, T19);
Chris@10: 			 T1l = VFNMS(LDK(KP707106781), T1a, T19);
Chris@10: 			 T13 = VFNMS(LDK(KP414213562), T12, T11);
Chris@10: 			 T1c = VFMA(LDK(KP414213562), T11, T12);
Chris@10: 			 TR = VADD(Tf, Tk);
Chris@10: 			 Tl = VSUB(Tf, Tk);
Chris@10: 			 T15 = VSUB(Tu, Ts);
Chris@10: 			 Tv = VADD(Ts, Tu);
Chris@10: 		    }
Chris@10: 	       }
Chris@10: 	       {
Chris@10: 		    V T1d, T16, TS, Tw, TU, TQ;
Chris@10: 		    T1d = VFMA(LDK(KP414213562), T14, T15);
Chris@10: 		    T16 = VFNMS(LDK(KP414213562), T15, T14);
Chris@10: 		    TS = VADD(Tq, Tv);
Chris@10: 		    Tw = VSUB(Tq, Tv);
Chris@10: 		    TU = VADD(TO, TP);
Chris@10: 		    TQ = VSUB(TO, TP);
Chris@10: 		    {
Chris@10: 			 V T1e, T1j, T17, T1m;
Chris@10: 			 T1e = VSUB(T1c, T1d);
Chris@10: 			 T1j = VADD(T1c, T1d);
Chris@10: 			 T17 = VADD(T13, T16);
Chris@10: 			 T1m = VSUB(T13, T16);
Chris@10: 			 {
Chris@10: 			      V TV, TT, TK, Tx;
Chris@10: 			      TV = VADD(TR, TS);
Chris@10: 			      TT = VSUB(TR, TS);
Chris@10: 			      TK = VSUB(Tl, Tw);
Chris@10: 			      Tx = VADD(Tl, Tw);
Chris@10: 			      {
Chris@10: 				   V T1h, T1f, T1o, T1k;
Chris@10: 				   T1h = VFMA(LDK(KP923879532), T1e, T1b);
Chris@10: 				   T1f = VFNMS(LDK(KP923879532), T1e, T1b);
Chris@10: 				   T1o = VFMA(LDK(KP923879532), T1j, T1i);
Chris@10: 				   T1k = VFNMS(LDK(KP923879532), T1j, T1i);
Chris@10: 				   {
Chris@10: 					V T1g, T18, T1p, T1n;
Chris@10: 					T1g = VFMA(LDK(KP923879532), T17, T10);
Chris@10: 					T18 = VFNMS(LDK(KP923879532), T17, T10);
Chris@10: 					T1p = VFNMS(LDK(KP923879532), T1m, T1l);
Chris@10: 					T1n = VFMA(LDK(KP923879532), T1m, T1l);
Chris@10: 					ST(&(x[WS(rs, 8)]), VSUB(TU, TV), ms, &(x[0]));
Chris@10: 					ST(&(x[0]), VADD(TU, TV), ms, &(x[0]));
Chris@10: 					ST(&(x[WS(rs, 4)]), VFMAI(TT, TQ), ms, &(x[0]));
Chris@10: 					ST(&(x[WS(rs, 12)]), VFNMSI(TT, TQ), ms, &(x[0]));
Chris@10: 					{
Chris@10: 					     V TN, TL, TM, Ty;
Chris@10: 					     TN = VFMA(LDK(KP707106781), TK, TJ);
Chris@10: 					     TL = VFNMS(LDK(KP707106781), TK, TJ);
Chris@10: 					     TM = VFMA(LDK(KP707106781), Tx, Ta);
Chris@10: 					     Ty = VFNMS(LDK(KP707106781), Tx, Ta);
Chris@10: 					     ST(&(x[WS(rs, 15)]), VFNMSI(T1h, T1g), ms, &(x[WS(rs, 1)]));
Chris@10: 					     ST(&(x[WS(rs, 1)]), VFMAI(T1h, T1g), ms, &(x[WS(rs, 1)]));
Chris@10: 					     ST(&(x[WS(rs, 9)]), VFMAI(T1f, T18), ms, &(x[WS(rs, 1)]));
Chris@10: 					     ST(&(x[WS(rs, 7)]), VFNMSI(T1f, T18), ms, &(x[WS(rs, 1)]));
Chris@10: 					     ST(&(x[WS(rs, 3)]), VFNMSI(T1p, T1o), ms, &(x[WS(rs, 1)]));
Chris@10: 					     ST(&(x[WS(rs, 13)]), VFMAI(T1p, T1o), ms, &(x[WS(rs, 1)]));
Chris@10: 					     ST(&(x[WS(rs, 11)]), VFNMSI(T1n, T1k), ms, &(x[WS(rs, 1)]));
Chris@10: 					     ST(&(x[WS(rs, 5)]), VFMAI(T1n, T1k), ms, &(x[WS(rs, 1)]));
Chris@10: 					     ST(&(x[WS(rs, 2)]), VFMAI(TN, TM), ms, &(x[0]));
Chris@10: 					     ST(&(x[WS(rs, 14)]), VFNMSI(TN, TM), ms, &(x[0]));
Chris@10: 					     ST(&(x[WS(rs, 10)]), VFMAI(TL, Ty), ms, &(x[0]));
Chris@10: 					     ST(&(x[WS(rs, 6)]), VFNMSI(TL, Ty), ms, &(x[0]));
Chris@10: 					}
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, 2),
Chris@10:      VTW(0, 3),
Chris@10:      VTW(0, 4),
Chris@10:      VTW(0, 5),
Chris@10:      VTW(0, 6),
Chris@10:      VTW(0, 7),
Chris@10:      VTW(0, 8),
Chris@10:      VTW(0, 9),
Chris@10:      VTW(0, 10),
Chris@10:      VTW(0, 11),
Chris@10:      VTW(0, 12),
Chris@10:      VTW(0, 13),
Chris@10:      VTW(0, 14),
Chris@10:      VTW(0, 15),
Chris@10:      {TW_NEXT, VL, 0}
Chris@10: };
Chris@10: 
Chris@10: static const ct_desc desc = { 16, XSIMD_STRING("t1bv_16"), twinstr, &GENUS, {53, 30, 34, 0}, 0, 0, 0 };
Chris@10: 
Chris@10: void XSIMD(codelet_t1bv_16) (planner *p) {
Chris@10:      X(kdft_dit_register) (p, t1bv_16, &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 16 -name t1bv_16 -include t1b.h -sign 1 */
Chris@10: 
Chris@10: /*
Chris@10:  * This function contains 87 FP additions, 42 FP multiplications,
Chris@10:  * (or, 83 additions, 38 multiplications, 4 fused multiply/add),
Chris@10:  * 36 stack variables, 3 constants, and 32 memory accesses
Chris@10:  */
Chris@10: #include "t1b.h"
Chris@10: 
Chris@10: static void t1bv_16(R *ri, R *ii, const R *W, stride rs, INT mb, INT me, INT ms)
Chris@10: {
Chris@10:      DVK(KP382683432, +0.382683432365089771728459984030398866761344562);
Chris@10:      DVK(KP923879532, +0.923879532511286756128183189396788286822416626);
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) * 30)); m < me; m = m + VL, x = x + (VL * ms), W = W + (TWVL * 30), MAKE_VOLATILE_STRIDE(16, rs)) {
Chris@10: 	       V TJ, T1b, TD, T1c, T17, T18, Ty, TK, T10, T11, T12, Tb, TM, T13, T14;
Chris@10: 	       V T15, Tm, TN, TG, TI, TH;
Chris@10: 	       TG = LD(&(x[0]), ms, &(x[0]));
Chris@10: 	       TH = LD(&(x[WS(rs, 8)]), ms, &(x[0]));
Chris@10: 	       TI = BYTW(&(W[TWVL * 14]), TH);
Chris@10: 	       TJ = VSUB(TG, TI);
Chris@10: 	       T1b = VADD(TG, TI);
Chris@10: 	       {
Chris@10: 		    V TA, TC, Tz, TB;
Chris@10: 		    Tz = LD(&(x[WS(rs, 4)]), ms, &(x[0]));
Chris@10: 		    TA = BYTW(&(W[TWVL * 6]), Tz);
Chris@10: 		    TB = LD(&(x[WS(rs, 12)]), ms, &(x[0]));
Chris@10: 		    TC = BYTW(&(W[TWVL * 22]), TB);
Chris@10: 		    TD = VSUB(TA, TC);
Chris@10: 		    T1c = VADD(TA, TC);
Chris@10: 	       }
Chris@10: 	       {
Chris@10: 		    V Tp, Tw, Tr, Tu, Ts, Tx;
Chris@10: 		    {
Chris@10: 			 V To, Tv, Tq, Tt;
Chris@10: 			 To = LD(&(x[WS(rs, 2)]), ms, &(x[0]));
Chris@10: 			 Tp = BYTW(&(W[TWVL * 2]), To);
Chris@10: 			 Tv = LD(&(x[WS(rs, 6)]), ms, &(x[0]));
Chris@10: 			 Tw = BYTW(&(W[TWVL * 10]), Tv);
Chris@10: 			 Tq = LD(&(x[WS(rs, 10)]), ms, &(x[0]));
Chris@10: 			 Tr = BYTW(&(W[TWVL * 18]), Tq);
Chris@10: 			 Tt = LD(&(x[WS(rs, 14)]), ms, &(x[0]));
Chris@10: 			 Tu = BYTW(&(W[TWVL * 26]), Tt);
Chris@10: 		    }
Chris@10: 		    T17 = VADD(Tp, Tr);
Chris@10: 		    T18 = VADD(Tu, Tw);
Chris@10: 		    Ts = VSUB(Tp, Tr);
Chris@10: 		    Tx = VSUB(Tu, Tw);
Chris@10: 		    Ty = VMUL(LDK(KP707106781), VSUB(Ts, Tx));
Chris@10: 		    TK = VMUL(LDK(KP707106781), VADD(Ts, Tx));
Chris@10: 	       }
Chris@10: 	       {
Chris@10: 		    V T2, T9, T4, T7, T5, Ta;
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 = BYTW(&(W[0]), T1);
Chris@10: 			 T8 = LD(&(x[WS(rs, 13)]), ms, &(x[WS(rs, 1)]));
Chris@10: 			 T9 = BYTW(&(W[TWVL * 24]), T8);
Chris@10: 			 T3 = LD(&(x[WS(rs, 9)]), ms, &(x[WS(rs, 1)]));
Chris@10: 			 T4 = BYTW(&(W[TWVL * 16]), T3);
Chris@10: 			 T6 = LD(&(x[WS(rs, 5)]), ms, &(x[WS(rs, 1)]));
Chris@10: 			 T7 = BYTW(&(W[TWVL * 8]), T6);
Chris@10: 		    }
Chris@10: 		    T10 = VADD(T2, T4);
Chris@10: 		    T11 = VADD(T7, T9);
Chris@10: 		    T12 = VSUB(T10, T11);
Chris@10: 		    T5 = VSUB(T2, T4);
Chris@10: 		    Ta = VSUB(T7, T9);
Chris@10: 		    Tb = VFNMS(LDK(KP382683432), Ta, VMUL(LDK(KP923879532), T5));
Chris@10: 		    TM = VFMA(LDK(KP382683432), T5, VMUL(LDK(KP923879532), Ta));
Chris@10: 	       }
Chris@10: 	       {
Chris@10: 		    V Td, Tk, Tf, Ti, Tg, Tl;
Chris@10: 		    {
Chris@10: 			 V Tc, Tj, Te, Th;
Chris@10: 			 Tc = LD(&(x[WS(rs, 15)]), ms, &(x[WS(rs, 1)]));
Chris@10: 			 Td = BYTW(&(W[TWVL * 28]), Tc);
Chris@10: 			 Tj = LD(&(x[WS(rs, 11)]), ms, &(x[WS(rs, 1)]));
Chris@10: 			 Tk = BYTW(&(W[TWVL * 20]), Tj);
Chris@10: 			 Te = LD(&(x[WS(rs, 7)]), ms, &(x[WS(rs, 1)]));
Chris@10: 			 Tf = BYTW(&(W[TWVL * 12]), Te);
Chris@10: 			 Th = LD(&(x[WS(rs, 3)]), ms, &(x[WS(rs, 1)]));
Chris@10: 			 Ti = BYTW(&(W[TWVL * 4]), Th);
Chris@10: 		    }
Chris@10: 		    T13 = VADD(Td, Tf);
Chris@10: 		    T14 = VADD(Ti, Tk);
Chris@10: 		    T15 = VSUB(T13, T14);
Chris@10: 		    Tg = VSUB(Td, Tf);
Chris@10: 		    Tl = VSUB(Ti, Tk);
Chris@10: 		    Tm = VFMA(LDK(KP923879532), Tg, VMUL(LDK(KP382683432), Tl));
Chris@10: 		    TN = VFNMS(LDK(KP382683432), Tg, VMUL(LDK(KP923879532), Tl));
Chris@10: 	       }
Chris@10: 	       {
Chris@10: 		    V T1a, T1g, T1f, T1h;
Chris@10: 		    {
Chris@10: 			 V T16, T19, T1d, T1e;
Chris@10: 			 T16 = VMUL(LDK(KP707106781), VSUB(T12, T15));
Chris@10: 			 T19 = VSUB(T17, T18);
Chris@10: 			 T1a = VBYI(VSUB(T16, T19));
Chris@10: 			 T1g = VBYI(VADD(T19, T16));
Chris@10: 			 T1d = VSUB(T1b, T1c);
Chris@10: 			 T1e = VMUL(LDK(KP707106781), VADD(T12, T15));
Chris@10: 			 T1f = VSUB(T1d, T1e);
Chris@10: 			 T1h = VADD(T1d, T1e);
Chris@10: 		    }
Chris@10: 		    ST(&(x[WS(rs, 6)]), VADD(T1a, T1f), ms, &(x[0]));
Chris@10: 		    ST(&(x[WS(rs, 14)]), VSUB(T1h, T1g), ms, &(x[0]));
Chris@10: 		    ST(&(x[WS(rs, 10)]), VSUB(T1f, T1a), ms, &(x[0]));
Chris@10: 		    ST(&(x[WS(rs, 2)]), VADD(T1g, T1h), ms, &(x[0]));
Chris@10: 	       }
Chris@10: 	       {
Chris@10: 		    V T1k, T1o, T1n, T1p;
Chris@10: 		    {
Chris@10: 			 V T1i, T1j, T1l, T1m;
Chris@10: 			 T1i = VADD(T1b, T1c);
Chris@10: 			 T1j = VADD(T17, T18);
Chris@10: 			 T1k = VSUB(T1i, T1j);
Chris@10: 			 T1o = VADD(T1i, T1j);
Chris@10: 			 T1l = VADD(T10, T11);
Chris@10: 			 T1m = VADD(T13, T14);
Chris@10: 			 T1n = VBYI(VSUB(T1l, T1m));
Chris@10: 			 T1p = VADD(T1l, T1m);
Chris@10: 		    }
Chris@10: 		    ST(&(x[WS(rs, 12)]), VSUB(T1k, T1n), ms, &(x[0]));
Chris@10: 		    ST(&(x[0]), VADD(T1o, T1p), ms, &(x[0]));
Chris@10: 		    ST(&(x[WS(rs, 4)]), VADD(T1k, T1n), ms, &(x[0]));
Chris@10: 		    ST(&(x[WS(rs, 8)]), VSUB(T1o, T1p), ms, &(x[0]));
Chris@10: 	       }
Chris@10: 	       {
Chris@10: 		    V TF, TQ, TP, TR;
Chris@10: 		    {
Chris@10: 			 V Tn, TE, TL, TO;
Chris@10: 			 Tn = VSUB(Tb, Tm);
Chris@10: 			 TE = VSUB(Ty, TD);
Chris@10: 			 TF = VBYI(VSUB(Tn, TE));
Chris@10: 			 TQ = VBYI(VADD(TE, Tn));
Chris@10: 			 TL = VSUB(TJ, TK);
Chris@10: 			 TO = VSUB(TM, TN);
Chris@10: 			 TP = VSUB(TL, TO);
Chris@10: 			 TR = VADD(TL, TO);
Chris@10: 		    }
Chris@10: 		    ST(&(x[WS(rs, 5)]), VADD(TF, TP), ms, &(x[WS(rs, 1)]));
Chris@10: 		    ST(&(x[WS(rs, 13)]), VSUB(TR, TQ), ms, &(x[WS(rs, 1)]));
Chris@10: 		    ST(&(x[WS(rs, 11)]), VSUB(TP, TF), ms, &(x[WS(rs, 1)]));
Chris@10: 		    ST(&(x[WS(rs, 3)]), VADD(TQ, TR), ms, &(x[WS(rs, 1)]));
Chris@10: 	       }
Chris@10: 	       {
Chris@10: 		    V TU, TY, TX, TZ;
Chris@10: 		    {
Chris@10: 			 V TS, TT, TV, TW;
Chris@10: 			 TS = VADD(TJ, TK);
Chris@10: 			 TT = VADD(Tb, Tm);
Chris@10: 			 TU = VADD(TS, TT);
Chris@10: 			 TY = VSUB(TS, TT);
Chris@10: 			 TV = VADD(TD, Ty);
Chris@10: 			 TW = VADD(TM, TN);
Chris@10: 			 TX = VBYI(VADD(TV, TW));
Chris@10: 			 TZ = VBYI(VSUB(TW, TV));
Chris@10: 		    }
Chris@10: 		    ST(&(x[WS(rs, 15)]), VSUB(TU, TX), ms, &(x[WS(rs, 1)]));
Chris@10: 		    ST(&(x[WS(rs, 7)]), VADD(TY, TZ), ms, &(x[WS(rs, 1)]));
Chris@10: 		    ST(&(x[WS(rs, 1)]), VADD(TU, TX), ms, &(x[WS(rs, 1)]));
Chris@10: 		    ST(&(x[WS(rs, 9)]), VSUB(TY, TZ), ms, &(x[WS(rs, 1)]));
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:      VTW(0, 5),
Chris@10:      VTW(0, 6),
Chris@10:      VTW(0, 7),
Chris@10:      VTW(0, 8),
Chris@10:      VTW(0, 9),
Chris@10:      VTW(0, 10),
Chris@10:      VTW(0, 11),
Chris@10:      VTW(0, 12),
Chris@10:      VTW(0, 13),
Chris@10:      VTW(0, 14),
Chris@10:      VTW(0, 15),
Chris@10:      {TW_NEXT, VL, 0}
Chris@10: };
Chris@10: 
Chris@10: static const ct_desc desc = { 16, XSIMD_STRING("t1bv_16"), twinstr, &GENUS, {83, 38, 4, 0}, 0, 0, 0 };
Chris@10: 
Chris@10: void XSIMD(codelet_t1bv_16) (planner *p) {
Chris@10:      X(kdft_dit_register) (p, t1bv_16, &desc);
Chris@10: }
Chris@10: #endif				/* HAVE_FMA */