cannam@95: /*
cannam@95:  * Copyright (c) 2003, 2007-11 Matteo Frigo
cannam@95:  * Copyright (c) 2003, 2007-11 Massachusetts Institute of Technology
cannam@95:  *
cannam@95:  * This program is free software; you can redistribute it and/or modify
cannam@95:  * it under the terms of the GNU General Public License as published by
cannam@95:  * the Free Software Foundation; either version 2 of the License, or
cannam@95:  * (at your option) any later version.
cannam@95:  *
cannam@95:  * This program is distributed in the hope that it will be useful,
cannam@95:  * but WITHOUT ANY WARRANTY; without even the implied warranty of
cannam@95:  * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
cannam@95:  * GNU General Public License for more details.
cannam@95:  *
cannam@95:  * You should have received a copy of the GNU General Public License
cannam@95:  * along with this program; if not, write to the Free Software
cannam@95:  * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA  02110-1301  USA
cannam@95:  *
cannam@95:  */
cannam@95: 
cannam@95: /* This file was automatically generated --- DO NOT EDIT */
cannam@95: /* Generated on Sun Nov 25 07:38:35 EST 2012 */
cannam@95: 
cannam@95: #include "codelet-dft.h"
cannam@95: 
cannam@95: #ifdef HAVE_FMA
cannam@95: 
cannam@95: /* Generated by: ../../../genfft/gen_twiddle_c.native -fma -reorder-insns -schedule-for-pipeline -simd -compact -variables 4 -pipeline-latency 8 -n 16 -name t2fv_16 -include t2f.h */
cannam@95: 
cannam@95: /*
cannam@95:  * This function contains 87 FP additions, 64 FP multiplications,
cannam@95:  * (or, 53 additions, 30 multiplications, 34 fused multiply/add),
cannam@95:  * 61 stack variables, 3 constants, and 32 memory accesses
cannam@95:  */
cannam@95: #include "t2f.h"
cannam@95: 
cannam@95: static void t2fv_16(R *ri, R *ii, const R *W, stride rs, INT mb, INT me, INT ms)
cannam@95: {
cannam@95:      DVK(KP923879532, +0.923879532511286756128183189396788286822416626);
cannam@95:      DVK(KP414213562, +0.414213562373095048801688724209698078569671875);
cannam@95:      DVK(KP707106781, +0.707106781186547524400844362104849039284835938);
cannam@95:      {
cannam@95: 	  INT m;
cannam@95: 	  R *x;
cannam@95: 	  x = ri;
cannam@95: 	  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)) {
cannam@95: 	       V TO, Ta, TJ, TP, T14, Tq, T1i, T10, T1b, T1l, T13, T1c, TR, Tl, T15;
cannam@95: 	       V Tv;
cannam@95: 	       {
cannam@95: 		    V Tc, TW, T4, T19, T9, TD, TI, Tj, TZ, T1a, Te, Th, Tn, Tr, Tu;
cannam@95: 		    V Tp;
cannam@95: 		    {
cannam@95: 			 V T1, T2, T5, T7;
cannam@95: 			 T1 = LD(&(x[0]), ms, &(x[0]));
cannam@95: 			 T2 = LD(&(x[WS(rs, 8)]), ms, &(x[0]));
cannam@95: 			 T5 = LD(&(x[WS(rs, 4)]), ms, &(x[0]));
cannam@95: 			 T7 = LD(&(x[WS(rs, 12)]), ms, &(x[0]));
cannam@95: 			 {
cannam@95: 			      V Tz, TG, TB, TE;
cannam@95: 			      Tz = LD(&(x[WS(rs, 14)]), ms, &(x[0]));
cannam@95: 			      TG = LD(&(x[WS(rs, 10)]), ms, &(x[0]));
cannam@95: 			      TB = LD(&(x[WS(rs, 6)]), ms, &(x[0]));
cannam@95: 			      TE = LD(&(x[WS(rs, 2)]), ms, &(x[0]));
cannam@95: 			      {
cannam@95: 				   V Ti, TY, TX, Td, Tg, Tm, Tt, To;
cannam@95: 				   {
cannam@95: 					V T3, T6, T8, TA, TH, TC, TF, Tb;
cannam@95: 					Tb = LD(&(x[WS(rs, 1)]), ms, &(x[WS(rs, 1)]));
cannam@95: 					T3 = BYTWJ(&(W[TWVL * 14]), T2);
cannam@95: 					T6 = BYTWJ(&(W[TWVL * 6]), T5);
cannam@95: 					T8 = BYTWJ(&(W[TWVL * 22]), T7);
cannam@95: 					TA = BYTWJ(&(W[TWVL * 26]), Tz);
cannam@95: 					TH = BYTWJ(&(W[TWVL * 18]), TG);
cannam@95: 					TC = BYTWJ(&(W[TWVL * 10]), TB);
cannam@95: 					TF = BYTWJ(&(W[TWVL * 2]), TE);
cannam@95: 					Tc = BYTWJ(&(W[0]), Tb);
cannam@95: 					TW = VSUB(T1, T3);
cannam@95: 					T4 = VADD(T1, T3);
cannam@95: 					T19 = VSUB(T6, T8);
cannam@95: 					T9 = VADD(T6, T8);
cannam@95: 					Ti = LD(&(x[WS(rs, 13)]), ms, &(x[WS(rs, 1)]));
cannam@95: 					TD = VADD(TA, TC);
cannam@95: 					TY = VSUB(TA, TC);
cannam@95: 					TI = VADD(TF, TH);
cannam@95: 					TX = VSUB(TF, TH);
cannam@95: 				   }
cannam@95: 				   Td = LD(&(x[WS(rs, 9)]), ms, &(x[WS(rs, 1)]));
cannam@95: 				   Tg = LD(&(x[WS(rs, 5)]), ms, &(x[WS(rs, 1)]));
cannam@95: 				   Tm = LD(&(x[WS(rs, 15)]), ms, &(x[WS(rs, 1)]));
cannam@95: 				   Tj = BYTWJ(&(W[TWVL * 24]), Ti);
cannam@95: 				   Tt = LD(&(x[WS(rs, 11)]), ms, &(x[WS(rs, 1)]));
cannam@95: 				   To = LD(&(x[WS(rs, 7)]), ms, &(x[WS(rs, 1)]));
cannam@95: 				   TZ = VADD(TX, TY);
cannam@95: 				   T1a = VSUB(TY, TX);
cannam@95: 				   Te = BYTWJ(&(W[TWVL * 16]), Td);
cannam@95: 				   Th = BYTWJ(&(W[TWVL * 8]), Tg);
cannam@95: 				   Tn = BYTWJ(&(W[TWVL * 28]), Tm);
cannam@95: 				   Tr = LD(&(x[WS(rs, 3)]), ms, &(x[WS(rs, 1)]));
cannam@95: 				   Tu = BYTWJ(&(W[TWVL * 20]), Tt);
cannam@95: 				   Tp = BYTWJ(&(W[TWVL * 12]), To);
cannam@95: 			      }
cannam@95: 			 }
cannam@95: 		    }
cannam@95: 		    {
cannam@95: 			 V Tf, T11, Tk, T12, Ts;
cannam@95: 			 TO = VADD(T4, T9);
cannam@95: 			 Ta = VSUB(T4, T9);
cannam@95: 			 TJ = VSUB(TD, TI);
cannam@95: 			 TP = VADD(TI, TD);
cannam@95: 			 Tf = VADD(Tc, Te);
cannam@95: 			 T11 = VSUB(Tc, Te);
cannam@95: 			 Tk = VADD(Th, Tj);
cannam@95: 			 T12 = VSUB(Th, Tj);
cannam@95: 			 Ts = BYTWJ(&(W[TWVL * 4]), Tr);
cannam@95: 			 T14 = VSUB(Tn, Tp);
cannam@95: 			 Tq = VADD(Tn, Tp);
cannam@95: 			 T1i = VFNMS(LDK(KP707106781), TZ, TW);
cannam@95: 			 T10 = VFMA(LDK(KP707106781), TZ, TW);
cannam@95: 			 T1b = VFNMS(LDK(KP707106781), T1a, T19);
cannam@95: 			 T1l = VFMA(LDK(KP707106781), T1a, T19);
cannam@95: 			 T13 = VFNMS(LDK(KP414213562), T12, T11);
cannam@95: 			 T1c = VFMA(LDK(KP414213562), T11, T12);
cannam@95: 			 TR = VADD(Tf, Tk);
cannam@95: 			 Tl = VSUB(Tf, Tk);
cannam@95: 			 T15 = VSUB(Tu, Ts);
cannam@95: 			 Tv = VADD(Ts, Tu);
cannam@95: 		    }
cannam@95: 	       }
cannam@95: 	       {
cannam@95: 		    V T1d, T16, TS, Tw, TU, TQ;
cannam@95: 		    T1d = VFMA(LDK(KP414213562), T14, T15);
cannam@95: 		    T16 = VFNMS(LDK(KP414213562), T15, T14);
cannam@95: 		    TS = VADD(Tq, Tv);
cannam@95: 		    Tw = VSUB(Tq, Tv);
cannam@95: 		    TU = VSUB(TO, TP);
cannam@95: 		    TQ = VADD(TO, TP);
cannam@95: 		    {
cannam@95: 			 V T1e, T1j, T17, T1m;
cannam@95: 			 T1e = VSUB(T1c, T1d);
cannam@95: 			 T1j = VADD(T1c, T1d);
cannam@95: 			 T17 = VADD(T13, T16);
cannam@95: 			 T1m = VSUB(T16, T13);
cannam@95: 			 {
cannam@95: 			      V TV, TT, TK, Tx;
cannam@95: 			      TV = VSUB(TS, TR);
cannam@95: 			      TT = VADD(TR, TS);
cannam@95: 			      TK = VSUB(Tw, Tl);
cannam@95: 			      Tx = VADD(Tl, Tw);
cannam@95: 			      {
cannam@95: 				   V T1h, T1f, T1o, T1k;
cannam@95: 				   T1h = VFMA(LDK(KP923879532), T1e, T1b);
cannam@95: 				   T1f = VFNMS(LDK(KP923879532), T1e, T1b);
cannam@95: 				   T1o = VFMA(LDK(KP923879532), T1j, T1i);
cannam@95: 				   T1k = VFNMS(LDK(KP923879532), T1j, T1i);
cannam@95: 				   {
cannam@95: 					V T1g, T18, T1p, T1n;
cannam@95: 					T1g = VFMA(LDK(KP923879532), T17, T10);
cannam@95: 					T18 = VFNMS(LDK(KP923879532), T17, T10);
cannam@95: 					T1p = VFMA(LDK(KP923879532), T1m, T1l);
cannam@95: 					T1n = VFNMS(LDK(KP923879532), T1m, T1l);
cannam@95: 					ST(&(x[WS(rs, 12)]), VFNMSI(TV, TU), ms, &(x[0]));
cannam@95: 					ST(&(x[WS(rs, 4)]), VFMAI(TV, TU), ms, &(x[0]));
cannam@95: 					ST(&(x[0]), VADD(TQ, TT), ms, &(x[0]));
cannam@95: 					ST(&(x[WS(rs, 8)]), VSUB(TQ, TT), ms, &(x[0]));
cannam@95: 					{
cannam@95: 					     V TN, TL, TM, Ty;
cannam@95: 					     TN = VFMA(LDK(KP707106781), TK, TJ);
cannam@95: 					     TL = VFNMS(LDK(KP707106781), TK, TJ);
cannam@95: 					     TM = VFMA(LDK(KP707106781), Tx, Ta);
cannam@95: 					     Ty = VFNMS(LDK(KP707106781), Tx, Ta);
cannam@95: 					     ST(&(x[WS(rs, 1)]), VFNMSI(T1h, T1g), ms, &(x[WS(rs, 1)]));
cannam@95: 					     ST(&(x[WS(rs, 15)]), VFMAI(T1h, T1g), ms, &(x[WS(rs, 1)]));
cannam@95: 					     ST(&(x[WS(rs, 7)]), VFMAI(T1f, T18), ms, &(x[WS(rs, 1)]));
cannam@95: 					     ST(&(x[WS(rs, 9)]), VFNMSI(T1f, T18), ms, &(x[WS(rs, 1)]));
cannam@95: 					     ST(&(x[WS(rs, 3)]), VFMAI(T1p, T1o), ms, &(x[WS(rs, 1)]));
cannam@95: 					     ST(&(x[WS(rs, 13)]), VFNMSI(T1p, T1o), ms, &(x[WS(rs, 1)]));
cannam@95: 					     ST(&(x[WS(rs, 11)]), VFMAI(T1n, T1k), ms, &(x[WS(rs, 1)]));
cannam@95: 					     ST(&(x[WS(rs, 5)]), VFNMSI(T1n, T1k), ms, &(x[WS(rs, 1)]));
cannam@95: 					     ST(&(x[WS(rs, 14)]), VFNMSI(TN, TM), ms, &(x[0]));
cannam@95: 					     ST(&(x[WS(rs, 2)]), VFMAI(TN, TM), ms, &(x[0]));
cannam@95: 					     ST(&(x[WS(rs, 10)]), VFMAI(TL, Ty), ms, &(x[0]));
cannam@95: 					     ST(&(x[WS(rs, 6)]), VFNMSI(TL, Ty), ms, &(x[0]));
cannam@95: 					}
cannam@95: 				   }
cannam@95: 			      }
cannam@95: 			 }
cannam@95: 		    }
cannam@95: 	       }
cannam@95: 	  }
cannam@95:      }
cannam@95:      VLEAVE();
cannam@95: }
cannam@95: 
cannam@95: static const tw_instr twinstr[] = {
cannam@95:      VTW(0, 1),
cannam@95:      VTW(0, 2),
cannam@95:      VTW(0, 3),
cannam@95:      VTW(0, 4),
cannam@95:      VTW(0, 5),
cannam@95:      VTW(0, 6),
cannam@95:      VTW(0, 7),
cannam@95:      VTW(0, 8),
cannam@95:      VTW(0, 9),
cannam@95:      VTW(0, 10),
cannam@95:      VTW(0, 11),
cannam@95:      VTW(0, 12),
cannam@95:      VTW(0, 13),
cannam@95:      VTW(0, 14),
cannam@95:      VTW(0, 15),
cannam@95:      {TW_NEXT, VL, 0}
cannam@95: };
cannam@95: 
cannam@95: static const ct_desc desc = { 16, XSIMD_STRING("t2fv_16"), twinstr, &GENUS, {53, 30, 34, 0}, 0, 0, 0 };
cannam@95: 
cannam@95: void XSIMD(codelet_t2fv_16) (planner *p) {
cannam@95:      X(kdft_dit_register) (p, t2fv_16, &desc);
cannam@95: }
cannam@95: #else				/* HAVE_FMA */
cannam@95: 
cannam@95: /* Generated by: ../../../genfft/gen_twiddle_c.native -simd -compact -variables 4 -pipeline-latency 8 -n 16 -name t2fv_16 -include t2f.h */
cannam@95: 
cannam@95: /*
cannam@95:  * This function contains 87 FP additions, 42 FP multiplications,
cannam@95:  * (or, 83 additions, 38 multiplications, 4 fused multiply/add),
cannam@95:  * 36 stack variables, 3 constants, and 32 memory accesses
cannam@95:  */
cannam@95: #include "t2f.h"
cannam@95: 
cannam@95: static void t2fv_16(R *ri, R *ii, const R *W, stride rs, INT mb, INT me, INT ms)
cannam@95: {
cannam@95:      DVK(KP923879532, +0.923879532511286756128183189396788286822416626);
cannam@95:      DVK(KP382683432, +0.382683432365089771728459984030398866761344562);
cannam@95:      DVK(KP707106781, +0.707106781186547524400844362104849039284835938);
cannam@95:      {
cannam@95: 	  INT m;
cannam@95: 	  R *x;
cannam@95: 	  x = ri;
cannam@95: 	  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)) {
cannam@95: 	       V TJ, T10, TD, T11, T1b, T1c, Ty, TK, T16, T17, T18, Tb, TN, T13, T14;
cannam@95: 	       V T15, Tm, TM, TG, TI, TH;
cannam@95: 	       TG = LD(&(x[0]), ms, &(x[0]));
cannam@95: 	       TH = LD(&(x[WS(rs, 8)]), ms, &(x[0]));
cannam@95: 	       TI = BYTWJ(&(W[TWVL * 14]), TH);
cannam@95: 	       TJ = VSUB(TG, TI);
cannam@95: 	       T10 = VADD(TG, TI);
cannam@95: 	       {
cannam@95: 		    V TA, TC, Tz, TB;
cannam@95: 		    Tz = LD(&(x[WS(rs, 4)]), ms, &(x[0]));
cannam@95: 		    TA = BYTWJ(&(W[TWVL * 6]), Tz);
cannam@95: 		    TB = LD(&(x[WS(rs, 12)]), ms, &(x[0]));
cannam@95: 		    TC = BYTWJ(&(W[TWVL * 22]), TB);
cannam@95: 		    TD = VSUB(TA, TC);
cannam@95: 		    T11 = VADD(TA, TC);
cannam@95: 	       }
cannam@95: 	       {
cannam@95: 		    V Tp, Tw, Tr, Tu, Ts, Tx;
cannam@95: 		    {
cannam@95: 			 V To, Tv, Tq, Tt;
cannam@95: 			 To = LD(&(x[WS(rs, 14)]), ms, &(x[0]));
cannam@95: 			 Tp = BYTWJ(&(W[TWVL * 26]), To);
cannam@95: 			 Tv = LD(&(x[WS(rs, 10)]), ms, &(x[0]));
cannam@95: 			 Tw = BYTWJ(&(W[TWVL * 18]), Tv);
cannam@95: 			 Tq = LD(&(x[WS(rs, 6)]), ms, &(x[0]));
cannam@95: 			 Tr = BYTWJ(&(W[TWVL * 10]), Tq);
cannam@95: 			 Tt = LD(&(x[WS(rs, 2)]), ms, &(x[0]));
cannam@95: 			 Tu = BYTWJ(&(W[TWVL * 2]), Tt);
cannam@95: 		    }
cannam@95: 		    T1b = VADD(Tp, Tr);
cannam@95: 		    T1c = VADD(Tu, Tw);
cannam@95: 		    Ts = VSUB(Tp, Tr);
cannam@95: 		    Tx = VSUB(Tu, Tw);
cannam@95: 		    Ty = VMUL(LDK(KP707106781), VSUB(Ts, Tx));
cannam@95: 		    TK = VMUL(LDK(KP707106781), VADD(Tx, Ts));
cannam@95: 	       }
cannam@95: 	       {
cannam@95: 		    V T2, T9, T4, T7, T5, Ta;
cannam@95: 		    {
cannam@95: 			 V T1, T8, T3, T6;
cannam@95: 			 T1 = LD(&(x[WS(rs, 15)]), ms, &(x[WS(rs, 1)]));
cannam@95: 			 T2 = BYTWJ(&(W[TWVL * 28]), T1);
cannam@95: 			 T8 = LD(&(x[WS(rs, 11)]), ms, &(x[WS(rs, 1)]));
cannam@95: 			 T9 = BYTWJ(&(W[TWVL * 20]), T8);
cannam@95: 			 T3 = LD(&(x[WS(rs, 7)]), ms, &(x[WS(rs, 1)]));
cannam@95: 			 T4 = BYTWJ(&(W[TWVL * 12]), T3);
cannam@95: 			 T6 = LD(&(x[WS(rs, 3)]), ms, &(x[WS(rs, 1)]));
cannam@95: 			 T7 = BYTWJ(&(W[TWVL * 4]), T6);
cannam@95: 		    }
cannam@95: 		    T16 = VADD(T2, T4);
cannam@95: 		    T17 = VADD(T7, T9);
cannam@95: 		    T18 = VSUB(T16, T17);
cannam@95: 		    T5 = VSUB(T2, T4);
cannam@95: 		    Ta = VSUB(T7, T9);
cannam@95: 		    Tb = VFNMS(LDK(KP923879532), Ta, VMUL(LDK(KP382683432), T5));
cannam@95: 		    TN = VFMA(LDK(KP923879532), T5, VMUL(LDK(KP382683432), Ta));
cannam@95: 	       }
cannam@95: 	       {
cannam@95: 		    V Td, Tk, Tf, Ti, Tg, Tl;
cannam@95: 		    {
cannam@95: 			 V Tc, Tj, Te, Th;
cannam@95: 			 Tc = LD(&(x[WS(rs, 1)]), ms, &(x[WS(rs, 1)]));
cannam@95: 			 Td = BYTWJ(&(W[0]), Tc);
cannam@95: 			 Tj = LD(&(x[WS(rs, 13)]), ms, &(x[WS(rs, 1)]));
cannam@95: 			 Tk = BYTWJ(&(W[TWVL * 24]), Tj);
cannam@95: 			 Te = LD(&(x[WS(rs, 9)]), ms, &(x[WS(rs, 1)]));
cannam@95: 			 Tf = BYTWJ(&(W[TWVL * 16]), Te);
cannam@95: 			 Th = LD(&(x[WS(rs, 5)]), ms, &(x[WS(rs, 1)]));
cannam@95: 			 Ti = BYTWJ(&(W[TWVL * 8]), Th);
cannam@95: 		    }
cannam@95: 		    T13 = VADD(Td, Tf);
cannam@95: 		    T14 = VADD(Ti, Tk);
cannam@95: 		    T15 = VSUB(T13, T14);
cannam@95: 		    Tg = VSUB(Td, Tf);
cannam@95: 		    Tl = VSUB(Ti, Tk);
cannam@95: 		    Tm = VFMA(LDK(KP382683432), Tg, VMUL(LDK(KP923879532), Tl));
cannam@95: 		    TM = VFNMS(LDK(KP382683432), Tl, VMUL(LDK(KP923879532), Tg));
cannam@95: 	       }
cannam@95: 	       {
cannam@95: 		    V T1a, T1g, T1f, T1h;
cannam@95: 		    {
cannam@95: 			 V T12, T19, T1d, T1e;
cannam@95: 			 T12 = VSUB(T10, T11);
cannam@95: 			 T19 = VMUL(LDK(KP707106781), VADD(T15, T18));
cannam@95: 			 T1a = VADD(T12, T19);
cannam@95: 			 T1g = VSUB(T12, T19);
cannam@95: 			 T1d = VSUB(T1b, T1c);
cannam@95: 			 T1e = VMUL(LDK(KP707106781), VSUB(T18, T15));
cannam@95: 			 T1f = VBYI(VADD(T1d, T1e));
cannam@95: 			 T1h = VBYI(VSUB(T1e, T1d));
cannam@95: 		    }
cannam@95: 		    ST(&(x[WS(rs, 14)]), VSUB(T1a, T1f), ms, &(x[0]));
cannam@95: 		    ST(&(x[WS(rs, 6)]), VADD(T1g, T1h), ms, &(x[0]));
cannam@95: 		    ST(&(x[WS(rs, 2)]), VADD(T1a, T1f), ms, &(x[0]));
cannam@95: 		    ST(&(x[WS(rs, 10)]), VSUB(T1g, T1h), ms, &(x[0]));
cannam@95: 	       }
cannam@95: 	       {
cannam@95: 		    V T1k, T1o, T1n, T1p;
cannam@95: 		    {
cannam@95: 			 V T1i, T1j, T1l, T1m;
cannam@95: 			 T1i = VADD(T10, T11);
cannam@95: 			 T1j = VADD(T1c, T1b);
cannam@95: 			 T1k = VADD(T1i, T1j);
cannam@95: 			 T1o = VSUB(T1i, T1j);
cannam@95: 			 T1l = VADD(T13, T14);
cannam@95: 			 T1m = VADD(T16, T17);
cannam@95: 			 T1n = VADD(T1l, T1m);
cannam@95: 			 T1p = VBYI(VSUB(T1m, T1l));
cannam@95: 		    }
cannam@95: 		    ST(&(x[WS(rs, 8)]), VSUB(T1k, T1n), ms, &(x[0]));
cannam@95: 		    ST(&(x[WS(rs, 4)]), VADD(T1o, T1p), ms, &(x[0]));
cannam@95: 		    ST(&(x[0]), VADD(T1k, T1n), ms, &(x[0]));
cannam@95: 		    ST(&(x[WS(rs, 12)]), VSUB(T1o, T1p), ms, &(x[0]));
cannam@95: 	       }
cannam@95: 	       {
cannam@95: 		    V TF, TQ, TP, TR;
cannam@95: 		    {
cannam@95: 			 V Tn, TE, TL, TO;
cannam@95: 			 Tn = VSUB(Tb, Tm);
cannam@95: 			 TE = VSUB(Ty, TD);
cannam@95: 			 TF = VBYI(VSUB(Tn, TE));
cannam@95: 			 TQ = VBYI(VADD(TE, Tn));
cannam@95: 			 TL = VADD(TJ, TK);
cannam@95: 			 TO = VADD(TM, TN);
cannam@95: 			 TP = VSUB(TL, TO);
cannam@95: 			 TR = VADD(TL, TO);
cannam@95: 		    }
cannam@95: 		    ST(&(x[WS(rs, 7)]), VADD(TF, TP), ms, &(x[WS(rs, 1)]));
cannam@95: 		    ST(&(x[WS(rs, 15)]), VSUB(TR, TQ), ms, &(x[WS(rs, 1)]));
cannam@95: 		    ST(&(x[WS(rs, 9)]), VSUB(TP, TF), ms, &(x[WS(rs, 1)]));
cannam@95: 		    ST(&(x[WS(rs, 1)]), VADD(TQ, TR), ms, &(x[WS(rs, 1)]));
cannam@95: 	       }
cannam@95: 	       {
cannam@95: 		    V TU, TY, TX, TZ;
cannam@95: 		    {
cannam@95: 			 V TS, TT, TV, TW;
cannam@95: 			 TS = VSUB(TJ, TK);
cannam@95: 			 TT = VADD(Tm, Tb);
cannam@95: 			 TU = VADD(TS, TT);
cannam@95: 			 TY = VSUB(TS, TT);
cannam@95: 			 TV = VADD(TD, Ty);
cannam@95: 			 TW = VSUB(TN, TM);
cannam@95: 			 TX = VBYI(VADD(TV, TW));
cannam@95: 			 TZ = VBYI(VSUB(TW, TV));
cannam@95: 		    }
cannam@95: 		    ST(&(x[WS(rs, 13)]), VSUB(TU, TX), ms, &(x[WS(rs, 1)]));
cannam@95: 		    ST(&(x[WS(rs, 5)]), VADD(TY, TZ), ms, &(x[WS(rs, 1)]));
cannam@95: 		    ST(&(x[WS(rs, 3)]), VADD(TU, TX), ms, &(x[WS(rs, 1)]));
cannam@95: 		    ST(&(x[WS(rs, 11)]), VSUB(TY, TZ), ms, &(x[WS(rs, 1)]));
cannam@95: 	       }
cannam@95: 	  }
cannam@95:      }
cannam@95:      VLEAVE();
cannam@95: }
cannam@95: 
cannam@95: static const tw_instr twinstr[] = {
cannam@95:      VTW(0, 1),
cannam@95:      VTW(0, 2),
cannam@95:      VTW(0, 3),
cannam@95:      VTW(0, 4),
cannam@95:      VTW(0, 5),
cannam@95:      VTW(0, 6),
cannam@95:      VTW(0, 7),
cannam@95:      VTW(0, 8),
cannam@95:      VTW(0, 9),
cannam@95:      VTW(0, 10),
cannam@95:      VTW(0, 11),
cannam@95:      VTW(0, 12),
cannam@95:      VTW(0, 13),
cannam@95:      VTW(0, 14),
cannam@95:      VTW(0, 15),
cannam@95:      {TW_NEXT, VL, 0}
cannam@95: };
cannam@95: 
cannam@95: static const ct_desc desc = { 16, XSIMD_STRING("t2fv_16"), twinstr, &GENUS, {83, 38, 4, 0}, 0, 0, 0 };
cannam@95: 
cannam@95: void XSIMD(codelet_t2fv_16) (planner *p) {
cannam@95:      X(kdft_dit_register) (p, t2fv_16, &desc);
cannam@95: }
cannam@95: #endif				/* HAVE_FMA */