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 07:39:26 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.native -fma -reorder-insns -schedule-for-pipeline -simd -compact -variables 4 -pipeline-latency 8 -twiddle-log3 -precompute-twiddles -n 8 -name t2sv_8 -include ts.h */
Chris@10: 
Chris@10: /*
Chris@10:  * This function contains 74 FP additions, 50 FP multiplications,
Chris@10:  * (or, 44 additions, 20 multiplications, 30 fused multiply/add),
Chris@10:  * 64 stack variables, 1 constants, and 32 memory accesses
Chris@10:  */
Chris@10: #include "ts.h"
Chris@10: 
Chris@10: static void t2sv_8(R *ri, R *ii, const R *W, stride rs, INT mb, INT me, INT ms)
Chris@10: {
Chris@10:      DVK(KP707106781, +0.707106781186547524400844362104849039284835938);
Chris@10:      {
Chris@10: 	  INT m;
Chris@10: 	  for (m = mb, W = W + (mb * 6); m < me; m = m + (2 * VL), ri = ri + ((2 * VL) * ms), ii = ii + ((2 * VL) * ms), W = W + ((2 * VL) * 6), MAKE_VOLATILE_STRIDE(16, rs)) {
Chris@10: 	       V T1m, T1l, T1k, T1u, T1n, T1o;
Chris@10: 	       {
Chris@10: 		    V T2, T3, Tl, Tn, T5, T6;
Chris@10: 		    T2 = LDW(&(W[0]));
Chris@10: 		    T3 = LDW(&(W[TWVL * 2]));
Chris@10: 		    Tl = LDW(&(W[TWVL * 4]));
Chris@10: 		    Tn = LDW(&(W[TWVL * 5]));
Chris@10: 		    T5 = LDW(&(W[TWVL * 1]));
Chris@10: 		    T6 = LDW(&(W[TWVL * 3]));
Chris@10: 		    {
Chris@10: 			 V T1, T1s, TK, T1r, Td, Tk, TG, TC, TY, Tu, TW, TL, TM, TO, TQ;
Chris@10: 			 V Tx, Tz, TD, TH;
Chris@10: 			 {
Chris@10: 			      V T8, T4, Tm, Tr, Tc, Ta;
Chris@10: 			      T1 = LD(&(ri[0]), ms, &(ri[0]));
Chris@10: 			      T1s = LD(&(ii[0]), ms, &(ii[0]));
Chris@10: 			      T8 = LD(&(ri[WS(rs, 4)]), ms, &(ri[0]));
Chris@10: 			      T4 = VMUL(T2, T3);
Chris@10: 			      Tm = VMUL(T2, Tl);
Chris@10: 			      Tr = VMUL(T2, Tn);
Chris@10: 			      Tc = LD(&(ii[WS(rs, 4)]), ms, &(ii[0]));
Chris@10: 			      Ta = VMUL(T2, T6);
Chris@10: 			      {
Chris@10: 				   V Tp, Tt, Tg, T7, Tf, To, Ts, Ti, Tb, Tj;
Chris@10: 				   Tp = LD(&(ri[WS(rs, 6)]), ms, &(ri[0]));
Chris@10: 				   Tt = LD(&(ii[WS(rs, 6)]), ms, &(ii[0]));
Chris@10: 				   Tg = LD(&(ri[WS(rs, 2)]), ms, &(ri[0]));
Chris@10: 				   T7 = VFNMS(T5, T6, T4);
Chris@10: 				   Tf = VFMA(T5, T6, T4);
Chris@10: 				   To = VFMA(T5, Tn, Tm);
Chris@10: 				   Ts = VFNMS(T5, Tl, Tr);
Chris@10: 				   Ti = VFNMS(T5, T3, Ta);
Chris@10: 				   Tb = VFMA(T5, T3, Ta);
Chris@10: 				   Tj = LD(&(ii[WS(rs, 2)]), ms, &(ii[0]));
Chris@10: 				   TK = LD(&(ri[WS(rs, 7)]), ms, &(ri[WS(rs, 1)]));
Chris@10: 				   {
Chris@10: 					V T1q, T9, Th, TF;
Chris@10: 					T1q = VMUL(T7, Tc);
Chris@10: 					T9 = VMUL(T7, T8);
Chris@10: 					Th = VMUL(Tf, Tg);
Chris@10: 					TF = VMUL(Tf, Tn);
Chris@10: 					{
Chris@10: 					     V TB, TX, Tq, TV;
Chris@10: 					     TB = VMUL(Tf, Tl);
Chris@10: 					     TX = VMUL(To, Tt);
Chris@10: 					     Tq = VMUL(To, Tp);
Chris@10: 					     TV = VMUL(Tf, Tj);
Chris@10: 					     T1r = VFNMS(Tb, T8, T1q);
Chris@10: 					     Td = VFMA(Tb, Tc, T9);
Chris@10: 					     Tk = VFMA(Ti, Tj, Th);
Chris@10: 					     TG = VFNMS(Ti, Tl, TF);
Chris@10: 					     TC = VFMA(Ti, Tn, TB);
Chris@10: 					     TY = VFNMS(Ts, Tp, TX);
Chris@10: 					     Tu = VFMA(Ts, Tt, Tq);
Chris@10: 					     TW = VFNMS(Ti, Tg, TV);
Chris@10: 					     TL = VMUL(Tl, TK);
Chris@10: 					}
Chris@10: 				   }
Chris@10: 				   TM = LD(&(ii[WS(rs, 7)]), ms, &(ii[WS(rs, 1)]));
Chris@10: 				   TO = LD(&(ri[WS(rs, 3)]), ms, &(ri[WS(rs, 1)]));
Chris@10: 				   TQ = LD(&(ii[WS(rs, 3)]), ms, &(ii[WS(rs, 1)]));
Chris@10: 				   Tx = LD(&(ri[WS(rs, 1)]), ms, &(ri[WS(rs, 1)]));
Chris@10: 				   Tz = LD(&(ii[WS(rs, 1)]), ms, &(ii[WS(rs, 1)]));
Chris@10: 				   TD = LD(&(ri[WS(rs, 5)]), ms, &(ri[WS(rs, 1)]));
Chris@10: 				   TH = LD(&(ii[WS(rs, 5)]), ms, &(ii[WS(rs, 1)]));
Chris@10: 			      }
Chris@10: 			 }
Chris@10: 			 {
Chris@10: 			      V Te, T1p, T1g, T10, TS, T18, T1d, T1t, T1x, T1y, Tv, TJ, T11, T16;
Chris@10: 			      {
Chris@10: 				   V TN, T1a, TR, T1c, TA, T13, TI, T15;
Chris@10: 				   {
Chris@10: 					V TU, T19, TP, T1b, Ty, T12, TE, T14, TZ;
Chris@10: 					TU = VSUB(T1, Td);
Chris@10: 					Te = VADD(T1, Td);
Chris@10: 					TN = VFMA(Tn, TM, TL);
Chris@10: 					T19 = VMUL(Tl, TM);
Chris@10: 					TP = VMUL(T3, TO);
Chris@10: 					T1b = VMUL(T3, TQ);
Chris@10: 					Ty = VMUL(T2, Tx);
Chris@10: 					T12 = VMUL(T2, Tz);
Chris@10: 					TE = VMUL(TC, TD);
Chris@10: 					T14 = VMUL(TC, TH);
Chris@10: 					T1p = VADD(TW, TY);
Chris@10: 					TZ = VSUB(TW, TY);
Chris@10: 					T1a = VFNMS(Tn, TK, T19);
Chris@10: 					TR = VFMA(T6, TQ, TP);
Chris@10: 					T1c = VFNMS(T6, TO, T1b);
Chris@10: 					TA = VFMA(T5, Tz, Ty);
Chris@10: 					T13 = VFNMS(T5, Tx, T12);
Chris@10: 					TI = VFMA(TG, TH, TE);
Chris@10: 					T15 = VFNMS(TG, TD, T14);
Chris@10: 					T1g = VSUB(TU, TZ);
Chris@10: 					T10 = VADD(TU, TZ);
Chris@10: 				   }
Chris@10: 				   TS = VADD(TN, TR);
Chris@10: 				   T18 = VSUB(TN, TR);
Chris@10: 				   T1d = VSUB(T1a, T1c);
Chris@10: 				   T1m = VADD(T1a, T1c);
Chris@10: 				   T1t = VADD(T1r, T1s);
Chris@10: 				   T1x = VSUB(T1s, T1r);
Chris@10: 				   T1y = VSUB(Tk, Tu);
Chris@10: 				   Tv = VADD(Tk, Tu);
Chris@10: 				   TJ = VADD(TA, TI);
Chris@10: 				   T11 = VSUB(TA, TI);
Chris@10: 				   T16 = VSUB(T13, T15);
Chris@10: 				   T1l = VADD(T13, T15);
Chris@10: 			      }
Chris@10: 			      {
Chris@10: 				   V Tw, T1w, T1v, TT;
Chris@10: 				   {
Chris@10: 					V T1i, T1e, T1B, T1z, T1h, T17;
Chris@10: 					T1i = VADD(T18, T1d);
Chris@10: 					T1e = VSUB(T18, T1d);
Chris@10: 					T1B = VADD(T1y, T1x);
Chris@10: 					T1z = VSUB(T1x, T1y);
Chris@10: 					T1h = VSUB(T16, T11);
Chris@10: 					T17 = VADD(T11, T16);
Chris@10: 					T1k = VSUB(Te, Tv);
Chris@10: 					Tw = VADD(Te, Tv);
Chris@10: 					{
Chris@10: 					     V T1A, T1j, T1C, T1f;
Chris@10: 					     T1A = VADD(T1h, T1i);
Chris@10: 					     T1j = VSUB(T1h, T1i);
Chris@10: 					     T1C = VSUB(T1e, T17);
Chris@10: 					     T1f = VADD(T17, T1e);
Chris@10: 					     T1w = VSUB(T1t, T1p);
Chris@10: 					     T1u = VADD(T1p, T1t);
Chris@10: 					     T1v = VSUB(TS, TJ);
Chris@10: 					     TT = VADD(TJ, TS);
Chris@10: 					     ST(&(ii[WS(rs, 1)]), VFMA(LDK(KP707106781), T1A, T1z), ms, &(ii[WS(rs, 1)]));
Chris@10: 					     ST(&(ii[WS(rs, 5)]), VFNMS(LDK(KP707106781), T1A, T1z), ms, &(ii[WS(rs, 1)]));
Chris@10: 					     ST(&(ri[WS(rs, 3)]), VFMA(LDK(KP707106781), T1j, T1g), ms, &(ri[WS(rs, 1)]));
Chris@10: 					     ST(&(ri[WS(rs, 7)]), VFNMS(LDK(KP707106781), T1j, T1g), ms, &(ri[WS(rs, 1)]));
Chris@10: 					     ST(&(ii[WS(rs, 3)]), VFMA(LDK(KP707106781), T1C, T1B), ms, &(ii[WS(rs, 1)]));
Chris@10: 					     ST(&(ii[WS(rs, 7)]), VFNMS(LDK(KP707106781), T1C, T1B), ms, &(ii[WS(rs, 1)]));
Chris@10: 					     ST(&(ri[WS(rs, 1)]), VFMA(LDK(KP707106781), T1f, T10), ms, &(ri[WS(rs, 1)]));
Chris@10: 					     ST(&(ri[WS(rs, 5)]), VFNMS(LDK(KP707106781), T1f, T10), ms, &(ri[WS(rs, 1)]));
Chris@10: 					}
Chris@10: 				   }
Chris@10: 				   ST(&(ri[WS(rs, 4)]), VSUB(Tw, TT), ms, &(ri[0]));
Chris@10: 				   ST(&(ri[0]), VADD(Tw, TT), ms, &(ri[0]));
Chris@10: 				   ST(&(ii[WS(rs, 6)]), VSUB(T1w, T1v), ms, &(ii[0]));
Chris@10: 				   ST(&(ii[WS(rs, 2)]), VADD(T1v, T1w), ms, &(ii[0]));
Chris@10: 			      }
Chris@10: 			 }
Chris@10: 		    }
Chris@10: 	       }
Chris@10: 	       T1n = VSUB(T1l, T1m);
Chris@10: 	       T1o = VADD(T1l, T1m);
Chris@10: 	       ST(&(ii[0]), VADD(T1o, T1u), ms, &(ii[0]));
Chris@10: 	       ST(&(ii[WS(rs, 4)]), VSUB(T1u, T1o), ms, &(ii[0]));
Chris@10: 	       ST(&(ri[WS(rs, 2)]), VADD(T1k, T1n), ms, &(ri[0]));
Chris@10: 	       ST(&(ri[WS(rs, 6)]), VSUB(T1k, T1n), ms, &(ri[0]));
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, 3),
Chris@10:      VTW(0, 7),
Chris@10:      {TW_NEXT, (2 * VL), 0}
Chris@10: };
Chris@10: 
Chris@10: static const ct_desc desc = { 8, XSIMD_STRING("t2sv_8"), twinstr, &GENUS, {44, 20, 30, 0}, 0, 0, 0 };
Chris@10: 
Chris@10: void XSIMD(codelet_t2sv_8) (planner *p) {
Chris@10:      X(kdft_dit_register) (p, t2sv_8, &desc);
Chris@10: }
Chris@10: #else				/* HAVE_FMA */
Chris@10: 
Chris@10: /* Generated by: ../../../genfft/gen_twiddle.native -simd -compact -variables 4 -pipeline-latency 8 -twiddle-log3 -precompute-twiddles -n 8 -name t2sv_8 -include ts.h */
Chris@10: 
Chris@10: /*
Chris@10:  * This function contains 74 FP additions, 44 FP multiplications,
Chris@10:  * (or, 56 additions, 26 multiplications, 18 fused multiply/add),
Chris@10:  * 42 stack variables, 1 constants, and 32 memory accesses
Chris@10:  */
Chris@10: #include "ts.h"
Chris@10: 
Chris@10: static void t2sv_8(R *ri, R *ii, const R *W, stride rs, INT mb, INT me, INT ms)
Chris@10: {
Chris@10:      DVK(KP707106781, +0.707106781186547524400844362104849039284835938);
Chris@10:      {
Chris@10: 	  INT m;
Chris@10: 	  for (m = mb, W = W + (mb * 6); m < me; m = m + (2 * VL), ri = ri + ((2 * VL) * ms), ii = ii + ((2 * VL) * ms), W = W + ((2 * VL) * 6), MAKE_VOLATILE_STRIDE(16, rs)) {
Chris@10: 	       V T2, T5, T3, T6, T8, Tc, Tg, Ti, Tl, Tm, Tn, Tz, Tp, Tx;
Chris@10: 	       {
Chris@10: 		    V T4, Tb, T7, Ta;
Chris@10: 		    T2 = LDW(&(W[0]));
Chris@10: 		    T5 = LDW(&(W[TWVL * 1]));
Chris@10: 		    T3 = LDW(&(W[TWVL * 2]));
Chris@10: 		    T6 = LDW(&(W[TWVL * 3]));
Chris@10: 		    T4 = VMUL(T2, T3);
Chris@10: 		    Tb = VMUL(T5, T3);
Chris@10: 		    T7 = VMUL(T5, T6);
Chris@10: 		    Ta = VMUL(T2, T6);
Chris@10: 		    T8 = VSUB(T4, T7);
Chris@10: 		    Tc = VADD(Ta, Tb);
Chris@10: 		    Tg = VADD(T4, T7);
Chris@10: 		    Ti = VSUB(Ta, Tb);
Chris@10: 		    Tl = LDW(&(W[TWVL * 4]));
Chris@10: 		    Tm = LDW(&(W[TWVL * 5]));
Chris@10: 		    Tn = VFMA(T2, Tl, VMUL(T5, Tm));
Chris@10: 		    Tz = VFNMS(Ti, Tl, VMUL(Tg, Tm));
Chris@10: 		    Tp = VFNMS(T5, Tl, VMUL(T2, Tm));
Chris@10: 		    Tx = VFMA(Tg, Tl, VMUL(Ti, Tm));
Chris@10: 	       }
Chris@10: 	       {
Chris@10: 		    V Tf, T1i, TL, T1d, TJ, T17, TV, TY, Ts, T1j, TO, T1a, TC, T16, TQ;
Chris@10: 		    V TT;
Chris@10: 		    {
Chris@10: 			 V T1, T1c, Te, T1b, T9, Td;
Chris@10: 			 T1 = LD(&(ri[0]), ms, &(ri[0]));
Chris@10: 			 T1c = LD(&(ii[0]), ms, &(ii[0]));
Chris@10: 			 T9 = LD(&(ri[WS(rs, 4)]), ms, &(ri[0]));
Chris@10: 			 Td = LD(&(ii[WS(rs, 4)]), ms, &(ii[0]));
Chris@10: 			 Te = VFMA(T8, T9, VMUL(Tc, Td));
Chris@10: 			 T1b = VFNMS(Tc, T9, VMUL(T8, Td));
Chris@10: 			 Tf = VADD(T1, Te);
Chris@10: 			 T1i = VSUB(T1c, T1b);
Chris@10: 			 TL = VSUB(T1, Te);
Chris@10: 			 T1d = VADD(T1b, T1c);
Chris@10: 		    }
Chris@10: 		    {
Chris@10: 			 V TF, TW, TI, TX;
Chris@10: 			 {
Chris@10: 			      V TD, TE, TG, TH;
Chris@10: 			      TD = LD(&(ri[WS(rs, 7)]), ms, &(ri[WS(rs, 1)]));
Chris@10: 			      TE = LD(&(ii[WS(rs, 7)]), ms, &(ii[WS(rs, 1)]));
Chris@10: 			      TF = VFMA(Tl, TD, VMUL(Tm, TE));
Chris@10: 			      TW = VFNMS(Tm, TD, VMUL(Tl, TE));
Chris@10: 			      TG = LD(&(ri[WS(rs, 3)]), ms, &(ri[WS(rs, 1)]));
Chris@10: 			      TH = LD(&(ii[WS(rs, 3)]), ms, &(ii[WS(rs, 1)]));
Chris@10: 			      TI = VFMA(T3, TG, VMUL(T6, TH));
Chris@10: 			      TX = VFNMS(T6, TG, VMUL(T3, TH));
Chris@10: 			 }
Chris@10: 			 TJ = VADD(TF, TI);
Chris@10: 			 T17 = VADD(TW, TX);
Chris@10: 			 TV = VSUB(TF, TI);
Chris@10: 			 TY = VSUB(TW, TX);
Chris@10: 		    }
Chris@10: 		    {
Chris@10: 			 V Tk, TM, Tr, TN;
Chris@10: 			 {
Chris@10: 			      V Th, Tj, To, Tq;
Chris@10: 			      Th = LD(&(ri[WS(rs, 2)]), ms, &(ri[0]));
Chris@10: 			      Tj = LD(&(ii[WS(rs, 2)]), ms, &(ii[0]));
Chris@10: 			      Tk = VFMA(Tg, Th, VMUL(Ti, Tj));
Chris@10: 			      TM = VFNMS(Ti, Th, VMUL(Tg, Tj));
Chris@10: 			      To = LD(&(ri[WS(rs, 6)]), ms, &(ri[0]));
Chris@10: 			      Tq = LD(&(ii[WS(rs, 6)]), ms, &(ii[0]));
Chris@10: 			      Tr = VFMA(Tn, To, VMUL(Tp, Tq));
Chris@10: 			      TN = VFNMS(Tp, To, VMUL(Tn, Tq));
Chris@10: 			 }
Chris@10: 			 Ts = VADD(Tk, Tr);
Chris@10: 			 T1j = VSUB(Tk, Tr);
Chris@10: 			 TO = VSUB(TM, TN);
Chris@10: 			 T1a = VADD(TM, TN);
Chris@10: 		    }
Chris@10: 		    {
Chris@10: 			 V Tw, TR, TB, TS;
Chris@10: 			 {
Chris@10: 			      V Tu, Tv, Ty, TA;
Chris@10: 			      Tu = LD(&(ri[WS(rs, 1)]), ms, &(ri[WS(rs, 1)]));
Chris@10: 			      Tv = LD(&(ii[WS(rs, 1)]), ms, &(ii[WS(rs, 1)]));
Chris@10: 			      Tw = VFMA(T2, Tu, VMUL(T5, Tv));
Chris@10: 			      TR = VFNMS(T5, Tu, VMUL(T2, Tv));
Chris@10: 			      Ty = LD(&(ri[WS(rs, 5)]), ms, &(ri[WS(rs, 1)]));
Chris@10: 			      TA = LD(&(ii[WS(rs, 5)]), ms, &(ii[WS(rs, 1)]));
Chris@10: 			      TB = VFMA(Tx, Ty, VMUL(Tz, TA));
Chris@10: 			      TS = VFNMS(Tz, Ty, VMUL(Tx, TA));
Chris@10: 			 }
Chris@10: 			 TC = VADD(Tw, TB);
Chris@10: 			 T16 = VADD(TR, TS);
Chris@10: 			 TQ = VSUB(Tw, TB);
Chris@10: 			 TT = VSUB(TR, TS);
Chris@10: 		    }
Chris@10: 		    {
Chris@10: 			 V Tt, TK, T1f, T1g;
Chris@10: 			 Tt = VADD(Tf, Ts);
Chris@10: 			 TK = VADD(TC, TJ);
Chris@10: 			 ST(&(ri[WS(rs, 4)]), VSUB(Tt, TK), ms, &(ri[0]));
Chris@10: 			 ST(&(ri[0]), VADD(Tt, TK), ms, &(ri[0]));
Chris@10: 			 {
Chris@10: 			      V T19, T1e, T15, T18;
Chris@10: 			      T19 = VADD(T16, T17);
Chris@10: 			      T1e = VADD(T1a, T1d);
Chris@10: 			      ST(&(ii[0]), VADD(T19, T1e), ms, &(ii[0]));
Chris@10: 			      ST(&(ii[WS(rs, 4)]), VSUB(T1e, T19), ms, &(ii[0]));
Chris@10: 			      T15 = VSUB(Tf, Ts);
Chris@10: 			      T18 = VSUB(T16, T17);
Chris@10: 			      ST(&(ri[WS(rs, 6)]), VSUB(T15, T18), ms, &(ri[0]));
Chris@10: 			      ST(&(ri[WS(rs, 2)]), VADD(T15, T18), ms, &(ri[0]));
Chris@10: 			 }
Chris@10: 			 T1f = VSUB(TJ, TC);
Chris@10: 			 T1g = VSUB(T1d, T1a);
Chris@10: 			 ST(&(ii[WS(rs, 2)]), VADD(T1f, T1g), ms, &(ii[0]));
Chris@10: 			 ST(&(ii[WS(rs, 6)]), VSUB(T1g, T1f), ms, &(ii[0]));
Chris@10: 			 {
Chris@10: 			      V T11, T1k, T14, T1h, T12, T13;
Chris@10: 			      T11 = VSUB(TL, TO);
Chris@10: 			      T1k = VSUB(T1i, T1j);
Chris@10: 			      T12 = VSUB(TT, TQ);
Chris@10: 			      T13 = VADD(TV, TY);
Chris@10: 			      T14 = VMUL(LDK(KP707106781), VSUB(T12, T13));
Chris@10: 			      T1h = VMUL(LDK(KP707106781), VADD(T12, T13));
Chris@10: 			      ST(&(ri[WS(rs, 7)]), VSUB(T11, T14), ms, &(ri[WS(rs, 1)]));
Chris@10: 			      ST(&(ii[WS(rs, 5)]), VSUB(T1k, T1h), ms, &(ii[WS(rs, 1)]));
Chris@10: 			      ST(&(ri[WS(rs, 3)]), VADD(T11, T14), ms, &(ri[WS(rs, 1)]));
Chris@10: 			      ST(&(ii[WS(rs, 1)]), VADD(T1h, T1k), ms, &(ii[WS(rs, 1)]));
Chris@10: 			 }
Chris@10: 			 {
Chris@10: 			      V TP, T1m, T10, T1l, TU, TZ;
Chris@10: 			      TP = VADD(TL, TO);
Chris@10: 			      T1m = VADD(T1j, T1i);
Chris@10: 			      TU = VADD(TQ, TT);
Chris@10: 			      TZ = VSUB(TV, TY);
Chris@10: 			      T10 = VMUL(LDK(KP707106781), VADD(TU, TZ));
Chris@10: 			      T1l = VMUL(LDK(KP707106781), VSUB(TZ, TU));
Chris@10: 			      ST(&(ri[WS(rs, 5)]), VSUB(TP, T10), ms, &(ri[WS(rs, 1)]));
Chris@10: 			      ST(&(ii[WS(rs, 7)]), VSUB(T1m, T1l), ms, &(ii[WS(rs, 1)]));
Chris@10: 			      ST(&(ri[WS(rs, 1)]), VADD(TP, T10), ms, &(ri[WS(rs, 1)]));
Chris@10: 			      ST(&(ii[WS(rs, 3)]), VADD(T1l, T1m), ms, &(ii[WS(rs, 1)]));
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, 3),
Chris@10:      VTW(0, 7),
Chris@10:      {TW_NEXT, (2 * VL), 0}
Chris@10: };
Chris@10: 
Chris@10: static const ct_desc desc = { 8, XSIMD_STRING("t2sv_8"), twinstr, &GENUS, {56, 26, 18, 0}, 0, 0, 0 };
Chris@10: 
Chris@10: void XSIMD(codelet_t2sv_8) (planner *p) {
Chris@10:      X(kdft_dit_register) (p, t2sv_8, &desc);
Chris@10: }
Chris@10: #endif				/* HAVE_FMA */