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 08:06:06 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 -twiddle-log3 -precompute-twiddles -no-generate-bytw -n 16 -name t3bv_16 -include dft/simd/t3b.h -sign 1 */
Chris@82: 
Chris@82: /*
Chris@82:  * This function contains 98 FP additions, 86 FP multiplications,
Chris@82:  * (or, 64 additions, 52 multiplications, 34 fused multiply/add),
Chris@82:  * 51 stack variables, 3 constants, and 32 memory accesses
Chris@82:  */
Chris@82: #include "dft/simd/t3b.h"
Chris@82: 
Chris@82: static void t3bv_16(R *ri, R *ii, const R *W, stride rs, INT mb, INT me, INT ms)
Chris@82: {
Chris@82:      DVK(KP923879532, +0.923879532511286756128183189396788286822416626);
Chris@82:      DVK(KP707106781, +0.707106781186547524400844362104849039284835938);
Chris@82:      DVK(KP414213562, +0.414213562373095048801688724209698078569671875);
Chris@82:      {
Chris@82: 	  INT m;
Chris@82: 	  R *x;
Chris@82: 	  x = ii;
Chris@82: 	  for (m = mb, W = W + (mb * ((TWVL / VL) * 8)); m < me; m = m + VL, x = x + (VL * ms), W = W + (TWVL * 8), MAKE_VOLATILE_STRIDE(16, rs)) {
Chris@82: 	       V T2, T8, T9, Tx, Tu, TR, T3, T4, TN, TU, Tc, Tm, Ty, TE, Tp;
Chris@82: 	       T2 = LDW(&(W[0]));
Chris@82: 	       T8 = LDW(&(W[TWVL * 2]));
Chris@82: 	       T9 = VZMUL(T2, T8);
Chris@82: 	       Tx = VZMULJ(T2, T8);
Chris@82: 	       Tu = LDW(&(W[TWVL * 6]));
Chris@82: 	       TR = VZMULJ(T2, Tu);
Chris@82: 	       T3 = LDW(&(W[TWVL * 4]));
Chris@82: 	       T4 = VZMULJ(T2, T3);
Chris@82: 	       TN = VZMUL(T2, T3);
Chris@82: 	       TU = VZMULJ(T8, T3);
Chris@82: 	       Tc = VZMUL(T8, T3);
Chris@82: 	       Tm = VZMULJ(T9, T3);
Chris@82: 	       Ty = VZMULJ(Tx, T3);
Chris@82: 	       TE = VZMUL(Tx, T3);
Chris@82: 	       Tp = VZMUL(T9, T3);
Chris@82: 	       {
Chris@82: 		    V T7, T1b, Tf, T1o, TQ, TX, T1e, T1p, Tl, Ts, Tt, T1i, T1r, TB, TH;
Chris@82: 		    V TI, T1l, T1s, T1, T6, T5;
Chris@82: 		    T1 = LD(&(x[0]), ms, &(x[0]));
Chris@82: 		    T5 = LD(&(x[WS(rs, 8)]), ms, &(x[0]));
Chris@82: 		    T6 = VZMUL(T4, T5);
Chris@82: 		    T7 = VADD(T1, T6);
Chris@82: 		    T1b = VSUB(T1, T6);
Chris@82: 		    {
Chris@82: 			 V Tb, Te, Ta, Td;
Chris@82: 			 Ta = LD(&(x[WS(rs, 4)]), ms, &(x[0]));
Chris@82: 			 Tb = VZMUL(T9, Ta);
Chris@82: 			 Td = LD(&(x[WS(rs, 12)]), ms, &(x[0]));
Chris@82: 			 Te = VZMUL(Tc, Td);
Chris@82: 			 Tf = VADD(Tb, Te);
Chris@82: 			 T1o = VSUB(Tb, Te);
Chris@82: 		    }
Chris@82: 		    {
Chris@82: 			 V TM, TW, TP, TT, T1c, T1d;
Chris@82: 			 {
Chris@82: 			      V TL, TV, TO, TS;
Chris@82: 			      TL = LD(&(x[WS(rs, 2)]), ms, &(x[0]));
Chris@82: 			      TM = VZMUL(Tx, TL);
Chris@82: 			      TV = LD(&(x[WS(rs, 6)]), ms, &(x[0]));
Chris@82: 			      TW = VZMUL(TU, TV);
Chris@82: 			      TO = LD(&(x[WS(rs, 10)]), ms, &(x[0]));
Chris@82: 			      TP = VZMUL(TN, TO);
Chris@82: 			      TS = LD(&(x[WS(rs, 14)]), ms, &(x[0]));
Chris@82: 			      TT = VZMUL(TR, TS);
Chris@82: 			 }
Chris@82: 			 TQ = VADD(TM, TP);
Chris@82: 			 TX = VADD(TT, TW);
Chris@82: 			 T1c = VSUB(TM, TP);
Chris@82: 			 T1d = VSUB(TT, TW);
Chris@82: 			 T1e = VADD(T1c, T1d);
Chris@82: 			 T1p = VSUB(T1c, T1d);
Chris@82: 		    }
Chris@82: 		    {
Chris@82: 			 V Ti, Tr, Tk, To, T1g, T1h;
Chris@82: 			 {
Chris@82: 			      V Th, Tq, Tj, Tn;
Chris@82: 			      Th = LD(&(x[WS(rs, 1)]), ms, &(x[WS(rs, 1)]));
Chris@82: 			      Ti = VZMUL(T2, Th);
Chris@82: 			      Tq = LD(&(x[WS(rs, 13)]), ms, &(x[WS(rs, 1)]));
Chris@82: 			      Tr = VZMUL(Tp, Tq);
Chris@82: 			      Tj = LD(&(x[WS(rs, 9)]), ms, &(x[WS(rs, 1)]));
Chris@82: 			      Tk = VZMUL(T3, Tj);
Chris@82: 			      Tn = LD(&(x[WS(rs, 5)]), ms, &(x[WS(rs, 1)]));
Chris@82: 			      To = VZMUL(Tm, Tn);
Chris@82: 			 }
Chris@82: 			 Tl = VADD(Ti, Tk);
Chris@82: 			 Ts = VADD(To, Tr);
Chris@82: 			 Tt = VSUB(Tl, Ts);
Chris@82: 			 T1g = VSUB(Ti, Tk);
Chris@82: 			 T1h = VSUB(To, Tr);
Chris@82: 			 T1i = VFNMS(LDK(KP414213562), T1h, T1g);
Chris@82: 			 T1r = VFMA(LDK(KP414213562), T1g, T1h);
Chris@82: 		    }
Chris@82: 		    {
Chris@82: 			 V Tw, TG, TA, TD, T1j, T1k;
Chris@82: 			 {
Chris@82: 			      V Tv, TF, Tz, TC;
Chris@82: 			      Tv = LD(&(x[WS(rs, 15)]), ms, &(x[WS(rs, 1)]));
Chris@82: 			      Tw = VZMUL(Tu, Tv);
Chris@82: 			      TF = LD(&(x[WS(rs, 11)]), ms, &(x[WS(rs, 1)]));
Chris@82: 			      TG = VZMUL(TE, TF);
Chris@82: 			      Tz = LD(&(x[WS(rs, 7)]), ms, &(x[WS(rs, 1)]));
Chris@82: 			      TA = VZMUL(Ty, Tz);
Chris@82: 			      TC = LD(&(x[WS(rs, 3)]), ms, &(x[WS(rs, 1)]));
Chris@82: 			      TD = VZMUL(T8, TC);
Chris@82: 			 }
Chris@82: 			 TB = VADD(Tw, TA);
Chris@82: 			 TH = VADD(TD, TG);
Chris@82: 			 TI = VSUB(TB, TH);
Chris@82: 			 T1j = VSUB(Tw, TA);
Chris@82: 			 T1k = VSUB(TG, TD);
Chris@82: 			 T1l = VFNMS(LDK(KP414213562), T1k, T1j);
Chris@82: 			 T1s = VFMA(LDK(KP414213562), T1j, T1k);
Chris@82: 		    }
Chris@82: 		    {
Chris@82: 			 V TK, T11, T10, T12;
Chris@82: 			 {
Chris@82: 			      V Tg, TJ, TY, TZ;
Chris@82: 			      Tg = VSUB(T7, Tf);
Chris@82: 			      TJ = VADD(Tt, TI);
Chris@82: 			      TK = VFNMS(LDK(KP707106781), TJ, Tg);
Chris@82: 			      T11 = VFMA(LDK(KP707106781), TJ, Tg);
Chris@82: 			      TY = VSUB(TQ, TX);
Chris@82: 			      TZ = VSUB(Tt, TI);
Chris@82: 			      T10 = VFNMS(LDK(KP707106781), TZ, TY);
Chris@82: 			      T12 = VFMA(LDK(KP707106781), TZ, TY);
Chris@82: 			 }
Chris@82: 			 ST(&(x[WS(rs, 6)]), VFNMSI(T10, TK), ms, &(x[0]));
Chris@82: 			 ST(&(x[WS(rs, 14)]), VFNMSI(T12, T11), ms, &(x[0]));
Chris@82: 			 ST(&(x[WS(rs, 10)]), VFMAI(T10, TK), ms, &(x[0]));
Chris@82: 			 ST(&(x[WS(rs, 2)]), VFMAI(T12, T11), ms, &(x[0]));
Chris@82: 		    }
Chris@82: 		    {
Chris@82: 			 V T1z, T1D, T1C, T1E;
Chris@82: 			 {
Chris@82: 			      V T1x, T1y, T1A, T1B;
Chris@82: 			      T1x = VFNMS(LDK(KP707106781), T1e, T1b);
Chris@82: 			      T1y = VADD(T1r, T1s);
Chris@82: 			      T1z = VFNMS(LDK(KP923879532), T1y, T1x);
Chris@82: 			      T1D = VFMA(LDK(KP923879532), T1y, T1x);
Chris@82: 			      T1A = VFNMS(LDK(KP707106781), T1p, T1o);
Chris@82: 			      T1B = VSUB(T1i, T1l);
Chris@82: 			      T1C = VFMA(LDK(KP923879532), T1B, T1A);
Chris@82: 			      T1E = VFNMS(LDK(KP923879532), T1B, T1A);
Chris@82: 			 }
Chris@82: 			 ST(&(x[WS(rs, 5)]), VFMAI(T1C, T1z), ms, &(x[WS(rs, 1)]));
Chris@82: 			 ST(&(x[WS(rs, 13)]), VFMAI(T1E, T1D), ms, &(x[WS(rs, 1)]));
Chris@82: 			 ST(&(x[WS(rs, 11)]), VFNMSI(T1C, T1z), ms, &(x[WS(rs, 1)]));
Chris@82: 			 ST(&(x[WS(rs, 3)]), VFNMSI(T1E, T1D), ms, &(x[WS(rs, 1)]));
Chris@82: 		    }
Chris@82: 		    {
Chris@82: 			 V T15, T19, T18, T1a;
Chris@82: 			 {
Chris@82: 			      V T13, T14, T16, T17;
Chris@82: 			      T13 = VADD(T7, Tf);
Chris@82: 			      T14 = VADD(TQ, TX);
Chris@82: 			      T15 = VSUB(T13, T14);
Chris@82: 			      T19 = VADD(T13, T14);
Chris@82: 			      T16 = VADD(Tl, Ts);
Chris@82: 			      T17 = VADD(TB, TH);
Chris@82: 			      T18 = VSUB(T16, T17);
Chris@82: 			      T1a = VADD(T16, T17);
Chris@82: 			 }
Chris@82: 			 ST(&(x[WS(rs, 12)]), VFNMSI(T18, T15), ms, &(x[0]));
Chris@82: 			 ST(&(x[0]), VADD(T19, T1a), ms, &(x[0]));
Chris@82: 			 ST(&(x[WS(rs, 4)]), VFMAI(T18, T15), ms, &(x[0]));
Chris@82: 			 ST(&(x[WS(rs, 8)]), VSUB(T19, T1a), ms, &(x[0]));
Chris@82: 		    }
Chris@82: 		    {
Chris@82: 			 V T1n, T1v, T1u, T1w;
Chris@82: 			 {
Chris@82: 			      V T1f, T1m, T1q, T1t;
Chris@82: 			      T1f = VFMA(LDK(KP707106781), T1e, T1b);
Chris@82: 			      T1m = VADD(T1i, T1l);
Chris@82: 			      T1n = VFNMS(LDK(KP923879532), T1m, T1f);
Chris@82: 			      T1v = VFMA(LDK(KP923879532), T1m, T1f);
Chris@82: 			      T1q = VFMA(LDK(KP707106781), T1p, T1o);
Chris@82: 			      T1t = VSUB(T1r, T1s);
Chris@82: 			      T1u = VFNMS(LDK(KP923879532), T1t, T1q);
Chris@82: 			      T1w = VFMA(LDK(KP923879532), T1t, T1q);
Chris@82: 			 }
Chris@82: 			 ST(&(x[WS(rs, 7)]), VFNMSI(T1u, T1n), ms, &(x[WS(rs, 1)]));
Chris@82: 			 ST(&(x[WS(rs, 1)]), VFMAI(T1w, T1v), ms, &(x[WS(rs, 1)]));
Chris@82: 			 ST(&(x[WS(rs, 9)]), VFMAI(T1u, T1n), ms, &(x[WS(rs, 1)]));
Chris@82: 			 ST(&(x[WS(rs, 15)]), VFNMSI(T1w, T1v), 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, 3),
Chris@82:      VTW(0, 9),
Chris@82:      VTW(0, 15),
Chris@82:      {TW_NEXT, VL, 0}
Chris@82: };
Chris@82: 
Chris@82: static const ct_desc desc = { 16, XSIMD_STRING("t3bv_16"), twinstr, &GENUS, {64, 52, 34, 0}, 0, 0, 0 };
Chris@82: 
Chris@82: void XSIMD(codelet_t3bv_16) (planner *p) {
Chris@82:      X(kdft_dit_register) (p, t3bv_16, &desc);
Chris@82: }
Chris@82: #else
Chris@82: 
Chris@82: /* Generated by: ../../../genfft/gen_twiddle_c.native -simd -compact -variables 4 -pipeline-latency 8 -twiddle-log3 -precompute-twiddles -no-generate-bytw -n 16 -name t3bv_16 -include dft/simd/t3b.h -sign 1 */
Chris@82: 
Chris@82: /*
Chris@82:  * This function contains 98 FP additions, 64 FP multiplications,
Chris@82:  * (or, 94 additions, 60 multiplications, 4 fused multiply/add),
Chris@82:  * 51 stack variables, 3 constants, and 32 memory accesses
Chris@82:  */
Chris@82: #include "dft/simd/t3b.h"
Chris@82: 
Chris@82: static void t3bv_16(R *ri, R *ii, const R *W, stride rs, INT mb, INT me, INT ms)
Chris@82: {
Chris@82:      DVK(KP382683432, +0.382683432365089771728459984030398866761344562);
Chris@82:      DVK(KP923879532, +0.923879532511286756128183189396788286822416626);
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) * 8)); m < me; m = m + VL, x = x + (VL * ms), W = W + (TWVL * 8), MAKE_VOLATILE_STRIDE(16, rs)) {
Chris@82: 	       V T1, T8, T9, Tl, Ti, TE, T4, Ta, TO, TV, Td, Tm, TA, TH, Ts;
Chris@82: 	       T1 = LDW(&(W[0]));
Chris@82: 	       T8 = LDW(&(W[TWVL * 2]));
Chris@82: 	       T9 = VZMUL(T1, T8);
Chris@82: 	       Tl = VZMULJ(T1, T8);
Chris@82: 	       Ti = LDW(&(W[TWVL * 6]));
Chris@82: 	       TE = VZMULJ(T1, Ti);
Chris@82: 	       T4 = LDW(&(W[TWVL * 4]));
Chris@82: 	       Ta = VZMULJ(T9, T4);
Chris@82: 	       TO = VZMUL(T8, T4);
Chris@82: 	       TV = VZMULJ(T1, T4);
Chris@82: 	       Td = VZMUL(T9, T4);
Chris@82: 	       Tm = VZMULJ(Tl, T4);
Chris@82: 	       TA = VZMUL(T1, T4);
Chris@82: 	       TH = VZMULJ(T8, T4);
Chris@82: 	       Ts = VZMUL(Tl, T4);
Chris@82: 	       {
Chris@82: 		    V TY, T1q, TR, T1r, T1m, T1n, TL, TZ, T1f, T1g, T1h, Th, T11, T1i, T1j;
Chris@82: 		    V T1k, Tw, T12, TU, TX, TW;
Chris@82: 		    TU = LD(&(x[0]), ms, &(x[0]));
Chris@82: 		    TW = LD(&(x[WS(rs, 8)]), ms, &(x[0]));
Chris@82: 		    TX = VZMUL(TV, TW);
Chris@82: 		    TY = VSUB(TU, TX);
Chris@82: 		    T1q = VADD(TU, TX);
Chris@82: 		    {
Chris@82: 			 V TN, TQ, TM, TP;
Chris@82: 			 TM = LD(&(x[WS(rs, 4)]), ms, &(x[0]));
Chris@82: 			 TN = VZMUL(T9, TM);
Chris@82: 			 TP = LD(&(x[WS(rs, 12)]), ms, &(x[0]));
Chris@82: 			 TQ = VZMUL(TO, TP);
Chris@82: 			 TR = VSUB(TN, TQ);
Chris@82: 			 T1r = VADD(TN, TQ);
Chris@82: 		    }
Chris@82: 		    {
Chris@82: 			 V Tz, TJ, TC, TG, TD, TK;
Chris@82: 			 {
Chris@82: 			      V Ty, TI, TB, TF;
Chris@82: 			      Ty = LD(&(x[WS(rs, 2)]), ms, &(x[0]));
Chris@82: 			      Tz = VZMUL(Tl, Ty);
Chris@82: 			      TI = LD(&(x[WS(rs, 6)]), ms, &(x[0]));
Chris@82: 			      TJ = VZMUL(TH, TI);
Chris@82: 			      TB = LD(&(x[WS(rs, 10)]), ms, &(x[0]));
Chris@82: 			      TC = VZMUL(TA, TB);
Chris@82: 			      TF = LD(&(x[WS(rs, 14)]), ms, &(x[0]));
Chris@82: 			      TG = VZMUL(TE, TF);
Chris@82: 			 }
Chris@82: 			 T1m = VADD(Tz, TC);
Chris@82: 			 T1n = VADD(TG, TJ);
Chris@82: 			 TD = VSUB(Tz, TC);
Chris@82: 			 TK = VSUB(TG, TJ);
Chris@82: 			 TL = VMUL(LDK(KP707106781), VSUB(TD, TK));
Chris@82: 			 TZ = VMUL(LDK(KP707106781), VADD(TD, TK));
Chris@82: 		    }
Chris@82: 		    {
Chris@82: 			 V T3, Tf, T6, Tc, T7, Tg;
Chris@82: 			 {
Chris@82: 			      V T2, Te, T5, Tb;
Chris@82: 			      T2 = LD(&(x[WS(rs, 1)]), ms, &(x[WS(rs, 1)]));
Chris@82: 			      T3 = VZMUL(T1, T2);
Chris@82: 			      Te = LD(&(x[WS(rs, 13)]), ms, &(x[WS(rs, 1)]));
Chris@82: 			      Tf = VZMUL(Td, Te);
Chris@82: 			      T5 = LD(&(x[WS(rs, 9)]), ms, &(x[WS(rs, 1)]));
Chris@82: 			      T6 = VZMUL(T4, T5);
Chris@82: 			      Tb = LD(&(x[WS(rs, 5)]), ms, &(x[WS(rs, 1)]));
Chris@82: 			      Tc = VZMUL(Ta, Tb);
Chris@82: 			 }
Chris@82: 			 T1f = VADD(T3, T6);
Chris@82: 			 T1g = VADD(Tc, Tf);
Chris@82: 			 T1h = VSUB(T1f, T1g);
Chris@82: 			 T7 = VSUB(T3, T6);
Chris@82: 			 Tg = VSUB(Tc, Tf);
Chris@82: 			 Th = VFNMS(LDK(KP382683432), Tg, VMUL(LDK(KP923879532), T7));
Chris@82: 			 T11 = VFMA(LDK(KP382683432), T7, VMUL(LDK(KP923879532), Tg));
Chris@82: 		    }
Chris@82: 		    {
Chris@82: 			 V Tk, Tu, To, Tr, Tp, Tv;
Chris@82: 			 {
Chris@82: 			      V Tj, Tt, Tn, Tq;
Chris@82: 			      Tj = LD(&(x[WS(rs, 15)]), ms, &(x[WS(rs, 1)]));
Chris@82: 			      Tk = VZMUL(Ti, Tj);
Chris@82: 			      Tt = LD(&(x[WS(rs, 11)]), ms, &(x[WS(rs, 1)]));
Chris@82: 			      Tu = VZMUL(Ts, Tt);
Chris@82: 			      Tn = LD(&(x[WS(rs, 7)]), ms, &(x[WS(rs, 1)]));
Chris@82: 			      To = VZMUL(Tm, Tn);
Chris@82: 			      Tq = LD(&(x[WS(rs, 3)]), ms, &(x[WS(rs, 1)]));
Chris@82: 			      Tr = VZMUL(T8, Tq);
Chris@82: 			 }
Chris@82: 			 T1i = VADD(Tk, To);
Chris@82: 			 T1j = VADD(Tr, Tu);
Chris@82: 			 T1k = VSUB(T1i, T1j);
Chris@82: 			 Tp = VSUB(Tk, To);
Chris@82: 			 Tv = VSUB(Tr, Tu);
Chris@82: 			 Tw = VFMA(LDK(KP923879532), Tp, VMUL(LDK(KP382683432), Tv));
Chris@82: 			 T12 = VFNMS(LDK(KP382683432), Tp, VMUL(LDK(KP923879532), Tv));
Chris@82: 		    }
Chris@82: 		    {
Chris@82: 			 V T1p, T1v, T1u, T1w;
Chris@82: 			 {
Chris@82: 			      V T1l, T1o, T1s, T1t;
Chris@82: 			      T1l = VMUL(LDK(KP707106781), VSUB(T1h, T1k));
Chris@82: 			      T1o = VSUB(T1m, T1n);
Chris@82: 			      T1p = VBYI(VSUB(T1l, T1o));
Chris@82: 			      T1v = VBYI(VADD(T1o, T1l));
Chris@82: 			      T1s = VSUB(T1q, T1r);
Chris@82: 			      T1t = VMUL(LDK(KP707106781), VADD(T1h, T1k));
Chris@82: 			      T1u = VSUB(T1s, T1t);
Chris@82: 			      T1w = VADD(T1s, T1t);
Chris@82: 			 }
Chris@82: 			 ST(&(x[WS(rs, 6)]), VADD(T1p, T1u), ms, &(x[0]));
Chris@82: 			 ST(&(x[WS(rs, 14)]), VSUB(T1w, T1v), ms, &(x[0]));
Chris@82: 			 ST(&(x[WS(rs, 10)]), VSUB(T1u, T1p), ms, &(x[0]));
Chris@82: 			 ST(&(x[WS(rs, 2)]), VADD(T1v, T1w), ms, &(x[0]));
Chris@82: 		    }
Chris@82: 		    {
Chris@82: 			 V T1z, T1D, T1C, T1E;
Chris@82: 			 {
Chris@82: 			      V T1x, T1y, T1A, T1B;
Chris@82: 			      T1x = VADD(T1q, T1r);
Chris@82: 			      T1y = VADD(T1m, T1n);
Chris@82: 			      T1z = VSUB(T1x, T1y);
Chris@82: 			      T1D = VADD(T1x, T1y);
Chris@82: 			      T1A = VADD(T1f, T1g);
Chris@82: 			      T1B = VADD(T1i, T1j);
Chris@82: 			      T1C = VBYI(VSUB(T1A, T1B));
Chris@82: 			      T1E = VADD(T1A, T1B);
Chris@82: 			 }
Chris@82: 			 ST(&(x[WS(rs, 12)]), VSUB(T1z, T1C), ms, &(x[0]));
Chris@82: 			 ST(&(x[0]), VADD(T1D, T1E), ms, &(x[0]));
Chris@82: 			 ST(&(x[WS(rs, 4)]), VADD(T1z, T1C), ms, &(x[0]));
Chris@82: 			 ST(&(x[WS(rs, 8)]), VSUB(T1D, T1E), ms, &(x[0]));
Chris@82: 		    }
Chris@82: 		    {
Chris@82: 			 V TT, T15, T14, T16;
Chris@82: 			 {
Chris@82: 			      V Tx, TS, T10, T13;
Chris@82: 			      Tx = VSUB(Th, Tw);
Chris@82: 			      TS = VSUB(TL, TR);
Chris@82: 			      TT = VBYI(VSUB(Tx, TS));
Chris@82: 			      T15 = VBYI(VADD(TS, Tx));
Chris@82: 			      T10 = VSUB(TY, TZ);
Chris@82: 			      T13 = VSUB(T11, T12);
Chris@82: 			      T14 = VSUB(T10, T13);
Chris@82: 			      T16 = VADD(T10, T13);
Chris@82: 			 }
Chris@82: 			 ST(&(x[WS(rs, 5)]), VADD(TT, T14), ms, &(x[WS(rs, 1)]));
Chris@82: 			 ST(&(x[WS(rs, 13)]), VSUB(T16, T15), ms, &(x[WS(rs, 1)]));
Chris@82: 			 ST(&(x[WS(rs, 11)]), VSUB(T14, TT), ms, &(x[WS(rs, 1)]));
Chris@82: 			 ST(&(x[WS(rs, 3)]), VADD(T15, T16), ms, &(x[WS(rs, 1)]));
Chris@82: 		    }
Chris@82: 		    {
Chris@82: 			 V T19, T1d, T1c, T1e;
Chris@82: 			 {
Chris@82: 			      V T17, T18, T1a, T1b;
Chris@82: 			      T17 = VADD(TY, TZ);
Chris@82: 			      T18 = VADD(Th, Tw);
Chris@82: 			      T19 = VADD(T17, T18);
Chris@82: 			      T1d = VSUB(T17, T18);
Chris@82: 			      T1a = VADD(TR, TL);
Chris@82: 			      T1b = VADD(T11, T12);
Chris@82: 			      T1c = VBYI(VADD(T1a, T1b));
Chris@82: 			      T1e = VBYI(VSUB(T1b, T1a));
Chris@82: 			 }
Chris@82: 			 ST(&(x[WS(rs, 15)]), VSUB(T19, T1c), ms, &(x[WS(rs, 1)]));
Chris@82: 			 ST(&(x[WS(rs, 7)]), VADD(T1d, T1e), ms, &(x[WS(rs, 1)]));
Chris@82: 			 ST(&(x[WS(rs, 1)]), VADD(T19, T1c), ms, &(x[WS(rs, 1)]));
Chris@82: 			 ST(&(x[WS(rs, 9)]), VSUB(T1d, T1e), 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, 3),
Chris@82:      VTW(0, 9),
Chris@82:      VTW(0, 15),
Chris@82:      {TW_NEXT, VL, 0}
Chris@82: };
Chris@82: 
Chris@82: static const ct_desc desc = { 16, XSIMD_STRING("t3bv_16"), twinstr, &GENUS, {94, 60, 4, 0}, 0, 0, 0 };
Chris@82: 
Chris@82: void XSIMD(codelet_t3bv_16) (planner *p) {
Chris@82:      X(kdft_dit_register) (p, t3bv_16, &desc);
Chris@82: }
Chris@82: #endif