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 08:04:14 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.native -fma -compact -variables 4 -pipeline-latency 4 -n 12 -name t1_12 -include dft/scalar/t.h */
Chris@82: 
Chris@82: /*
Chris@82:  * This function contains 118 FP additions, 68 FP multiplications,
Chris@82:  * (or, 72 additions, 22 multiplications, 46 fused multiply/add),
Chris@82:  * 47 stack variables, 2 constants, and 48 memory accesses
Chris@82:  */
Chris@82: #include "dft/scalar/t.h"
Chris@82: 
Chris@82: static void t1_12(R *ri, R *ii, const R *W, stride rs, INT mb, INT me, INT ms)
Chris@82: {
Chris@82:      DK(KP866025403, +0.866025403784438646763723170752936183471402627);
Chris@82:      DK(KP500000000, +0.500000000000000000000000000000000000000000000);
Chris@82:      {
Chris@82: 	  INT m;
Chris@82: 	  for (m = mb, W = W + (mb * 22); m < me; m = m + 1, ri = ri + ms, ii = ii + ms, W = W + 22, MAKE_VOLATILE_STRIDE(24, rs)) {
Chris@82: 	       E T1, T2i, Tl, T2e, T10, T1Y, TG, T1S, Ty, T2r, T1s, T2f, T1d, T21, T1H;
Chris@82: 	       E T1Z, Te, T2o, T1l, T2h, TT, T1V, T1A, T1T;
Chris@82: 	       T1 = ri[0];
Chris@82: 	       T2i = ii[0];
Chris@82: 	       {
Chris@82: 		    E Th, Tk, Ti, T2d, Tg, Tj;
Chris@82: 		    Th = ri[WS(rs, 6)];
Chris@82: 		    Tk = ii[WS(rs, 6)];
Chris@82: 		    Tg = W[10];
Chris@82: 		    Ti = Tg * Th;
Chris@82: 		    T2d = Tg * Tk;
Chris@82: 		    Tj = W[11];
Chris@82: 		    Tl = FMA(Tj, Tk, Ti);
Chris@82: 		    T2e = FNMS(Tj, Th, T2d);
Chris@82: 	       }
Chris@82: 	       {
Chris@82: 		    E TW, TZ, TX, T1X, TV, TY;
Chris@82: 		    TW = ri[WS(rs, 9)];
Chris@82: 		    TZ = ii[WS(rs, 9)];
Chris@82: 		    TV = W[16];
Chris@82: 		    TX = TV * TW;
Chris@82: 		    T1X = TV * TZ;
Chris@82: 		    TY = W[17];
Chris@82: 		    T10 = FMA(TY, TZ, TX);
Chris@82: 		    T1Y = FNMS(TY, TW, T1X);
Chris@82: 	       }
Chris@82: 	       {
Chris@82: 		    E TC, TF, TD, T1R, TB, TE;
Chris@82: 		    TC = ri[WS(rs, 3)];
Chris@82: 		    TF = ii[WS(rs, 3)];
Chris@82: 		    TB = W[4];
Chris@82: 		    TD = TB * TC;
Chris@82: 		    T1R = TB * TF;
Chris@82: 		    TE = W[5];
Chris@82: 		    TG = FMA(TE, TF, TD);
Chris@82: 		    T1S = FNMS(TE, TC, T1R);
Chris@82: 	       }
Chris@82: 	       {
Chris@82: 		    E Tn, Tq, To, T1o, Tt, Tw, Tu, T1q, Tm, Ts;
Chris@82: 		    Tn = ri[WS(rs, 10)];
Chris@82: 		    Tq = ii[WS(rs, 10)];
Chris@82: 		    Tm = W[18];
Chris@82: 		    To = Tm * Tn;
Chris@82: 		    T1o = Tm * Tq;
Chris@82: 		    Tt = ri[WS(rs, 2)];
Chris@82: 		    Tw = ii[WS(rs, 2)];
Chris@82: 		    Ts = W[2];
Chris@82: 		    Tu = Ts * Tt;
Chris@82: 		    T1q = Ts * Tw;
Chris@82: 		    {
Chris@82: 			 E Tr, T1p, Tx, T1r, Tp, Tv;
Chris@82: 			 Tp = W[19];
Chris@82: 			 Tr = FMA(Tp, Tq, To);
Chris@82: 			 T1p = FNMS(Tp, Tn, T1o);
Chris@82: 			 Tv = W[3];
Chris@82: 			 Tx = FMA(Tv, Tw, Tu);
Chris@82: 			 T1r = FNMS(Tv, Tt, T1q);
Chris@82: 			 Ty = Tr + Tx;
Chris@82: 			 T2r = Tx - Tr;
Chris@82: 			 T1s = T1p - T1r;
Chris@82: 			 T2f = T1p + T1r;
Chris@82: 		    }
Chris@82: 	       }
Chris@82: 	       {
Chris@82: 		    E T12, T15, T13, T1D, T18, T1b, T19, T1F, T11, T17;
Chris@82: 		    T12 = ri[WS(rs, 1)];
Chris@82: 		    T15 = ii[WS(rs, 1)];
Chris@82: 		    T11 = W[0];
Chris@82: 		    T13 = T11 * T12;
Chris@82: 		    T1D = T11 * T15;
Chris@82: 		    T18 = ri[WS(rs, 5)];
Chris@82: 		    T1b = ii[WS(rs, 5)];
Chris@82: 		    T17 = W[8];
Chris@82: 		    T19 = T17 * T18;
Chris@82: 		    T1F = T17 * T1b;
Chris@82: 		    {
Chris@82: 			 E T16, T1E, T1c, T1G, T14, T1a;
Chris@82: 			 T14 = W[1];
Chris@82: 			 T16 = FMA(T14, T15, T13);
Chris@82: 			 T1E = FNMS(T14, T12, T1D);
Chris@82: 			 T1a = W[9];
Chris@82: 			 T1c = FMA(T1a, T1b, T19);
Chris@82: 			 T1G = FNMS(T1a, T18, T1F);
Chris@82: 			 T1d = T16 + T1c;
Chris@82: 			 T21 = T1c - T16;
Chris@82: 			 T1H = T1E - T1G;
Chris@82: 			 T1Z = T1E + T1G;
Chris@82: 		    }
Chris@82: 	       }
Chris@82: 	       {
Chris@82: 		    E T3, T6, T4, T1h, T9, Tc, Ta, T1j, T2, T8;
Chris@82: 		    T3 = ri[WS(rs, 4)];
Chris@82: 		    T6 = ii[WS(rs, 4)];
Chris@82: 		    T2 = W[6];
Chris@82: 		    T4 = T2 * T3;
Chris@82: 		    T1h = T2 * T6;
Chris@82: 		    T9 = ri[WS(rs, 8)];
Chris@82: 		    Tc = ii[WS(rs, 8)];
Chris@82: 		    T8 = W[14];
Chris@82: 		    Ta = T8 * T9;
Chris@82: 		    T1j = T8 * Tc;
Chris@82: 		    {
Chris@82: 			 E T7, T1i, Td, T1k, T5, Tb;
Chris@82: 			 T5 = W[7];
Chris@82: 			 T7 = FMA(T5, T6, T4);
Chris@82: 			 T1i = FNMS(T5, T3, T1h);
Chris@82: 			 Tb = W[15];
Chris@82: 			 Td = FMA(Tb, Tc, Ta);
Chris@82: 			 T1k = FNMS(Tb, T9, T1j);
Chris@82: 			 Te = T7 + Td;
Chris@82: 			 T2o = Td - T7;
Chris@82: 			 T1l = T1i - T1k;
Chris@82: 			 T2h = T1i + T1k;
Chris@82: 		    }
Chris@82: 	       }
Chris@82: 	       {
Chris@82: 		    E TI, TL, TJ, T1w, TO, TR, TP, T1y, TH, TN;
Chris@82: 		    TI = ri[WS(rs, 7)];
Chris@82: 		    TL = ii[WS(rs, 7)];
Chris@82: 		    TH = W[12];
Chris@82: 		    TJ = TH * TI;
Chris@82: 		    T1w = TH * TL;
Chris@82: 		    TO = ri[WS(rs, 11)];
Chris@82: 		    TR = ii[WS(rs, 11)];
Chris@82: 		    TN = W[20];
Chris@82: 		    TP = TN * TO;
Chris@82: 		    T1y = TN * TR;
Chris@82: 		    {
Chris@82: 			 E TM, T1x, TS, T1z, TK, TQ;
Chris@82: 			 TK = W[13];
Chris@82: 			 TM = FMA(TK, TL, TJ);
Chris@82: 			 T1x = FNMS(TK, TI, T1w);
Chris@82: 			 TQ = W[21];
Chris@82: 			 TS = FMA(TQ, TR, TP);
Chris@82: 			 T1z = FNMS(TQ, TO, T1y);
Chris@82: 			 TT = TM + TS;
Chris@82: 			 T1V = TS - TM;
Chris@82: 			 T1A = T1x - T1z;
Chris@82: 			 T1T = T1x + T1z;
Chris@82: 		    }
Chris@82: 	       }
Chris@82: 	       {
Chris@82: 		    E TA, T28, T2k, T2m, T1f, T2l, T2b, T2c;
Chris@82: 		    {
Chris@82: 			 E Tf, Tz, T2g, T2j;
Chris@82: 			 Tf = T1 + Te;
Chris@82: 			 Tz = Tl + Ty;
Chris@82: 			 TA = Tf + Tz;
Chris@82: 			 T28 = Tf - Tz;
Chris@82: 			 T2g = T2e + T2f;
Chris@82: 			 T2j = T2h + T2i;
Chris@82: 			 T2k = T2g + T2j;
Chris@82: 			 T2m = T2j - T2g;
Chris@82: 		    }
Chris@82: 		    {
Chris@82: 			 E TU, T1e, T29, T2a;
Chris@82: 			 TU = TG + TT;
Chris@82: 			 T1e = T10 + T1d;
Chris@82: 			 T1f = TU + T1e;
Chris@82: 			 T2l = TU - T1e;
Chris@82: 			 T29 = T1S + T1T;
Chris@82: 			 T2a = T1Y + T1Z;
Chris@82: 			 T2b = T29 - T2a;
Chris@82: 			 T2c = T29 + T2a;
Chris@82: 		    }
Chris@82: 		    ri[WS(rs, 6)] = TA - T1f;
Chris@82: 		    ii[WS(rs, 6)] = T2k - T2c;
Chris@82: 		    ri[0] = TA + T1f;
Chris@82: 		    ii[0] = T2c + T2k;
Chris@82: 		    ri[WS(rs, 3)] = T28 - T2b;
Chris@82: 		    ii[WS(rs, 3)] = T2l + T2m;
Chris@82: 		    ri[WS(rs, 9)] = T28 + T2b;
Chris@82: 		    ii[WS(rs, 9)] = T2m - T2l;
Chris@82: 	       }
Chris@82: 	       {
Chris@82: 		    E T1m, T1K, T2p, T2y, T2s, T2x, T1t, T1L, T1B, T1N, T1W, T25, T22, T26, T1I;
Chris@82: 		    E T1O;
Chris@82: 		    {
Chris@82: 			 E T1g, T2n, T2q, T1n;
Chris@82: 			 T1g = FNMS(KP500000000, Te, T1);
Chris@82: 			 T1m = FNMS(KP866025403, T1l, T1g);
Chris@82: 			 T1K = FMA(KP866025403, T1l, T1g);
Chris@82: 			 T2n = FNMS(KP500000000, T2h, T2i);
Chris@82: 			 T2p = FMA(KP866025403, T2o, T2n);
Chris@82: 			 T2y = FNMS(KP866025403, T2o, T2n);
Chris@82: 			 T2q = FNMS(KP500000000, T2f, T2e);
Chris@82: 			 T2s = FMA(KP866025403, T2r, T2q);
Chris@82: 			 T2x = FNMS(KP866025403, T2r, T2q);
Chris@82: 			 T1n = FNMS(KP500000000, Ty, Tl);
Chris@82: 			 T1t = FNMS(KP866025403, T1s, T1n);
Chris@82: 			 T1L = FMA(KP866025403, T1s, T1n);
Chris@82: 		    }
Chris@82: 		    {
Chris@82: 			 E T1v, T1U, T20, T1C;
Chris@82: 			 T1v = FNMS(KP500000000, TT, TG);
Chris@82: 			 T1B = FNMS(KP866025403, T1A, T1v);
Chris@82: 			 T1N = FMA(KP866025403, T1A, T1v);
Chris@82: 			 T1U = FNMS(KP500000000, T1T, T1S);
Chris@82: 			 T1W = FMA(KP866025403, T1V, T1U);
Chris@82: 			 T25 = FNMS(KP866025403, T1V, T1U);
Chris@82: 			 T20 = FNMS(KP500000000, T1Z, T1Y);
Chris@82: 			 T22 = FMA(KP866025403, T21, T20);
Chris@82: 			 T26 = FNMS(KP866025403, T21, T20);
Chris@82: 			 T1C = FNMS(KP500000000, T1d, T10);
Chris@82: 			 T1I = FNMS(KP866025403, T1H, T1C);
Chris@82: 			 T1O = FMA(KP866025403, T1H, T1C);
Chris@82: 		    }
Chris@82: 		    {
Chris@82: 			 E T1u, T1J, T2z, T2A;
Chris@82: 			 T1u = T1m + T1t;
Chris@82: 			 T1J = T1B + T1I;
Chris@82: 			 ri[WS(rs, 2)] = T1u - T1J;
Chris@82: 			 ri[WS(rs, 8)] = T1u + T1J;
Chris@82: 			 T2z = T2x + T2y;
Chris@82: 			 T2A = T25 + T26;
Chris@82: 			 ii[WS(rs, 2)] = T2z - T2A;
Chris@82: 			 ii[WS(rs, 8)] = T2A + T2z;
Chris@82: 		    }
Chris@82: 		    {
Chris@82: 			 E T1M, T1P, T2v, T2w;
Chris@82: 			 T1M = T1K + T1L;
Chris@82: 			 T1P = T1N + T1O;
Chris@82: 			 ri[WS(rs, 10)] = T1M - T1P;
Chris@82: 			 ri[WS(rs, 4)] = T1M + T1P;
Chris@82: 			 T2v = T1W + T22;
Chris@82: 			 T2w = T2s + T2p;
Chris@82: 			 ii[WS(rs, 4)] = T2v + T2w;
Chris@82: 			 ii[WS(rs, 10)] = T2w - T2v;
Chris@82: 		    }
Chris@82: 		    {
Chris@82: 			 E T1Q, T23, T2t, T2u;
Chris@82: 			 T1Q = T1K - T1L;
Chris@82: 			 T23 = T1W - T22;
Chris@82: 			 ri[WS(rs, 7)] = T1Q - T23;
Chris@82: 			 ri[WS(rs, 1)] = T1Q + T23;
Chris@82: 			 T2t = T2p - T2s;
Chris@82: 			 T2u = T1N - T1O;
Chris@82: 			 ii[WS(rs, 1)] = T2t - T2u;
Chris@82: 			 ii[WS(rs, 7)] = T2u + T2t;
Chris@82: 		    }
Chris@82: 		    {
Chris@82: 			 E T24, T27, T2B, T2C;
Chris@82: 			 T24 = T1m - T1t;
Chris@82: 			 T27 = T25 - T26;
Chris@82: 			 ri[WS(rs, 11)] = T24 - T27;
Chris@82: 			 ri[WS(rs, 5)] = T24 + T27;
Chris@82: 			 T2B = T2y - T2x;
Chris@82: 			 T2C = T1B - T1I;
Chris@82: 			 ii[WS(rs, 5)] = T2B - T2C;
Chris@82: 			 ii[WS(rs, 11)] = T2C + T2B;
Chris@82: 		    }
Chris@82: 	       }
Chris@82: 	  }
Chris@82:      }
Chris@82: }
Chris@82: 
Chris@82: static const tw_instr twinstr[] = {
Chris@82:      {TW_FULL, 0, 12},
Chris@82:      {TW_NEXT, 1, 0}
Chris@82: };
Chris@82: 
Chris@82: static const ct_desc desc = { 12, "t1_12", twinstr, &GENUS, {72, 22, 46, 0}, 0, 0, 0 };
Chris@82: 
Chris@82: void X(codelet_t1_12) (planner *p) {
Chris@82:      X(kdft_dit_register) (p, t1_12, &desc);
Chris@82: }
Chris@82: #else
Chris@82: 
Chris@82: /* Generated by: ../../../genfft/gen_twiddle.native -compact -variables 4 -pipeline-latency 4 -n 12 -name t1_12 -include dft/scalar/t.h */
Chris@82: 
Chris@82: /*
Chris@82:  * This function contains 118 FP additions, 60 FP multiplications,
Chris@82:  * (or, 88 additions, 30 multiplications, 30 fused multiply/add),
Chris@82:  * 47 stack variables, 2 constants, and 48 memory accesses
Chris@82:  */
Chris@82: #include "dft/scalar/t.h"
Chris@82: 
Chris@82: static void t1_12(R *ri, R *ii, const R *W, stride rs, INT mb, INT me, INT ms)
Chris@82: {
Chris@82:      DK(KP500000000, +0.500000000000000000000000000000000000000000000);
Chris@82:      DK(KP866025403, +0.866025403784438646763723170752936183471402627);
Chris@82:      {
Chris@82: 	  INT m;
Chris@82: 	  for (m = mb, W = W + (mb * 22); m < me; m = m + 1, ri = ri + ms, ii = ii + ms, W = W + 22, MAKE_VOLATILE_STRIDE(24, rs)) {
Chris@82: 	       E T1, T1W, T18, T21, Tc, T15, T1V, T22, TR, T1E, T1o, T1D, T12, T1l, T1F;
Chris@82: 	       E T1G, Ti, T1S, T1d, T24, Tt, T1a, T1T, T25, TA, T1z, T1j, T1y, TL, T1g;
Chris@82: 	       E T1A, T1B;
Chris@82: 	       {
Chris@82: 		    E T6, T16, Tb, T17;
Chris@82: 		    T1 = ri[0];
Chris@82: 		    T1W = ii[0];
Chris@82: 		    {
Chris@82: 			 E T3, T5, T2, T4;
Chris@82: 			 T3 = ri[WS(rs, 4)];
Chris@82: 			 T5 = ii[WS(rs, 4)];
Chris@82: 			 T2 = W[6];
Chris@82: 			 T4 = W[7];
Chris@82: 			 T6 = FMA(T2, T3, T4 * T5);
Chris@82: 			 T16 = FNMS(T4, T3, T2 * T5);
Chris@82: 		    }
Chris@82: 		    {
Chris@82: 			 E T8, Ta, T7, T9;
Chris@82: 			 T8 = ri[WS(rs, 8)];
Chris@82: 			 Ta = ii[WS(rs, 8)];
Chris@82: 			 T7 = W[14];
Chris@82: 			 T9 = W[15];
Chris@82: 			 Tb = FMA(T7, T8, T9 * Ta);
Chris@82: 			 T17 = FNMS(T9, T8, T7 * Ta);
Chris@82: 		    }
Chris@82: 		    T18 = KP866025403 * (T16 - T17);
Chris@82: 		    T21 = KP866025403 * (Tb - T6);
Chris@82: 		    Tc = T6 + Tb;
Chris@82: 		    T15 = FNMS(KP500000000, Tc, T1);
Chris@82: 		    T1V = T16 + T17;
Chris@82: 		    T22 = FNMS(KP500000000, T1V, T1W);
Chris@82: 	       }
Chris@82: 	       {
Chris@82: 		    E T11, T1n, TW, T1m;
Chris@82: 		    {
Chris@82: 			 E TO, TQ, TN, TP;
Chris@82: 			 TO = ri[WS(rs, 9)];
Chris@82: 			 TQ = ii[WS(rs, 9)];
Chris@82: 			 TN = W[16];
Chris@82: 			 TP = W[17];
Chris@82: 			 TR = FMA(TN, TO, TP * TQ);
Chris@82: 			 T1E = FNMS(TP, TO, TN * TQ);
Chris@82: 		    }
Chris@82: 		    {
Chris@82: 			 E TY, T10, TX, TZ;
Chris@82: 			 TY = ri[WS(rs, 5)];
Chris@82: 			 T10 = ii[WS(rs, 5)];
Chris@82: 			 TX = W[8];
Chris@82: 			 TZ = W[9];
Chris@82: 			 T11 = FMA(TX, TY, TZ * T10);
Chris@82: 			 T1n = FNMS(TZ, TY, TX * T10);
Chris@82: 		    }
Chris@82: 		    {
Chris@82: 			 E TT, TV, TS, TU;
Chris@82: 			 TT = ri[WS(rs, 1)];
Chris@82: 			 TV = ii[WS(rs, 1)];
Chris@82: 			 TS = W[0];
Chris@82: 			 TU = W[1];
Chris@82: 			 TW = FMA(TS, TT, TU * TV);
Chris@82: 			 T1m = FNMS(TU, TT, TS * TV);
Chris@82: 		    }
Chris@82: 		    T1o = KP866025403 * (T1m - T1n);
Chris@82: 		    T1D = KP866025403 * (T11 - TW);
Chris@82: 		    T12 = TW + T11;
Chris@82: 		    T1l = FNMS(KP500000000, T12, TR);
Chris@82: 		    T1F = T1m + T1n;
Chris@82: 		    T1G = FNMS(KP500000000, T1F, T1E);
Chris@82: 	       }
Chris@82: 	       {
Chris@82: 		    E Ts, T1c, Tn, T1b;
Chris@82: 		    {
Chris@82: 			 E Tf, Th, Te, Tg;
Chris@82: 			 Tf = ri[WS(rs, 6)];
Chris@82: 			 Th = ii[WS(rs, 6)];
Chris@82: 			 Te = W[10];
Chris@82: 			 Tg = W[11];
Chris@82: 			 Ti = FMA(Te, Tf, Tg * Th);
Chris@82: 			 T1S = FNMS(Tg, Tf, Te * Th);
Chris@82: 		    }
Chris@82: 		    {
Chris@82: 			 E Tp, Tr, To, Tq;
Chris@82: 			 Tp = ri[WS(rs, 2)];
Chris@82: 			 Tr = ii[WS(rs, 2)];
Chris@82: 			 To = W[2];
Chris@82: 			 Tq = W[3];
Chris@82: 			 Ts = FMA(To, Tp, Tq * Tr);
Chris@82: 			 T1c = FNMS(Tq, Tp, To * Tr);
Chris@82: 		    }
Chris@82: 		    {
Chris@82: 			 E Tk, Tm, Tj, Tl;
Chris@82: 			 Tk = ri[WS(rs, 10)];
Chris@82: 			 Tm = ii[WS(rs, 10)];
Chris@82: 			 Tj = W[18];
Chris@82: 			 Tl = W[19];
Chris@82: 			 Tn = FMA(Tj, Tk, Tl * Tm);
Chris@82: 			 T1b = FNMS(Tl, Tk, Tj * Tm);
Chris@82: 		    }
Chris@82: 		    T1d = KP866025403 * (T1b - T1c);
Chris@82: 		    T24 = KP866025403 * (Ts - Tn);
Chris@82: 		    Tt = Tn + Ts;
Chris@82: 		    T1a = FNMS(KP500000000, Tt, Ti);
Chris@82: 		    T1T = T1b + T1c;
Chris@82: 		    T25 = FNMS(KP500000000, T1T, T1S);
Chris@82: 	       }
Chris@82: 	       {
Chris@82: 		    E TK, T1i, TF, T1h;
Chris@82: 		    {
Chris@82: 			 E Tx, Tz, Tw, Ty;
Chris@82: 			 Tx = ri[WS(rs, 3)];
Chris@82: 			 Tz = ii[WS(rs, 3)];
Chris@82: 			 Tw = W[4];
Chris@82: 			 Ty = W[5];
Chris@82: 			 TA = FMA(Tw, Tx, Ty * Tz);
Chris@82: 			 T1z = FNMS(Ty, Tx, Tw * Tz);
Chris@82: 		    }
Chris@82: 		    {
Chris@82: 			 E TH, TJ, TG, TI;
Chris@82: 			 TH = ri[WS(rs, 11)];
Chris@82: 			 TJ = ii[WS(rs, 11)];
Chris@82: 			 TG = W[20];
Chris@82: 			 TI = W[21];
Chris@82: 			 TK = FMA(TG, TH, TI * TJ);
Chris@82: 			 T1i = FNMS(TI, TH, TG * TJ);
Chris@82: 		    }
Chris@82: 		    {
Chris@82: 			 E TC, TE, TB, TD;
Chris@82: 			 TC = ri[WS(rs, 7)];
Chris@82: 			 TE = ii[WS(rs, 7)];
Chris@82: 			 TB = W[12];
Chris@82: 			 TD = W[13];
Chris@82: 			 TF = FMA(TB, TC, TD * TE);
Chris@82: 			 T1h = FNMS(TD, TC, TB * TE);
Chris@82: 		    }
Chris@82: 		    T1j = KP866025403 * (T1h - T1i);
Chris@82: 		    T1y = KP866025403 * (TK - TF);
Chris@82: 		    TL = TF + TK;
Chris@82: 		    T1g = FNMS(KP500000000, TL, TA);
Chris@82: 		    T1A = T1h + T1i;
Chris@82: 		    T1B = FNMS(KP500000000, T1A, T1z);
Chris@82: 	       }
Chris@82: 	       {
Chris@82: 		    E Tv, T1N, T1Y, T20, T14, T1Z, T1Q, T1R;
Chris@82: 		    {
Chris@82: 			 E Td, Tu, T1U, T1X;
Chris@82: 			 Td = T1 + Tc;
Chris@82: 			 Tu = Ti + Tt;
Chris@82: 			 Tv = Td + Tu;
Chris@82: 			 T1N = Td - Tu;
Chris@82: 			 T1U = T1S + T1T;
Chris@82: 			 T1X = T1V + T1W;
Chris@82: 			 T1Y = T1U + T1X;
Chris@82: 			 T20 = T1X - T1U;
Chris@82: 		    }
Chris@82: 		    {
Chris@82: 			 E TM, T13, T1O, T1P;
Chris@82: 			 TM = TA + TL;
Chris@82: 			 T13 = TR + T12;
Chris@82: 			 T14 = TM + T13;
Chris@82: 			 T1Z = TM - T13;
Chris@82: 			 T1O = T1z + T1A;
Chris@82: 			 T1P = T1E + T1F;
Chris@82: 			 T1Q = T1O - T1P;
Chris@82: 			 T1R = T1O + T1P;
Chris@82: 		    }
Chris@82: 		    ri[WS(rs, 6)] = Tv - T14;
Chris@82: 		    ii[WS(rs, 6)] = T1Y - T1R;
Chris@82: 		    ri[0] = Tv + T14;
Chris@82: 		    ii[0] = T1R + T1Y;
Chris@82: 		    ri[WS(rs, 3)] = T1N - T1Q;
Chris@82: 		    ii[WS(rs, 3)] = T1Z + T20;
Chris@82: 		    ri[WS(rs, 9)] = T1N + T1Q;
Chris@82: 		    ii[WS(rs, 9)] = T20 - T1Z;
Chris@82: 	       }
Chris@82: 	       {
Chris@82: 		    E T1t, T1x, T27, T2a, T1w, T28, T1I, T29;
Chris@82: 		    {
Chris@82: 			 E T1r, T1s, T23, T26;
Chris@82: 			 T1r = T15 + T18;
Chris@82: 			 T1s = T1a + T1d;
Chris@82: 			 T1t = T1r + T1s;
Chris@82: 			 T1x = T1r - T1s;
Chris@82: 			 T23 = T21 + T22;
Chris@82: 			 T26 = T24 + T25;
Chris@82: 			 T27 = T23 - T26;
Chris@82: 			 T2a = T26 + T23;
Chris@82: 		    }
Chris@82: 		    {
Chris@82: 			 E T1u, T1v, T1C, T1H;
Chris@82: 			 T1u = T1g + T1j;
Chris@82: 			 T1v = T1l + T1o;
Chris@82: 			 T1w = T1u + T1v;
Chris@82: 			 T28 = T1u - T1v;
Chris@82: 			 T1C = T1y + T1B;
Chris@82: 			 T1H = T1D + T1G;
Chris@82: 			 T1I = T1C - T1H;
Chris@82: 			 T29 = T1C + T1H;
Chris@82: 		    }
Chris@82: 		    ri[WS(rs, 10)] = T1t - T1w;
Chris@82: 		    ii[WS(rs, 10)] = T2a - T29;
Chris@82: 		    ri[WS(rs, 4)] = T1t + T1w;
Chris@82: 		    ii[WS(rs, 4)] = T29 + T2a;
Chris@82: 		    ri[WS(rs, 7)] = T1x - T1I;
Chris@82: 		    ii[WS(rs, 7)] = T28 + T27;
Chris@82: 		    ri[WS(rs, 1)] = T1x + T1I;
Chris@82: 		    ii[WS(rs, 1)] = T27 - T28;
Chris@82: 	       }
Chris@82: 	       {
Chris@82: 		    E T1f, T1J, T2d, T2f, T1q, T2g, T1M, T2e;
Chris@82: 		    {
Chris@82: 			 E T19, T1e, T2b, T2c;
Chris@82: 			 T19 = T15 - T18;
Chris@82: 			 T1e = T1a - T1d;
Chris@82: 			 T1f = T19 + T1e;
Chris@82: 			 T1J = T19 - T1e;
Chris@82: 			 T2b = T25 - T24;
Chris@82: 			 T2c = T22 - T21;
Chris@82: 			 T2d = T2b + T2c;
Chris@82: 			 T2f = T2c - T2b;
Chris@82: 		    }
Chris@82: 		    {
Chris@82: 			 E T1k, T1p, T1K, T1L;
Chris@82: 			 T1k = T1g - T1j;
Chris@82: 			 T1p = T1l - T1o;
Chris@82: 			 T1q = T1k + T1p;
Chris@82: 			 T2g = T1k - T1p;
Chris@82: 			 T1K = T1B - T1y;
Chris@82: 			 T1L = T1G - T1D;
Chris@82: 			 T1M = T1K - T1L;
Chris@82: 			 T2e = T1K + T1L;
Chris@82: 		    }
Chris@82: 		    ri[WS(rs, 2)] = T1f - T1q;
Chris@82: 		    ii[WS(rs, 2)] = T2d - T2e;
Chris@82: 		    ri[WS(rs, 8)] = T1f + T1q;
Chris@82: 		    ii[WS(rs, 8)] = T2e + T2d;
Chris@82: 		    ri[WS(rs, 11)] = T1J - T1M;
Chris@82: 		    ii[WS(rs, 11)] = T2g + T2f;
Chris@82: 		    ri[WS(rs, 5)] = T1J + T1M;
Chris@82: 		    ii[WS(rs, 5)] = T2f - T2g;
Chris@82: 	       }
Chris@82: 	  }
Chris@82:      }
Chris@82: }
Chris@82: 
Chris@82: static const tw_instr twinstr[] = {
Chris@82:      {TW_FULL, 0, 12},
Chris@82:      {TW_NEXT, 1, 0}
Chris@82: };
Chris@82: 
Chris@82: static const ct_desc desc = { 12, "t1_12", twinstr, &GENUS, {88, 30, 30, 0}, 0, 0, 0 };
Chris@82: 
Chris@82: void X(codelet_t1_12) (planner *p) {
Chris@82:      X(kdft_dit_register) (p, t1_12, &desc);
Chris@82: }
Chris@82: #endif