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 08:04:15 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 16 -name t1_16 -include dft/scalar/t.h */
Chris@82: 
Chris@82: /*
Chris@82:  * This function contains 174 FP additions, 100 FP multiplications,
Chris@82:  * (or, 104 additions, 30 multiplications, 70 fused multiply/add),
Chris@82:  * 60 stack variables, 3 constants, and 64 memory accesses
Chris@82:  */
Chris@82: #include "dft/scalar/t.h"
Chris@82: 
Chris@82: static void t1_16(R *ri, R *ii, const R *W, stride rs, INT mb, INT me, INT ms)
Chris@82: {
Chris@82:      DK(KP923879532, +0.923879532511286756128183189396788286822416626);
Chris@82:      DK(KP414213562, +0.414213562373095048801688724209698078569671875);
Chris@82:      DK(KP707106781, +0.707106781186547524400844362104849039284835938);
Chris@82:      {
Chris@82: 	  INT m;
Chris@82: 	  for (m = mb, W = W + (mb * 30); m < me; m = m + 1, ri = ri + ms, ii = ii + ms, W = W + 30, MAKE_VOLATILE_STRIDE(32, rs)) {
Chris@82: 	       E T8, T3z, T1I, T3o, T1s, T35, T2o, T2r, T1F, T36, T2p, T2w, Tl, T3A, T1N;
Chris@82: 	       E T3k, Tz, T2V, T1T, T1U, T11, T30, T29, T2c, T1e, T31, T2a, T2h, TM, T2W;
Chris@82: 	       E T1W, T21;
Chris@82: 	       {
Chris@82: 		    E T1, T3n, T3, T6, T4, T3l, T2, T7, T3m, T5;
Chris@82: 		    T1 = ri[0];
Chris@82: 		    T3n = ii[0];
Chris@82: 		    T3 = ri[WS(rs, 8)];
Chris@82: 		    T6 = ii[WS(rs, 8)];
Chris@82: 		    T2 = W[14];
Chris@82: 		    T4 = T2 * T3;
Chris@82: 		    T3l = T2 * T6;
Chris@82: 		    T5 = W[15];
Chris@82: 		    T7 = FMA(T5, T6, T4);
Chris@82: 		    T3m = FNMS(T5, T3, T3l);
Chris@82: 		    T8 = T1 + T7;
Chris@82: 		    T3z = T3n - T3m;
Chris@82: 		    T1I = T1 - T7;
Chris@82: 		    T3o = T3m + T3n;
Chris@82: 	       }
Chris@82: 	       {
Chris@82: 		    E T1h, T1k, T1i, T2k, T1n, T1q, T1o, T2m, T1g, T1m;
Chris@82: 		    T1h = ri[WS(rs, 15)];
Chris@82: 		    T1k = ii[WS(rs, 15)];
Chris@82: 		    T1g = W[28];
Chris@82: 		    T1i = T1g * T1h;
Chris@82: 		    T2k = T1g * T1k;
Chris@82: 		    T1n = ri[WS(rs, 7)];
Chris@82: 		    T1q = ii[WS(rs, 7)];
Chris@82: 		    T1m = W[12];
Chris@82: 		    T1o = T1m * T1n;
Chris@82: 		    T2m = T1m * T1q;
Chris@82: 		    {
Chris@82: 			 E T1l, T2l, T1r, T2n, T1j, T1p;
Chris@82: 			 T1j = W[29];
Chris@82: 			 T1l = FMA(T1j, T1k, T1i);
Chris@82: 			 T2l = FNMS(T1j, T1h, T2k);
Chris@82: 			 T1p = W[13];
Chris@82: 			 T1r = FMA(T1p, T1q, T1o);
Chris@82: 			 T2n = FNMS(T1p, T1n, T2m);
Chris@82: 			 T1s = T1l + T1r;
Chris@82: 			 T35 = T2l + T2n;
Chris@82: 			 T2o = T2l - T2n;
Chris@82: 			 T2r = T1l - T1r;
Chris@82: 		    }
Chris@82: 	       }
Chris@82: 	       {
Chris@82: 		    E T1u, T1x, T1v, T2s, T1A, T1D, T1B, T2u, T1t, T1z;
Chris@82: 		    T1u = ri[WS(rs, 3)];
Chris@82: 		    T1x = ii[WS(rs, 3)];
Chris@82: 		    T1t = W[4];
Chris@82: 		    T1v = T1t * T1u;
Chris@82: 		    T2s = T1t * T1x;
Chris@82: 		    T1A = ri[WS(rs, 11)];
Chris@82: 		    T1D = ii[WS(rs, 11)];
Chris@82: 		    T1z = W[20];
Chris@82: 		    T1B = T1z * T1A;
Chris@82: 		    T2u = T1z * T1D;
Chris@82: 		    {
Chris@82: 			 E T1y, T2t, T1E, T2v, T1w, T1C;
Chris@82: 			 T1w = W[5];
Chris@82: 			 T1y = FMA(T1w, T1x, T1v);
Chris@82: 			 T2t = FNMS(T1w, T1u, T2s);
Chris@82: 			 T1C = W[21];
Chris@82: 			 T1E = FMA(T1C, T1D, T1B);
Chris@82: 			 T2v = FNMS(T1C, T1A, T2u);
Chris@82: 			 T1F = T1y + T1E;
Chris@82: 			 T36 = T2t + T2v;
Chris@82: 			 T2p = T1y - T1E;
Chris@82: 			 T2w = T2t - T2v;
Chris@82: 		    }
Chris@82: 	       }
Chris@82: 	       {
Chris@82: 		    E Ta, Td, Tb, T1J, Tg, Tj, Th, T1L, T9, Tf;
Chris@82: 		    Ta = ri[WS(rs, 4)];
Chris@82: 		    Td = ii[WS(rs, 4)];
Chris@82: 		    T9 = W[6];
Chris@82: 		    Tb = T9 * Ta;
Chris@82: 		    T1J = T9 * Td;
Chris@82: 		    Tg = ri[WS(rs, 12)];
Chris@82: 		    Tj = ii[WS(rs, 12)];
Chris@82: 		    Tf = W[22];
Chris@82: 		    Th = Tf * Tg;
Chris@82: 		    T1L = Tf * Tj;
Chris@82: 		    {
Chris@82: 			 E Te, T1K, Tk, T1M, Tc, Ti;
Chris@82: 			 Tc = W[7];
Chris@82: 			 Te = FMA(Tc, Td, Tb);
Chris@82: 			 T1K = FNMS(Tc, Ta, T1J);
Chris@82: 			 Ti = W[23];
Chris@82: 			 Tk = FMA(Ti, Tj, Th);
Chris@82: 			 T1M = FNMS(Ti, Tg, T1L);
Chris@82: 			 Tl = Te + Tk;
Chris@82: 			 T3A = Te - Tk;
Chris@82: 			 T1N = T1K - T1M;
Chris@82: 			 T3k = T1K + T1M;
Chris@82: 		    }
Chris@82: 	       }
Chris@82: 	       {
Chris@82: 		    E To, Tr, Tp, T1P, Tu, Tx, Tv, T1R, Tn, Tt;
Chris@82: 		    To = ri[WS(rs, 2)];
Chris@82: 		    Tr = ii[WS(rs, 2)];
Chris@82: 		    Tn = W[2];
Chris@82: 		    Tp = Tn * To;
Chris@82: 		    T1P = Tn * Tr;
Chris@82: 		    Tu = ri[WS(rs, 10)];
Chris@82: 		    Tx = ii[WS(rs, 10)];
Chris@82: 		    Tt = W[18];
Chris@82: 		    Tv = Tt * Tu;
Chris@82: 		    T1R = Tt * Tx;
Chris@82: 		    {
Chris@82: 			 E Ts, T1Q, Ty, T1S, Tq, Tw;
Chris@82: 			 Tq = W[3];
Chris@82: 			 Ts = FMA(Tq, Tr, Tp);
Chris@82: 			 T1Q = FNMS(Tq, To, T1P);
Chris@82: 			 Tw = W[19];
Chris@82: 			 Ty = FMA(Tw, Tx, Tv);
Chris@82: 			 T1S = FNMS(Tw, Tu, T1R);
Chris@82: 			 Tz = Ts + Ty;
Chris@82: 			 T2V = T1Q + T1S;
Chris@82: 			 T1T = T1Q - T1S;
Chris@82: 			 T1U = Ts - Ty;
Chris@82: 		    }
Chris@82: 	       }
Chris@82: 	       {
Chris@82: 		    E TQ, TT, TR, T25, TW, TZ, TX, T27, TP, TV;
Chris@82: 		    TQ = ri[WS(rs, 1)];
Chris@82: 		    TT = ii[WS(rs, 1)];
Chris@82: 		    TP = W[0];
Chris@82: 		    TR = TP * TQ;
Chris@82: 		    T25 = TP * TT;
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: 		    T27 = TV * TZ;
Chris@82: 		    {
Chris@82: 			 E TU, T26, T10, T28, TS, TY;
Chris@82: 			 TS = W[1];
Chris@82: 			 TU = FMA(TS, TT, TR);
Chris@82: 			 T26 = FNMS(TS, TQ, T25);
Chris@82: 			 TY = W[17];
Chris@82: 			 T10 = FMA(TY, TZ, TX);
Chris@82: 			 T28 = FNMS(TY, TW, T27);
Chris@82: 			 T11 = TU + T10;
Chris@82: 			 T30 = T26 + T28;
Chris@82: 			 T29 = T26 - T28;
Chris@82: 			 T2c = TU - T10;
Chris@82: 		    }
Chris@82: 	       }
Chris@82: 	       {
Chris@82: 		    E T13, T16, T14, T2d, T19, T1c, T1a, T2f, T12, T18;
Chris@82: 		    T13 = ri[WS(rs, 5)];
Chris@82: 		    T16 = ii[WS(rs, 5)];
Chris@82: 		    T12 = W[8];
Chris@82: 		    T14 = T12 * T13;
Chris@82: 		    T2d = T12 * T16;
Chris@82: 		    T19 = ri[WS(rs, 13)];
Chris@82: 		    T1c = ii[WS(rs, 13)];
Chris@82: 		    T18 = W[24];
Chris@82: 		    T1a = T18 * T19;
Chris@82: 		    T2f = T18 * T1c;
Chris@82: 		    {
Chris@82: 			 E T17, T2e, T1d, T2g, T15, T1b;
Chris@82: 			 T15 = W[9];
Chris@82: 			 T17 = FMA(T15, T16, T14);
Chris@82: 			 T2e = FNMS(T15, T13, T2d);
Chris@82: 			 T1b = W[25];
Chris@82: 			 T1d = FMA(T1b, T1c, T1a);
Chris@82: 			 T2g = FNMS(T1b, T19, T2f);
Chris@82: 			 T1e = T17 + T1d;
Chris@82: 			 T31 = T2e + T2g;
Chris@82: 			 T2a = T17 - T1d;
Chris@82: 			 T2h = T2e - T2g;
Chris@82: 		    }
Chris@82: 	       }
Chris@82: 	       {
Chris@82: 		    E TB, TE, TC, T1X, TH, TK, TI, T1Z, TA, TG;
Chris@82: 		    TB = ri[WS(rs, 14)];
Chris@82: 		    TE = ii[WS(rs, 14)];
Chris@82: 		    TA = W[26];
Chris@82: 		    TC = TA * TB;
Chris@82: 		    T1X = TA * TE;
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: 		    T1Z = TG * TK;
Chris@82: 		    {
Chris@82: 			 E TF, T1Y, TL, T20, TD, TJ;
Chris@82: 			 TD = W[27];
Chris@82: 			 TF = FMA(TD, TE, TC);
Chris@82: 			 T1Y = FNMS(TD, TB, T1X);
Chris@82: 			 TJ = W[11];
Chris@82: 			 TL = FMA(TJ, TK, TI);
Chris@82: 			 T20 = FNMS(TJ, TH, T1Z);
Chris@82: 			 TM = TF + TL;
Chris@82: 			 T2W = T1Y + T20;
Chris@82: 			 T1W = TF - TL;
Chris@82: 			 T21 = T1Y - T20;
Chris@82: 		    }
Chris@82: 	       }
Chris@82: 	       {
Chris@82: 		    E TO, T3e, T3q, T3s, T1H, T3r, T3h, T3i;
Chris@82: 		    {
Chris@82: 			 E Tm, TN, T3j, T3p;
Chris@82: 			 Tm = T8 + Tl;
Chris@82: 			 TN = Tz + TM;
Chris@82: 			 TO = Tm + TN;
Chris@82: 			 T3e = Tm - TN;
Chris@82: 			 T3j = T2V + T2W;
Chris@82: 			 T3p = T3k + T3o;
Chris@82: 			 T3q = T3j + T3p;
Chris@82: 			 T3s = T3p - T3j;
Chris@82: 		    }
Chris@82: 		    {
Chris@82: 			 E T1f, T1G, T3f, T3g;
Chris@82: 			 T1f = T11 + T1e;
Chris@82: 			 T1G = T1s + T1F;
Chris@82: 			 T1H = T1f + T1G;
Chris@82: 			 T3r = T1G - T1f;
Chris@82: 			 T3f = T30 + T31;
Chris@82: 			 T3g = T35 + T36;
Chris@82: 			 T3h = T3f - T3g;
Chris@82: 			 T3i = T3f + T3g;
Chris@82: 		    }
Chris@82: 		    ri[WS(rs, 8)] = TO - T1H;
Chris@82: 		    ii[WS(rs, 8)] = T3q - T3i;
Chris@82: 		    ri[0] = TO + T1H;
Chris@82: 		    ii[0] = T3i + T3q;
Chris@82: 		    ri[WS(rs, 12)] = T3e - T3h;
Chris@82: 		    ii[WS(rs, 12)] = T3s - T3r;
Chris@82: 		    ri[WS(rs, 4)] = T3e + T3h;
Chris@82: 		    ii[WS(rs, 4)] = T3r + T3s;
Chris@82: 	       }
Chris@82: 	       {
Chris@82: 		    E T2Y, T3a, T3v, T3x, T33, T3b, T38, T3c;
Chris@82: 		    {
Chris@82: 			 E T2U, T2X, T3t, T3u;
Chris@82: 			 T2U = T8 - Tl;
Chris@82: 			 T2X = T2V - T2W;
Chris@82: 			 T2Y = T2U + T2X;
Chris@82: 			 T3a = T2U - T2X;
Chris@82: 			 T3t = TM - Tz;
Chris@82: 			 T3u = T3o - T3k;
Chris@82: 			 T3v = T3t + T3u;
Chris@82: 			 T3x = T3u - T3t;
Chris@82: 		    }
Chris@82: 		    {
Chris@82: 			 E T2Z, T32, T34, T37;
Chris@82: 			 T2Z = T11 - T1e;
Chris@82: 			 T32 = T30 - T31;
Chris@82: 			 T33 = T2Z + T32;
Chris@82: 			 T3b = T32 - T2Z;
Chris@82: 			 T34 = T1s - T1F;
Chris@82: 			 T37 = T35 - T36;
Chris@82: 			 T38 = T34 - T37;
Chris@82: 			 T3c = T34 + T37;
Chris@82: 		    }
Chris@82: 		    {
Chris@82: 			 E T39, T3w, T3d, T3y;
Chris@82: 			 T39 = T33 + T38;
Chris@82: 			 ri[WS(rs, 10)] = FNMS(KP707106781, T39, T2Y);
Chris@82: 			 ri[WS(rs, 2)] = FMA(KP707106781, T39, T2Y);
Chris@82: 			 T3w = T3b + T3c;
Chris@82: 			 ii[WS(rs, 2)] = FMA(KP707106781, T3w, T3v);
Chris@82: 			 ii[WS(rs, 10)] = FNMS(KP707106781, T3w, T3v);
Chris@82: 			 T3d = T3b - T3c;
Chris@82: 			 ri[WS(rs, 14)] = FNMS(KP707106781, T3d, T3a);
Chris@82: 			 ri[WS(rs, 6)] = FMA(KP707106781, T3d, T3a);
Chris@82: 			 T3y = T38 - T33;
Chris@82: 			 ii[WS(rs, 6)] = FMA(KP707106781, T3y, T3x);
Chris@82: 			 ii[WS(rs, 14)] = FNMS(KP707106781, T3y, T3x);
Chris@82: 		    }
Chris@82: 	       }
Chris@82: 	       {
Chris@82: 		    E T1O, T3B, T3H, T2E, T23, T3C, T2O, T2S, T2H, T3I, T2j, T2B, T2L, T2R, T2y;
Chris@82: 		    E T2C;
Chris@82: 		    {
Chris@82: 			 E T1V, T22, T2b, T2i;
Chris@82: 			 T1O = T1I - T1N;
Chris@82: 			 T3B = T3z - T3A;
Chris@82: 			 T3H = T3A + T3z;
Chris@82: 			 T2E = T1I + T1N;
Chris@82: 			 T1V = T1T - T1U;
Chris@82: 			 T22 = T1W + T21;
Chris@82: 			 T23 = T1V - T22;
Chris@82: 			 T3C = T1V + T22;
Chris@82: 			 {
Chris@82: 			      E T2M, T2N, T2F, T2G;
Chris@82: 			      T2M = T2r + T2w;
Chris@82: 			      T2N = T2o - T2p;
Chris@82: 			      T2O = FNMS(KP414213562, T2N, T2M);
Chris@82: 			      T2S = FMA(KP414213562, T2M, T2N);
Chris@82: 			      T2F = T1U + T1T;
Chris@82: 			      T2G = T1W - T21;
Chris@82: 			      T2H = T2F + T2G;
Chris@82: 			      T3I = T2G - T2F;
Chris@82: 			 }
Chris@82: 			 T2b = T29 + T2a;
Chris@82: 			 T2i = T2c - T2h;
Chris@82: 			 T2j = FMA(KP414213562, T2i, T2b);
Chris@82: 			 T2B = FNMS(KP414213562, T2b, T2i);
Chris@82: 			 {
Chris@82: 			      E T2J, T2K, T2q, T2x;
Chris@82: 			      T2J = T2c + T2h;
Chris@82: 			      T2K = T29 - T2a;
Chris@82: 			      T2L = FMA(KP414213562, T2K, T2J);
Chris@82: 			      T2R = FNMS(KP414213562, T2J, T2K);
Chris@82: 			      T2q = T2o + T2p;
Chris@82: 			      T2x = T2r - T2w;
Chris@82: 			      T2y = FNMS(KP414213562, T2x, T2q);
Chris@82: 			      T2C = FMA(KP414213562, T2q, T2x);
Chris@82: 			 }
Chris@82: 		    }
Chris@82: 		    {
Chris@82: 			 E T24, T2z, T3J, T3K;
Chris@82: 			 T24 = FMA(KP707106781, T23, T1O);
Chris@82: 			 T2z = T2j - T2y;
Chris@82: 			 ri[WS(rs, 11)] = FNMS(KP923879532, T2z, T24);
Chris@82: 			 ri[WS(rs, 3)] = FMA(KP923879532, T2z, T24);
Chris@82: 			 T3J = FMA(KP707106781, T3I, T3H);
Chris@82: 			 T3K = T2C - T2B;
Chris@82: 			 ii[WS(rs, 3)] = FMA(KP923879532, T3K, T3J);
Chris@82: 			 ii[WS(rs, 11)] = FNMS(KP923879532, T3K, T3J);
Chris@82: 		    }
Chris@82: 		    {
Chris@82: 			 E T2A, T2D, T3L, T3M;
Chris@82: 			 T2A = FNMS(KP707106781, T23, T1O);
Chris@82: 			 T2D = T2B + T2C;
Chris@82: 			 ri[WS(rs, 7)] = FNMS(KP923879532, T2D, T2A);
Chris@82: 			 ri[WS(rs, 15)] = FMA(KP923879532, T2D, T2A);
Chris@82: 			 T3L = FNMS(KP707106781, T3I, T3H);
Chris@82: 			 T3M = T2j + T2y;
Chris@82: 			 ii[WS(rs, 7)] = FNMS(KP923879532, T3M, T3L);
Chris@82: 			 ii[WS(rs, 15)] = FMA(KP923879532, T3M, T3L);
Chris@82: 		    }
Chris@82: 		    {
Chris@82: 			 E T2I, T2P, T3D, T3E;
Chris@82: 			 T2I = FMA(KP707106781, T2H, T2E);
Chris@82: 			 T2P = T2L + T2O;
Chris@82: 			 ri[WS(rs, 9)] = FNMS(KP923879532, T2P, T2I);
Chris@82: 			 ri[WS(rs, 1)] = FMA(KP923879532, T2P, T2I);
Chris@82: 			 T3D = FMA(KP707106781, T3C, T3B);
Chris@82: 			 T3E = T2R + T2S;
Chris@82: 			 ii[WS(rs, 1)] = FMA(KP923879532, T3E, T3D);
Chris@82: 			 ii[WS(rs, 9)] = FNMS(KP923879532, T3E, T3D);
Chris@82: 		    }
Chris@82: 		    {
Chris@82: 			 E T2Q, T2T, T3F, T3G;
Chris@82: 			 T2Q = FNMS(KP707106781, T2H, T2E);
Chris@82: 			 T2T = T2R - T2S;
Chris@82: 			 ri[WS(rs, 13)] = FNMS(KP923879532, T2T, T2Q);
Chris@82: 			 ri[WS(rs, 5)] = FMA(KP923879532, T2T, T2Q);
Chris@82: 			 T3F = FNMS(KP707106781, T3C, T3B);
Chris@82: 			 T3G = T2O - T2L;
Chris@82: 			 ii[WS(rs, 5)] = FMA(KP923879532, T3G, T3F);
Chris@82: 			 ii[WS(rs, 13)] = FNMS(KP923879532, T3G, T3F);
Chris@82: 		    }
Chris@82: 	       }
Chris@82: 	  }
Chris@82:      }
Chris@82: }
Chris@82: 
Chris@82: static const tw_instr twinstr[] = {
Chris@82:      {TW_FULL, 0, 16},
Chris@82:      {TW_NEXT, 1, 0}
Chris@82: };
Chris@82: 
Chris@82: static const ct_desc desc = { 16, "t1_16", twinstr, &GENUS, {104, 30, 70, 0}, 0, 0, 0 };
Chris@82: 
Chris@82: void X(codelet_t1_16) (planner *p) {
Chris@82:      X(kdft_dit_register) (p, t1_16, &desc);
Chris@82: }
Chris@82: #else
Chris@82: 
Chris@82: /* Generated by: ../../../genfft/gen_twiddle.native -compact -variables 4 -pipeline-latency 4 -n 16 -name t1_16 -include dft/scalar/t.h */
Chris@82: 
Chris@82: /*
Chris@82:  * This function contains 174 FP additions, 84 FP multiplications,
Chris@82:  * (or, 136 additions, 46 multiplications, 38 fused multiply/add),
Chris@82:  * 52 stack variables, 3 constants, and 64 memory accesses
Chris@82:  */
Chris@82: #include "dft/scalar/t.h"
Chris@82: 
Chris@82: static void t1_16(R *ri, R *ii, const R *W, stride rs, INT mb, INT me, INT ms)
Chris@82: {
Chris@82:      DK(KP382683432, +0.382683432365089771728459984030398866761344562);
Chris@82:      DK(KP923879532, +0.923879532511286756128183189396788286822416626);
Chris@82:      DK(KP707106781, +0.707106781186547524400844362104849039284835938);
Chris@82:      {
Chris@82: 	  INT m;
Chris@82: 	  for (m = mb, W = W + (mb * 30); m < me; m = m + 1, ri = ri + ms, ii = ii + ms, W = W + 30, MAKE_VOLATILE_STRIDE(32, rs)) {
Chris@82: 	       E T7, T37, T1t, T2U, Ti, T38, T1w, T2R, Tu, T2s, T1C, T2c, TF, T2t, T1H;
Chris@82: 	       E T2d, T1f, T1q, T2B, T2C, T2D, T2E, T1Z, T2j, T24, T2k, TS, T13, T2w, T2x;
Chris@82: 	       E T2y, T2z, T1O, T2g, T1T, T2h;
Chris@82: 	       {
Chris@82: 		    E T1, T2T, T6, T2S;
Chris@82: 		    T1 = ri[0];
Chris@82: 		    T2T = ii[0];
Chris@82: 		    {
Chris@82: 			 E T3, T5, T2, T4;
Chris@82: 			 T3 = ri[WS(rs, 8)];
Chris@82: 			 T5 = ii[WS(rs, 8)];
Chris@82: 			 T2 = W[14];
Chris@82: 			 T4 = W[15];
Chris@82: 			 T6 = FMA(T2, T3, T4 * T5);
Chris@82: 			 T2S = FNMS(T4, T3, T2 * T5);
Chris@82: 		    }
Chris@82: 		    T7 = T1 + T6;
Chris@82: 		    T37 = T2T - T2S;
Chris@82: 		    T1t = T1 - T6;
Chris@82: 		    T2U = T2S + T2T;
Chris@82: 	       }
Chris@82: 	       {
Chris@82: 		    E Tc, T1u, Th, T1v;
Chris@82: 		    {
Chris@82: 			 E T9, Tb, T8, Ta;
Chris@82: 			 T9 = ri[WS(rs, 4)];
Chris@82: 			 Tb = ii[WS(rs, 4)];
Chris@82: 			 T8 = W[6];
Chris@82: 			 Ta = W[7];
Chris@82: 			 Tc = FMA(T8, T9, Ta * Tb);
Chris@82: 			 T1u = FNMS(Ta, T9, T8 * Tb);
Chris@82: 		    }
Chris@82: 		    {
Chris@82: 			 E Te, Tg, Td, Tf;
Chris@82: 			 Te = ri[WS(rs, 12)];
Chris@82: 			 Tg = ii[WS(rs, 12)];
Chris@82: 			 Td = W[22];
Chris@82: 			 Tf = W[23];
Chris@82: 			 Th = FMA(Td, Te, Tf * Tg);
Chris@82: 			 T1v = FNMS(Tf, Te, Td * Tg);
Chris@82: 		    }
Chris@82: 		    Ti = Tc + Th;
Chris@82: 		    T38 = Tc - Th;
Chris@82: 		    T1w = T1u - T1v;
Chris@82: 		    T2R = T1u + T1v;
Chris@82: 	       }
Chris@82: 	       {
Chris@82: 		    E To, T1y, Tt, T1z, T1A, T1B;
Chris@82: 		    {
Chris@82: 			 E Tl, Tn, Tk, Tm;
Chris@82: 			 Tl = ri[WS(rs, 2)];
Chris@82: 			 Tn = ii[WS(rs, 2)];
Chris@82: 			 Tk = W[2];
Chris@82: 			 Tm = W[3];
Chris@82: 			 To = FMA(Tk, Tl, Tm * Tn);
Chris@82: 			 T1y = FNMS(Tm, Tl, Tk * Tn);
Chris@82: 		    }
Chris@82: 		    {
Chris@82: 			 E Tq, Ts, Tp, Tr;
Chris@82: 			 Tq = ri[WS(rs, 10)];
Chris@82: 			 Ts = ii[WS(rs, 10)];
Chris@82: 			 Tp = W[18];
Chris@82: 			 Tr = W[19];
Chris@82: 			 Tt = FMA(Tp, Tq, Tr * Ts);
Chris@82: 			 T1z = FNMS(Tr, Tq, Tp * Ts);
Chris@82: 		    }
Chris@82: 		    Tu = To + Tt;
Chris@82: 		    T2s = T1y + T1z;
Chris@82: 		    T1A = T1y - T1z;
Chris@82: 		    T1B = To - Tt;
Chris@82: 		    T1C = T1A - T1B;
Chris@82: 		    T2c = T1B + T1A;
Chris@82: 	       }
Chris@82: 	       {
Chris@82: 		    E Tz, T1E, TE, T1F, T1D, T1G;
Chris@82: 		    {
Chris@82: 			 E Tw, Ty, Tv, Tx;
Chris@82: 			 Tw = ri[WS(rs, 14)];
Chris@82: 			 Ty = ii[WS(rs, 14)];
Chris@82: 			 Tv = W[26];
Chris@82: 			 Tx = W[27];
Chris@82: 			 Tz = FMA(Tv, Tw, Tx * Ty);
Chris@82: 			 T1E = FNMS(Tx, Tw, Tv * Ty);
Chris@82: 		    }
Chris@82: 		    {
Chris@82: 			 E TB, TD, TA, TC;
Chris@82: 			 TB = ri[WS(rs, 6)];
Chris@82: 			 TD = ii[WS(rs, 6)];
Chris@82: 			 TA = W[10];
Chris@82: 			 TC = W[11];
Chris@82: 			 TE = FMA(TA, TB, TC * TD);
Chris@82: 			 T1F = FNMS(TC, TB, TA * TD);
Chris@82: 		    }
Chris@82: 		    TF = Tz + TE;
Chris@82: 		    T2t = T1E + T1F;
Chris@82: 		    T1D = Tz - TE;
Chris@82: 		    T1G = T1E - T1F;
Chris@82: 		    T1H = T1D + T1G;
Chris@82: 		    T2d = T1D - T1G;
Chris@82: 	       }
Chris@82: 	       {
Chris@82: 		    E T19, T20, T1p, T1X, T1e, T21, T1k, T1W;
Chris@82: 		    {
Chris@82: 			 E T16, T18, T15, T17;
Chris@82: 			 T16 = ri[WS(rs, 15)];
Chris@82: 			 T18 = ii[WS(rs, 15)];
Chris@82: 			 T15 = W[28];
Chris@82: 			 T17 = W[29];
Chris@82: 			 T19 = FMA(T15, T16, T17 * T18);
Chris@82: 			 T20 = FNMS(T17, T16, T15 * T18);
Chris@82: 		    }
Chris@82: 		    {
Chris@82: 			 E T1m, T1o, T1l, T1n;
Chris@82: 			 T1m = ri[WS(rs, 11)];
Chris@82: 			 T1o = ii[WS(rs, 11)];
Chris@82: 			 T1l = W[20];
Chris@82: 			 T1n = W[21];
Chris@82: 			 T1p = FMA(T1l, T1m, T1n * T1o);
Chris@82: 			 T1X = FNMS(T1n, T1m, T1l * T1o);
Chris@82: 		    }
Chris@82: 		    {
Chris@82: 			 E T1b, T1d, T1a, T1c;
Chris@82: 			 T1b = ri[WS(rs, 7)];
Chris@82: 			 T1d = ii[WS(rs, 7)];
Chris@82: 			 T1a = W[12];
Chris@82: 			 T1c = W[13];
Chris@82: 			 T1e = FMA(T1a, T1b, T1c * T1d);
Chris@82: 			 T21 = FNMS(T1c, T1b, T1a * T1d);
Chris@82: 		    }
Chris@82: 		    {
Chris@82: 			 E T1h, T1j, T1g, T1i;
Chris@82: 			 T1h = ri[WS(rs, 3)];
Chris@82: 			 T1j = ii[WS(rs, 3)];
Chris@82: 			 T1g = W[4];
Chris@82: 			 T1i = W[5];
Chris@82: 			 T1k = FMA(T1g, T1h, T1i * T1j);
Chris@82: 			 T1W = FNMS(T1i, T1h, T1g * T1j);
Chris@82: 		    }
Chris@82: 		    T1f = T19 + T1e;
Chris@82: 		    T1q = T1k + T1p;
Chris@82: 		    T2B = T1f - T1q;
Chris@82: 		    T2C = T20 + T21;
Chris@82: 		    T2D = T1W + T1X;
Chris@82: 		    T2E = T2C - T2D;
Chris@82: 		    {
Chris@82: 			 E T1V, T1Y, T22, T23;
Chris@82: 			 T1V = T19 - T1e;
Chris@82: 			 T1Y = T1W - T1X;
Chris@82: 			 T1Z = T1V - T1Y;
Chris@82: 			 T2j = T1V + T1Y;
Chris@82: 			 T22 = T20 - T21;
Chris@82: 			 T23 = T1k - T1p;
Chris@82: 			 T24 = T22 + T23;
Chris@82: 			 T2k = T22 - T23;
Chris@82: 		    }
Chris@82: 	       }
Chris@82: 	       {
Chris@82: 		    E TM, T1K, T12, T1R, TR, T1L, TX, T1Q;
Chris@82: 		    {
Chris@82: 			 E TJ, TL, TI, TK;
Chris@82: 			 TJ = ri[WS(rs, 1)];
Chris@82: 			 TL = ii[WS(rs, 1)];
Chris@82: 			 TI = W[0];
Chris@82: 			 TK = W[1];
Chris@82: 			 TM = FMA(TI, TJ, TK * TL);
Chris@82: 			 T1K = FNMS(TK, TJ, TI * TL);
Chris@82: 		    }
Chris@82: 		    {
Chris@82: 			 E TZ, T11, TY, T10;
Chris@82: 			 TZ = ri[WS(rs, 13)];
Chris@82: 			 T11 = ii[WS(rs, 13)];
Chris@82: 			 TY = W[24];
Chris@82: 			 T10 = W[25];
Chris@82: 			 T12 = FMA(TY, TZ, T10 * T11);
Chris@82: 			 T1R = FNMS(T10, TZ, TY * T11);
Chris@82: 		    }
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: 			 T1L = FNMS(TP, TO, TN * TQ);
Chris@82: 		    }
Chris@82: 		    {
Chris@82: 			 E TU, TW, TT, TV;
Chris@82: 			 TU = ri[WS(rs, 5)];
Chris@82: 			 TW = ii[WS(rs, 5)];
Chris@82: 			 TT = W[8];
Chris@82: 			 TV = W[9];
Chris@82: 			 TX = FMA(TT, TU, TV * TW);
Chris@82: 			 T1Q = FNMS(TV, TU, TT * TW);
Chris@82: 		    }
Chris@82: 		    TS = TM + TR;
Chris@82: 		    T13 = TX + T12;
Chris@82: 		    T2w = TS - T13;
Chris@82: 		    T2x = T1K + T1L;
Chris@82: 		    T2y = T1Q + T1R;
Chris@82: 		    T2z = T2x - T2y;
Chris@82: 		    {
Chris@82: 			 E T1M, T1N, T1P, T1S;
Chris@82: 			 T1M = T1K - T1L;
Chris@82: 			 T1N = TX - T12;
Chris@82: 			 T1O = T1M + T1N;
Chris@82: 			 T2g = T1M - T1N;
Chris@82: 			 T1P = TM - TR;
Chris@82: 			 T1S = T1Q - T1R;
Chris@82: 			 T1T = T1P - T1S;
Chris@82: 			 T2h = T1P + T1S;
Chris@82: 		    }
Chris@82: 	       }
Chris@82: 	       {
Chris@82: 		    E T1J, T27, T3g, T3i, T26, T3h, T2a, T3d;
Chris@82: 		    {
Chris@82: 			 E T1x, T1I, T3e, T3f;
Chris@82: 			 T1x = T1t - T1w;
Chris@82: 			 T1I = KP707106781 * (T1C - T1H);
Chris@82: 			 T1J = T1x + T1I;
Chris@82: 			 T27 = T1x - T1I;
Chris@82: 			 T3e = KP707106781 * (T2d - T2c);
Chris@82: 			 T3f = T38 + T37;
Chris@82: 			 T3g = T3e + T3f;
Chris@82: 			 T3i = T3f - T3e;
Chris@82: 		    }
Chris@82: 		    {
Chris@82: 			 E T1U, T25, T28, T29;
Chris@82: 			 T1U = FMA(KP923879532, T1O, KP382683432 * T1T);
Chris@82: 			 T25 = FNMS(KP923879532, T24, KP382683432 * T1Z);
Chris@82: 			 T26 = T1U + T25;
Chris@82: 			 T3h = T25 - T1U;
Chris@82: 			 T28 = FNMS(KP923879532, T1T, KP382683432 * T1O);
Chris@82: 			 T29 = FMA(KP382683432, T24, KP923879532 * T1Z);
Chris@82: 			 T2a = T28 - T29;
Chris@82: 			 T3d = T28 + T29;
Chris@82: 		    }
Chris@82: 		    ri[WS(rs, 11)] = T1J - T26;
Chris@82: 		    ii[WS(rs, 11)] = T3g - T3d;
Chris@82: 		    ri[WS(rs, 3)] = T1J + T26;
Chris@82: 		    ii[WS(rs, 3)] = T3d + T3g;
Chris@82: 		    ri[WS(rs, 15)] = T27 - T2a;
Chris@82: 		    ii[WS(rs, 15)] = T3i - T3h;
Chris@82: 		    ri[WS(rs, 7)] = T27 + T2a;
Chris@82: 		    ii[WS(rs, 7)] = T3h + T3i;
Chris@82: 	       }
Chris@82: 	       {
Chris@82: 		    E T2v, T2H, T32, T34, T2G, T33, T2K, T2Z;
Chris@82: 		    {
Chris@82: 			 E T2r, T2u, T30, T31;
Chris@82: 			 T2r = T7 - Ti;
Chris@82: 			 T2u = T2s - T2t;
Chris@82: 			 T2v = T2r + T2u;
Chris@82: 			 T2H = T2r - T2u;
Chris@82: 			 T30 = TF - Tu;
Chris@82: 			 T31 = T2U - T2R;
Chris@82: 			 T32 = T30 + T31;
Chris@82: 			 T34 = T31 - T30;
Chris@82: 		    }
Chris@82: 		    {
Chris@82: 			 E T2A, T2F, T2I, T2J;
Chris@82: 			 T2A = T2w + T2z;
Chris@82: 			 T2F = T2B - T2E;
Chris@82: 			 T2G = KP707106781 * (T2A + T2F);
Chris@82: 			 T33 = KP707106781 * (T2F - T2A);
Chris@82: 			 T2I = T2z - T2w;
Chris@82: 			 T2J = T2B + T2E;
Chris@82: 			 T2K = KP707106781 * (T2I - T2J);
Chris@82: 			 T2Z = KP707106781 * (T2I + T2J);
Chris@82: 		    }
Chris@82: 		    ri[WS(rs, 10)] = T2v - T2G;
Chris@82: 		    ii[WS(rs, 10)] = T32 - T2Z;
Chris@82: 		    ri[WS(rs, 2)] = T2v + T2G;
Chris@82: 		    ii[WS(rs, 2)] = T2Z + T32;
Chris@82: 		    ri[WS(rs, 14)] = T2H - T2K;
Chris@82: 		    ii[WS(rs, 14)] = T34 - T33;
Chris@82: 		    ri[WS(rs, 6)] = T2H + T2K;
Chris@82: 		    ii[WS(rs, 6)] = T33 + T34;
Chris@82: 	       }
Chris@82: 	       {
Chris@82: 		    E T2f, T2n, T3a, T3c, T2m, T3b, T2q, T35;
Chris@82: 		    {
Chris@82: 			 E T2b, T2e, T36, T39;
Chris@82: 			 T2b = T1t + T1w;
Chris@82: 			 T2e = KP707106781 * (T2c + T2d);
Chris@82: 			 T2f = T2b + T2e;
Chris@82: 			 T2n = T2b - T2e;
Chris@82: 			 T36 = KP707106781 * (T1C + T1H);
Chris@82: 			 T39 = T37 - T38;
Chris@82: 			 T3a = T36 + T39;
Chris@82: 			 T3c = T39 - T36;
Chris@82: 		    }
Chris@82: 		    {
Chris@82: 			 E T2i, T2l, T2o, T2p;
Chris@82: 			 T2i = FMA(KP382683432, T2g, KP923879532 * T2h);
Chris@82: 			 T2l = FNMS(KP382683432, T2k, KP923879532 * T2j);
Chris@82: 			 T2m = T2i + T2l;
Chris@82: 			 T3b = T2l - T2i;
Chris@82: 			 T2o = FNMS(KP382683432, T2h, KP923879532 * T2g);
Chris@82: 			 T2p = FMA(KP923879532, T2k, KP382683432 * T2j);
Chris@82: 			 T2q = T2o - T2p;
Chris@82: 			 T35 = T2o + T2p;
Chris@82: 		    }
Chris@82: 		    ri[WS(rs, 9)] = T2f - T2m;
Chris@82: 		    ii[WS(rs, 9)] = T3a - T35;
Chris@82: 		    ri[WS(rs, 1)] = T2f + T2m;
Chris@82: 		    ii[WS(rs, 1)] = T35 + T3a;
Chris@82: 		    ri[WS(rs, 13)] = T2n - T2q;
Chris@82: 		    ii[WS(rs, 13)] = T3c - T3b;
Chris@82: 		    ri[WS(rs, 5)] = T2n + T2q;
Chris@82: 		    ii[WS(rs, 5)] = T3b + T3c;
Chris@82: 	       }
Chris@82: 	       {
Chris@82: 		    E TH, T2L, T2W, T2Y, T1s, T2X, T2O, T2P;
Chris@82: 		    {
Chris@82: 			 E Tj, TG, T2Q, T2V;
Chris@82: 			 Tj = T7 + Ti;
Chris@82: 			 TG = Tu + TF;
Chris@82: 			 TH = Tj + TG;
Chris@82: 			 T2L = Tj - TG;
Chris@82: 			 T2Q = T2s + T2t;
Chris@82: 			 T2V = T2R + T2U;
Chris@82: 			 T2W = T2Q + T2V;
Chris@82: 			 T2Y = T2V - T2Q;
Chris@82: 		    }
Chris@82: 		    {
Chris@82: 			 E T14, T1r, T2M, T2N;
Chris@82: 			 T14 = TS + T13;
Chris@82: 			 T1r = T1f + T1q;
Chris@82: 			 T1s = T14 + T1r;
Chris@82: 			 T2X = T1r - T14;
Chris@82: 			 T2M = T2x + T2y;
Chris@82: 			 T2N = T2C + T2D;
Chris@82: 			 T2O = T2M - T2N;
Chris@82: 			 T2P = T2M + T2N;
Chris@82: 		    }
Chris@82: 		    ri[WS(rs, 8)] = TH - T1s;
Chris@82: 		    ii[WS(rs, 8)] = T2W - T2P;
Chris@82: 		    ri[0] = TH + T1s;
Chris@82: 		    ii[0] = T2P + T2W;
Chris@82: 		    ri[WS(rs, 12)] = T2L - T2O;
Chris@82: 		    ii[WS(rs, 12)] = T2Y - T2X;
Chris@82: 		    ri[WS(rs, 4)] = T2L + T2O;
Chris@82: 		    ii[WS(rs, 4)] = T2X + T2Y;
Chris@82: 	       }
Chris@82: 	  }
Chris@82:      }
Chris@82: }
Chris@82: 
Chris@82: static const tw_instr twinstr[] = {
Chris@82:      {TW_FULL, 0, 16},
Chris@82:      {TW_NEXT, 1, 0}
Chris@82: };
Chris@82: 
Chris@82: static const ct_desc desc = { 16, "t1_16", twinstr, &GENUS, {136, 46, 38, 0}, 0, 0, 0 };
Chris@82: 
Chris@82: void X(codelet_t1_16) (planner *p) {
Chris@82:      X(kdft_dit_register) (p, t1_16, &desc);
Chris@82: }
Chris@82: #endif