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 08:04:13 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 8 -name t1_8 -include dft/scalar/t.h */
Chris@82: 
Chris@82: /*
Chris@82:  * This function contains 66 FP additions, 36 FP multiplications,
Chris@82:  * (or, 44 additions, 14 multiplications, 22 fused multiply/add),
Chris@82:  * 34 stack variables, 1 constants, and 32 memory accesses
Chris@82:  */
Chris@82: #include "dft/scalar/t.h"
Chris@82: 
Chris@82: static void t1_8(R *ri, R *ii, const R *W, stride rs, INT mb, INT me, INT ms)
Chris@82: {
Chris@82:      DK(KP707106781, +0.707106781186547524400844362104849039284835938);
Chris@82:      {
Chris@82: 	  INT m;
Chris@82: 	  for (m = mb, W = W + (mb * 14); m < me; m = m + 1, ri = ri + ms, ii = ii + ms, W = W + 14, MAKE_VOLATILE_STRIDE(16, rs)) {
Chris@82: 	       E T1, T1m, T7, T1l, Tk, TS, Te, TQ, TF, T14, TL, T16, T12, T17, Ts;
Chris@82: 	       E TX, Ty, TZ, TV, T10;
Chris@82: 	       T1 = ri[0];
Chris@82: 	       T1m = ii[0];
Chris@82: 	       {
Chris@82: 		    E T3, T6, T4, T1k, T2, T5;
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: 		    T1k = T2 * T6;
Chris@82: 		    T5 = W[7];
Chris@82: 		    T7 = FMA(T5, T6, T4);
Chris@82: 		    T1l = FNMS(T5, T3, T1k);
Chris@82: 	       }
Chris@82: 	       {
Chris@82: 		    E Tg, Tj, Th, TR, Tf, Ti;
Chris@82: 		    Tg = ri[WS(rs, 6)];
Chris@82: 		    Tj = ii[WS(rs, 6)];
Chris@82: 		    Tf = W[10];
Chris@82: 		    Th = Tf * Tg;
Chris@82: 		    TR = Tf * Tj;
Chris@82: 		    Ti = W[11];
Chris@82: 		    Tk = FMA(Ti, Tj, Th);
Chris@82: 		    TS = FNMS(Ti, Tg, TR);
Chris@82: 	       }
Chris@82: 	       {
Chris@82: 		    E Ta, Td, Tb, TP, T9, Tc;
Chris@82: 		    Ta = ri[WS(rs, 2)];
Chris@82: 		    Td = ii[WS(rs, 2)];
Chris@82: 		    T9 = W[2];
Chris@82: 		    Tb = T9 * Ta;
Chris@82: 		    TP = T9 * Td;
Chris@82: 		    Tc = W[3];
Chris@82: 		    Te = FMA(Tc, Td, Tb);
Chris@82: 		    TQ = FNMS(Tc, Ta, TP);
Chris@82: 	       }
Chris@82: 	       {
Chris@82: 		    E TB, TE, TC, T13, TH, TK, TI, T15, TA, TG, TD, TJ;
Chris@82: 		    TB = ri[WS(rs, 7)];
Chris@82: 		    TE = ii[WS(rs, 7)];
Chris@82: 		    TA = W[12];
Chris@82: 		    TC = TA * TB;
Chris@82: 		    T13 = TA * TE;
Chris@82: 		    TH = ri[WS(rs, 3)];
Chris@82: 		    TK = ii[WS(rs, 3)];
Chris@82: 		    TG = W[4];
Chris@82: 		    TI = TG * TH;
Chris@82: 		    T15 = TG * TK;
Chris@82: 		    TD = W[13];
Chris@82: 		    TF = FMA(TD, TE, TC);
Chris@82: 		    T14 = FNMS(TD, TB, T13);
Chris@82: 		    TJ = W[5];
Chris@82: 		    TL = FMA(TJ, TK, TI);
Chris@82: 		    T16 = FNMS(TJ, TH, T15);
Chris@82: 		    T12 = TF - TL;
Chris@82: 		    T17 = T14 - T16;
Chris@82: 	       }
Chris@82: 	       {
Chris@82: 		    E To, Tr, Tp, TW, Tu, Tx, Tv, TY, Tn, Tt, Tq, Tw;
Chris@82: 		    To = ri[WS(rs, 1)];
Chris@82: 		    Tr = ii[WS(rs, 1)];
Chris@82: 		    Tn = W[0];
Chris@82: 		    Tp = Tn * To;
Chris@82: 		    TW = Tn * Tr;
Chris@82: 		    Tu = ri[WS(rs, 5)];
Chris@82: 		    Tx = ii[WS(rs, 5)];
Chris@82: 		    Tt = W[8];
Chris@82: 		    Tv = Tt * Tu;
Chris@82: 		    TY = Tt * Tx;
Chris@82: 		    Tq = W[1];
Chris@82: 		    Ts = FMA(Tq, Tr, Tp);
Chris@82: 		    TX = FNMS(Tq, To, TW);
Chris@82: 		    Tw = W[9];
Chris@82: 		    Ty = FMA(Tw, Tx, Tv);
Chris@82: 		    TZ = FNMS(Tw, Tu, TY);
Chris@82: 		    TV = Ts - Ty;
Chris@82: 		    T10 = TX - TZ;
Chris@82: 	       }
Chris@82: 	       {
Chris@82: 		    E TU, T1a, T1t, T1v, T19, T1w, T1d, T1u;
Chris@82: 		    {
Chris@82: 			 E TO, TT, T1r, T1s;
Chris@82: 			 TO = T1 - T7;
Chris@82: 			 TT = TQ - TS;
Chris@82: 			 TU = TO + TT;
Chris@82: 			 T1a = TO - TT;
Chris@82: 			 T1r = T1m - T1l;
Chris@82: 			 T1s = Te - Tk;
Chris@82: 			 T1t = T1r - T1s;
Chris@82: 			 T1v = T1s + T1r;
Chris@82: 		    }
Chris@82: 		    {
Chris@82: 			 E T11, T18, T1b, T1c;
Chris@82: 			 T11 = TV + T10;
Chris@82: 			 T18 = T12 - T17;
Chris@82: 			 T19 = T11 + T18;
Chris@82: 			 T1w = T18 - T11;
Chris@82: 			 T1b = T10 - TV;
Chris@82: 			 T1c = T12 + T17;
Chris@82: 			 T1d = T1b - T1c;
Chris@82: 			 T1u = T1b + T1c;
Chris@82: 		    }
Chris@82: 		    ri[WS(rs, 5)] = FNMS(KP707106781, T19, TU);
Chris@82: 		    ii[WS(rs, 5)] = FNMS(KP707106781, T1u, T1t);
Chris@82: 		    ri[WS(rs, 1)] = FMA(KP707106781, T19, TU);
Chris@82: 		    ii[WS(rs, 1)] = FMA(KP707106781, T1u, T1t);
Chris@82: 		    ri[WS(rs, 7)] = FNMS(KP707106781, T1d, T1a);
Chris@82: 		    ii[WS(rs, 7)] = FNMS(KP707106781, T1w, T1v);
Chris@82: 		    ri[WS(rs, 3)] = FMA(KP707106781, T1d, T1a);
Chris@82: 		    ii[WS(rs, 3)] = FMA(KP707106781, T1w, T1v);
Chris@82: 	       }
Chris@82: 	       {
Chris@82: 		    E Tm, T1e, T1o, T1q, TN, T1p, T1h, T1i;
Chris@82: 		    {
Chris@82: 			 E T8, Tl, T1j, T1n;
Chris@82: 			 T8 = T1 + T7;
Chris@82: 			 Tl = Te + Tk;
Chris@82: 			 Tm = T8 + Tl;
Chris@82: 			 T1e = T8 - Tl;
Chris@82: 			 T1j = TQ + TS;
Chris@82: 			 T1n = T1l + T1m;
Chris@82: 			 T1o = T1j + T1n;
Chris@82: 			 T1q = T1n - T1j;
Chris@82: 		    }
Chris@82: 		    {
Chris@82: 			 E Tz, TM, T1f, T1g;
Chris@82: 			 Tz = Ts + Ty;
Chris@82: 			 TM = TF + TL;
Chris@82: 			 TN = Tz + TM;
Chris@82: 			 T1p = TM - Tz;
Chris@82: 			 T1f = TX + TZ;
Chris@82: 			 T1g = T14 + T16;
Chris@82: 			 T1h = T1f - T1g;
Chris@82: 			 T1i = T1f + T1g;
Chris@82: 		    }
Chris@82: 		    ri[WS(rs, 4)] = Tm - TN;
Chris@82: 		    ii[WS(rs, 4)] = T1o - T1i;
Chris@82: 		    ri[0] = Tm + TN;
Chris@82: 		    ii[0] = T1i + T1o;
Chris@82: 		    ri[WS(rs, 6)] = T1e - T1h;
Chris@82: 		    ii[WS(rs, 6)] = T1q - T1p;
Chris@82: 		    ri[WS(rs, 2)] = T1e + T1h;
Chris@82: 		    ii[WS(rs, 2)] = T1p + T1q;
Chris@82: 	       }
Chris@82: 	  }
Chris@82:      }
Chris@82: }
Chris@82: 
Chris@82: static const tw_instr twinstr[] = {
Chris@82:      {TW_FULL, 0, 8},
Chris@82:      {TW_NEXT, 1, 0}
Chris@82: };
Chris@82: 
Chris@82: static const ct_desc desc = { 8, "t1_8", twinstr, &GENUS, {44, 14, 22, 0}, 0, 0, 0 };
Chris@82: 
Chris@82: void X(codelet_t1_8) (planner *p) {
Chris@82:      X(kdft_dit_register) (p, t1_8, &desc);
Chris@82: }
Chris@82: #else
Chris@82: 
Chris@82: /* Generated by: ../../../genfft/gen_twiddle.native -compact -variables 4 -pipeline-latency 4 -n 8 -name t1_8 -include dft/scalar/t.h */
Chris@82: 
Chris@82: /*
Chris@82:  * This function contains 66 FP additions, 32 FP multiplications,
Chris@82:  * (or, 52 additions, 18 multiplications, 14 fused multiply/add),
Chris@82:  * 28 stack variables, 1 constants, and 32 memory accesses
Chris@82:  */
Chris@82: #include "dft/scalar/t.h"
Chris@82: 
Chris@82: static void t1_8(R *ri, R *ii, const R *W, stride rs, INT mb, INT me, INT ms)
Chris@82: {
Chris@82:      DK(KP707106781, +0.707106781186547524400844362104849039284835938);
Chris@82:      {
Chris@82: 	  INT m;
Chris@82: 	  for (m = mb, W = W + (mb * 14); m < me; m = m + 1, ri = ri + ms, ii = ii + ms, W = W + 14, MAKE_VOLATILE_STRIDE(16, rs)) {
Chris@82: 	       E T7, T1e, TH, T19, TF, T13, TR, TU, Ti, T1f, TK, T16, Tu, T12, TM;
Chris@82: 	       E TP;
Chris@82: 	       {
Chris@82: 		    E T1, T18, T6, T17;
Chris@82: 		    T1 = ri[0];
Chris@82: 		    T18 = 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: 			 T17 = FNMS(T4, T3, T2 * T5);
Chris@82: 		    }
Chris@82: 		    T7 = T1 + T6;
Chris@82: 		    T1e = T18 - T17;
Chris@82: 		    TH = T1 - T6;
Chris@82: 		    T19 = T17 + T18;
Chris@82: 	       }
Chris@82: 	       {
Chris@82: 		    E Tz, TS, TE, TT;
Chris@82: 		    {
Chris@82: 			 E Tw, Ty, Tv, Tx;
Chris@82: 			 Tw = ri[WS(rs, 7)];
Chris@82: 			 Ty = ii[WS(rs, 7)];
Chris@82: 			 Tv = W[12];
Chris@82: 			 Tx = W[13];
Chris@82: 			 Tz = FMA(Tv, Tw, Tx * Ty);
Chris@82: 			 TS = FNMS(Tx, Tw, Tv * Ty);
Chris@82: 		    }
Chris@82: 		    {
Chris@82: 			 E TB, TD, TA, TC;
Chris@82: 			 TB = ri[WS(rs, 3)];
Chris@82: 			 TD = ii[WS(rs, 3)];
Chris@82: 			 TA = W[4];
Chris@82: 			 TC = W[5];
Chris@82: 			 TE = FMA(TA, TB, TC * TD);
Chris@82: 			 TT = FNMS(TC, TB, TA * TD);
Chris@82: 		    }
Chris@82: 		    TF = Tz + TE;
Chris@82: 		    T13 = TS + TT;
Chris@82: 		    TR = Tz - TE;
Chris@82: 		    TU = TS - TT;
Chris@82: 	       }
Chris@82: 	       {
Chris@82: 		    E Tc, TI, Th, TJ;
Chris@82: 		    {
Chris@82: 			 E T9, Tb, T8, Ta;
Chris@82: 			 T9 = ri[WS(rs, 2)];
Chris@82: 			 Tb = ii[WS(rs, 2)];
Chris@82: 			 T8 = W[2];
Chris@82: 			 Ta = W[3];
Chris@82: 			 Tc = FMA(T8, T9, Ta * Tb);
Chris@82: 			 TI = FNMS(Ta, T9, T8 * Tb);
Chris@82: 		    }
Chris@82: 		    {
Chris@82: 			 E Te, Tg, Td, Tf;
Chris@82: 			 Te = ri[WS(rs, 6)];
Chris@82: 			 Tg = ii[WS(rs, 6)];
Chris@82: 			 Td = W[10];
Chris@82: 			 Tf = W[11];
Chris@82: 			 Th = FMA(Td, Te, Tf * Tg);
Chris@82: 			 TJ = FNMS(Tf, Te, Td * Tg);
Chris@82: 		    }
Chris@82: 		    Ti = Tc + Th;
Chris@82: 		    T1f = Tc - Th;
Chris@82: 		    TK = TI - TJ;
Chris@82: 		    T16 = TI + TJ;
Chris@82: 	       }
Chris@82: 	       {
Chris@82: 		    E To, TN, Tt, TO;
Chris@82: 		    {
Chris@82: 			 E Tl, Tn, Tk, Tm;
Chris@82: 			 Tl = ri[WS(rs, 1)];
Chris@82: 			 Tn = ii[WS(rs, 1)];
Chris@82: 			 Tk = W[0];
Chris@82: 			 Tm = W[1];
Chris@82: 			 To = FMA(Tk, Tl, Tm * Tn);
Chris@82: 			 TN = FNMS(Tm, Tl, Tk * Tn);
Chris@82: 		    }
Chris@82: 		    {
Chris@82: 			 E Tq, Ts, Tp, Tr;
Chris@82: 			 Tq = ri[WS(rs, 5)];
Chris@82: 			 Ts = ii[WS(rs, 5)];
Chris@82: 			 Tp = W[8];
Chris@82: 			 Tr = W[9];
Chris@82: 			 Tt = FMA(Tp, Tq, Tr * Ts);
Chris@82: 			 TO = FNMS(Tr, Tq, Tp * Ts);
Chris@82: 		    }
Chris@82: 		    Tu = To + Tt;
Chris@82: 		    T12 = TN + TO;
Chris@82: 		    TM = To - Tt;
Chris@82: 		    TP = TN - TO;
Chris@82: 	       }
Chris@82: 	       {
Chris@82: 		    E Tj, TG, T1b, T1c;
Chris@82: 		    Tj = T7 + Ti;
Chris@82: 		    TG = Tu + TF;
Chris@82: 		    ri[WS(rs, 4)] = Tj - TG;
Chris@82: 		    ri[0] = Tj + TG;
Chris@82: 		    {
Chris@82: 			 E T15, T1a, T11, T14;
Chris@82: 			 T15 = T12 + T13;
Chris@82: 			 T1a = T16 + T19;
Chris@82: 			 ii[0] = T15 + T1a;
Chris@82: 			 ii[WS(rs, 4)] = T1a - T15;
Chris@82: 			 T11 = T7 - Ti;
Chris@82: 			 T14 = T12 - T13;
Chris@82: 			 ri[WS(rs, 6)] = T11 - T14;
Chris@82: 			 ri[WS(rs, 2)] = T11 + T14;
Chris@82: 		    }
Chris@82: 		    T1b = TF - Tu;
Chris@82: 		    T1c = T19 - T16;
Chris@82: 		    ii[WS(rs, 2)] = T1b + T1c;
Chris@82: 		    ii[WS(rs, 6)] = T1c - T1b;
Chris@82: 		    {
Chris@82: 			 E TX, T1g, T10, T1d, TY, TZ;
Chris@82: 			 TX = TH - TK;
Chris@82: 			 T1g = T1e - T1f;
Chris@82: 			 TY = TP - TM;
Chris@82: 			 TZ = TR + TU;
Chris@82: 			 T10 = KP707106781 * (TY - TZ);
Chris@82: 			 T1d = KP707106781 * (TY + TZ);
Chris@82: 			 ri[WS(rs, 7)] = TX - T10;
Chris@82: 			 ii[WS(rs, 5)] = T1g - T1d;
Chris@82: 			 ri[WS(rs, 3)] = TX + T10;
Chris@82: 			 ii[WS(rs, 1)] = T1d + T1g;
Chris@82: 		    }
Chris@82: 		    {
Chris@82: 			 E TL, T1i, TW, T1h, TQ, TV;
Chris@82: 			 TL = TH + TK;
Chris@82: 			 T1i = T1f + T1e;
Chris@82: 			 TQ = TM + TP;
Chris@82: 			 TV = TR - TU;
Chris@82: 			 TW = KP707106781 * (TQ + TV);
Chris@82: 			 T1h = KP707106781 * (TV - TQ);
Chris@82: 			 ri[WS(rs, 5)] = TL - TW;
Chris@82: 			 ii[WS(rs, 7)] = T1i - T1h;
Chris@82: 			 ri[WS(rs, 1)] = TL + TW;
Chris@82: 			 ii[WS(rs, 3)] = T1h + T1i;
Chris@82: 		    }
Chris@82: 	       }
Chris@82: 	  }
Chris@82:      }
Chris@82: }
Chris@82: 
Chris@82: static const tw_instr twinstr[] = {
Chris@82:      {TW_FULL, 0, 8},
Chris@82:      {TW_NEXT, 1, 0}
Chris@82: };
Chris@82: 
Chris@82: static const ct_desc desc = { 8, "t1_8", twinstr, &GENUS, {52, 18, 14, 0}, 0, 0, 0 };
Chris@82: 
Chris@82: void X(codelet_t1_8) (planner *p) {
Chris@82:      X(kdft_dit_register) (p, t1_8, &desc);
Chris@82: }
Chris@82: #endif