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 08:04:19 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 -twiddle-log3 -precompute-twiddles -n 8 -name t2_8 -include dft/scalar/t.h */
Chris@82: 
Chris@82: /*
Chris@82:  * This function contains 74 FP additions, 50 FP multiplications,
Chris@82:  * (or, 44 additions, 20 multiplications, 30 fused multiply/add),
Chris@82:  * 48 stack variables, 1 constants, and 32 memory accesses
Chris@82:  */
Chris@82: #include "dft/scalar/t.h"
Chris@82: 
Chris@82: static void t2_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 * 6); m < me; m = m + 1, ri = ri + ms, ii = ii + ms, W = W + 6, MAKE_VOLATILE_STRIDE(16, rs)) {
Chris@82: 	       E T2, T3, Tl, Tn, T5, T6, Tf, T7, Ts, Tb, To, Ti, TC, TG;
Chris@82: 	       {
Chris@82: 		    E T4, Tm, Tr, Ta, TB, TF;
Chris@82: 		    T2 = W[0];
Chris@82: 		    T3 = W[2];
Chris@82: 		    T4 = T2 * T3;
Chris@82: 		    Tl = W[4];
Chris@82: 		    Tm = T2 * Tl;
Chris@82: 		    Tn = W[5];
Chris@82: 		    Tr = T2 * Tn;
Chris@82: 		    T5 = W[1];
Chris@82: 		    T6 = W[3];
Chris@82: 		    Ta = T2 * T6;
Chris@82: 		    Tf = FMA(T5, T6, T4);
Chris@82: 		    T7 = FNMS(T5, T6, T4);
Chris@82: 		    Ts = FNMS(T5, Tl, Tr);
Chris@82: 		    Tb = FMA(T5, T3, Ta);
Chris@82: 		    To = FMA(T5, Tn, Tm);
Chris@82: 		    TB = Tf * Tl;
Chris@82: 		    TF = Tf * Tn;
Chris@82: 		    Ti = FNMS(T5, T3, Ta);
Chris@82: 		    TC = FMA(Ti, Tn, TB);
Chris@82: 		    TG = FNMS(Ti, Tl, TF);
Chris@82: 	       }
Chris@82: 	       {
Chris@82: 		    E T1, T1s, Td, T1r, Tu, TY, Tk, TW, TN, TR, T18, T1a, T1c, T1d, TA;
Chris@82: 		    E TI, T11, T13, T15, T16;
Chris@82: 		    T1 = ri[0];
Chris@82: 		    T1s = ii[0];
Chris@82: 		    {
Chris@82: 			 E T8, T9, Tc, T1q;
Chris@82: 			 T8 = ri[WS(rs, 4)];
Chris@82: 			 T9 = T7 * T8;
Chris@82: 			 Tc = ii[WS(rs, 4)];
Chris@82: 			 T1q = T7 * Tc;
Chris@82: 			 Td = FMA(Tb, Tc, T9);
Chris@82: 			 T1r = FNMS(Tb, T8, T1q);
Chris@82: 		    }
Chris@82: 		    {
Chris@82: 			 E Tp, Tq, Tt, TX;
Chris@82: 			 Tp = ri[WS(rs, 6)];
Chris@82: 			 Tq = To * Tp;
Chris@82: 			 Tt = ii[WS(rs, 6)];
Chris@82: 			 TX = To * Tt;
Chris@82: 			 Tu = FMA(Ts, Tt, Tq);
Chris@82: 			 TY = FNMS(Ts, Tp, TX);
Chris@82: 		    }
Chris@82: 		    {
Chris@82: 			 E Tg, Th, Tj, TV;
Chris@82: 			 Tg = ri[WS(rs, 2)];
Chris@82: 			 Th = Tf * Tg;
Chris@82: 			 Tj = ii[WS(rs, 2)];
Chris@82: 			 TV = Tf * Tj;
Chris@82: 			 Tk = FMA(Ti, Tj, Th);
Chris@82: 			 TW = FNMS(Ti, Tg, TV);
Chris@82: 		    }
Chris@82: 		    {
Chris@82: 			 E TK, TL, TM, T19, TO, TP, TQ, T1b;
Chris@82: 			 TK = ri[WS(rs, 7)];
Chris@82: 			 TL = Tl * TK;
Chris@82: 			 TM = ii[WS(rs, 7)];
Chris@82: 			 T19 = Tl * TM;
Chris@82: 			 TO = ri[WS(rs, 3)];
Chris@82: 			 TP = T3 * TO;
Chris@82: 			 TQ = ii[WS(rs, 3)];
Chris@82: 			 T1b = T3 * TQ;
Chris@82: 			 TN = FMA(Tn, TM, TL);
Chris@82: 			 TR = FMA(T6, TQ, TP);
Chris@82: 			 T18 = TN - TR;
Chris@82: 			 T1a = FNMS(Tn, TK, T19);
Chris@82: 			 T1c = FNMS(T6, TO, T1b);
Chris@82: 			 T1d = T1a - T1c;
Chris@82: 		    }
Chris@82: 		    {
Chris@82: 			 E Tx, Ty, Tz, T12, TD, TE, TH, T14;
Chris@82: 			 Tx = ri[WS(rs, 1)];
Chris@82: 			 Ty = T2 * Tx;
Chris@82: 			 Tz = ii[WS(rs, 1)];
Chris@82: 			 T12 = T2 * Tz;
Chris@82: 			 TD = ri[WS(rs, 5)];
Chris@82: 			 TE = TC * TD;
Chris@82: 			 TH = ii[WS(rs, 5)];
Chris@82: 			 T14 = TC * TH;
Chris@82: 			 TA = FMA(T5, Tz, Ty);
Chris@82: 			 TI = FMA(TG, TH, TE);
Chris@82: 			 T11 = TA - TI;
Chris@82: 			 T13 = FNMS(T5, Tx, T12);
Chris@82: 			 T15 = FNMS(TG, TD, T14);
Chris@82: 			 T16 = T13 - T15;
Chris@82: 		    }
Chris@82: 		    {
Chris@82: 			 E T10, T1g, T1z, T1B, T1f, T1C, T1j, T1A;
Chris@82: 			 {
Chris@82: 			      E TU, TZ, T1x, T1y;
Chris@82: 			      TU = T1 - Td;
Chris@82: 			      TZ = TW - TY;
Chris@82: 			      T10 = TU + TZ;
Chris@82: 			      T1g = TU - TZ;
Chris@82: 			      T1x = T1s - T1r;
Chris@82: 			      T1y = Tk - Tu;
Chris@82: 			      T1z = T1x - T1y;
Chris@82: 			      T1B = T1y + T1x;
Chris@82: 			 }
Chris@82: 			 {
Chris@82: 			      E T17, T1e, T1h, T1i;
Chris@82: 			      T17 = T11 + T16;
Chris@82: 			      T1e = T18 - T1d;
Chris@82: 			      T1f = T17 + T1e;
Chris@82: 			      T1C = T1e - T17;
Chris@82: 			      T1h = T16 - T11;
Chris@82: 			      T1i = T18 + T1d;
Chris@82: 			      T1j = T1h - T1i;
Chris@82: 			      T1A = T1h + T1i;
Chris@82: 			 }
Chris@82: 			 ri[WS(rs, 5)] = FNMS(KP707106781, T1f, T10);
Chris@82: 			 ii[WS(rs, 5)] = FNMS(KP707106781, T1A, T1z);
Chris@82: 			 ri[WS(rs, 1)] = FMA(KP707106781, T1f, T10);
Chris@82: 			 ii[WS(rs, 1)] = FMA(KP707106781, T1A, T1z);
Chris@82: 			 ri[WS(rs, 7)] = FNMS(KP707106781, T1j, T1g);
Chris@82: 			 ii[WS(rs, 7)] = FNMS(KP707106781, T1C, T1B);
Chris@82: 			 ri[WS(rs, 3)] = FMA(KP707106781, T1j, T1g);
Chris@82: 			 ii[WS(rs, 3)] = FMA(KP707106781, T1C, T1B);
Chris@82: 		    }
Chris@82: 		    {
Chris@82: 			 E Tw, T1k, T1u, T1w, TT, T1v, T1n, T1o;
Chris@82: 			 {
Chris@82: 			      E Te, Tv, T1p, T1t;
Chris@82: 			      Te = T1 + Td;
Chris@82: 			      Tv = Tk + Tu;
Chris@82: 			      Tw = Te + Tv;
Chris@82: 			      T1k = Te - Tv;
Chris@82: 			      T1p = TW + TY;
Chris@82: 			      T1t = T1r + T1s;
Chris@82: 			      T1u = T1p + T1t;
Chris@82: 			      T1w = T1t - T1p;
Chris@82: 			 }
Chris@82: 			 {
Chris@82: 			      E TJ, TS, T1l, T1m;
Chris@82: 			      TJ = TA + TI;
Chris@82: 			      TS = TN + TR;
Chris@82: 			      TT = TJ + TS;
Chris@82: 			      T1v = TS - TJ;
Chris@82: 			      T1l = T13 + T15;
Chris@82: 			      T1m = T1a + T1c;
Chris@82: 			      T1n = T1l - T1m;
Chris@82: 			      T1o = T1l + T1m;
Chris@82: 			 }
Chris@82: 			 ri[WS(rs, 4)] = Tw - TT;
Chris@82: 			 ii[WS(rs, 4)] = T1u - T1o;
Chris@82: 			 ri[0] = Tw + TT;
Chris@82: 			 ii[0] = T1o + T1u;
Chris@82: 			 ri[WS(rs, 6)] = T1k - T1n;
Chris@82: 			 ii[WS(rs, 6)] = T1w - T1v;
Chris@82: 			 ri[WS(rs, 2)] = T1k + T1n;
Chris@82: 			 ii[WS(rs, 2)] = T1v + T1w;
Chris@82: 		    }
Chris@82: 	       }
Chris@82: 	  }
Chris@82:      }
Chris@82: }
Chris@82: 
Chris@82: static const tw_instr twinstr[] = {
Chris@82:      {TW_CEXP, 0, 1},
Chris@82:      {TW_CEXP, 0, 3},
Chris@82:      {TW_CEXP, 0, 7},
Chris@82:      {TW_NEXT, 1, 0}
Chris@82: };
Chris@82: 
Chris@82: static const ct_desc desc = { 8, "t2_8", twinstr, &GENUS, {44, 20, 30, 0}, 0, 0, 0 };
Chris@82: 
Chris@82: void X(codelet_t2_8) (planner *p) {
Chris@82:      X(kdft_dit_register) (p, t2_8, &desc);
Chris@82: }
Chris@82: #else
Chris@82: 
Chris@82: /* Generated by: ../../../genfft/gen_twiddle.native -compact -variables 4 -pipeline-latency 4 -twiddle-log3 -precompute-twiddles -n 8 -name t2_8 -include dft/scalar/t.h */
Chris@82: 
Chris@82: /*
Chris@82:  * This function contains 74 FP additions, 44 FP multiplications,
Chris@82:  * (or, 56 additions, 26 multiplications, 18 fused multiply/add),
Chris@82:  * 42 stack variables, 1 constants, and 32 memory accesses
Chris@82:  */
Chris@82: #include "dft/scalar/t.h"
Chris@82: 
Chris@82: static void t2_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 * 6); m < me; m = m + 1, ri = ri + ms, ii = ii + ms, W = W + 6, MAKE_VOLATILE_STRIDE(16, rs)) {
Chris@82: 	       E T2, T5, T3, T6, T8, Tc, Tg, Ti, Tl, Tm, Tn, Tz, Tp, Tx;
Chris@82: 	       {
Chris@82: 		    E T4, Tb, T7, Ta;
Chris@82: 		    T2 = W[0];
Chris@82: 		    T5 = W[1];
Chris@82: 		    T3 = W[2];
Chris@82: 		    T6 = W[3];
Chris@82: 		    T4 = T2 * T3;
Chris@82: 		    Tb = T5 * T3;
Chris@82: 		    T7 = T5 * T6;
Chris@82: 		    Ta = T2 * T6;
Chris@82: 		    T8 = T4 - T7;
Chris@82: 		    Tc = Ta + Tb;
Chris@82: 		    Tg = T4 + T7;
Chris@82: 		    Ti = Ta - Tb;
Chris@82: 		    Tl = W[4];
Chris@82: 		    Tm = W[5];
Chris@82: 		    Tn = FMA(T2, Tl, T5 * Tm);
Chris@82: 		    Tz = FNMS(Ti, Tl, Tg * Tm);
Chris@82: 		    Tp = FNMS(T5, Tl, T2 * Tm);
Chris@82: 		    Tx = FMA(Tg, Tl, Ti * Tm);
Chris@82: 	       }
Chris@82: 	       {
Chris@82: 		    E Tf, T1i, TL, T1d, TJ, T17, TV, TY, Ts, T1j, TO, T1a, TC, T16, TQ;
Chris@82: 		    E TT;
Chris@82: 		    {
Chris@82: 			 E T1, T1c, Te, T1b, T9, Td;
Chris@82: 			 T1 = ri[0];
Chris@82: 			 T1c = ii[0];
Chris@82: 			 T9 = ri[WS(rs, 4)];
Chris@82: 			 Td = ii[WS(rs, 4)];
Chris@82: 			 Te = FMA(T8, T9, Tc * Td);
Chris@82: 			 T1b = FNMS(Tc, T9, T8 * Td);
Chris@82: 			 Tf = T1 + Te;
Chris@82: 			 T1i = T1c - T1b;
Chris@82: 			 TL = T1 - Te;
Chris@82: 			 T1d = T1b + T1c;
Chris@82: 		    }
Chris@82: 		    {
Chris@82: 			 E TF, TW, TI, TX;
Chris@82: 			 {
Chris@82: 			      E TD, TE, TG, TH;
Chris@82: 			      TD = ri[WS(rs, 7)];
Chris@82: 			      TE = ii[WS(rs, 7)];
Chris@82: 			      TF = FMA(Tl, TD, Tm * TE);
Chris@82: 			      TW = FNMS(Tm, TD, Tl * TE);
Chris@82: 			      TG = ri[WS(rs, 3)];
Chris@82: 			      TH = ii[WS(rs, 3)];
Chris@82: 			      TI = FMA(T3, TG, T6 * TH);
Chris@82: 			      TX = FNMS(T6, TG, T3 * TH);
Chris@82: 			 }
Chris@82: 			 TJ = TF + TI;
Chris@82: 			 T17 = TW + TX;
Chris@82: 			 TV = TF - TI;
Chris@82: 			 TY = TW - TX;
Chris@82: 		    }
Chris@82: 		    {
Chris@82: 			 E Tk, TM, Tr, TN;
Chris@82: 			 {
Chris@82: 			      E Th, Tj, To, Tq;
Chris@82: 			      Th = ri[WS(rs, 2)];
Chris@82: 			      Tj = ii[WS(rs, 2)];
Chris@82: 			      Tk = FMA(Tg, Th, Ti * Tj);
Chris@82: 			      TM = FNMS(Ti, Th, Tg * Tj);
Chris@82: 			      To = ri[WS(rs, 6)];
Chris@82: 			      Tq = ii[WS(rs, 6)];
Chris@82: 			      Tr = FMA(Tn, To, Tp * Tq);
Chris@82: 			      TN = FNMS(Tp, To, Tn * Tq);
Chris@82: 			 }
Chris@82: 			 Ts = Tk + Tr;
Chris@82: 			 T1j = Tk - Tr;
Chris@82: 			 TO = TM - TN;
Chris@82: 			 T1a = TM + TN;
Chris@82: 		    }
Chris@82: 		    {
Chris@82: 			 E Tw, TR, TB, TS;
Chris@82: 			 {
Chris@82: 			      E Tu, Tv, Ty, TA;
Chris@82: 			      Tu = ri[WS(rs, 1)];
Chris@82: 			      Tv = ii[WS(rs, 1)];
Chris@82: 			      Tw = FMA(T2, Tu, T5 * Tv);
Chris@82: 			      TR = FNMS(T5, Tu, T2 * Tv);
Chris@82: 			      Ty = ri[WS(rs, 5)];
Chris@82: 			      TA = ii[WS(rs, 5)];
Chris@82: 			      TB = FMA(Tx, Ty, Tz * TA);
Chris@82: 			      TS = FNMS(Tz, Ty, Tx * TA);
Chris@82: 			 }
Chris@82: 			 TC = Tw + TB;
Chris@82: 			 T16 = TR + TS;
Chris@82: 			 TQ = Tw - TB;
Chris@82: 			 TT = TR - TS;
Chris@82: 		    }
Chris@82: 		    {
Chris@82: 			 E Tt, TK, T1f, T1g;
Chris@82: 			 Tt = Tf + Ts;
Chris@82: 			 TK = TC + TJ;
Chris@82: 			 ri[WS(rs, 4)] = Tt - TK;
Chris@82: 			 ri[0] = Tt + TK;
Chris@82: 			 {
Chris@82: 			      E T19, T1e, T15, T18;
Chris@82: 			      T19 = T16 + T17;
Chris@82: 			      T1e = T1a + T1d;
Chris@82: 			      ii[0] = T19 + T1e;
Chris@82: 			      ii[WS(rs, 4)] = T1e - T19;
Chris@82: 			      T15 = Tf - Ts;
Chris@82: 			      T18 = T16 - T17;
Chris@82: 			      ri[WS(rs, 6)] = T15 - T18;
Chris@82: 			      ri[WS(rs, 2)] = T15 + T18;
Chris@82: 			 }
Chris@82: 			 T1f = TJ - TC;
Chris@82: 			 T1g = T1d - T1a;
Chris@82: 			 ii[WS(rs, 2)] = T1f + T1g;
Chris@82: 			 ii[WS(rs, 6)] = T1g - T1f;
Chris@82: 			 {
Chris@82: 			      E T11, T1k, T14, T1h, T12, T13;
Chris@82: 			      T11 = TL - TO;
Chris@82: 			      T1k = T1i - T1j;
Chris@82: 			      T12 = TT - TQ;
Chris@82: 			      T13 = TV + TY;
Chris@82: 			      T14 = KP707106781 * (T12 - T13);
Chris@82: 			      T1h = KP707106781 * (T12 + T13);
Chris@82: 			      ri[WS(rs, 7)] = T11 - T14;
Chris@82: 			      ii[WS(rs, 5)] = T1k - T1h;
Chris@82: 			      ri[WS(rs, 3)] = T11 + T14;
Chris@82: 			      ii[WS(rs, 1)] = T1h + T1k;
Chris@82: 			 }
Chris@82: 			 {
Chris@82: 			      E TP, T1m, T10, T1l, TU, TZ;
Chris@82: 			      TP = TL + TO;
Chris@82: 			      T1m = T1j + T1i;
Chris@82: 			      TU = TQ + TT;
Chris@82: 			      TZ = TV - TY;
Chris@82: 			      T10 = KP707106781 * (TU + TZ);
Chris@82: 			      T1l = KP707106781 * (TZ - TU);
Chris@82: 			      ri[WS(rs, 5)] = TP - T10;
Chris@82: 			      ii[WS(rs, 7)] = T1m - T1l;
Chris@82: 			      ri[WS(rs, 1)] = TP + T10;
Chris@82: 			      ii[WS(rs, 3)] = T1l + T1m;
Chris@82: 			 }
Chris@82: 		    }
Chris@82: 	       }
Chris@82: 	  }
Chris@82:      }
Chris@82: }
Chris@82: 
Chris@82: static const tw_instr twinstr[] = {
Chris@82:      {TW_CEXP, 0, 1},
Chris@82:      {TW_CEXP, 0, 3},
Chris@82:      {TW_CEXP, 0, 7},
Chris@82:      {TW_NEXT, 1, 0}
Chris@82: };
Chris@82: 
Chris@82: static const ct_desc desc = { 8, "t2_8", twinstr, &GENUS, {56, 26, 18, 0}, 0, 0, 0 };
Chris@82: 
Chris@82: void X(codelet_t2_8) (planner *p) {
Chris@82:      X(kdft_dit_register) (p, t2_8, &desc);
Chris@82: }
Chris@82: #endif