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 08:07:10 EDT 2018 */
Chris@82: 
Chris@82: #include "rdft/codelet-rdft.h"
Chris@82: 
Chris@82: #if defined(ARCH_PREFERS_FMA) || defined(ISA_EXTENSION_PREFERS_FMA)
Chris@82: 
Chris@82: /* Generated by: ../../../genfft/gen_hc2cdft.native -fma -compact -variables 4 -pipeline-latency 4 -n 8 -dit -name hc2cfdft_8 -include rdft/scalar/hc2cf.h */
Chris@82: 
Chris@82: /*
Chris@82:  * This function contains 82 FP additions, 52 FP multiplications,
Chris@82:  * (or, 60 additions, 30 multiplications, 22 fused multiply/add),
Chris@82:  * 31 stack variables, 2 constants, and 32 memory accesses
Chris@82:  */
Chris@82: #include "rdft/scalar/hc2cf.h"
Chris@82: 
Chris@82: static void hc2cfdft_8(R *Rp, R *Ip, R *Rm, R *Im, const R *W, stride rs, INT mb, INT me, INT ms)
Chris@82: {
Chris@82:      DK(KP707106781, +0.707106781186547524400844362104849039284835938);
Chris@82:      DK(KP500000000, +0.500000000000000000000000000000000000000000000);
Chris@82:      {
Chris@82: 	  INT m;
Chris@82: 	  for (m = mb, W = W + ((mb - 1) * 14); m < me; m = m + 1, Rp = Rp + ms, Ip = Ip + ms, Rm = Rm - ms, Im = Im - ms, W = W + 14, MAKE_VOLATILE_STRIDE(32, rs)) {
Chris@82: 	       E Ty, T14, TO, T1o, Tv, T16, TG, T1m, Ta, T19, TV, T1h, Tk, T1b, T11;
Chris@82: 	       E T1j;
Chris@82: 	       {
Chris@82: 		    E Tw, Tx, TN, TI, TJ, TK;
Chris@82: 		    Tw = Ip[0];
Chris@82: 		    Tx = Im[0];
Chris@82: 		    TN = Tw + Tx;
Chris@82: 		    TI = Rm[0];
Chris@82: 		    TJ = Rp[0];
Chris@82: 		    TK = TI - TJ;
Chris@82: 		    Ty = Tw - Tx;
Chris@82: 		    T14 = TJ + TI;
Chris@82: 		    {
Chris@82: 			 E TH, TL, TM, T1n;
Chris@82: 			 TH = W[0];
Chris@82: 			 TL = TH * TK;
Chris@82: 			 TM = W[1];
Chris@82: 			 T1n = TM * TK;
Chris@82: 			 TO = FNMS(TM, TN, TL);
Chris@82: 			 T1o = FMA(TH, TN, T1n);
Chris@82: 		    }
Chris@82: 	       }
Chris@82: 	       {
Chris@82: 		    E Tp, TF, Tu, TC;
Chris@82: 		    {
Chris@82: 			 E Tn, To, Ts, Tt;
Chris@82: 			 Tn = Ip[WS(rs, 2)];
Chris@82: 			 To = Im[WS(rs, 2)];
Chris@82: 			 Tp = Tn - To;
Chris@82: 			 TF = Tn + To;
Chris@82: 			 Ts = Rp[WS(rs, 2)];
Chris@82: 			 Tt = Rm[WS(rs, 2)];
Chris@82: 			 Tu = Ts + Tt;
Chris@82: 			 TC = Tt - Ts;
Chris@82: 		    }
Chris@82: 		    {
Chris@82: 			 E Tq, T15, Tm, Tr;
Chris@82: 			 Tm = W[6];
Chris@82: 			 Tq = Tm * Tp;
Chris@82: 			 T15 = Tm * Tu;
Chris@82: 			 Tr = W[7];
Chris@82: 			 Tv = FNMS(Tr, Tu, Tq);
Chris@82: 			 T16 = FMA(Tr, Tp, T15);
Chris@82: 		    }
Chris@82: 		    {
Chris@82: 			 E TB, TD, TE, T1l;
Chris@82: 			 TB = W[8];
Chris@82: 			 TD = TB * TC;
Chris@82: 			 TE = W[9];
Chris@82: 			 T1l = TE * TC;
Chris@82: 			 TG = FNMS(TE, TF, TD);
Chris@82: 			 T1m = FMA(TB, TF, T1l);
Chris@82: 		    }
Chris@82: 	       }
Chris@82: 	       {
Chris@82: 		    E T4, TU, T9, TR;
Chris@82: 		    {
Chris@82: 			 E T2, T3, T7, T8;
Chris@82: 			 T2 = Ip[WS(rs, 1)];
Chris@82: 			 T3 = Im[WS(rs, 1)];
Chris@82: 			 T4 = T2 - T3;
Chris@82: 			 TU = T2 + T3;
Chris@82: 			 T7 = Rp[WS(rs, 1)];
Chris@82: 			 T8 = Rm[WS(rs, 1)];
Chris@82: 			 T9 = T7 + T8;
Chris@82: 			 TR = T7 - T8;
Chris@82: 		    }
Chris@82: 		    {
Chris@82: 			 E T5, T18, T1, T6;
Chris@82: 			 T1 = W[2];
Chris@82: 			 T5 = T1 * T4;
Chris@82: 			 T18 = T1 * T9;
Chris@82: 			 T6 = W[3];
Chris@82: 			 Ta = FNMS(T6, T9, T5);
Chris@82: 			 T19 = FMA(T6, T4, T18);
Chris@82: 		    }
Chris@82: 		    {
Chris@82: 			 E TS, T1g, TQ, TT;
Chris@82: 			 TQ = W[4];
Chris@82: 			 TS = TQ * TR;
Chris@82: 			 T1g = TQ * TU;
Chris@82: 			 TT = W[5];
Chris@82: 			 TV = FMA(TT, TU, TS);
Chris@82: 			 T1h = FNMS(TT, TR, T1g);
Chris@82: 		    }
Chris@82: 	       }
Chris@82: 	       {
Chris@82: 		    E Te, T10, Tj, TX;
Chris@82: 		    {
Chris@82: 			 E Tc, Td, Th, Ti;
Chris@82: 			 Tc = Ip[WS(rs, 3)];
Chris@82: 			 Td = Im[WS(rs, 3)];
Chris@82: 			 Te = Tc - Td;
Chris@82: 			 T10 = Tc + Td;
Chris@82: 			 Th = Rp[WS(rs, 3)];
Chris@82: 			 Ti = Rm[WS(rs, 3)];
Chris@82: 			 Tj = Th + Ti;
Chris@82: 			 TX = Th - Ti;
Chris@82: 		    }
Chris@82: 		    {
Chris@82: 			 E Tf, T1a, Tb, Tg;
Chris@82: 			 Tb = W[10];
Chris@82: 			 Tf = Tb * Te;
Chris@82: 			 T1a = Tb * Tj;
Chris@82: 			 Tg = W[11];
Chris@82: 			 Tk = FNMS(Tg, Tj, Tf);
Chris@82: 			 T1b = FMA(Tg, Te, T1a);
Chris@82: 		    }
Chris@82: 		    {
Chris@82: 			 E TY, T1i, TW, TZ;
Chris@82: 			 TW = W[12];
Chris@82: 			 TY = TW * TX;
Chris@82: 			 T1i = TW * T10;
Chris@82: 			 TZ = W[13];
Chris@82: 			 T11 = FMA(TZ, T10, TY);
Chris@82: 			 T1j = FNMS(TZ, TX, T1i);
Chris@82: 		    }
Chris@82: 	       }
Chris@82: 	       {
Chris@82: 		    E TA, T1f, T1q, T1s, T13, T1e, T1d, T1r;
Chris@82: 		    {
Chris@82: 			 E Tl, Tz, T1k, T1p;
Chris@82: 			 Tl = Ta + Tk;
Chris@82: 			 Tz = Tv + Ty;
Chris@82: 			 TA = Tl + Tz;
Chris@82: 			 T1f = Tz - Tl;
Chris@82: 			 T1k = T1h + T1j;
Chris@82: 			 T1p = T1m + T1o;
Chris@82: 			 T1q = T1k - T1p;
Chris@82: 			 T1s = T1k + T1p;
Chris@82: 		    }
Chris@82: 		    {
Chris@82: 			 E TP, T12, T17, T1c;
Chris@82: 			 TP = TG + TO;
Chris@82: 			 T12 = TV + T11;
Chris@82: 			 T13 = TP - T12;
Chris@82: 			 T1e = T12 + TP;
Chris@82: 			 T17 = T14 + T16;
Chris@82: 			 T1c = T19 + T1b;
Chris@82: 			 T1d = T17 - T1c;
Chris@82: 			 T1r = T17 + T1c;
Chris@82: 		    }
Chris@82: 		    Ip[0] = KP500000000 * (TA + T13);
Chris@82: 		    Rp[0] = KP500000000 * (T1r + T1s);
Chris@82: 		    Im[WS(rs, 3)] = KP500000000 * (T13 - TA);
Chris@82: 		    Rm[WS(rs, 3)] = KP500000000 * (T1r - T1s);
Chris@82: 		    Rm[WS(rs, 1)] = KP500000000 * (T1d - T1e);
Chris@82: 		    Im[WS(rs, 1)] = KP500000000 * (T1q - T1f);
Chris@82: 		    Rp[WS(rs, 2)] = KP500000000 * (T1d + T1e);
Chris@82: 		    Ip[WS(rs, 2)] = KP500000000 * (T1f + T1q);
Chris@82: 	       }
Chris@82: 	       {
Chris@82: 		    E T1v, T1H, T1F, T1L, T1y, T1I, T1B, T1J;
Chris@82: 		    {
Chris@82: 			 E T1t, T1u, T1D, T1E;
Chris@82: 			 T1t = Ty - Tv;
Chris@82: 			 T1u = T19 - T1b;
Chris@82: 			 T1v = T1t - T1u;
Chris@82: 			 T1H = T1u + T1t;
Chris@82: 			 T1D = T14 - T16;
Chris@82: 			 T1E = Ta - Tk;
Chris@82: 			 T1F = T1D - T1E;
Chris@82: 			 T1L = T1D + T1E;
Chris@82: 		    }
Chris@82: 		    {
Chris@82: 			 E T1w, T1x, T1z, T1A;
Chris@82: 			 T1w = T1j - T1h;
Chris@82: 			 T1x = TV - T11;
Chris@82: 			 T1y = T1w + T1x;
Chris@82: 			 T1I = T1w - T1x;
Chris@82: 			 T1z = TO - TG;
Chris@82: 			 T1A = T1o - T1m;
Chris@82: 			 T1B = T1z - T1A;
Chris@82: 			 T1J = T1z + T1A;
Chris@82: 		    }
Chris@82: 		    {
Chris@82: 			 E T1C, T1M, T1G, T1K;
Chris@82: 			 T1C = T1y + T1B;
Chris@82: 			 Ip[WS(rs, 1)] = KP500000000 * (FMA(KP707106781, T1C, T1v));
Chris@82: 			 Im[WS(rs, 2)] = -(KP500000000 * (FNMS(KP707106781, T1C, T1v)));
Chris@82: 			 T1M = T1I + T1J;
Chris@82: 			 Rm[WS(rs, 2)] = KP500000000 * (FNMS(KP707106781, T1M, T1L));
Chris@82: 			 Rp[WS(rs, 1)] = KP500000000 * (FMA(KP707106781, T1M, T1L));
Chris@82: 			 T1G = T1B - T1y;
Chris@82: 			 Rm[0] = KP500000000 * (FNMS(KP707106781, T1G, T1F));
Chris@82: 			 Rp[WS(rs, 3)] = KP500000000 * (FMA(KP707106781, T1G, T1F));
Chris@82: 			 T1K = T1I - T1J;
Chris@82: 			 Ip[WS(rs, 3)] = KP500000000 * (FMA(KP707106781, T1K, T1H));
Chris@82: 			 Im[0] = -(KP500000000 * (FNMS(KP707106781, T1K, T1H)));
Chris@82: 		    }
Chris@82: 	       }
Chris@82: 	  }
Chris@82:      }
Chris@82: }
Chris@82: 
Chris@82: static const tw_instr twinstr[] = {
Chris@82:      {TW_FULL, 1, 8},
Chris@82:      {TW_NEXT, 1, 0}
Chris@82: };
Chris@82: 
Chris@82: static const hc2c_desc desc = { 8, "hc2cfdft_8", twinstr, &GENUS, {60, 30, 22, 0} };
Chris@82: 
Chris@82: void X(codelet_hc2cfdft_8) (planner *p) {
Chris@82:      X(khc2c_register) (p, hc2cfdft_8, &desc, HC2C_VIA_DFT);
Chris@82: }
Chris@82: #else
Chris@82: 
Chris@82: /* Generated by: ../../../genfft/gen_hc2cdft.native -compact -variables 4 -pipeline-latency 4 -n 8 -dit -name hc2cfdft_8 -include rdft/scalar/hc2cf.h */
Chris@82: 
Chris@82: /*
Chris@82:  * This function contains 82 FP additions, 44 FP multiplications,
Chris@82:  * (or, 68 additions, 30 multiplications, 14 fused multiply/add),
Chris@82:  * 39 stack variables, 2 constants, and 32 memory accesses
Chris@82:  */
Chris@82: #include "rdft/scalar/hc2cf.h"
Chris@82: 
Chris@82: static void hc2cfdft_8(R *Rp, R *Ip, R *Rm, R *Im, const R *W, stride rs, INT mb, INT me, INT ms)
Chris@82: {
Chris@82:      DK(KP353553390, +0.353553390593273762200422181052424519642417969);
Chris@82:      DK(KP500000000, +0.500000000000000000000000000000000000000000000);
Chris@82:      {
Chris@82: 	  INT m;
Chris@82: 	  for (m = mb, W = W + ((mb - 1) * 14); m < me; m = m + 1, Rp = Rp + ms, Ip = Ip + ms, Rm = Rm - ms, Im = Im - ms, W = W + 14, MAKE_VOLATILE_STRIDE(32, rs)) {
Chris@82: 	       E Tv, TX, Ts, TY, TE, T1a, TJ, T19, T1l, T1m, T9, T10, Ti, T11, TP;
Chris@82: 	       E T16, TU, T17, T1i, T1j;
Chris@82: 	       {
Chris@82: 		    E Tt, Tu, TD, Tz, TA, TB, Tn, TI, Tr, TG, Tk, To;
Chris@82: 		    Tt = Ip[0];
Chris@82: 		    Tu = Im[0];
Chris@82: 		    TD = Tt + Tu;
Chris@82: 		    Tz = Rm[0];
Chris@82: 		    TA = Rp[0];
Chris@82: 		    TB = Tz - TA;
Chris@82: 		    {
Chris@82: 			 E Tl, Tm, Tp, Tq;
Chris@82: 			 Tl = Ip[WS(rs, 2)];
Chris@82: 			 Tm = Im[WS(rs, 2)];
Chris@82: 			 Tn = Tl - Tm;
Chris@82: 			 TI = Tl + Tm;
Chris@82: 			 Tp = Rp[WS(rs, 2)];
Chris@82: 			 Tq = Rm[WS(rs, 2)];
Chris@82: 			 Tr = Tp + Tq;
Chris@82: 			 TG = Tp - Tq;
Chris@82: 		    }
Chris@82: 		    Tv = Tt - Tu;
Chris@82: 		    TX = TA + Tz;
Chris@82: 		    Tk = W[6];
Chris@82: 		    To = W[7];
Chris@82: 		    Ts = FNMS(To, Tr, Tk * Tn);
Chris@82: 		    TY = FMA(Tk, Tr, To * Tn);
Chris@82: 		    {
Chris@82: 			 E Ty, TC, TF, TH;
Chris@82: 			 Ty = W[0];
Chris@82: 			 TC = W[1];
Chris@82: 			 TE = FNMS(TC, TD, Ty * TB);
Chris@82: 			 T1a = FMA(TC, TB, Ty * TD);
Chris@82: 			 TF = W[8];
Chris@82: 			 TH = W[9];
Chris@82: 			 TJ = FMA(TF, TG, TH * TI);
Chris@82: 			 T19 = FNMS(TH, TG, TF * TI);
Chris@82: 		    }
Chris@82: 		    T1l = TJ + TE;
Chris@82: 		    T1m = T1a - T19;
Chris@82: 	       }
Chris@82: 	       {
Chris@82: 		    E T4, TO, T8, TM, Td, TT, Th, TR;
Chris@82: 		    {
Chris@82: 			 E T2, T3, T6, T7;
Chris@82: 			 T2 = Ip[WS(rs, 1)];
Chris@82: 			 T3 = Im[WS(rs, 1)];
Chris@82: 			 T4 = T2 - T3;
Chris@82: 			 TO = T2 + T3;
Chris@82: 			 T6 = Rp[WS(rs, 1)];
Chris@82: 			 T7 = Rm[WS(rs, 1)];
Chris@82: 			 T8 = T6 + T7;
Chris@82: 			 TM = T6 - T7;
Chris@82: 		    }
Chris@82: 		    {
Chris@82: 			 E Tb, Tc, Tf, Tg;
Chris@82: 			 Tb = Ip[WS(rs, 3)];
Chris@82: 			 Tc = Im[WS(rs, 3)];
Chris@82: 			 Td = Tb - Tc;
Chris@82: 			 TT = Tb + Tc;
Chris@82: 			 Tf = Rp[WS(rs, 3)];
Chris@82: 			 Tg = Rm[WS(rs, 3)];
Chris@82: 			 Th = Tf + Tg;
Chris@82: 			 TR = Tf - Tg;
Chris@82: 		    }
Chris@82: 		    {
Chris@82: 			 E T1, T5, Ta, Te;
Chris@82: 			 T1 = W[2];
Chris@82: 			 T5 = W[3];
Chris@82: 			 T9 = FNMS(T5, T8, T1 * T4);
Chris@82: 			 T10 = FMA(T1, T8, T5 * T4);
Chris@82: 			 Ta = W[10];
Chris@82: 			 Te = W[11];
Chris@82: 			 Ti = FNMS(Te, Th, Ta * Td);
Chris@82: 			 T11 = FMA(Ta, Th, Te * Td);
Chris@82: 			 {
Chris@82: 			      E TL, TN, TQ, TS;
Chris@82: 			      TL = W[4];
Chris@82: 			      TN = W[5];
Chris@82: 			      TP = FMA(TL, TM, TN * TO);
Chris@82: 			      T16 = FNMS(TN, TM, TL * TO);
Chris@82: 			      TQ = W[12];
Chris@82: 			      TS = W[13];
Chris@82: 			      TU = FMA(TQ, TR, TS * TT);
Chris@82: 			      T17 = FNMS(TS, TR, TQ * TT);
Chris@82: 			 }
Chris@82: 			 T1i = T17 - T16;
Chris@82: 			 T1j = TP - TU;
Chris@82: 		    }
Chris@82: 	       }
Chris@82: 	       {
Chris@82: 		    E T1h, T1t, T1w, T1y, T1o, T1s, T1r, T1x;
Chris@82: 		    {
Chris@82: 			 E T1f, T1g, T1u, T1v;
Chris@82: 			 T1f = Tv - Ts;
Chris@82: 			 T1g = T10 - T11;
Chris@82: 			 T1h = KP500000000 * (T1f - T1g);
Chris@82: 			 T1t = KP500000000 * (T1g + T1f);
Chris@82: 			 T1u = T1i - T1j;
Chris@82: 			 T1v = T1l + T1m;
Chris@82: 			 T1w = KP353553390 * (T1u - T1v);
Chris@82: 			 T1y = KP353553390 * (T1u + T1v);
Chris@82: 		    }
Chris@82: 		    {
Chris@82: 			 E T1k, T1n, T1p, T1q;
Chris@82: 			 T1k = T1i + T1j;
Chris@82: 			 T1n = T1l - T1m;
Chris@82: 			 T1o = KP353553390 * (T1k + T1n);
Chris@82: 			 T1s = KP353553390 * (T1n - T1k);
Chris@82: 			 T1p = TX - TY;
Chris@82: 			 T1q = T9 - Ti;
Chris@82: 			 T1r = KP500000000 * (T1p - T1q);
Chris@82: 			 T1x = KP500000000 * (T1p + T1q);
Chris@82: 		    }
Chris@82: 		    Ip[WS(rs, 1)] = T1h + T1o;
Chris@82: 		    Rp[WS(rs, 1)] = T1x + T1y;
Chris@82: 		    Im[WS(rs, 2)] = T1o - T1h;
Chris@82: 		    Rm[WS(rs, 2)] = T1x - T1y;
Chris@82: 		    Rm[0] = T1r - T1s;
Chris@82: 		    Im[0] = T1w - T1t;
Chris@82: 		    Rp[WS(rs, 3)] = T1r + T1s;
Chris@82: 		    Ip[WS(rs, 3)] = T1t + T1w;
Chris@82: 	       }
Chris@82: 	       {
Chris@82: 		    E Tx, T15, T1c, T1e, TW, T14, T13, T1d;
Chris@82: 		    {
Chris@82: 			 E Tj, Tw, T18, T1b;
Chris@82: 			 Tj = T9 + Ti;
Chris@82: 			 Tw = Ts + Tv;
Chris@82: 			 Tx = Tj + Tw;
Chris@82: 			 T15 = Tw - Tj;
Chris@82: 			 T18 = T16 + T17;
Chris@82: 			 T1b = T19 + T1a;
Chris@82: 			 T1c = T18 - T1b;
Chris@82: 			 T1e = T18 + T1b;
Chris@82: 		    }
Chris@82: 		    {
Chris@82: 			 E TK, TV, TZ, T12;
Chris@82: 			 TK = TE - TJ;
Chris@82: 			 TV = TP + TU;
Chris@82: 			 TW = TK - TV;
Chris@82: 			 T14 = TV + TK;
Chris@82: 			 TZ = TX + TY;
Chris@82: 			 T12 = T10 + T11;
Chris@82: 			 T13 = TZ - T12;
Chris@82: 			 T1d = TZ + T12;
Chris@82: 		    }
Chris@82: 		    Ip[0] = KP500000000 * (Tx + TW);
Chris@82: 		    Rp[0] = KP500000000 * (T1d + T1e);
Chris@82: 		    Im[WS(rs, 3)] = KP500000000 * (TW - Tx);
Chris@82: 		    Rm[WS(rs, 3)] = KP500000000 * (T1d - T1e);
Chris@82: 		    Rm[WS(rs, 1)] = KP500000000 * (T13 - T14);
Chris@82: 		    Im[WS(rs, 1)] = KP500000000 * (T1c - T15);
Chris@82: 		    Rp[WS(rs, 2)] = KP500000000 * (T13 + T14);
Chris@82: 		    Ip[WS(rs, 2)] = KP500000000 * (T15 + T1c);
Chris@82: 	       }
Chris@82: 	  }
Chris@82:      }
Chris@82: }
Chris@82: 
Chris@82: static const tw_instr twinstr[] = {
Chris@82:      {TW_FULL, 1, 8},
Chris@82:      {TW_NEXT, 1, 0}
Chris@82: };
Chris@82: 
Chris@82: static const hc2c_desc desc = { 8, "hc2cfdft_8", twinstr, &GENUS, {68, 30, 14, 0} };
Chris@82: 
Chris@82: void X(codelet_hc2cfdft_8) (planner *p) {
Chris@82:      X(khc2c_register) (p, hc2cfdft_8, &desc, HC2C_VIA_DFT);
Chris@82: }
Chris@82: #endif