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 08:07:51 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_hc2c.native -fma -compact -variables 4 -pipeline-latency 4 -sign 1 -n 8 -dif -name hc2cb_8 -include rdft/scalar/hc2cb.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:  * 33 stack variables, 1 constants, and 32 memory accesses
Chris@82:  */
Chris@82: #include "rdft/scalar/hc2cb.h"
Chris@82: 
Chris@82: static void hc2cb_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:      {
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 T7, T1i, T1n, Tk, TD, TV, T1b, TQ, Te, T1e, T1o, T1j, TE, TF, TR;
Chris@82: 	       E Tv, TW;
Chris@82: 	       {
Chris@82: 		    E T3, Tg, TC, T19, T6, Tz, Tj, T1a;
Chris@82: 		    {
Chris@82: 			 E T1, T2, TA, TB;
Chris@82: 			 T1 = Rp[0];
Chris@82: 			 T2 = Rm[WS(rs, 3)];
Chris@82: 			 T3 = T1 + T2;
Chris@82: 			 Tg = T1 - T2;
Chris@82: 			 TA = Ip[0];
Chris@82: 			 TB = Im[WS(rs, 3)];
Chris@82: 			 TC = TA + TB;
Chris@82: 			 T19 = TA - TB;
Chris@82: 		    }
Chris@82: 		    {
Chris@82: 			 E T4, T5, Th, Ti;
Chris@82: 			 T4 = Rp[WS(rs, 2)];
Chris@82: 			 T5 = Rm[WS(rs, 1)];
Chris@82: 			 T6 = T4 + T5;
Chris@82: 			 Tz = T4 - T5;
Chris@82: 			 Th = Ip[WS(rs, 2)];
Chris@82: 			 Ti = Im[WS(rs, 1)];
Chris@82: 			 Tj = Th + Ti;
Chris@82: 			 T1a = Th - Ti;
Chris@82: 		    }
Chris@82: 		    T7 = T3 + T6;
Chris@82: 		    T1i = T3 - T6;
Chris@82: 		    T1n = T19 - T1a;
Chris@82: 		    Tk = Tg - Tj;
Chris@82: 		    TD = Tz + TC;
Chris@82: 		    TV = TC - Tz;
Chris@82: 		    T1b = T19 + T1a;
Chris@82: 		    TQ = Tg + Tj;
Chris@82: 	       }
Chris@82: 	       {
Chris@82: 		    E Ta, Tl, To, T1c, Td, Tq, Tt, T1d, Tp, Tu;
Chris@82: 		    {
Chris@82: 			 E T8, T9, Tm, Tn;
Chris@82: 			 T8 = Rp[WS(rs, 1)];
Chris@82: 			 T9 = Rm[WS(rs, 2)];
Chris@82: 			 Ta = T8 + T9;
Chris@82: 			 Tl = T8 - T9;
Chris@82: 			 Tm = Ip[WS(rs, 1)];
Chris@82: 			 Tn = Im[WS(rs, 2)];
Chris@82: 			 To = Tm + Tn;
Chris@82: 			 T1c = Tm - Tn;
Chris@82: 		    }
Chris@82: 		    {
Chris@82: 			 E Tb, Tc, Tr, Ts;
Chris@82: 			 Tb = Rm[0];
Chris@82: 			 Tc = Rp[WS(rs, 3)];
Chris@82: 			 Td = Tb + Tc;
Chris@82: 			 Tq = Tb - Tc;
Chris@82: 			 Tr = Ip[WS(rs, 3)];
Chris@82: 			 Ts = Im[0];
Chris@82: 			 Tt = Tr + Ts;
Chris@82: 			 T1d = Tr - Ts;
Chris@82: 		    }
Chris@82: 		    Te = Ta + Td;
Chris@82: 		    T1e = T1c + T1d;
Chris@82: 		    T1o = Ta - Td;
Chris@82: 		    T1j = T1d - T1c;
Chris@82: 		    TE = Tl + To;
Chris@82: 		    TF = Tq + Tt;
Chris@82: 		    TR = TE + TF;
Chris@82: 		    Tp = Tl - To;
Chris@82: 		    Tu = Tq - Tt;
Chris@82: 		    Tv = Tp + Tu;
Chris@82: 		    TW = Tp - Tu;
Chris@82: 	       }
Chris@82: 	       Rp[0] = T7 + Te;
Chris@82: 	       Rm[0] = T1b + T1e;
Chris@82: 	       {
Chris@82: 		    E TS, TX, TT, TY, TP, TU;
Chris@82: 		    TS = FNMS(KP707106781, TR, TQ);
Chris@82: 		    TX = FMA(KP707106781, TW, TV);
Chris@82: 		    TP = W[4];
Chris@82: 		    TT = TP * TS;
Chris@82: 		    TY = TP * TX;
Chris@82: 		    TU = W[5];
Chris@82: 		    Ip[WS(rs, 1)] = FNMS(TU, TX, TT);
Chris@82: 		    Im[WS(rs, 1)] = FMA(TU, TS, TY);
Chris@82: 	       }
Chris@82: 	       {
Chris@82: 		    E T1s, T1v, T1t, T1w, T1r, T1u;
Chris@82: 		    T1s = T1i + T1j;
Chris@82: 		    T1v = T1o + T1n;
Chris@82: 		    T1r = W[2];
Chris@82: 		    T1t = T1r * T1s;
Chris@82: 		    T1w = T1r * T1v;
Chris@82: 		    T1u = W[3];
Chris@82: 		    Rp[WS(rs, 1)] = FNMS(T1u, T1v, T1t);
Chris@82: 		    Rm[WS(rs, 1)] = FMA(T1u, T1s, T1w);
Chris@82: 	       }
Chris@82: 	       {
Chris@82: 		    E T10, T13, T11, T14, TZ, T12;
Chris@82: 		    T10 = FMA(KP707106781, TR, TQ);
Chris@82: 		    T13 = FNMS(KP707106781, TW, TV);
Chris@82: 		    TZ = W[12];
Chris@82: 		    T11 = TZ * T10;
Chris@82: 		    T14 = TZ * T13;
Chris@82: 		    T12 = W[13];
Chris@82: 		    Ip[WS(rs, 3)] = FNMS(T12, T13, T11);
Chris@82: 		    Im[WS(rs, 3)] = FMA(T12, T10, T14);
Chris@82: 	       }
Chris@82: 	       {
Chris@82: 		    E T1f, T15, T17, T18, T1g, T16;
Chris@82: 		    T1f = T1b - T1e;
Chris@82: 		    T16 = T7 - Te;
Chris@82: 		    T15 = W[6];
Chris@82: 		    T17 = T15 * T16;
Chris@82: 		    T18 = W[7];
Chris@82: 		    T1g = T18 * T16;
Chris@82: 		    Rp[WS(rs, 2)] = FNMS(T18, T1f, T17);
Chris@82: 		    Rm[WS(rs, 2)] = FMA(T15, T1f, T1g);
Chris@82: 	       }
Chris@82: 	       {
Chris@82: 		    E T1k, T1p, T1l, T1q, T1h, T1m;
Chris@82: 		    T1k = T1i - T1j;
Chris@82: 		    T1p = T1n - T1o;
Chris@82: 		    T1h = W[10];
Chris@82: 		    T1l = T1h * T1k;
Chris@82: 		    T1q = T1h * T1p;
Chris@82: 		    T1m = W[11];
Chris@82: 		    Rp[WS(rs, 3)] = FNMS(T1m, T1p, T1l);
Chris@82: 		    Rm[WS(rs, 3)] = FMA(T1m, T1k, T1q);
Chris@82: 	       }
Chris@82: 	       {
Chris@82: 		    E TH, TN, TJ, TL, TM, TO, Tf, Tx, Ty, TI, TG, TK, Tw;
Chris@82: 		    TG = TE - TF;
Chris@82: 		    TH = FNMS(KP707106781, TG, TD);
Chris@82: 		    TN = FMA(KP707106781, TG, TD);
Chris@82: 		    TK = FMA(KP707106781, Tv, Tk);
Chris@82: 		    TJ = W[0];
Chris@82: 		    TL = TJ * TK;
Chris@82: 		    TM = W[1];
Chris@82: 		    TO = TM * TK;
Chris@82: 		    Tw = FNMS(KP707106781, Tv, Tk);
Chris@82: 		    Tf = W[8];
Chris@82: 		    Tx = Tf * Tw;
Chris@82: 		    Ty = W[9];
Chris@82: 		    TI = Ty * Tw;
Chris@82: 		    Ip[WS(rs, 2)] = FNMS(Ty, TH, Tx);
Chris@82: 		    Im[WS(rs, 2)] = FMA(Tf, TH, TI);
Chris@82: 		    Ip[0] = FNMS(TM, TN, TL);
Chris@82: 		    Im[0] = FMA(TJ, TN, TO);
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, "hc2cb_8", twinstr, &GENUS, {44, 14, 22, 0} };
Chris@82: 
Chris@82: void X(codelet_hc2cb_8) (planner *p) {
Chris@82:      X(khc2c_register) (p, hc2cb_8, &desc, HC2C_VIA_RDFT);
Chris@82: }
Chris@82: #else
Chris@82: 
Chris@82: /* Generated by: ../../../genfft/gen_hc2c.native -compact -variables 4 -pipeline-latency 4 -sign 1 -n 8 -dif -name hc2cb_8 -include rdft/scalar/hc2cb.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:  * 30 stack variables, 1 constants, and 32 memory accesses
Chris@82:  */
Chris@82: #include "rdft/scalar/hc2cb.h"
Chris@82: 
Chris@82: static void hc2cb_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:      {
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 T7, T18, T1c, To, Ty, TM, TY, TC, Te, TZ, T10, Tv, Tz, TP, TS;
Chris@82: 	       E TD;
Chris@82: 	       {
Chris@82: 		    E T3, TK, Tk, TX, T6, TW, Tn, TL;
Chris@82: 		    {
Chris@82: 			 E T1, T2, Ti, Tj;
Chris@82: 			 T1 = Rp[0];
Chris@82: 			 T2 = Rm[WS(rs, 3)];
Chris@82: 			 T3 = T1 + T2;
Chris@82: 			 TK = T1 - T2;
Chris@82: 			 Ti = Ip[0];
Chris@82: 			 Tj = Im[WS(rs, 3)];
Chris@82: 			 Tk = Ti - Tj;
Chris@82: 			 TX = Ti + Tj;
Chris@82: 		    }
Chris@82: 		    {
Chris@82: 			 E T4, T5, Tl, Tm;
Chris@82: 			 T4 = Rp[WS(rs, 2)];
Chris@82: 			 T5 = Rm[WS(rs, 1)];
Chris@82: 			 T6 = T4 + T5;
Chris@82: 			 TW = T4 - T5;
Chris@82: 			 Tl = Ip[WS(rs, 2)];
Chris@82: 			 Tm = Im[WS(rs, 1)];
Chris@82: 			 Tn = Tl - Tm;
Chris@82: 			 TL = Tl + Tm;
Chris@82: 		    }
Chris@82: 		    T7 = T3 + T6;
Chris@82: 		    T18 = TK + TL;
Chris@82: 		    T1c = TX - TW;
Chris@82: 		    To = Tk + Tn;
Chris@82: 		    Ty = T3 - T6;
Chris@82: 		    TM = TK - TL;
Chris@82: 		    TY = TW + TX;
Chris@82: 		    TC = Tk - Tn;
Chris@82: 	       }
Chris@82: 	       {
Chris@82: 		    E Ta, TN, Tr, TO, Td, TQ, Tu, TR;
Chris@82: 		    {
Chris@82: 			 E T8, T9, Tp, Tq;
Chris@82: 			 T8 = Rp[WS(rs, 1)];
Chris@82: 			 T9 = Rm[WS(rs, 2)];
Chris@82: 			 Ta = T8 + T9;
Chris@82: 			 TN = T8 - T9;
Chris@82: 			 Tp = Ip[WS(rs, 1)];
Chris@82: 			 Tq = Im[WS(rs, 2)];
Chris@82: 			 Tr = Tp - Tq;
Chris@82: 			 TO = Tp + Tq;
Chris@82: 		    }
Chris@82: 		    {
Chris@82: 			 E Tb, Tc, Ts, Tt;
Chris@82: 			 Tb = Rm[0];
Chris@82: 			 Tc = Rp[WS(rs, 3)];
Chris@82: 			 Td = Tb + Tc;
Chris@82: 			 TQ = Tb - Tc;
Chris@82: 			 Ts = Ip[WS(rs, 3)];
Chris@82: 			 Tt = Im[0];
Chris@82: 			 Tu = Ts - Tt;
Chris@82: 			 TR = Ts + Tt;
Chris@82: 		    }
Chris@82: 		    Te = Ta + Td;
Chris@82: 		    TZ = TN + TO;
Chris@82: 		    T10 = TQ + TR;
Chris@82: 		    Tv = Tr + Tu;
Chris@82: 		    Tz = Tu - Tr;
Chris@82: 		    TP = TN - TO;
Chris@82: 		    TS = TQ - TR;
Chris@82: 		    TD = Ta - Td;
Chris@82: 	       }
Chris@82: 	       Rp[0] = T7 + Te;
Chris@82: 	       Rm[0] = To + Tv;
Chris@82: 	       {
Chris@82: 		    E Tg, Tw, Tf, Th;
Chris@82: 		    Tg = T7 - Te;
Chris@82: 		    Tw = To - Tv;
Chris@82: 		    Tf = W[6];
Chris@82: 		    Th = W[7];
Chris@82: 		    Rp[WS(rs, 2)] = FNMS(Th, Tw, Tf * Tg);
Chris@82: 		    Rm[WS(rs, 2)] = FMA(Th, Tg, Tf * Tw);
Chris@82: 	       }
Chris@82: 	       {
Chris@82: 		    E TG, TI, TF, TH;
Chris@82: 		    TG = Ty + Tz;
Chris@82: 		    TI = TD + TC;
Chris@82: 		    TF = W[2];
Chris@82: 		    TH = W[3];
Chris@82: 		    Rp[WS(rs, 1)] = FNMS(TH, TI, TF * TG);
Chris@82: 		    Rm[WS(rs, 1)] = FMA(TF, TI, TH * TG);
Chris@82: 	       }
Chris@82: 	       {
Chris@82: 		    E TA, TE, Tx, TB;
Chris@82: 		    TA = Ty - Tz;
Chris@82: 		    TE = TC - TD;
Chris@82: 		    Tx = W[10];
Chris@82: 		    TB = W[11];
Chris@82: 		    Rp[WS(rs, 3)] = FNMS(TB, TE, Tx * TA);
Chris@82: 		    Rm[WS(rs, 3)] = FMA(Tx, TE, TB * TA);
Chris@82: 	       }
Chris@82: 	       {
Chris@82: 		    E T1a, T1g, T1e, T1i, T19, T1d;
Chris@82: 		    T19 = KP707106781 * (TZ + T10);
Chris@82: 		    T1a = T18 - T19;
Chris@82: 		    T1g = T18 + T19;
Chris@82: 		    T1d = KP707106781 * (TP - TS);
Chris@82: 		    T1e = T1c + T1d;
Chris@82: 		    T1i = T1c - T1d;
Chris@82: 		    {
Chris@82: 			 E T17, T1b, T1f, T1h;
Chris@82: 			 T17 = W[4];
Chris@82: 			 T1b = W[5];
Chris@82: 			 Ip[WS(rs, 1)] = FNMS(T1b, T1e, T17 * T1a);
Chris@82: 			 Im[WS(rs, 1)] = FMA(T17, T1e, T1b * T1a);
Chris@82: 			 T1f = W[12];
Chris@82: 			 T1h = W[13];
Chris@82: 			 Ip[WS(rs, 3)] = FNMS(T1h, T1i, T1f * T1g);
Chris@82: 			 Im[WS(rs, 3)] = FMA(T1f, T1i, T1h * T1g);
Chris@82: 		    }
Chris@82: 	       }
Chris@82: 	       {
Chris@82: 		    E TU, T14, T12, T16, TT, T11;
Chris@82: 		    TT = KP707106781 * (TP + TS);
Chris@82: 		    TU = TM - TT;
Chris@82: 		    T14 = TM + TT;
Chris@82: 		    T11 = KP707106781 * (TZ - T10);
Chris@82: 		    T12 = TY - T11;
Chris@82: 		    T16 = TY + T11;
Chris@82: 		    {
Chris@82: 			 E TJ, TV, T13, T15;
Chris@82: 			 TJ = W[8];
Chris@82: 			 TV = W[9];
Chris@82: 			 Ip[WS(rs, 2)] = FNMS(TV, T12, TJ * TU);
Chris@82: 			 Im[WS(rs, 2)] = FMA(TV, TU, TJ * T12);
Chris@82: 			 T13 = W[0];
Chris@82: 			 T15 = W[1];
Chris@82: 			 Ip[0] = FNMS(T15, T16, T13 * T14);
Chris@82: 			 Im[0] = FMA(T15, T14, T13 * T16);
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, "hc2cb_8", twinstr, &GENUS, {52, 18, 14, 0} };
Chris@82: 
Chris@82: void X(codelet_hc2cb_8) (planner *p) {
Chris@82:      X(khc2c_register) (p, hc2cb_8, &desc, HC2C_VIA_RDFT);
Chris@82: }
Chris@82: #endif