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 08:08:00 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 -sign 1 -n 4 -dif -name hc2cbdft2_4 -include rdft/scalar/hc2cb.h */
Chris@82: 
Chris@82: /*
Chris@82:  * This function contains 30 FP additions, 12 FP multiplications,
Chris@82:  * (or, 24 additions, 6 multiplications, 6 fused multiply/add),
Chris@82:  * 23 stack variables, 0 constants, and 16 memory accesses
Chris@82:  */
Chris@82: #include "rdft/scalar/hc2cb.h"
Chris@82: 
Chris@82: static void hc2cbdft2_4(R *Rp, R *Ip, R *Rm, R *Im, const R *W, stride rs, INT mb, INT me, INT ms)
Chris@82: {
Chris@82:      {
Chris@82: 	  INT m;
Chris@82: 	  for (m = mb, W = W + ((mb - 1) * 6); m < me; m = m + 1, Rp = Rp + ms, Ip = Ip + ms, Rm = Rm - ms, Im = Im - ms, W = W + 6, MAKE_VOLATILE_STRIDE(16, rs)) {
Chris@82: 	       E T3, Tm, T6, Tn, Td, Tk, TB, Ty, Tv, Ts;
Chris@82: 	       {
Chris@82: 		    E Tg, Tc, T9, Tj;
Chris@82: 		    {
Chris@82: 			 E T1, T2, Ta, Tb;
Chris@82: 			 T1 = Rp[0];
Chris@82: 			 T2 = Rm[WS(rs, 1)];
Chris@82: 			 T3 = T1 + T2;
Chris@82: 			 Tg = T1 - T2;
Chris@82: 			 Ta = Ip[0];
Chris@82: 			 Tb = Im[WS(rs, 1)];
Chris@82: 			 Tc = Ta + Tb;
Chris@82: 			 Tm = Ta - Tb;
Chris@82: 		    }
Chris@82: 		    {
Chris@82: 			 E T4, T5, Th, Ti;
Chris@82: 			 T4 = Rp[WS(rs, 1)];
Chris@82: 			 T5 = Rm[0];
Chris@82: 			 T6 = T4 + T5;
Chris@82: 			 T9 = T4 - T5;
Chris@82: 			 Th = Ip[WS(rs, 1)];
Chris@82: 			 Ti = Im[0];
Chris@82: 			 Tj = Th + Ti;
Chris@82: 			 Tn = Th - Ti;
Chris@82: 		    }
Chris@82: 		    Td = T9 + Tc;
Chris@82: 		    Tk = Tg - Tj;
Chris@82: 		    TB = Tg + Tj;
Chris@82: 		    Ty = Tc - T9;
Chris@82: 		    Tv = Tm - Tn;
Chris@82: 		    Ts = T3 - T6;
Chris@82: 	       }
Chris@82: 	       {
Chris@82: 		    E T7, To, Te, Tp, T8, Tl, Tq, Tf;
Chris@82: 		    T7 = T3 + T6;
Chris@82: 		    To = Tm + Tn;
Chris@82: 		    T8 = W[0];
Chris@82: 		    Te = T8 * Td;
Chris@82: 		    Tp = T8 * Tk;
Chris@82: 		    Tf = W[1];
Chris@82: 		    Tl = FMA(Tf, Tk, Te);
Chris@82: 		    Tq = FNMS(Tf, Td, Tp);
Chris@82: 		    Rp[0] = T7 - Tl;
Chris@82: 		    Ip[0] = To + Tq;
Chris@82: 		    Rm[0] = T7 + Tl;
Chris@82: 		    Im[0] = Tq - To;
Chris@82: 	       }
Chris@82: 	       {
Chris@82: 		    E Tr, Tt, Tu, TD, Tz, TF, Tx;
Chris@82: 		    Tr = W[2];
Chris@82: 		    Tt = Tr * Ts;
Chris@82: 		    Tu = W[3];
Chris@82: 		    TD = Tu * Ts;
Chris@82: 		    Tx = W[4];
Chris@82: 		    Tz = Tx * Ty;
Chris@82: 		    TF = Tx * TB;
Chris@82: 		    {
Chris@82: 			 E Tw, TE, TC, TG, TA;
Chris@82: 			 Tw = FNMS(Tu, Tv, Tt);
Chris@82: 			 TE = FMA(Tr, Tv, TD);
Chris@82: 			 TA = W[5];
Chris@82: 			 TC = FMA(TA, TB, Tz);
Chris@82: 			 TG = FNMS(TA, Ty, TF);
Chris@82: 			 Rp[WS(rs, 1)] = Tw - TC;
Chris@82: 			 Ip[WS(rs, 1)] = TE + TG;
Chris@82: 			 Rm[WS(rs, 1)] = Tw + TC;
Chris@82: 			 Im[WS(rs, 1)] = TG - TE;
Chris@82: 		    }
Chris@82: 	       }
Chris@82: 	  }
Chris@82:      }
Chris@82: }
Chris@82: 
Chris@82: static const tw_instr twinstr[] = {
Chris@82:      {TW_FULL, 1, 4},
Chris@82:      {TW_NEXT, 1, 0}
Chris@82: };
Chris@82: 
Chris@82: static const hc2c_desc desc = { 4, "hc2cbdft2_4", twinstr, &GENUS, {24, 6, 6, 0} };
Chris@82: 
Chris@82: void X(codelet_hc2cbdft2_4) (planner *p) {
Chris@82:      X(khc2c_register) (p, hc2cbdft2_4, &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 -sign 1 -n 4 -dif -name hc2cbdft2_4 -include rdft/scalar/hc2cb.h */
Chris@82: 
Chris@82: /*
Chris@82:  * This function contains 30 FP additions, 12 FP multiplications,
Chris@82:  * (or, 24 additions, 6 multiplications, 6 fused multiply/add),
Chris@82:  * 19 stack variables, 0 constants, and 16 memory accesses
Chris@82:  */
Chris@82: #include "rdft/scalar/hc2cb.h"
Chris@82: 
Chris@82: static void hc2cbdft2_4(R *Rp, R *Ip, R *Rm, R *Im, const R *W, stride rs, INT mb, INT me, INT ms)
Chris@82: {
Chris@82:      {
Chris@82: 	  INT m;
Chris@82: 	  for (m = mb, W = W + ((mb - 1) * 6); m < me; m = m + 1, Rp = Rp + ms, Ip = Ip + ms, Rm = Rm - ms, Im = Im - ms, W = W + 6, MAKE_VOLATILE_STRIDE(16, rs)) {
Chris@82: 	       E T3, Tl, T6, Tm, Td, Tj, Tx, Tv, Ts, Tq;
Chris@82: 	       {
Chris@82: 		    E Tf, Tc, T9, Ti;
Chris@82: 		    {
Chris@82: 			 E T1, T2, Ta, Tb;
Chris@82: 			 T1 = Rp[0];
Chris@82: 			 T2 = Rm[WS(rs, 1)];
Chris@82: 			 T3 = T1 + T2;
Chris@82: 			 Tf = T1 - T2;
Chris@82: 			 Ta = Ip[0];
Chris@82: 			 Tb = Im[WS(rs, 1)];
Chris@82: 			 Tc = Ta + Tb;
Chris@82: 			 Tl = Ta - Tb;
Chris@82: 		    }
Chris@82: 		    {
Chris@82: 			 E T4, T5, Tg, Th;
Chris@82: 			 T4 = Rp[WS(rs, 1)];
Chris@82: 			 T5 = Rm[0];
Chris@82: 			 T6 = T4 + T5;
Chris@82: 			 T9 = T4 - T5;
Chris@82: 			 Tg = Ip[WS(rs, 1)];
Chris@82: 			 Th = Im[0];
Chris@82: 			 Ti = Tg + Th;
Chris@82: 			 Tm = Tg - Th;
Chris@82: 		    }
Chris@82: 		    Td = T9 + Tc;
Chris@82: 		    Tj = Tf - Ti;
Chris@82: 		    Tx = Tf + Ti;
Chris@82: 		    Tv = Tc - T9;
Chris@82: 		    Ts = Tl - Tm;
Chris@82: 		    Tq = T3 - T6;
Chris@82: 	       }
Chris@82: 	       {
Chris@82: 		    E T7, Tn, Tk, To, T8, Te;
Chris@82: 		    T7 = T3 + T6;
Chris@82: 		    Tn = Tl + Tm;
Chris@82: 		    T8 = W[0];
Chris@82: 		    Te = W[1];
Chris@82: 		    Tk = FMA(T8, Td, Te * Tj);
Chris@82: 		    To = FNMS(Te, Td, T8 * Tj);
Chris@82: 		    Rp[0] = T7 - Tk;
Chris@82: 		    Ip[0] = Tn + To;
Chris@82: 		    Rm[0] = T7 + Tk;
Chris@82: 		    Im[0] = To - Tn;
Chris@82: 	       }
Chris@82: 	       {
Chris@82: 		    E Tt, Tz, Ty, TA;
Chris@82: 		    {
Chris@82: 			 E Tp, Tr, Tu, Tw;
Chris@82: 			 Tp = W[2];
Chris@82: 			 Tr = W[3];
Chris@82: 			 Tt = FNMS(Tr, Ts, Tp * Tq);
Chris@82: 			 Tz = FMA(Tr, Tq, Tp * Ts);
Chris@82: 			 Tu = W[4];
Chris@82: 			 Tw = W[5];
Chris@82: 			 Ty = FMA(Tu, Tv, Tw * Tx);
Chris@82: 			 TA = FNMS(Tw, Tv, Tu * Tx);
Chris@82: 		    }
Chris@82: 		    Rp[WS(rs, 1)] = Tt - Ty;
Chris@82: 		    Ip[WS(rs, 1)] = Tz + TA;
Chris@82: 		    Rm[WS(rs, 1)] = Tt + Ty;
Chris@82: 		    Im[WS(rs, 1)] = TA - Tz;
Chris@82: 	       }
Chris@82: 	  }
Chris@82:      }
Chris@82: }
Chris@82: 
Chris@82: static const tw_instr twinstr[] = {
Chris@82:      {TW_FULL, 1, 4},
Chris@82:      {TW_NEXT, 1, 0}
Chris@82: };
Chris@82: 
Chris@82: static const hc2c_desc desc = { 4, "hc2cbdft2_4", twinstr, &GENUS, {24, 6, 6, 0} };
Chris@82: 
Chris@82: void X(codelet_hc2cbdft2_4) (planner *p) {
Chris@82:      X(khc2c_register) (p, hc2cbdft2_4, &desc, HC2C_VIA_DFT);
Chris@82: }
Chris@82: #endif