d@0: /*
d@0:  * Copyright (c) 2003, 2007-8 Matteo Frigo
d@0:  * Copyright (c) 2003, 2007-8 Massachusetts Institute of Technology
d@0:  *
d@0:  * This program is free software; you can redistribute it and/or modify
d@0:  * it under the terms of the GNU General Public License as published by
d@0:  * the Free Software Foundation; either version 2 of the License, or
d@0:  * (at your option) any later version.
d@0:  *
d@0:  * This program is distributed in the hope that it will be useful,
d@0:  * but WITHOUT ANY WARRANTY; without even the implied warranty of
d@0:  * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
d@0:  * GNU General Public License for more details.
d@0:  *
d@0:  * You should have received a copy of the GNU General Public License
d@0:  * along with this program; if not, write to the Free Software
d@0:  * Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA  02111-1307  USA
d@0:  *
d@0:  */
d@0: /* Generated by: ../../genfft/gen_twiddle_c -standalone -fma -reorder-insns -simd -compact -variables 100000 -include fftw-spu.h -trivial-stores -n 8 -name X(spu_t1fv_8) */
d@0: 
d@0: /*
d@0:  * This function contains 33 FP additions, 24 FP multiplications,
d@0:  * (or, 23 additions, 14 multiplications, 10 fused multiply/add),
d@0:  * 44 stack variables, 1 constants, and 16 memory accesses
d@0:  */
d@0: #include "fftw-spu.h"
d@0: 
d@0: void X(spu_t1fv_8) (R *ri, R *ii, const R *W, stride rs, INT mb, INT me, INT ms) {
d@0:      DVK(KP707106781, +0.707106781186547524400844362104849039284835938);
d@0:      INT m;
d@0:      R *x;
d@0:      x = ri;
d@0:      for (m = mb, W = W + (mb * ((TWVL / VL) * 14)); m < me; m = m + VL, x = x + (VL * ms), W = W + (TWVL * 14), MAKE_VOLATILE_STRIDE(rs)) {
d@0: 	  V T4, Tu, Tl, Tv, Tx, Ty, Tf, Tm, T1, T3, T2, Ti, Tk, Th, Tj;
d@0: 	  V T6, Td, T8, Tb, T5, Tc, T7, Ta, T9, Te, Tq, Tr, Tw, Tz, To;
d@0: 	  V Tp, Tg, Tn, TE, TF, TC, TD, Ts, Tt, TA, TB;
d@0: 	  T1 = LD(&(x[0]), ms, &(x[0]));
d@0: 	  T2 = LD(&(x[WS(rs, 4)]), ms, &(x[0]));
d@0: 	  T3 = BYTWJ(&(W[TWVL * 6]), T2);
d@0: 	  T4 = VSUB(T1, T3);
d@0: 	  Tu = VADD(T1, T3);
d@0: 	  Th = LD(&(x[WS(rs, 2)]), ms, &(x[0]));
d@0: 	  Ti = BYTWJ(&(W[TWVL * 2]), Th);
d@0: 	  Tj = LD(&(x[WS(rs, 6)]), ms, &(x[0]));
d@0: 	  Tk = BYTWJ(&(W[TWVL * 10]), Tj);
d@0: 	  Tl = VSUB(Ti, Tk);
d@0: 	  Tv = VADD(Ti, Tk);
d@0: 	  T5 = LD(&(x[WS(rs, 1)]), ms, &(x[WS(rs, 1)]));
d@0: 	  T6 = BYTWJ(&(W[0]), T5);
d@0: 	  Tc = LD(&(x[WS(rs, 3)]), ms, &(x[WS(rs, 1)]));
d@0: 	  Td = BYTWJ(&(W[TWVL * 4]), Tc);
d@0: 	  T7 = LD(&(x[WS(rs, 5)]), ms, &(x[WS(rs, 1)]));
d@0: 	  T8 = BYTWJ(&(W[TWVL * 8]), T7);
d@0: 	  Ta = LD(&(x[WS(rs, 7)]), ms, &(x[WS(rs, 1)]));
d@0: 	  Tb = BYTWJ(&(W[TWVL * 12]), Ta);
d@0: 	  Tx = VADD(T6, T8);
d@0: 	  T9 = VSUB(T6, T8);
d@0: 	  Te = VSUB(Tb, Td);
d@0: 	  Ty = VADD(Tb, Td);
d@0: 	  Tf = VADD(T9, Te);
d@0: 	  Tm = VSUB(Te, T9);
d@0: 	  Tg = VFMA(LDK(KP707106781), Tf, T4);
d@0: 	  Tq = VFNMS(LDK(KP707106781), Tf, T4);
d@0: 	  Tr = VFMA(LDK(KP707106781), Tm, Tl);
d@0: 	  Tn = VFNMS(LDK(KP707106781), Tm, Tl);
d@0: 	  To = VFNMSI(Tn, Tg);
d@0: 	  Tp = VFMAI(Tn, Tg);
d@0: 	  ST(&(x[WS(rs, 1)]), To, ms, &(x[WS(rs, 1)]));
d@0: 	  ST(&(x[WS(rs, 7)]), Tp, ms, &(x[WS(rs, 1)]));
d@0: 	  TC = VSUB(Tu, Tv);
d@0: 	  Tw = VADD(Tu, Tv);
d@0: 	  Tz = VADD(Tx, Ty);
d@0: 	  TD = VSUB(Ty, Tx);
d@0: 	  TE = VFNMSI(TD, TC);
d@0: 	  TF = VFMAI(TD, TC);
d@0: 	  ST(&(x[WS(rs, 6)]), TE, ms, &(x[0]));
d@0: 	  ST(&(x[WS(rs, 2)]), TF, ms, &(x[0]));
d@0: 	  Ts = VFNMSI(Tr, Tq);
d@0: 	  Tt = VFMAI(Tr, Tq);
d@0: 	  ST(&(x[WS(rs, 5)]), Ts, ms, &(x[WS(rs, 1)]));
d@0: 	  ST(&(x[WS(rs, 3)]), Tt, ms, &(x[WS(rs, 1)]));
d@0: 	  TA = VSUB(Tw, Tz);
d@0: 	  TB = VADD(Tw, Tz);
d@0: 	  ST(&(x[WS(rs, 4)]), TA, ms, &(x[0]));
d@0: 	  ST(&(x[0]), TB, ms, &(x[0]));
d@0:      }
d@0: }