> > +/* Copy rules */ > > +#define fLSBOLD(VAL) (fGETBIT(0, VAL)) > > +#define fSATH(VAL) fSATN(16, VAL) > > +#define fSATUH(VAL) fSATUN(16, VAL) > > +#define fVSATH(VAL) fVSATN(16, VAL) > > +#define fVSATUH(VAL) fVSATUN(16, VAL) > > +#define fSATUB(VAL) fSATUN(8, VAL) > > +#define fSATB(VAL) fSATN(8, VAL) > > +#define fVSATUB(VAL) fVSATUN(8, VAL) > > +#define fVSATB(VAL) fVSATN(8, VAL) > > +#define fCALL(A) fWRITE_LR(fREAD_NPC()); fWRITE_NPC(A); > > +#define fCALLR(A) fWRITE_LR(fREAD_NPC()); fWRITE_NPC(A); > > +#define fCAST2_8s(A) fSXTN(16, 64, A) > > +#define fCAST2_8u(A) fZXTN(16, 64, A) > > +#define fCAST8S_16S(A) (fSXTN(64, 128, A)) > > +#define fCAST16S_8S(A) (fSXTN(128, 64, A)) > > +#define fVSATW(A) fVSATN(32, fCAST8_8s(A)) > > +#define fSATW(A) fSATN(32, fCAST8_8s(A)) > > +#define fVSAT(A) fVSATN(32, A) > > +#define fSAT(A) fSATN(32, A) > > + > > +/* Ease parsing */ > > +#define f8BITSOF(VAL) ((VAL) ? 0xff : 0x00) > > +#define fREAD_GP() (Constant_extended ? (0) : GP) > > +#define fCLIP(DST, SRC, U) (DST = fMIN((1 << U) - 1, fMAX(SRC, -(1 << U)))) > > I guess this is what's in the manual, but my reading of this expression is > "saturate", not "clip". How does it differ from > > fSATN(U, SRC) > > ?
Their semantics are different. E.g.: - fCLIP(dst, 0x80, 8) => dst = 0x80 - fSATN(8, 0x80) => 0x7F Take a look at `target/hexagon/macros.h` to see their slow (macro/helper-based) implementation, which is basically our reference. As for the naming, it's not up to us. Paolo
