Le 12/08/2015 06:07, Richard Henderson a écrit : > On 08/09/2015 01:13 PM, Laurent Vivier wrote: >> -#define OS_BYTE 0 >> -#define OS_WORD 1 >> -#define OS_LONG 2 >> -#define OS_SINGLE 4 >> -#define OS_DOUBLE 5 >> +#define OS_BYTE 1 >> +#define OS_WORD 2 >> +#define OS_LONG 3 >> +#define OS_SINGLE 4 >> +#define OS_DOUBLE 5 >> +#define OS_EXTENDED 6 >> +#define OS_PACKED 7 >> > > Is there a reason you've skipped the 0 value when adding the new values?
I think there is no reason, but if I change the value I have to update abdc_mem, sbcd_mem instructions as they use it as an incrementer/decrementer. I agree, it's a strange idea. > >> +static inline int insn_opsize(int insn, int pos) >> +{ >> + switch ((insn >> pos) & 3) { > > > In particular, that change means that insn_opsize is more complicated > than needed. Further, is there any reason for POS to be a varable? > Isn't it at the same place for all insns? > >> +static inline int ext_opsize(int ext, int pos) > > This should probably wait until the fp insns get added. Yes. Laurent