Re: [m68k] arch flag cleanup v2

[Ben Elliston <bje+dated+1142068537 dot 22a799 at air dot net dot au>]

Hi Nathan

> div (an isa-a insn), except for 5200, 5202, 5204 & 5206.  Rather
> than have a separate ELF flag for hwdiv, this patch creates a
> variant isa 'a-minus', which is all of isa-a except for the hwdiv
> insns.  similarly all isa-a+ & isa-b chips implement USP support
> except for 5407.  Thus I enumerated an isa-b-minus variant in the
> ELF flags for that.

> These new isa names are user visible, in that objdump and readelf
> will show them.  The user cannot select them directly (use
> isaa,nohwdiv and isab,nousp respectively).  I could rework the
> display so that what I've termed isa-a-minus is shown as '[isa-a]
> [nohwdiv]' for instance.  I have no strong feeling about that,
> except a negative capability flag might be a bit odd.

I think you really should produce objdump/readelf output similar to
what you've proposed -- a user is going to have no clue what an
"isa-a-minus" is.  They should not need to refer to the documentation
to find out, either!

> This elf flag relabling is binary incompatible with the previous
> version, but as that's only been committed for a couple of weeks, I
> think that is ok.

Agreed; more importantly, there have been no official binutils
releases.

> built & tested on m68k-elf with a cf simulator, ok?

OK with the changes above.  I wouldn't mind some test cases, either!
(In particlar for the merging logic).

Cheers, Ben

Attachment: pgp00000.pgp
Description: PGP signature