This is the mail archive of the
binutils@sources.redhat.com
mailing list for the binutils project.
Re: introduce SH 2a support
- From: Alexandre Oliva <aoliva at redhat dot com>
- To: "Stubbs, Andrew" <andrew dot stubbs at superh dot com>
- Cc: <binutils at sources dot redhat dot com>
- Date: 30 Jul 2004 07:01:41 -0300
- Subject: Re: introduce SH 2a support
- Organization: Red Hat Global Engineering Services Compiler Team
- References: <9FF3133289A7A84E81E2ED8F5E56B379E68AD4@sh-uk-ex01.uk.w2k.superh.com>
On Jul 29, 2004, "Stubbs, Andrew" <andrew.stubbs@superh.com> wrote:
> I have not tested or checked this patch in detail, but there are two
> issues I notice straight off:
> 1) The inheritance tree is now badly broken.
> You have not updated the inheritance diagram in sh-opc.h so it is
> difficult to see exactly what you have done, but some analysis reveals
> the problems. This diagram was put in to try to make sense of the code,
I added a comment explaining what SH2a did. I didn't have the energy
needed to update the diagram. I was actually unsure as to what to do
about it.
> arch_sh4_up | arch_sh2a_up
> which introduces ambiguity when an instruction was introduced into two
> architectures independently
> The only solution is to invent one or more imaginary architectures (c.f
> sh3-nommu) in order to maintain a proper inheritance tree in which each
> and every instruction may be traced to a common ancestor.
Ugh. Oh, well, I'll see what we can come up with.
> It looks like this new monster will cause some pretty nasty changes.
Yup. I tried to warn you in advance about this upcoming part, but
unfortunately I couldn't share much information about it back then :-(
> Once this is properly resolved there will no longer be any need to
> always pass -isa=sh2a from the compiler
Excellent.
> I assume that, since you have added new test cases elsewhere, you have
> run the testsuite
I'm 100% sure I ran the assembler testsuite, because I know I had
failures in it that I had to fix before posting the final patch. I'm
afraid I missed the linker testsuite, but I don't see any regressions
there either. I see an ERROR compiling ld-sh/sh2.c, but IIRC it was
already there before.
That said, I don't see any evidence that the arch tests you mention
ran. Do I have to do anything other than `make -k check-gas check-ld'
to get them to run?
Thanks for your feedback, it's very appreciated. I'll try to get your
comments addressed as soon as possible.
--
Alexandre Oliva http://www.ic.unicamp.br/~oliva/
Red Hat Compiler Engineer aoliva@{redhat.com, gcc.gnu.org}
Free Software Evangelist oliva@{lsd.ic.unicamp.br, gnu.org}