This is the mail archive of the gcc-patches@gcc.gnu.org mailing list for the GCC project.


Index Nav: [Date Index] [Subject Index] [Author Index] [Thread Index]
Message Nav: [Date Prev] [Date Next] [Thread Prev] [Thread Next]
Other format: [Raw text]

Re: fix m68k-linux bootstrap (PR9093)


Hi,

On Wed, 5 Feb 2003, Richard Henderson wrote:

> > Fine, but could you give me hint how?
> > Every user of note_uses() I found wants only constant expressions, should 
> > I really fix all users (currently only validate_replace_src_1() and 
> > find_used_regs()?
> 
> Um, I don't think either of these should be changed.  Since
> we're talking about an auto-inc, this isn't an issue with the
> use, it's more likely an issue with not registering the kill
> somewhere.
> 
> In general gcse doesn't expect to have to deal with auto-inc,
> since that's supposed to be generated later.  Clearly someone
> forgot about stack pushes.  My guess is that the proper solution
> is to add code to compute_sets to notice these.

Hmm, I don't see how compute_sets fits in here. It goes wrong during 
local_cprop_pass which AFAICT doesn't makes use of the reg_set info and 
record_set_info ignores the stack pointer. local_cprop_pass is new in 
3.[34] and was added by Jan, so I just Cc him :).

> Can you attach your preprocessed test case to the PR?

How do I add that to the PR?
Anyway, Andreas added one with PR9226.

bye, Roman


Index Nav: [Date Index] [Subject Index] [Author Index] [Thread Index]
Message Nav: [Date Prev] [Date Next] [Thread Prev] [Thread Next]