Re: [Open64-devel] Big documentation push

[Sun C. <sun...@gm...>]

so far, your warning messages are just that. I don't see any bug, just
that the routine got bigger (likely due to inlining) that had
triggered Olimit to kick in. If you are willing to wait, at Olimit=0
to your comannd line (and pass that down to the backends)
Sun

2009/6/9 "C. Bergström" <cod...@os...>:
> Sun Chan wrote:
>>
>> Wenguang,
>> What a heavy suggestion. Unfortunately, there is no cook book for
>> debugging open64 IPA. Used to, back then, mostly 4 person in SGI does that.
>> Somehow, the IPA team (3 of these 4) doesn't believe in automatic triaging.
>> I can help write something up. For debugging ipa, I suggest we should make
>> it being able to binary search. But that in itself is not as easy since the
>> design has not considered that to start out with. That said, Intel managed
>> to accomplish that, so I guess it can be done. Let's talk early next week.
>> Sun
>>
> Hi Sun,
>
> This thread went quiet a while back, but wanted to follow-up since I may
> have to do some IPA debugging in the very near future.
>
> Postgresql with IPA and without -OPT:Olimit=0 fails to build..  (ipl has
> been spinning on one object for about an hour and hoping it will finish
> soon)
>
> "1.I": Warning: Variable anon61 in get_hash_entry..postgresLNCB might be
> used uninitialized
> ...
> and "25.I": Warning: Olimit was exceeded on function base_yyparse; will not
> perform function-scope optimization.
>
> Any hints which code I should look at or debugging pointers?
>
> Thanks
>
> ./C
>
>