This is the mail archive of the
gcc-patches@gcc.gnu.org
mailing list for the GCC project.
Re: [CFT, try 7] Emulated tls rewrite
On Thu, Jul 15, 2010 at 08:59:35AM +0100, IainS wrote:
>
> On 15 Jul 2010, at 08:48, Richard Guenther wrote:
>
>> On Thu, Jul 15, 2010 at 1:50 AM, Richard Henderson <rth@redhat.com>
>> wrote:
>>> This go-around handles phi nodes, as requested by Richi. This was
>>> much more
>>> irritating than it should have been, due to immediate use
>>> maintainence. It
>>
>> Hm, interesting. I skimmed around the tree and SET_PHI_ARG_DEF
>> should do the job.
>>
>>> also gets re-arranged to use walk_gimple_ops, which I should have
>>> been using
>>> in the first place.
>>>
>>> The testsuite is still running here, amd64-linux --disable-tls, but
>>> I'm
>>> quitting for today. I never got around to looking at those non-
>>> tls.exp
>>> testsuite failures, Iain. Hopefully they went away magically. ;-)
>>
>> Overall it looks reasonable, minus the various ???s.
>
> unfortunately, although it applies cleanly, it doesn't bootstrap for me
> (against either 162180 or 162193).
> (fails in building libgomp/parallel.c with 'caller edge frequency 1000
> does not match BB frequency of 61')...
> ... I will look a little more - but this is uncharted territory for me ;)
Bootstraps on x86_64-apple-darwin10 fine here. Regression testresults at...
http://gcc.gnu.org/ml/gcc-testresults/2010-07/msg01432.html
Did you start with a fresh tree before applying the new patch?
Jack
>
> Iain