[llvm] r294264 - [LVI] Switch from BFS to DFS exploration order

Mehdi Amini via llvm-commits llvm-commits at lists.llvm.org
Fri Feb 24 09:30:03 PST 2017


> On Feb 24, 2017, at 9:27 AM, Hans Wennborg <hans at chromium.org> wrote:
> 
> On Fri, Feb 24, 2017 at 9:04 AM, Mehdi Amini <mehdi.amini at apple.com> wrote:
>> 
>>> On Feb 21, 2017, at 11:30 AM, Hans Wennborg via llvm-commits <llvm-commits at lists.llvm.org> wrote:
>>> 
>>> I'm not saying there's anything wrong with the change; it sounds great
>>> to me. It's just that since it's a non-trivial change, and not fixing
>>> a regression from 3.9, I don't want to risk merging it this far into
>>> the release process (according to the schedule we should tag 'final'
>>> today).
>>> 
>>> We'll get to enjoy this improvement in llvm 5 instead.
>> 
>> Why not 4.1?
> 
> Because there's no plan for a 4.1
> (http://blog.llvm.org/2016/12/llvms-new-versioning-scheme.html).
> 
> If you mean 4.0.1, it's up to Tom if he wants to take it. (I
> personally think of the stable releases mostly as bug-fix releases
> though.)

Well, one can see “exploding the memory consumption” as a bug :)

—
Mehdi
 
> 
> 
>>> On Tue, Feb 21, 2017 at 11:17 AM, Daniel Berlin <dberlin at dberlin.org> wrote:
>>>> FWIW: I mentioned he was welcome to try, but i did not have time to try to
>>>> verify the sanity of doing it.
>>>> As for riskiness, at this point, it's been tested on billions of lines of
>>>> code, (including entire os distro compiles).
>>>> 
>>>> I'm not horribly worried about the percent chance it will cause regressions
>>>> in *compile time* (In fact, it'll likely be a huge net positive in almost
>>>> all cases), but i mostly agree with Hal that we should try to stick to stuff
>>>> that maybe hasn't had to have followups, and isn't basically a week or two
>>>> old.
>>>> 
>>>> 
>>>> 
>>>> On Tue, Feb 21, 2017 at 10:52 AM, Hans Wennborg via llvm-commits
>>>> <llvm-commits at lists.llvm.org> wrote:
>>>>> 
>>>>> ,
>>>>> 
>>>>> On Sat, Feb 18, 2017 at 3:29 PM, Joerg Sonnenberger <joerg at bec.de> wrote:
>>>>>> On Mon, Feb 06, 2017 at 04:49:57PM -0800, Hans Wennborg wrote:
>>>>>>> On Mon, Feb 6, 2017 at 4:42 PM, Joerg Sonnenberger via llvm-commits
>>>>>>> <llvm-commits at lists.llvm.org> wrote:
>>>>>>>> On Tue, Feb 07, 2017 at 12:25:24AM -0000, Philip Reames via
>>>>>>>> llvm-commits wrote:
>>>>>>>>> Author: reames
>>>>>>>>> Date: Mon Feb  6 18:25:24 2017
>>>>>>>>> New Revision: 294264
>>>>>>>>> 
>>>>>>>>> URL: http://llvm.org/viewvc/llvm-project?rev=294264&view=rev
>>>>>>>>> Log:
>>>>>>>>> [LVI] Switch from BFS to DFS exploration order
>>>>>>>> 
>>>>>>>> Thanks! Hans, any comment on merging this into 4.0? It fixes one of
>>>>>>>> the
>>>>>>>> two remaining cases in pkgsrc where clang with normal optimiser
>>>>>>>> settings
>>>>>>>> hits 2GB VA.
>>>>>>> 
>>>>>>> Well.. This just landed, and I don't think it's fixing a regression
>>>>>>> from 3.9. While the change sounds great, I think we should just let it
>>>>>>> go into the next release.
>>>>>> 
>>>>>> It's been a while and I still would like to see this and the follow-up
>>>>>> r294463 merged. That was discussed with Daniel Berlin already.
>>>>> 
>>>>> I'm sorry, but I really do think this kind of change is too risky so
>>>>> late in the release process.
>>>>> 
>>>>> - Hans
>>>>> _______________________________________________
>>>>> llvm-commits mailing list
>>>>> llvm-commits at lists.llvm.org
>>>>> http://lists.llvm.org/cgi-bin/mailman/listinfo/llvm-commits
>>>> 
>>>> 
>>> _______________________________________________
>>> llvm-commits mailing list
>>> llvm-commits at lists.llvm.org
>>> http://lists.llvm.org/cgi-bin/mailman/listinfo/llvm-commits
>> 



More information about the llvm-commits mailing list