> On Feb 29, 2016, at 10:39 AM, Griffith, Boyce Eugene <[email protected]> 
> wrote:
> 
>> 
>> On Feb 29, 2016, at 12:15 AM, Barry Smith <[email protected]> wrote:
>> 
>> 
>>> On Feb 27, 2016, at 8:22 PM, Boyce Griffith <[email protected]> wrote:
>>> 
>>> Barry —
>>> 
>>> I haven’t been following master, and so maybe these have already been 
>>> addressed, but is there a chance that fixes for the performance issues 
>>> Amneet found (options in setting up lots of KSPs in small-block ASM,
>> 
>>  Thanks for the reminder. I have added a new ./configure 
>> option--with-debugging=0 which will remove all of that overhead and likely 
>> perform better than the "hack" I suggested to apply earlier. This is in the 
>> branch barry/fixes-for-release-1 and will be in master if it passes all the 
>> test.
>> 
>>  I'll also remind you of the option --with-errorchecking=0 which turns off 
>> additional overhead that in theory is not needed in production runs and 
>> could help a bit more when using many small solvers.
> 
> Thanks Barry! Is --with-errorchecking=0 implied by --with-debugging=0 (or 
> vice versa), or are these somewhat independent? We were running the timing 
> tests with --with-debugging=0 but not with --with-errorchecking=0.

   The three options are all independent. Turn them all on for faster time 
--with-debugging=0  --with-viewfromoptions=0 --with-errorchecking=0

> 
> Amneet, can you please verify that the performance issues you uncovered are 
> fixed with a clean checkout of the PETSc branch Barry mentions above along 
> with the foregoing PETSc configure options?

   You could send me a trace file (like you did before) and I'll see if there 
is anything else that might be improved.

   Barry

> 
> Thanks,
> 
> -- Boyce
> 
>> 
>> Please let us know if there are any difficulties.
>> 
>> Barry
>> 
>> 
>>> and logging of IS objects) will make it into 3.7?
>>> 
>>> Thanks,
>>> 
>>> — Boyce
>>> 
>>>> On Feb 27, 2016, at 3:36 PM, Barry Smith <[email protected]> wrote:
>>>> 
>>>> 
>>>> PETSc Users,
>>>> 
>>>> We are planning the PETSc release 3.7 shortly. If you know of any bugs 
>>>> that need to be fixed or enhancements added before the release please let 
>>>> us know.
>>>> 
>>>> You can think of the master branch of the PETSc repository obtainable with 
>>>> 
>>>> git clone https://bitbucket.org/petsc/petsc petsc
>>>> 
>>>> as a release candidate for 3.7. Changes for the release are listed at 
>>>> http://www.mcs.anl.gov/petsc/documentation/changes/dev.html
>>>> 
>>>> Thanks
>>>> 
>>>> Barry

Reply via email to