[petsc-dev] Current status: GPUs for PETSc

[Matthew Knepley]

On Sun, Nov 4, 2012 at 9:51 PM, Karl Rupp <rupp at mcs.anl.gov> wrote:

> Hi Matt,
>
>
>  Very cool. I did not expect the large AMD overheads, but the results
>> below make the current CUDA strategy look
>> pretty good, and an OpenCL strategy look fine for large problems.
>>
>> I have a more basic question. I saw the previous implementation in very
>> simple terms:
>>
>>    1) Replication of CPU vectors on the GPU
>>
>>    2) A coherence policy
>>
>> The nice thing here is how robust it is. We really do not have to commit
>> to any implementation because
>> the CPU part can always pick up the slack. From what is written below, I
>> cannot understand how the
>> "coherence policy" works.
>>
>> Lets use an example to explain it to me. Say that you have a CUDA
>> vector, but you want to execute
>> VecPointwiseMult() with another Vec, but that operation is not part of
>> your CUDA implementation.
>> What happens?
>>
>>
> It works in the same way as it does now:
>
>   if (memory_handle.active_handle != main_memory_flag) {
>     copy_to_cpu();
>     memory_handle.active_handle = main_memory_flag;
>   }
>   process_on_cpu();
>
> If each of the memory regions carries a 'valid-flag', the flag for the
> CUDA part is set to invalid after the processing.
>

Great. Then I understand much more.


> The copy_to_cpu-part could be managed via page-locked memory, yet I still
> have to investigate its robustness. It is certainly a nice option for APUs,
> because it has zero overhead. The only drawback is that APUs just lack
> performance in general...
>
> Overall, I don't want to give away the robustness you described above.
> There will always be some operations that work better on the CPU, while
> others work better on accelerators, so hopping between them is
> (unfortunately) rather the rule than the exception in real world
> applications.
>

Okay, then I have a procedural question. Should we write about this for the
GPU book, or just describe the
current state of affairs?

  Thanks,

      Matt


> Best regards,
> Karli
>
>


-- 
What most experimenters take for granted before they begin their
experiments is infinitely more interesting than any results to which their
experiments lead.
-- Norbert Wiener
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.mcs.anl.gov/pipermail/petsc-dev/attachments/20121104/4750a670/attachment.html>