[petsc-dev] preallocation on GPU

[mary sweat]

So on the GPU I should use  MatSeqAIJSetPreallocation and MatMPIAIJSetPreallocation, just like on the CPU.
Reading you answer I have a doubt: using preallocation on GPU is he speedup high?



Il Domenica 1 Giugno 2014 12:23, Jed Brown <jed at jedbrown.org> ha scritto:
 


mary sweat <mary.sweat78 at yahoo.it> writes:


> Hi all
> are there any functions such as MatSeqAIJSetPreallocation to
> preallocate MatSeqAIJCusp and MatMPIAIJCusp matrices? 

Yeah, it's name is MatSeqAIJSetPreallocation, just like the CPU version.
Matrix assembly is an irregular process that currently happens on the
CPU and is copied to the GPU.  The insertion itself is generally not a
bottleneck.
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.mcs.anl.gov/pipermail/petsc-dev/attachments/20140601/c96d4b74/attachment.html>