[petsc-dev] Part II: Basic Parallel Communication Performance

[Lawrence Mitchell]

Hi Junchao, et al.,

> On 22 Sep 2020, at 02:24, Junchao Zhang <junchao.zhang at gmail.com> wrote:
> 
>   We wrote a tech-report on the design and performance of PetscSF on OLCF Summit. SF is petsc's MPI communication module. The report focuses on synchronization and concurrency. The report does not give a general communication performance model.

I attach some comments on the general exposition and presentation of the data. I'm not sure of the target audience, so take them with a pinch of salt.

WRT not addressing the general communication performance model, you do provide some modelling for simple patterns, and I think it would be nice to validate that against your experimental results.

Thanks,

Lawrence

-------------- next part --------------
A non-text attachment was scrubbed...
Name: Summit_performance__basic_communication_DRAFT-comments.pdf
Type: application/pdf
Size: 5473458 bytes
Desc: not available
URL: <http://lists.mcs.anl.gov/pipermail/petsc-dev/attachments/20200922/ddf034a9/attachment-0001.pdf>