[petsc-dev] -vecscatter_window

Barry Smith bsmith at mcs.anl.gov
Sun Mar 28 21:56:58 CDT 2010


      My guess is that it is a fine patch.

      I put the MPI2 vecscatter together without much thought or  
testing (it is probably no properly tested in make alltests).


On Mar 27, 2010, at 2:55 PM, Jed Brown wrote:

> I fixed this up before the release (6ae96dca2adc), but I didn't  
> address
> this point.  Why is the second MPI_Win_fence called immediately after
> the MPI_Put in VecScatterBegin.  I would expect better latency  
> tolerance
> if this was called in VecScatterEnd, as in the attached patch.  Is  
> this
> patch somehow incorrect?
> Jed
> <window.patch>

More information about the petsc-dev mailing list