Difference between SstMarshalBP and SstMarshalBP5?

[caitlinross]

I'm running into an issue with the tutorial example using SST. This is running gray scott, with ADIOS2.9.1 and running a ParaView script that does a contour. I've tracked it down to the change in commit 93dbc3f that changes the default marshalling for SST.
With the following images, I ran with 4 ranks on the write side and 2 on the read side, but I've also gotten the same when using a single rank on each side.

With SstMarshalBP5, I get the following image:

But if I either revert back to v2.8.3 or in v2.9.1 change the MarshalMethod back to SstMarshalBP, I get the correct image:

I don't really know much about the details of SST, so I'm not sure how to go about fixing this. Is there something that I'm misunderstanding about using SST that was possible before, but not really the correct way of doing things?

@eisenhauer @pnorbert

[eisenhauer]

@pnorbert, could this be the memory selection problem? @caitlinross, could you please try this with ADIOS master? PR #3823 merged an update to SST to allow it to do memory selection (writing only a subset of the block it is presented with, useful if the simulation is using ghost cells). I believe that gray-scott uses this at least in some modes.

[caitlinross]

Using master fixed the problem! Thanks!