-
Notifications
You must be signed in to change notification settings - Fork 563
New issue
Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.
By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.
Already on GitHub? Sign in to your account
Ifpack2: column map of OverlappingRowMatrix is not complete #7017
Comments
Related to #6877 |
I am wondering this could be related to issues discussed in #558. |
This issue has had no activity for 365 days and is marked for closure. It will be closed after an additional 30 days of inactivity. |
This issue has had no activity for 365 days and is marked for closure. It will be closed after an additional 30 days of inactivity. |
This issue was closed due to inactivity for 395 days. |
Bug Report (maybe?)
@trilinos/ifpack2 @csiefer2
Description
Halo rows imported into an OverlappingRowMatrix appear to exclude column indices corresponding to rows that 1) are not in the original local matrix and 2) are not halo rows themselves.
Implications: In order to perform K matvecs with a single communication, the halo must be K layers deep. If this can fixed, then the halo would only have to be K-1 layers deep.
Example
Consider a 1D Poisson matrix on two ranks, 5 rows per process, with overlap 2.
Here is the original "local" matrix:
And here's the halo matrix with just the imported overlap rows:
Global row 6 on proc 0 doesn't have global column 7, and global row 3 on proc 1 doesn't have global column 2. So it is only possible to do two matvecs (not three) that agree with the standard SPMV that does an import at each step.
The text was updated successfully, but these errors were encountered: