Closed rsln-s closed 3 years ago
You are correct that the driver program in Zoltan does not read symmetric matrix-market files. It is written to read general files or pattern files. While not a good long-term solution, a quick short-term solution would be to convert the symmetric matrix-market file to a general file, including the transposed nonzeros and provide that file to zdrive.
Enabling the reader to read symmetric files would not be difficult. It would require some attention to maximum array sizes and computation of global number of pins, as well as new testing. Unfortunately, we are short-handed at this time.
@trilinos/zoltan
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.
If you would like to keep this issue open please add a comment and/or remove the MARKED_FOR_CLOSURE
label.
If this issue should be kept open even with no activity beyond the time limits you can add the label DO_NOT_AUTOCLOSE
.
If it is ok for this issue to be closed, feel free to go ahead and close it. Please do not add any comments or change any labels or otherwise touch this issue unless your intention is to reset the inactivity counter for an additional year.
This issue was closed due to inactivity for 395 days.
@trilinos/zoltan
Expectations
Zoltan should correctly read MatrixMarket format matrices and interpret them in row-net fashion (i.e. each row in the matrix corresponds to a hyperedge and each column to a vertex).
Current Behavior
As pointed out by @SebastianSchlag, on matrix VDOL/hangGlider_3 after reading in the hypergraph the number of edges in it is 10259, vertices -- 10260 and pins (nonzeros) -- 44643. However, the matrix contains 10260 vertices, 10260 edges and 92703 nonzeros. @SebastianSchlag points out that this results in incorrect cut computations when compared to hMetis, for example. Please note that this problem has been tested in serial mode (i.e. on one processor, without MPI).
Motivation and Context
I have previously attempted to fix this problem, but recently discovered (thanks, again, to @SebastianSchlag) that my fix is indeed incorrect. I will close the corresponding pull request #1198
Definition of Done
After the fix Zoltan should read the matrix in correctly. For example, for VDOL/hangGlider_3 the following should be true at the most fine level:
Steps to Reproduce
To check the number of vertices, hyperedges and nonzeros I have introduced the following three lines in 'phg/phg.c' (after line 475):
Then to build and run:
Where
zdrive.inp.hangGlider_3.agg.2.110
: