bugGNU Octave - Bugs: bug #49519, SparseComplexMatrix::solve numeric...

 
 

bug #49519: SparseComplexMatrix::solve numeric factorization

Submitter:  None
Submitted:  Wed 02 Nov 2016 08:11:27 PM UTC
   
 
Category:  Libraries Severity:  3 - Normal
Priority:  5 - Normal Item Group:  Matlab Compatibility
Status:  Confirmed Assigned to:  None
Originator Name:  Mark A. Horn Originator Email:  -email is unavailable-
Open/Closed:  * Open Release:  * 4.0.0
Operating System:  * GNU/Linux Fixed Release:  None
Planned Release:  None
* Mandatory Fields

Add a New Comment Rich Markup
   

Jump to the original submission

Fri 04 Nov 2016 07:37:41 PM UTC, comment #6: 

That is some very good diagnostics and advice.  Please let us think about it for a bit and decide next steps.  Thank you kindly for your hard work.

Anonymous
Fri 04 Nov 2016 07:30:17 PM UTC, comment #5: 

I received your files correctly and can reproduce the error.  From what I can see this is related to a factorization call in UMFPACK.

I notice that your sparse matrix, if it were converted to a full matrix, would have number of elements greater than the 32-bit index type used by Octave.  This may be a problem.  Just for grins, I truncated the left-hand matrix to 46e3 x 46e3.  The left division operator now works.

So, one possibility is to compile Octave from source code and use the '--enable-64' configure option which will use 64-bit indices (and you must be running on 64-bit HW as well).  The issue is that you also need to re-build all of your support libraries like BLAS and LAPACK to also use 64-bit indices.  The instructions for building a 64-bit Octave are in the manual in Appendix E.3 if you want to try.

Since you are a commercial enterprise, you might just want to pay someone to resolve the situation.  The founder of Octave, John W. Eaton, does this kind of work.  See https://www.gnu.org/software/octave/commercial-support.html.

I've saved the matrices in Octave's own binary format and by using lzip I have reduced the file size to ~7.5 MB.  This still won't upload to this report so I've asked on the Octave-Maintainer's list as to why this is not working.

Rik <rik5>
Group administrator
Fri 04 Nov 2016 05:01:08 PM UTC, comment #4: 

I have split it to two ~ 8 MB zip files.  Still the same error.  There is something wrong with the upload site.  It just spins for roughly a minute or so and then gives me an error.  I am more than happy to e-mail or dump to ftp site if possible.

Anonymous
Fri 04 Nov 2016 04:49:53 PM UTC, comment #3: 

There is a limit on the file size that can be uploaded.  In the Attachment section of the bug report it states


(Note: upload size limit is set to 16384 kB, after insertion of the required escape characters.)


That would be about 16.4 MB.  My guess is that your 15 MB file is probably just over this limit.

Can you split the file in some manner and upload each part separately?

Rik <rik5>
Group administrator
Fri 04 Nov 2016 03:28:37 PM UTC, comment #2: 

I have tried 10 times to add a 15.0 MB zip file that completely runs in Matlab but not Octave.  The web page fails every time with error.  I can provide the files by e-mail if possible.  We are desperate for Octave to work.  Please help.

Anonymous
Wed 02 Nov 2016 09:44:43 PM UTC, comment #1: 

Could you attach analyze.m to this bug report.  I don't see it in the list of files (although there are two attachments of a .mat file with the same name).

Rik <rik5>
Group administrator
Wed 02 Nov 2016 08:11:27 PM UTC, original submission:  

Recieving SparseComplexMatrix::solve numeric factorization failed in Octave but not in Matlab.  I loaded up a sparse system I am trying to solve with "/" operator.  I also attached analyze.m to help demonstrate failure.

I ran my application in Matlab with no problems.  I run in Octave 4.0 and I get the error.  I dumped the matrices from both Matlab and Octave runs and they compare to within 1e-14.  I dump the matrics from Octave and rebuilt in Matlab with sparse commands and Matlab solves the dumped Octave matrices.  So Matlab can solve this all possible ways.  Octave cannot.

Files too large, but I can provide matrix (Matlab mat files) if someone needs them.

Anonymous

 

(Note: upload size limit is set to 16384 kB, after insertion of the required escape characters.)

Attach Files:
   
   
Comment:
   

Attached Files
file #38887:  analyze.m added by None (987B - text/x-objcsrc)
file #38868:  matlab_0_1_RHSf.mat added by None (6KiB - application/octet-stream)

 

Depends on the following items: None found

Items that depend on this one: None found

 

Carbon-Copy List
  • -email is unavailable- added by rik5 (Posted a comment)
  • -email is unavailable- added by None (Submitted the item)
  •  

    There are 0 votes so far. Votes easily highlight which items people would like to see resolved in priority, independently of the priority of the item set by tracker managers.

    Only group members can vote.

     

    Follow 6 latest changes.

    Date Changed by Updated Field Previous Value => Replaced by
    2016-11-04 rik5 Item GroupWTF, Matlab?!? Matlab Compatibility
        StatusNone Confirmed
    2016-11-04 rik5 Attached File#38867 Removed
    2016-11-04 None Attached File- Added analyze.m, #38887
    2016-11-02 None Attached File- Added matlab_0_1_RHSf.mat, #38868
    2016-11-02 None Attached File- Added matlab_0_1_RHSf.mat, #38867

    Back to the top

    Powered by Savane 3.13-02a9.
    Corresponding source code