Public bug reported:

Octave produces erroneous output for even slightly large matrix-vector
multiplications:

$ octave
GNU Octave, version 8.4.0
...
octave:1> c = rand( 120 ); x = rand( 120, 1 ); ( c * x ) - ( c * x )
ans =

  -19.442
  -23.669
  -21.009
  -24.441

1. The above result is erroneous. The correct answer is a column vector
of all zeros.

2. octave:3> c = rand( 110 ); x = rand( 110, 1 ); ( c * x ) - ( c * x )
produces the correct answer. 

3. $ export OMP_NUM_THREADS=1; octave 
octave:1> c = rand( 120 ); x = rand( 120, 1 ); ( c * x ) - ( c * x )
ans =

   0
   0 ...

produces the correct answer. This suggests multi-threading may be
responsible. The BLAS version is:

octave:1> version( "-blas" )
ans = Intel(R) Math Kernel Library Version 2020.0.4 Product Build 20200917 for 
Intel(R) 64 architecture applications

4. The present bug occurs in a computer with "13th Gen Intel(R) Core(TM) 
i9-13900K" CPU. 
It does not occur in another machine with "Intel(R) Core(TM) i9-14900K", with 
the same Ubuntu and octave versions. 


Thank you for your kind attention to this bug report. 

Mahesh

ProblemType: Bug
DistroRelease: Ubuntu 24.04
Package: octave 8.4.0-1build5
ProcVersionSignature: Ubuntu 6.8.0-55.57-generic 6.8.12
Uname: Linux 6.8.0-55-generic x86_64
ApportVersion: 2.28.1-0ubuntu3.5
Architecture: amd64
CasperMD5CheckResult: pass
Date: Tue Mar 18 14:59:53 2025
InstallationDate: Installed on 2024-02-15 (397 days ago)
InstallationMedia: Ubuntu 22.04.3 LTS "Jammy Jellyfish" - Release amd64 
(20230807.2)
SourcePackage: octave
UpgradeStatus: Upgraded to noble on 2024-09-20 (179 days ago)

** Affects: octave (Ubuntu)
     Importance: Undecided
         Status: New


** Tags: amd64 apport-bug noble

** Attachment added: "apport.octave.wbp4g3px.apport"
   
https://bugs.launchpad.net/bugs/2103502/+attachment/5865445/+files/apport.octave.wbp4g3px.apport

-- 
You received this bug notification because you are a member of Ubuntu
Bugs, which is subscribed to Ubuntu.
https://bugs.launchpad.net/bugs/2103502

Title:
  Erroneous result from matrix-vector multiplication

To manage notifications about this bug go to:
https://bugs.launchpad.net/ubuntu/+source/octave/+bug/2103502/+subscriptions


-- 
ubuntu-bugs mailing list
ubuntu-bugs@lists.ubuntu.com
https://lists.ubuntu.com/mailman/listinfo/ubuntu-bugs

Reply via email to