[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]
[Octave-bug-tracker] [bug #63939] strange performance of sparse matrix o
From: |
Andreas Stahel |
Subject: |
[Octave-bug-tracker] [bug #63939] strange performance of sparse matrix operations |
Date: |
Fri, 17 Mar 2023 11:54:27 -0400 (EDT) |
URL:
<https://savannah.gnu.org/bugs/?63939>
Summary: strange performance of sparse matrix operations
Group: GNU Octave
Submitter: andreasstahel
Submitted: Fri 17 Mar 2023 03:54:25 PM UTC
Category: Performance
Severity: 3 - Normal
Priority: 5 - Normal
Item Group: Performance
Status: None
Assigned to: None
Originator Name: Andreas Stahel
Originator Email:
Open/Closed: Open
Release: 8.1.0
Discussion Lock: Any
Operating System: GNU/Linux
Fixed Release: None
Planned Release: None
_______________________________________________________
Follow-up Comments:
-------------------------------------------------------
Date: Fri 17 Mar 2023 03:54:25 PM UTC By: Andreas Stahel <andreasstahel>
On a tight loop with sparse matrices
(one matrix*vector, one system solve by \, both with the same matrix)
the results depends in a stange way on value OMP_NUM_THREADS=32
Octave with GUI started.
export OMP_NUM_THREADS=32
leads to 25 seconds of wall time
export OMP_NUM_THREADS=16
leads to 5 seconds of wall time
CPU: AMD Ryzen 9 3950X 16-Core Processor,
hyperthreading enabled, i.e. the OS sees 32 cores, 64GB RAM
Xubuntu 22.04
The development version and 8.1.0 lead to the same results
_______________________________________________________
File Attachments:
-------------------------------------------------------
Date: Fri 17 Mar 2023 03:54:25 PM UTC Name: TestSparse.m Size: 485B By:
andreasstahel
<http://savannah.gnu.org/bugs/download.php?file_id=54507>
_______________________________________________________
Reply to this item at:
<https://savannah.gnu.org/bugs/?63939>
_______________________________________________
Message sent via Savannah
https://savannah.gnu.org/
- [Octave-bug-tracker] [bug #63939] strange performance of sparse matrix operations,
Andreas Stahel <=