[Top][All Lists]

[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

[Octave-bug-tracker] [bug #54422] Declining Octave performance over time

From: Rik
Subject: [Octave-bug-tracker] [bug #54422] Declining Octave performance over time
Date: Wed, 1 Aug 2018 17:01:34 -0400 (EDT)
User-agent: Mozilla/5.0 (X11; Ubuntu; Linux x86_64; rv:61.0) Gecko/20100101 Firefox/61.0


                 Summary: Declining Octave performance over time
                 Project: GNU Octave
            Submitted by: rik5
            Submitted on: Wed 01 Aug 2018 02:01:32 PM PDT
                Category: Interpreter
                Severity: 3 - Normal
                Priority: 5 - Normal
              Item Group: Performance
                  Status: None
             Assigned to: None
         Originator Name: 
        Originator Email: 
             Open/Closed: Open
         Discussion Lock: Any
                 Release: dev
        Operating System: GNU/Linux



I happened to be looking at an old bug which had examples of three for loops. 
I have attached the script as bm_accum.m.  When I run the script in an old
release (3.4.3) of Octave I get

octave-3.4.3:1> bm_accum
Elapsed time is 0.353004 seconds.
Elapsed time is 0.172448 seconds.
Elapsed time is 0.15855 seconds.

When I run it with the current development branch I get

Elapsed time is 1.54277 seconds.
Elapsed time is 0.675958 seconds.
Elapsed time is 1.24968 seconds.

There are slowdowns of from 4X-7.5X.

To make sure this wasn't a measurement artifiact within Octave I tried using
the tcsh shell's time command.

time run-octave -f bm_accum.m
Elapsed time is 1.53665 seconds.
Elapsed time is 0.683419 seconds.
Elapsed time is 1.28853 seconds.
3.704u 0.072s 0:03.87 97.4%     0+0k 160+16io 2pf+0w


time octave-3.4.3 -fq bm_accum.m
Elapsed time is 0.332402 seconds.
Elapsed time is 0.154823 seconds.
Elapsed time is 0.154164 seconds.

0.684u 0.024s 0:00.71 98.5%     0+0k 0+16io 0pf+0w

Again, a significant 5.5X difference.


Reply to this item at:


  Message sent via Savannah

reply via email to

[Prev in Thread] Current Thread [Next in Thread]