Komodo performance issue

Tom SAVELL <[email protected]> Thu, 11 Feb 2010 12:02:41 -0800
Newsgroups gmane.comp.ide.komodo.general
Message-ID <OFA80A827F.F0F243C4-ON882576C5.006B71E8-882576C7.006E1C19@cli.creaf.com>
Hi,

I've been using Komodo 5.x for some time on Windows, and recently was able 
to get it installed on a Linux machine -- I had to wait for an O/S release 
upgrade to be installed (now at CentOS 5.4).  I noticed a substantial 
difference in the performance of Komodo 5.2.4 on Linux versus Windows.  It 
is _much_ slower on the Linux machine.

I checked with our IT department, and the Linux machine is:

Dual-CPU Xeon 2.4 GHz (P4-era single core)  w/ hyper-threading.

My Windows machine is:

Pentium D 3.4 GHz.  The effective number of CPUs is basically equivalent, 
but actually irrelevant because the Linux machine is completely unloaded 
-- there are no other users and the only application running is Komodo and 
a single xterm instance.  So I suppose the performance difference could be 
attributed to GHz, but then I would expect a 2.4:3.4 ratio, or 
approximately 70% on Linux machine. 

However, it is perceptably worse than that.  It is nearly unusable on the 
slower machine.  For example, clicking the "Project Tools" GUI button 
takes almost a whole *second* to reflect the change of state.  And in that 
same drop-down menu, the mouse has to be over a menu option for almost a 
second before the menu option changes state to highlighted.  On the faster 
machine, it is practically instantaneous on the same button (there is a 
slight perceptable delay, which as a musician I would estimate to be in 
the range of 15-50 milliseconds, or 10x to 50x faster).

Is there anything you can suggest, or are there any known issues with 
respect to performance?

BTW, I did a little "interpreted language" test combined with GUI updates 
by running the "tkcon" application on the same two machines.  First I 
noted that the response to mouse button clicks is acceptable on the slower 
Linux machine.  Then I ran the following Tcl script on both, which uses 
"puts" to write to the console window:

time {for {set i 0} {$i < 100} {incr i} {puts $i}} 10

The results were surprising: it was about 2x *faster* on the slower Linux 
machine:

Slower Linux Machine Result:    65118.2 microseconds per iteration
Faster Windows Machine Result: 118818.9 microseconds per iteration

I ran it multiple times and got very similar results each time.  The 
paradoxical performance difference in this particular example may have to 
do with using a remote display (X-Window Server) rather than a local 
display (Windows local GUI).

-Tom

_______________________________________________
Komodo-discuss mailing list
[email protected]
To unsubscribe: http://listserv.ActiveState.com/mailman/mysubs
Other options: http://listserv.ActiveState.com/mailman/listinfo/Komodo-discuss