[PLUG] CPU Usage

[John Jason Jordan]

On Thu, 3 Sep 2009 05:16:35 -0700
drew wymore <drew.wymore at gmail.com> dijo:

> On Wed, Sep 2, 2009 at 10:20 PM, John Jason Jordan <johnxj at comcast.net>wrote:
> 
> > I have a Thinkpad T61 with Intel Core2 Duo at 2.00 GHz on which I have
> > Jaunty x86_64. At the moment I am trying to print a couple of pages of
> > a PDF from Okular and it is taking forever to image. While I watch the
> > progress in System Monitor I note that the CPUs switch back and forth.
> > That is, for a while CPU1 will be at or near 100% and CPU2 will be
> > around 30%. After a while they swap and CPU1 will be at 30% or so and
> > CPU2 will be at or near 100%. I get the same results from Adobe Reader
> > 9.1. However, Adobe Reader is so slow that I killed it after waiting 20
> > minutes.
> >
> > I am wondering how Linux decides which CPU to use for which process.

> I don't know which apps are multi-threaded but if they are they can use both
> CPU's which could be why it's bouncing back and forth between the 2. How big
> are the PDF's you're working with?

The PDFs were exported from Scribus. The Scribus document is 97 MB, and
contains no bitmap images at all - just text and vector graphics. The
first time I exported the file I embedded all nine fonts used in the
document and the resulting PDF was 207 MB. The second time I converted
all fonts to outlines and the resulting PDF was 203 MB. That surprised
me, because converting to outlines usually results in a much smaller
PDF. 

According to System Monitor Scribus is using 2 GB of RAM with the
document open, Adobe Reader uses 500 MB with the document open, but
Okular uses less than 100 MB. I was trying to print just two pages.
Okular took ten minutes to send the job to CUPS, but I never found out
how long Adobe Reader would take because I gave up on it after 20
minutes. 

I have no idea whether either program is multithreaded, but I very much
doubt that Adobe Reader is because it is still only 32-bit. I was
thinking that neither was multithreaded and the behavior of the CPUs
was nearly 100% on one CPU for the program that I was printing from,
and the other CPU at 20-40% because it was handling the OS. I just
found it strange that the roles of the CPUs were swapping back and
forth.