PLplot

Alan W. Irwin wrote:
> Hi Hazen:
> 
> I have been testing the efficiency of text clipping for the cairo devices
> for a reasonably modern libcairo (version 1.6.4) and relatively modest
> 2.4GHz PC, and unlike older libcairo versions it is looking good.
> 
> Here is one example:
> 
> software@...> time c/x09c -dev pngcairo -drvopt text_clipping=0 -fam
> -fflen 2 -o cairotest%n.png
> 
> real    0m0.391s
> user    0m0.360s
> sys     0m0.008s
> software@...> time c/x09c -dev pngcairo -drvopt text_clipping=1 -fam
> -fflen 2 -o cairotest%n.png
> 
> real    0m0.485s
> user    0m0.468s
> sys     0m0.012s
> 
> There is a relatively large amount of text in that example (the contour
> labels), and also quite a bit of graphics.  In fact it is that example that
> shows (on the first page) what happens if you don't have text clipped
> properly.  But the total difference in elapsed time is trivial, just 0.094
> seconds.

Sounds okay to me. I'm a little worried that many people will have older 
versions of the library, but perhaps we can have, or at least recommend 
a minimum libcairo version during the build process?

-Hazen

On 2009-03-26 14:06-0400 Hazen Babcock wrote:

> Alan W. Irwin wrote:
>> Hi Hazen:
>> 
>> I have been testing the efficiency of text clipping for the cairo devices
>> for a reasonably modern libcairo (version 1.6.4) and relatively modest
>> 2.4GHz PC, and unlike older libcairo versions it is looking good.
>> 
>> Here is one example:
>> 
>> software@...> time c/x09c -dev pngcairo -drvopt text_clipping=0 -fam
>> -fflen 2 -o cairotest%n.png
>> 
>> real    0m0.391s
>> user    0m0.360s
>> sys     0m0.008s
>> software@...> time c/x09c -dev pngcairo -drvopt text_clipping=1 -fam
>> -fflen 2 -o cairotest%n.png
>> 
>> real    0m0.485s
>> user    0m0.468s
>> sys     0m0.012s
>> 
>> There is a relatively large amount of text in that example (the contour
>> labels), and also quite a bit of graphics.  In fact it is that example that
>> shows (on the first page) what happens if you don't have text clipped
>> properly.  But the total difference in elapsed time is trivial, just 0.094
>> seconds.
>
> Sounds okay to me. I'm a little worried that many people will have older 
> versions of the library, but perhaps we can have, or at least recommend a 
> minimum libcairo version during the build process?

Done.  See the updated (revision 9775) README.release.

Alan
__________________________
Alan W. Irwin

Astronomical research affiliation with Department of Physics and Astronomy,
University of Victoria (astrowww.phys.uvic.ca).

Programming affiliations with the FreeEOS equation-of-state implementation
for stellar interiors (freeeos.sf.net); PLplot scientific plotting software
package (plplot.org); the libLASi project (unifont.org/lasi); the Loads of
Linux Links project (loll.sf.net); and the Linux Brochure Project
(lbproject.sf.net).
__________________________

Linux-powered Science
__________________________

Alan W. Irwin wrote:
> On 2009-03-26 14:06-0400 Hazen Babcock wrote:
> 
>> Alan W. Irwin wrote:
>>> Hi Hazen:
>>>
>>> I have been testing the efficiency of text clipping for the cairo devices
>>> for a reasonably modern libcairo (version 1.6.4) and relatively modest
>>> 2.4GHz PC, and unlike older libcairo versions it is looking good.
>>>
>>> Here is one example:
>>>
>>> software@...> time c/x09c -dev pngcairo -drvopt text_clipping=0 -fam
>>> -fflen 2 -o cairotest%n.png
>>>
>>> real    0m0.391s
>>> user    0m0.360s
>>> sys     0m0.008s
>>> software@...> time c/x09c -dev pngcairo -drvopt text_clipping=1 -fam
>>> -fflen 2 -o cairotest%n.png
>>>
>>> real    0m0.485s
>>> user    0m0.468s
>>> sys     0m0.012s
>>>
>>> There is a relatively large amount of text in that example (the contour
>>> labels), and also quite a bit of graphics.  In fact it is that example that
>>> shows (on the first page) what happens if you don't have text clipped
>>> properly.  But the total difference in elapsed time is trivial, just 0.094
>>> seconds.
>> Sounds okay to me. I'm a little worried that many people will have older 
>> versions of the library, but perhaps we can have, or at least recommend a 
>> minimum libcairo version during the build process?
> 
> Done.  See the updated (revision 9775) README.release.

I was thinking about putting the warning into the cairo.cmake. Is that 
possible? Any pointers on where to look for an example of how to do this?

-Hazen

Hazen Babcock wrote:
> Alan W. Irwin wrote:
>> On 2009-03-26 14:06-0400 Hazen Babcock wrote:
>>
>>> Alan W. Irwin wrote:
>>>> Hi Hazen:
>>>>
>>>> I have been testing the efficiency of text clipping for the cairo devices
>>>> for a reasonably modern libcairo (version 1.6.4) and relatively modest
>>>> 2.4GHz PC, and unlike older libcairo versions it is looking good.
>>>>
>>>> Here is one example:
>>>>
>>>> software@...> time c/x09c -dev pngcairo -drvopt text_clipping=0 -fam
>>>> -fflen 2 -o cairotest%n.png
>>>>
>>>> real    0m0.391s
>>>> user    0m0.360s
>>>> sys     0m0.008s
>>>> software@...> time c/x09c -dev pngcairo -drvopt text_clipping=1 -fam
>>>> -fflen 2 -o cairotest%n.png
>>>>
>>>> real    0m0.485s
>>>> user    0m0.468s
>>>> sys     0m0.012s
>>>>
>>>> There is a relatively large amount of text in that example (the contour
>>>> labels), and also quite a bit of graphics.  In fact it is that example that
>>>> shows (on the first page) what happens if you don't have text clipped
>>>> properly.  But the total difference in elapsed time is trivial, just 0.094
>>>> seconds.
>>> Sounds okay to me. I'm a little worried that many people will have older 
>>> versions of the library, but perhaps we can have, or at least recommend a 
>>> minimum libcairo version during the build process?
>> Done.  See the updated (revision 9775) README.release.
> 
> I was thinking about putting the warning into the cairo.cmake. Is that 
> possible? Any pointers on where to look for an example of how to do this?

Thank you Alan for your help with this.

I've added a version check in cairo.cmake. You'll get a warning if your 
version is less than 1.20.5. To do this I added a version argument to 
the pkg_check_pkgconfig macro and propagated this change to the other 
driver cmake files. I don't think I broke anything...

I have a slightly more recent version of pangocairo (1.22.2) and I also 
get pretty much identical rendering speeds with text_clipping on or off.

best,
-Hazen

Hazen Babcock wrote:
> Hazen Babcock wrote:
>> Alan W. Irwin wrote:
>>> On 2009-03-26 14:06-0400 Hazen Babcock wrote:
>>>
>>>> Alan W. Irwin wrote:
>>>>> Hi Hazen:
>>>>>
>>>>> I have been testing the efficiency of text clipping for the cairo devices
>>>>> for a reasonably modern libcairo (version 1.6.4) and relatively modest
>>>>> 2.4GHz PC, and unlike older libcairo versions it is looking good.
>>>>>
>>>>> Here is one example:
>>>>>
>>>>> software@...> time c/x09c -dev pngcairo -drvopt text_clipping=0 -fam
>>>>> -fflen 2 -o cairotest%n.png
>>>>>
>>>>> real    0m0.391s
>>>>> user    0m0.360s
>>>>> sys     0m0.008s
>>>>> software@...> time c/x09c -dev pngcairo -drvopt text_clipping=1 -fam
>>>>> -fflen 2 -o cairotest%n.png
>>>>>
>>>>> real    0m0.485s
>>>>> user    0m0.468s
>>>>> sys     0m0.012s
>>>>>
>>>>> There is a relatively large amount of text in that example (the contour
>>>>> labels), and also quite a bit of graphics.  In fact it is that example that
>>>>> shows (on the first page) what happens if you don't have text clipped
>>>>> properly.  But the total difference in elapsed time is trivial, just 0.094
>>>>> seconds.
>>>> Sounds okay to me. I'm a little worried that many people will have older 
>>>> versions of the library, but perhaps we can have, or at least recommend a 
>>>> minimum libcairo version during the build process?
>>> Done.  See the updated (revision 9775) README.release.
>> I was thinking about putting the warning into the cairo.cmake. Is that 
>> possible? Any pointers on where to look for an example of how to do this?
> 
> Thank you Alan for your help with this.
> 
> I've added a version check in cairo.cmake. You'll get a warning if your 
> version is less than 1.20.5. To do this I added a version argument to 
> the pkg_check_pkgconfig macro and propagated this change to the other 
> driver cmake files. I don't think I broke anything...
> 
> I have a slightly more recent version of pangocairo (1.22.2) and I also 
> get pretty much identical rendering speeds with text_clipping on or off.

And now I'm not so sure about this result. Alan, can you try with the 
xcairo driver? It seems noticeably slower to me now with text_clipping 
on. Slow enough that I can almost see it drawing each string.

-Hazen

On 2009-04-04 22:49-0400 Hazen Babcock wrote:

>> I have a slightly more recent version of pangocairo (1.22.2) and I also get 
>> pretty much identical rendering speeds with text_clipping on or off.
>
> And now I'm not so sure about this result. Alan, can you try with the xcairo 
> driver? It seems noticeably slower to me now with text_clipping on. Slow 
> enough that I can almost see it drawing each string.

I measure no significant difference.  I held down the return key (to go
through the pages as quickly as possible) and got consistent timings (using
the time command) with text clipping and consistent timings without text
clipping.  And without was consistently faster.  But the difference was only
0.01 seconds, i.e., negligible.  So device xcairo is consistent with the
other cairo devices (none of which showed large changes in speed between
text clipping on or off).

I wonder if your recent rewrite of text handling in cairo is a factor.  The
above test was without your recent changes, i.e., revision 9774 of cairo.c.
You might want to repeat your test with that revision.  Also, I would highly
advise using the time command (or whatever is equivalent on Mac OS X) to
measure your timing results.

Alan
__________________________
Alan W. Irwin

Astronomical research affiliation with Department of Physics and Astronomy,
University of Victoria (astrowww.phys.uvic.ca).

Programming affiliations with the FreeEOS equation-of-state implementation
for stellar interiors (freeeos.sf.net); PLplot scientific plotting software
package (plplot.org); the libLASi project (unifont.org/lasi); the Loads of
Linux Links project (loll.sf.net); and the Linux Brochure Project
(lbproject.sf.net).
__________________________

Linux-powered Science
__________________________

Alan W. Irwin wrote:
> On 2009-04-04 22:49-0400 Hazen Babcock wrote:
> 
>>> I have a slightly more recent version of pangocairo (1.22.2) and I also get 
>>> pretty much identical rendering speeds with text_clipping on or off.
>> And now I'm not so sure about this result. Alan, can you try with the xcairo 
>> driver? It seems noticeably slower to me now with text_clipping on. Slow 
>> enough that I can almost see it drawing each string.
> 
> I measure no significant difference.  I held down the return key (to go
> through the pages as quickly as possible) and got consistent timings (using
> the time command) with text clipping and consistent timings without text
> clipping.  And without was consistently faster.  But the difference was only
> 0.01 seconds, i.e., negligible.  So device xcairo is consistent with the
> other cairo devices (none of which showed large changes in speed between
> text clipping on or off).
> 
> I wonder if your recent rewrite of text handling in cairo is a factor.  The
> above test was without your recent changes, i.e., revision 9774 of cairo.c.
> You might want to repeat your test with that revision.  Also, I would highly
> advise using the time command (or whatever is equivalent on Mac OS X) to
> measure your timing results.

I thought that might be the issue but the speed was the same with & 
without this change. Instead it looks like the issue is somehow related 
to X forwarding with SSH. If I run the examples directly on my linux box 
the text rendering speed is pretty much the same with and without 
clipping, but if I use X forwarding to display the examples on my OS-X 
box then I can see a significant difference. A little puzzling, but 
probably not a reason to go back to the no text clipping default.

-Hazen

Hi Hazen,

>
> I thought that might be the issue but the speed was the same with &
> without this change. Instead it looks like the issue is somehow  
> related
> to X forwarding with SSH. If I run the examples directly on my linux  
> box
> the text rendering speed is pretty much the same with and without
> clipping, but if I use X forwarding to display the examples on my OS-X
> box then I can see a significant difference. A little puzzling, but
> probably not a reason to go back to the no text clipping default.

Have you tried to speed up your ssh connection? Maybe less safe, but  
especially for showing plots this is needed:

http://www.miscdebris.net/blog/2007/06/01/speed-up-ssh-x11-forwarding/

Regards,
Werner

>
> -Hazen
>
>
> ------------------------------------------------------------------------------
> _______________________________________________
> Plplot-devel mailing list
> Plplot-devel@...
> https://lists.sourceforge.net/lists/listinfo/plplot-devel

--
Dr. Werner Smekal
Institut fuer Allgemeine Physik
Technische Universitaet Wien
Wiedner Hauptstr 8-10
A-1040 Wien
Austria

email: smekal@...
web: http://www.iap.tuwien.ac.at/~smekal
phone: +43-(0)1-58801-13463 (office), +43-(0)1-58801-13469 (laboratory)
fax: +43-(0)1-58801-13499