PulpCorePlayer performance vs in browser performance (me too)

[axel]

Hi,

decided to give PulpCore a try and installed it today using Eclipse.
I'm new to Eclipse too, and the "how to" for Eclipse at
http://code.google.com/p/pulpcore/wiki/Eclipse obviously refers to an
older version of Eclipse, but I'm moderately confident I could follow
it nonetheless. The renamed BubbleMark compiles and runs. But very
slow.

When I started it in browser from the pulp core web pages, 60 fps was
no problem. When I managed to switch off "Cap Frame Rate" (did not
work for Firefox, but worked in IE and Chrome) 250 fps was no problem.

When I start it from Eclipse I get < 30 fps. So about 1/10th of the
performance.

When I start it in debugger, it is another 10 times slower: about 1
frame per second. (But in-game display still claims to run at ~30
fps???).

I've been searching for the problem for hours.

Using debug or release version of PulpCore didn't seem to make any
difference.

Using default player instead of PulpCore Player made no difference.

Another interesting observation: when I turn on View/"Dirty
Rectangles" in the player started from Eclipse, it runs much faster,
60 fps no problem. But when launched with debugger attached, no speed
difference is visible; both ways it runs with about one visible update
per second.

When run in debugger, and I press break, the active call stack is
almost always something like:

CompositeSrcOver(Composite).getPixelBilinearTranslucent(int[], int,
int, int, int, int, int) line: 250
CompositeSrcOver.blend(int[], int, boolean, int, int, int, int, int,
int, int, int, int, boolean, boolean, int, int[], int, int, int, int)
line: 313
CoreGraphics.internalDrawScaledImage(CoreImage, int, int, int, int,
int, int, int, int) line: 1169
CoreGraphics.internalDrawScaledImage(CoreImage, int, int, int, int)
line: 1078
CoreGraphics.drawImage(CoreImage, int, int, int, int) line: 807
CoreGraphics.drawImage(CoreImage) line: 786
BubbleMark$BallSprite(ImageSprite).drawSprite(CoreGraphics) line: 274

In other words: it looks like the program is busy drawing sprites
pixel by pixel. At first I suspected this to be the problem, but the
FAQ says this is normal and "the software renderer has consistent
performance across platforms and JREs". In this case it looks like it
does not? But why?

Finally I found an older post here complaining about the same problem,
but without solution;
http://groups.google.com/group/pulpcore/browse_thread/thread/26a60e00efd1f1eb/374f30a8ee4f0b34?lnk=gst&q=performance#374f30a8ee4f0b34
"PulpCorePlayer performance vs in browser performance"

Also I wonder if this post is related to my problem
http://groups.google.com/group/pulpcore/browse_thread/thread/ebaab17271323e02/bb98ba1e3f5d65e6?lnk=gst&q=performance#bb98ba1e3f5d65e6

PulpCore looks like a cool piece of smart software, worth using very
much. But not being able to run it from the IDE with debugger (at a
realistic speed) is quite a show stopper. And programming with the
constant threat of unexplainable severe performance degradation is
something I'd rather like to avoid... :-(

Anyone any idea what I might do wrong? Or what I could do better?
Axel

[David Brackeen]

Sorry, I don't think anyone ever found out why performance was poor in Eclipse. I'm guessing it's an AWT/SWT conflict. If it helps, you can turn off dirty rectangles programmatically. 

Axel

--
You received this message because you are subscribed to the Google Groups "PulpCore" group.
To post to this group, send email to pulp...@googlegroups.com.
To unsubscribe from this group, send email to pulpcore+u...@googlegroups.com.
For more options, visit this group at http://groups.google.com/group/pulpcore?hl=en.

[Jaime Hablutzel Egoavil]

not familiar to PulpCore but for gaining performance in debugging you can add Step Filters, so you exclude java packages and classes that shouldn't be inspected as awt pulpcore packages... and letting only your own.

--
Jaime Hablutzel - 9-9956-3299
 
(tildes omitidas intencionalmente)

[Troy Cox]

This won't help with Eclipse. If you need free, I remember having a
good experience with NetBeans when I first was playing around with
PulpCore. I had a little trouble learning how to profile with it.

Currently, I personally wouldn't want to use anything but IntelliJ for
java. I've never noticed this problem with it. And it is far easier to
setup (for me).

On Apr 9, 6:13 pm, axel <goo...@walthelm.net> wrote:
> Hi,
>
> decided to give PulpCore a try and installed it today using Eclipse.

> I'm new to Eclipse too, and the "how to" for Eclipse athttp://code.google.com/p/pulpcore/wiki/Eclipseobviously refers to an

> but without solution;http://groups.google.com/group/pulpcore/browse_thread/thread/26a60e00...

> "PulpCorePlayer performance vs in browser performance"
>

> Also I wonder if this post is related to my problemhttp://groups.google.com/group/pulpcore/browse_thread/thread/ebaab172...

[Jaime Hablutzel Egoavil]

step filtering doesn't help improve performance when debugging?? hmmm... ohh you are right, it is just for avoiding stepping into some code, my apologies...

--
You received this message because you are subscribed to the Google Groups "PulpCore" group.
To post to this group, send email to pulp...@googlegroups.com.
To unsubscribe from this group, send email to pulpcore+u...@googlegroups.com.
For more options, visit this group at http://groups.google.com/group/pulpcore?hl=en.