Re: [JPP-Devel] Streamlining Java2DConverter decimation

2007-06-05 Thread Larry Becker
Setting -Dsun.java2d.d3d=true is also slightly slower than the default. Larry On 6/5/07, Sascha L. Teichmann <[EMAIL PROTECTED]> wrote: > Have you also tried the Direct3D path (-Dsun.java2d.d3d=true)? > > I think the results depend on the graphics hardware and driver quality. > > Sascha > > > Lar

Re: [JPP-Devel] Streamlining Java2DConverter decimation

2007-06-05 Thread Sascha L. Teichmann
Have you also tried the Direct3D path (-Dsun.java2d.d3d=true)? I think the results depend on the graphics hardware and driver quality. Sascha Larry Becker schrieb: > I did try the -Dsun.java2d.opengl=true on Windows XP and it rendered > just a little slower than the default method. > > Larry >

Re: [JPP-Devel] Streamlining Java2DConverter decimation

2007-06-05 Thread Larry Becker
I did try the -Dsun.java2d.opengl=true on Windows XP and it rendered just a little slower than the default method. Larry On 6/5/07, Larry Becker <[EMAIL PROTECTED]> wrote: > Hi Sascha, > >I've done experiments with doing inline transformations (no > AffineTransform), and found that it was no

Re: [JPP-Devel] Streamlining Java2DConverter decimation

2007-06-05 Thread Larry Becker
Hi Sascha, I've done experiments with doing inline transformations (no AffineTransform), and found that it was no faster either. One theory to explain this is that the bottleneck is in the graphics system itself. However, that is contradicted by other experiments that I have done which show t

Re: [JPP-Devel] Streamlining Java2DConverter decimation

2007-06-05 Thread Sascha L. Teichmann
Hi! It's only that temp array, there are actually two. In case of decimation the first allocated is copied over to a second to archive tight fitting. Not to forget the Coordinate->Point2D->Coordinate transformations with the PointConverter, which also introduces a lot of temporal object allocation

Re: [JPP-Devel] Streamlining Java2DConverter decimation

2007-06-04 Thread Larry Becker
I don't have time to look at it closely right now, but it sounds like a logical simplification. I hate that temp array too. regards, Larry On 6/4/07, Michaël Michaud <[EMAIL PROTECTED]> wrote: > Hi Sascha; > > Sounds interesting. > Please, let me some more time to have a closer look and see how

Re: [JPP-Devel] Streamlining Java2DConverter decimation

2007-06-03 Thread Michaël Michaud
Hi Sascha; Sounds interesting. Please, let me some more time to have a closer look and see how your code compare to the one in CVS. Note : I made a recent change in CVS to have the resolution as a property and modify it as needed (default=1/2 pixel) for special renderers. Michael Sascha L. Tei

[JPP-Devel] Streamlining Java2DConverter decimation

2007-06-03 Thread Sascha L. Teichmann
Hi Larry, hi Michaël, I had a look at the decimation code in Java2DConverter. This is awesome! Congratulations! :-) But way not go step further and streamline the model to view coordination transform. Why to create all this temporary Coordinate[] stuff? In the end all what matters is a PathIterat