[sldev] OpenJPEG usuable now
jhurliman at wsu.edu
Fri Feb 9 13:50:14 PST 2007
Jason Giglio wrote:
> David Fries wrote:
>> I'm a bit excited.
>> I have some patches that I'm working on, but I was just able to fly a
>> loop around a three region area just now with them and it seemed
>> comparable to the binary client. This is with OpenJPEG and the first
> Good catch! I think we should still pursue the other OpenJPEG
> optimizations, but it sounds like you have solved a major part of it!
> The way I understand it, since all of the CPU slack is taken up
> pushing out frames, you may be able to compare framerates to determine
> relative texture decoding performance.
> Of course a full profile like bushing did would be the ideal. Maybe
> you could get with him on profiling your patches. I think he said he
> has some crazy setup that can compile in a few minutes, so it might
> not be too hard for him. :P
The problem with Ben's setup is that it is built to only analyze raw
decoding times, it doesn't take in to account how SL is actually using
the decoder and doesn't gauge the performance of metadata decoding at
all. To get an accurate picture you would need some sort of usage
statistics for SL (such as: "in the image decoding, 80% of the time is
spent decompressing images in to GL textures and 20% of the time is
spent decoding metadata" or something like that). But it is kind of a
moot point since this is a "Pareto" optimization; it improves
performance without decreasing performance anywhere, and there is no
current better alternative.
More information about the SLDev