Quote:
Originally Posted by [ACL]
I have the details on that xda thread i wrote a while back for dev. No one replied :-\..
But alot of the discovery work has been done. Bzo and myself compared notes and math so i dont think there will be any-more changes on the gralloc and lib ebi allocation.
Unfortunately i havent had time to get the gralloc and gles to work on the raph/diam properly which is the only thing preventing it from getting committed.
Right now we need help doing some tests on wince. Also if anyone has time to run tests on 2D fps that would help. I have kernel msgs that claim our 2D can does 30fps. However it does not feel like 30 fps when android is up but only during the boot/bootanim portion. Maybe its because we are pushing less pixels at those times so we can go balls to the wall?
I'm interested to see what fps results we get on 2d in wince. This way we know the limits. Also how does vsync affect our fps? a test with and without vsync enabled can give us a ceiling to shoot for.
Thanks dood.. also feel free to get more mem dumps so we can finally fix that panel down issue for nand  |
I took a look at your post, but it didn't make any sense to me without knowing where the source files were to look at. If you want a third pair of eyes for any of the math, let me know.
I'm getting ready to head home here in 15 minutes and I'll start in on the mmu tracing. The full memory dumps actually caused crashing on my machine when I tried to do diffs within the editor, so I have to find a better way of getting subsets. It's crazy that something so simple could crash a Core i7 950 with 12 GB of RAM...