Hi all,

I'm trying to find out exactly where the problem is in my app that causes the frame-rate to drop.
I assume perhaps I am not using instruments correctly.
Currently most of my textures are direct outputs from gimp as .png's which I assume are therefore 8888 format.
I briefly tried texture packer's conversion to 1555; but the effect wasn't noticeable (BTW; where can I get a good mac texture format conversion util from?; Imagemagic seems to be the one but no nice gui, and no pre-mac built dmg).
Currently (it's a 2D scroller) I have GL_DEPTH_TEST disabled.
So I assume although I draw front to back *everything* gets drawn regardless of whether it's visible or not.
Currently (as it's 2d) all sprites just have an x/y coord. Do I need to add a Z just so I can use the depth buffer and adjust my vertex draw to take that into account. Would this give a significant speed boost?
Don't really want to go down the PVR root as the texture quality will degrade (and as they are sprites I assume it will be noticeable).

Any other suggestions, or perhaps things to check in instruments appreciated

Cheers

If you're in the iOS paid dev program, you have access to the Xcode 4 betas, which include improved OpenGL ES performance tools.

I am! :-)))
Presumably I just need to 'check for updates' in Xcode and grab the new version?
What improved things do you actually get??

Cheers

You need to download the new version from the developer portal. The changes should be described in the release notes or documentation, but are currently under NDA.

It would seem any alpha (and I mean like 1555 not 4444) cripples my iphone 3g (not 's').
I mean the frame rate dives!
Any advice from anyone as I try foolishly to try and stay at 60fps.?!?!?

Cheers

turn off blending if you don't need it...

I set GL_BLEND to FALSE is this all I need?

//Not sure if this helps or makes no difference either??
glDisable(GL_ALPHA_TEST);

In my quest for 60 fps I took out the z buffer creation from eagle thus:


//glGenRenderbuffersOES(1, &depthRenderbuffer);
//glBindRenderbufferOES(GL_RENDERBUFFER_OES, depthRenderbuffer);
//glRenderbufferStorageOES(GL_RENDERBUFFER_OES, GL_DEPTH_COMPONENT16_OES, backingWidth, backingHeight);
//glFramebufferRenderbufferOES(GL_FRAMEBUFFER_OES, GL_DEPTH_ATTACHMENT_OES, GL_RENDERBUFFER_OES, depthRenderbuffer);


Firstly; is this safe??? (obviously I don't use the z buffer lol).
Secondly; now at a firm 60fps!!!!!!!!!!!!!!!!!!!!!!!

Anyone care to support my method?!? :-)

Cheers

p.s I am guessing as no z-buffer is allocated there is more texture ram; and texture thrashing occurs less?????

p.s.s if this is true then the theory of putting say all player animation frames into 1 texture atlas may be flawed. (i.e. the texture could be quite large). Would it not possibly be better to put all 'walking left' in 1 atlas; walking up' in another etc. etc.
then you will only require a smaller texture per call for the player.
Can anyone confirm or deny my logic(!)??? :-)

p.s.s.s CRUMBS!, my render utilization is down from 95% to 57%!!!

Cheers