What can be disabled for iPhone Optimisation

[Garfinkle]

Hi All,

Just wondering, as my project is only targeting the iPhone, are there any sections of the irrlicht engine that I can completely disable or remove from the libraries to keep them smaller and faster?

Thanks in advance.

[kvakvs]

In irrcompileconfig.h (if i remember correct the file name, but you should have no trouble finding it anyway) there is plenty of options to be disabled. Few of them are depending on others, but you will figure that fast, when it will refuse to link without something, you recently have disabled.

[Garfinkle]

Thanks kvakvs, I'll have a look through

[Garfinkle]

By commenting out the functions that I won't be using, psd image support etc...i have managed to jump from approx 30 fps to 50 - 60 fps.... definately an improvement in my books

[slavik262]

o.0

What does the size of the library have to do with execution speed?

[Garfinkle]

By disabling particular functions, then specific checks are not made during the execution of other methods, such as file format checking, graphics rendered checking etc.

Also by having a smaller library then less code needs to be loaded in to memory at run time allowing for more memory to be allocated to other areas resulting in fast loaded and processing.

[slavik262]

By disabling particular functions, then specific checks are not made during the execution of other methods, such as file format checking, graphics rendered checking etc.

Sure, but I don't think these checks happen as much as you think - if you look at the source, you'll find that most of it is abstracted (for example, image file format is completely irrelevant once the image is loaded thanks to the IImage interface).

Also by having a smaller library then less code needs to be loaded in to memory at run time allowing for more memory to be allocated to other areas resulting in fast loaded and processing.

Faster loading time, yes. Faster processing, not necessarily, unless you start talking about cache coherency. RAM is just that - random access. You can access anywhere in the RAM in the same amount of time.

[Garfinkle]

Faster loading time, yes. Faster running, not necessarily, unless you start talking about cache coherency.

Faster running is more noticable on devices such as the iPhone due to the limited memory capacity. By having less running natively in the engine allows you that bit extra memory capacity to load more objects in cache (which in turn allows you to write code that can use this cached information). Therefore you gain those few extra Frames.

[slavik262]

By having less running natively in the engine

The size of the library does not change the number of processor registers. The same "amount" is always running.

...that bit extra memory capacity to load more objects in cache (which in turn allows you to write code that can use this cached information). Therefore you gain those few extra Frames.

What on earth do you mean? The processor cache isn't something you can directly control; you can only try to increase cache coherency by writing efficient code. You can't "load more objects" into it. Are you just talking about the RAM itself? I know iPhones and other embedded devices have a RAM limit, and I'm not disputing that Irrlicht runs faster on the iPhone when you strip unneeded stuff out, but you're not making sense.

[ent1ty]

I think he's trying to say that smaller library leaves more free RAM