Game Development Community

LLVM Compiler 4.0 can't set fast small.optimization

by Andrea Fraboni · in iTorque 2D · 08/04/2012 (1:46 am) · 5 replies

New XCode 4.4 use new compiler LLVM Compiler 4.0 and when create app with fast and small optimization ... the app crash on all device ....

if i use compiler without optimization app compiled run on all devices.but present freeze on all devices ...

hope apple in the new update modify XCode for good optimization ...

i read online the same.problem was present in th past with compiler from 3.3.2 to 4.0 ..... i think is bad bug ....

is there possible method to compile with optimization ?????

#1
08/05/2012 (4:43 am)
Nope, I have had all sorts of issues with optimisation switched on and the app crashing. I just leave it off.
#2
08/05/2012 (5:40 am)
your app work fine on all devices ???

or freeze ???
#3
08/05/2012 (7:15 am)
It crashes it a number of places on different devices. The only way I stopped the crashing was to set optmization to none.
#4
08/08/2012 (11:22 pm)

Ok we have resolved all by using LLVM GCC 4.2 instead of LLVM compiler 4.0 !!!
#5
08/09/2012 (9:56 am)
I just wouldn't expect LLVM GCC to remain around forever. While Apple has upgraded it, its as a matter of fact more a deprecated project I would guess as it misses some points apple added on their end, some major ones (the LLVM one supports C++11 for example among many other features).
Especially when we keep in mind that iOS 5.1 SDK will likely be the last one that has to support ARMV6, a thing that at least partially requires GCC 4.2 to be present.


also what was read on 3.2.2 -> 4 was WAY back in the days when GCC 3.2.x was upgraded to GCC 4. That was in iOS 2 / 3 days ;)


PS: The optimizations are not remotely as needed anymore. Many of them date back to pre 3GS and the field has really changed with the ARMV7 cpus, their caches and much much higher execution speed and string handling
As such I would look for the PUAP optimize blocks and see which ones you really need to be present so you can enable optimization and go with LLVM