Originally Posted by rcfa
One big thing, I'd guess the entire OS and all apps will be compiled with LLVM/CLANG, which likely means smaller, faster code.
Well sometimes it produces smaller and faster code. Actually consider how young CLang is it produces very good code.
Possibly they will deliver apps as LLVM "machine" code, meaning the end of fat binaries (aka universal binaries), because in many cases it's faster to translate LLVM code into target code than to link traditional binaries.
Isn't the need for fat binaries pretty much gone? Support for PPC is already gone and 32 bit x86 is dieing real fast.
That is not to dismiss some of the positives that might come from the distribution of LLVM bitcode, i just don't see apple rushing to another CPU architecture.
Add to that the space savings, and the fact that iPhone apps also are going to be LLVM compiled, and we might be looking at truly CPU independent code.
No more need for upgrading any app just because the CPU changed.
Where this will be a big advantage is with GPU code. Apparently this is ptetty much what Apple is doing now. It does deal with the rapid change in GPU architecture better. For your x86 code though I see few advantages.
There is the interesting possibility that your apps could get faster over time as LLVM's processors improve in quality. I'd be surprised if it would be a significant gain though. I'd like to be wrong but I suspect that the mainstream app code will continue to be compiled directly to x86 machine code. They will likely be using the LLVM tools for 10.7, I just don't see a jump to bitcode.