[LLVMdev] intended use/structure of AllocA/local variables

[Duncan Sands]

Hi,

On 02/07/13 05:14, edA-qa mort-ora-y wrote:
> I'm trying to determine what is the best approach to using AllocA.
> Currently I just allocate all my variables at the beginning of the
> function but wondering if they should rather be done closer to the scope
> they are used. This would also require a way to do a free on the
> allocated structure, but there doesn't appear to be such a function.

just allocate them at the start and use lifetime intrinsics to tell LLVM
when they are actually in use.

>
> Is it the intent that all stack variables are alloc'd at the beginning
> of the function? I guess this has the advantage that LLVM then knows the
> full size of the function's stack at entry and can just use fixed
> offsets for all variables. It has the disadvantage that all potentially
> used stack variables are always allocated, which uses more stack memory
> than is actually required.
>
> What's the intended practice with AllocA?
>
> Can the optimizer determine when the use of two AllocA spaces is
> exclusive and share the stack space?

The code generators will reuse stack space if it can.  You can help it out
using the lifetime intrinsics mentioned above.

Ciao, Duncan.