[PATCH] Don't use runtime bounds check between different address space pointers

[Matt Arsenault]

On 10/03/2013 03:37 AM, Duncan Sands wrote:
> I don't know what this check is doing exactly, but if two pointers are in
> different address spaces, doesn't that automatically mean they don't 
> alias?
> If so, then maybe here they should be considered to compare as "not 
> equal".
No, it was decided that address spaces in LLVM aren't required to be 
non-overlapping. For GPU address spaces it used to be true that they 
would never alias, but more recently there is now a flat address space 
which can alias all others.


See these threads:
http://lists.cs.uiuc.edu/pipermail/llvm-commits/Week-of-Mon-20111010/129615.html
http://lists.cs.uiuc.edu/pipermail/llvmdev/2013-August/064620.html