[PATCH] Change objectsize intrinsic to accept different address spaces.

Eric Christopher echristo at gmail.com
Thu Oct 3 18:27:50 PDT 2013

Right. I was trying to figure out if it's because we actually expected
the memory coming in to be something other than addrspace 0. I.e. does
the fortify checking really make sense for that case?


On Thu, Oct 3, 2013 at 6:16 PM, Matt Arsenault
<Matthew.Arsenault at amd.com> wrote:
> On 10/03/2013 06:03 PM, Eric Christopher wrote:
>> I'm curious how you're generating the intrinsics that would need the
>> addrspace cast? I'm not necessarily objecting, just figuring out if
>> it's a change that necessarily needs to be made or if a change in use
>> would suffice.
>> Thanks!
> I'm working on fixing everywhere that can introduce bitcasts between address
> spaces. You need this if you want objectsizes on multiple objects in
> different address spaces to coexist. Since these are compile time constants
> that just try to find object definitions and then disappear, it might work
> to addrspacecast everything to address space 0? I'm not 100% sure that would
> always be OK though.

More information about the llvm-commits mailing list