[llvm-dev] Existing studies on the benefits of pointer analysis

[Jia Chen via llvm-dev]

On 03/25/2016 08:08 PM, Chris Lattner wrote:
> I’m still a big fan of context sensitive, flow insensitive, 
> unification based models.

Interestingly I find the unification approach quite unsatisfactory 
sometime. What happens there is pointers with the same "depth" are too 
often clobbered together unless they are really unrelated to each other.

> Contrary to your claim, context sensitivity *is* useful for mod-ref 
> analysis, e.g. “can I hoist a load across this call”?  Context 
> sensitivity improves the precision of the mod/ref set of the call.
>
I'm not sure about that. How often does mod-ref information change 
across callsites? Isn't a good context-insensitive function summary good 
enough?

-Jia

-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.llvm.org/pipermail/llvm-dev/attachments/20160325/2473e70e/attachment.html>