[cfe-dev] C++ devirtualization

[Domagoj Šarić]

On 23.4.2014. 18:02, Matthieu Monrocq wrote:
> The only downside that I could  fathom are the redundant store
 > instructions: they are the witnesses that LLVM did not consider them
 > redundant and therefore could not have devirtualized calls without
 > them. Proper meta-data would solve this issue I suppose.

When Clang finally adds support for (something like) MSVC's __assume 
perhaps you could replace the store with something like __assume( vfptr 
== something )...


-- 
"What Huxley teaches is that in the age of advanced technology, 
spiritual devastation is more likely to come from an enemy with a 
smiling face than from one whose countenance exudes suspicion and hate."
Neil Postman
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.llvm.org/pipermail/cfe-dev/attachments/20140426/4788f5a9/attachment.html>