andykaylor wrote:

> Hmm... perhaps this approach is misguided then. I'm beginning to question the 
> idea of hoisting at all, what is the purpose here if we end up not being able 
> to keep its read/write consistency? Should we just skip hoisting for 
> constants?

As I mentioned offline earlier, I think we need to hoist the allocas for 
consistency with classic codegen and to have the stack space for the variables 
allocated at function entry. So, I think hoisting the allocas is good, but I'm 
not so certain about the invariant handling. 

https://github.com/llvm/llvm-project/pull/175037
_______________________________________________
cfe-commits mailing list
[email protected]
https://lists.llvm.org/cgi-bin/mailman/listinfo/cfe-commits

Reply via email to