[llvm-commits] [llvm] r48940 - /llvm/trunk/lib/Transforms/Scalar/GVN.cpp

[Chris Lattner]

On Mar 29, 2008, at 12:13 AM, Owen Anderson wrote:
> What's the threshold for splitting this and other memory optzns into  
> a separate pass?  The core functionality of GVN is starting to get  
> overwhelmed by the volume of extra things we've tacked onto it?

I agree with you.  Can you take care of this?  It seems like this  
should go in a 'memcpy optimizer' pass, which is scheduled right next  
to GVN.  The scalarstore->memset and returnmemcpy elimination stuff  
should go into it.  If I ever get around to writing a "scalar load 
+store -> memcpy" optimizer, it would also go into this pass.

-Chris