On Jan 29, 2009, at 1:54 PM, Douglas Gregor wrote: > > On Jan 29, 2009, at 10:09 AM, Steve Naroff wrote: > >> Author: snaroff >> Date: Thu Jan 29 12:09:31 2009 >> New Revision: 63320 >> >> URL: http://llvm.org/viewvc/llvm-project?rev=63320&view=rev >> Log: >> Hack Sema::LookupDeclInScope() to avoid calling Sema::LookupName() >> when parsing C/ObjC. >> >> This results in a 1.7% improvement for "Cocoa.h". If we can figure >> out how to return a "Decl *", rather than a Sema::LookupResult(), >> we will likely bump the speedup from 1.7%->2.5%. I verified this, >> however couldn't get it to work without breaking a fair number of C+ >> + test cases. Will discuss with Doug offline. > > > That's a lot of code repetition for 1.7%. Let's discuss this offline. >
I would have preferred to optimize LookupName(), however it is more complex and needs to support more C++ features (LookupCriteria, LookupResult, etc.). Since C++ support is a long way off, it would be really nice if name lookup cost (to C/ObjC) were minimal. Do you agree? If so, I'm hopeful we can improve the performance *without* resorting to code copying. From my perspective, a 2.5% penalty for parsing C/ObjC is unacceptable. snaroff > - Doug _______________________________________________ cfe-commits mailing list [email protected] http://lists.cs.uiuc.edu/mailman/listinfo/cfe-commits
