[cfe-commits] r138220 - in /cfe/trunk: include/clang/StaticAnalyzer/Core/PathSensitive/ExprEngine.h lib/StaticAnalyzer/Core/CFRefCount.cpp lib/StaticAnalyzer/Core/ExprEngineCallAndReturn.cpp lib/StaticAnalyzer/Core/ExprEngineObjC.cpp

[Jordy Rose]

On Aug 21, 2011, at 22:58, Ted Kremenek wrote:

> On Aug 21, 2011, at 6:03 PM, Jordy Rose <jediknil at belkadan.com> wrote:
> 
>> This is running a Debug build and watching the total time of the run, without attempting to quiet the rest of my system -- in other words, entirely anecdotal. I feel like the total time has crept up by a second or so, though.
>> 
> 
> Ah, you are running the tests.  That wasn't clear.
> 
> I'd be very surprised if these changes resulted in a noticeable performance change, even in debug mode.  It's probably worth investigating where the time is going, or actually measuring the time (even 'time make test') to make sure this isn't a real regression.  This really isn't the hot path in the analyzer, so 1 second would be possibly a big deal.

Well, any changes to evalCall are in the hot path, even if they're doing the same work before and after.

I ran 'time TESTDIRS=Analysis make test' a couple times before and after the change, and while the usertime average does seem to have gone up by a second or two, the error margin is something like 10s just with Mail.app running in the background. So never mind.