review request (S) - 6849716 BitMap performance regression

[John Coomes]

I (John.Coomes at sun.com) wrote:
> Andrey Petrusenko (Andrey.Petrusenko at Sun.COM) wrote:
> > You could do that with PRODUCT_RETURN, but I like the way Jonh fixed it 
> > - explicitly showing that the verification is only in debug.
> > 
> > P.S.: looks good to me.
> 
> Thanks.
> 
> Ramki also mentioned PRODUCT_RETURN.  That could work, but it makes
> the verification code visible in optimized and profiled builds, which
> could lead to a similar problem of calls to empty debug methods in
> those builds.  So here's an alternative
> 
> 	http://cr.openjdk.java.net/~jcoomes/6849716-not-debug-return/
> 
> This adds a NOT_DEBUG_RETURN macro that returns in builds without
> asserts, and I use that for the BitMap verify routines. ...

After some feedback from Ramki (thanks!), I decided to go with
NOT_DEBUG_RETURN and have updated the webrev in the original location:

	http://cr.openjdk.java.net/~jcoomes/6849716-bitmap-verify/

The only difference from the prior one is that the definitions of
verify_index() and verify_range() are now in bitMap.inline.hpp.  As a
result, I changed an entry in includeDB_compiler1 to keep them
visible.

-John

> > john cuthbertson - Sun Microsystems wrote:
> > > Hi John,
> > > 
> > > Looks good to me. One question: was there a problem using PRODUCT_RETURN?
> > > 
> > > JohnC
> > > 
> > > John Coomes wrote:
> > >> Ensure verification code does not affect the product build.
> > >>
> > >> http://cr.openjdk.java.net/~jcoomes/6849716-bitmap-verify/
> > >>
> > >> -John
> > >>
> > >>   
> > >