Some code interpreted with -Xcomp

Gary Benson gbenson at
Tue Apr 15 11:39:58 PDT 2008

Steve Goldman wrote:
> Gary Benson wrote:
> > I'm writing a new compiler for HotSpot.  I've been using -Xcomp
> > to stress it, but I'm even so I'm seeing code being interpreted,
> > some 378,000 instructions or so that happen before
> > CompileBroker::compilation_init() runs.  Is this normal or did
> > I mess something up?
> That's normal, the compilers are delayed. In addition with c2 if you
> run -Xcomp you'll see lots of uncommon traps and interpretation even
> after the compilers come up. Not that it probably applies to you but
> for c2 setting a low compile threshold is more stressful than Xcomp
> unless you are trying to stress deoptimization.

Thanks for that.  I was worried I was doing something odd :)



More information about the hotspot-compiler-dev mailing list