This is the mail archive of the
java@gcc.gnu.org
mailing list for the Java project.
Re: need to focus on java performance?
Per Bothner writes:
> David Daney wrote:
> > In x86 code, a call instruction has the same size as a mov to eax. So
> > instead of trapping, initially encode all static accesses as a call to a
> > stub. When patching the code, rewrite it to a load (mov). This avoids
> > all trapping. Non-starter eliminated
>
> We really need to look at this in the context of shared libraries and
> the dynamic linker: Shared libraries add their own overheads and
> indirection. The question is can we make use of the shared library
> indirection without adding extra indirection? I'm assuming the people
> who designed the binary-compatible ABI have looked closely at this,
Yes, but the question always was "what can we do portably?" The part
of the dynamic linker that does fixups is not portable, and if we
don't mind that we can use some of the same tricks.
We can certainly do runtime fixups if we really want to, but at the
expense of libraries not being sharable. This would certainly improve
our benchmark scores.
Andrew.