[Webkit-unassigned] [Bug 67460] Add JSVALUE32_64 support to DFG JIT

bugzilla-daemon at webkit.org bugzilla-daemon at webkit.org
Wed Sep 14 21:39:41 PDT 2011


--- Comment #5 from Filip Pizlo <fpizlo at apple.com>  2011-09-14 21:39:41 PST ---
(In reply to comment #4)
> (In reply to comment #3)
> > Hi,
> > 
> > I noticed big changes recently with the introduction of "tierd compilation" in JSC, which supports switching between speculative DFG JIT and the old JIT.
> > Will this become the future direction of JSC? And will the non-speculative JIT be abandoned (and the estimated timeline)? I ask this because it will help us make right choices in our development work on enabling DFG JIT on 32-bit platforms.
> > 
> > Thanks a lot.
> Hi!
> Tiered compilation is definitely the future direction of JSC.  This means that the non-speculative JIT will definitely be removed.
> Performing this switch is predicated upon tiered compilation being a pure win on all benchmark suites, and having zero major regressions in individual benchmarks (though we expect small regressions on very short-running benchmarks).
> Right now this is already true when running on the command-line (jsc), but is not yet true in browser.  These in-browser bugs are being fixed as we speak.
> My optimistic estimate is that tiered compilation will be turned on in trunk in the next week.  My pessimistic estimate is that this will happen in two weeks.  The non-speculative JIT, and all code for static prediction and speculation, will likely be removed shortly thereafter.
> -Filip

Note that our performance target has been met as of today.  As soon as all of the bots are happy we will enable tiered compilation in trunk by default (for platforms that support DFG) and proceed with removing the non-speculative JIT and static prediction logic.


Configure bugmail: https://bugs.webkit.org/userprefs.cgi?tab=email
------- You are receiving this mail because: -------
You are the assignee for the bug.

More information about the webkit-unassigned mailing list