Updates:
Status: Assigned
Owner: [email protected]
Comment #3 on issue 2096 by [email protected]: Fixing Navier-Stokes
performance on ARM.
http://code.google.com/p/v8/issues/detail?id=2096
Assigning to myself as I have started to look at this on Friday as well in
relation to our deopts on some Box2D code.
I think this optimistic assumption fails us on any benchmark that has stale
integer feedback coming from a rarely executed conditional branch.
I'll try to repro it on ARM hardware. It can also be that something is
wrong with ICs type-feedback on ARM as it closely tied to BinaryOpIC
implementation which is arch dependent.
--
v8-dev mailing list
[email protected]
http://groups.google.com/group/v8-dev
