Updates:
Summary: Share optimized code across closure instances
Status: Assigned
Owner: [email protected]
Comment #1 on issue 2087 by [email protected]: Share optimized code
across closure instances
http://code.google.com/p/v8/issues/detail?id=2087
It's not a decision to deoptimize function, but a decision not to optimize
this function.
V8 atm does not share optimized code across different instances of the same
function literal and does not distinguish between reoptimization after
deoptimization and optimization of a newly created function literal.
To avoid optimizing the same function many times we artificially limit
number of optimizations (see Compiler::kDefaultMaxOptCount).
It's a known problem and we have been discussing optimized code sharing or
decoupling of deopt counter from opt counter. But both approaches have some
inherent problems (e.g. memory leakage through pointers embedded into the
code).
--
v8-dev mailing list
[email protected]
http://groups.google.com/group/v8-dev
