Request for review (L): 7087838: JSR 292: add throttling logic for optimistic call site optimizations
john.r.rose at oracle.com
Fri Sep 16 11:39:58 PDT 2011
Good work; reviewed!
This assert in dependencies is puzzling to read, since it depends on an undocumented invariant:
assert(target == changes->method_handle(), "must be");
I suggest adding an assert on Compile_lock (like many of the other methods of that kind), or at least a comment to the effect that the CS is stable under the Compile_lock.
Note: We may eventually have scaling problems if there are lots of CS's bound into code, *and* lots of new CS's being created. This may require marking a CS with a bit which guarantees (under Compile_lock) there are no dependencies on that particular CS, so that the code cache does not need to be searched when a new CS is reconfigured. This is not an immediate problem, because CallSite.<init> does not use the native method; it just initializes the field.
On Sep 16, 2011, at 5:02 AM, Christian Thalinger wrote:
> [This change will be pushed after 7087357. So ignore the code removal in src/share/vm/classfile/javaClasses.cpp.]
> 7087838: JSR 292: add throttling logic for optimistic call site optimizations
More information about the hotspot-compiler-dev