Joseph Bowbeer wrote:
> > Are we talking about fusing synch blocks at the bytecode level
> > or at runtime in a dynamic compiler? --Steve K
> This is what I was getting at with my comment about RTJ having its own
> javac: if (non-realtime) javac were permitted to fuse locks then it would be
> "ruining" the code for subsequent use on a RT JVM.
> "javac" should always perform a faithful translation (right?), but what does
> "javac -O" do?
> I'd think "javac -O" would either need to know the target VM, or as you
> suggest it would need to annotate the .class file with information subject
> to interpretation by the VM.
> (B.t.w, does javac -O lead to better performance on HotSpot?)
javac attempts a faithful translation.
So does javac -O; it used to do more but nearly all that it did was buggy
in one way or another so it's optimizations got backed out. -O does not
help HotSpot; I doubt it helps any of the newer jits.
JavaMemoryModel mailing list - http://www.cs.umd.edu/~pugh/java/memoryModel
This archive was generated by hypermail 2b29 : Thu Oct 13 2005 - 07:00:30 EDT