On Feb 10, 2012, at 4:52 AM, Jochen Theodorou wrote: > I still have to figure out when exactly I need that, but currently I have an > catchException on each resulting handle for the call site. of course this is > not good by itself, but I wonder if other have comparable numbers.. According > to my tests I see a 50% performance decrease if I use the exception handler > compared to when not. > > Now I thought this is something the JIT would be able to optimize away > mostly. At least in Java code this normally has no such big impact. Can > somebody confirm that? Maybe even explain why that is so?
I would guess that there is an inlining failure for MethodHandleImpl$GuardWithCatch.invoke_L1 or similar. Can't tell without a trace of inlining decisions. To investigate, I would make a microbenchmark that demonstrates the performance differential, and then run with +LogCompilation, etc. There may also be boxing overheads, since the very special invokers used by GuardWithCatch are not fully signature-polymorphic. -- John -- You received this message because you are subscribed to the Google Groups "JVM Languages" group. To post to this group, send email to jvm-languages@googlegroups.com. To unsubscribe from this group, send email to jvm-languages+unsubscr...@googlegroups.com. For more options, visit this group at http://groups.google.com/group/jvm-languages?hl=en.
