RFR (S): 8024067: Missing replace_in_map() calls following null checks

Vladimir Kozlov vladimir.kozlov at oracle.com
Tue Sep 3 13:36:29 PDT 2013

Why you need this when null_check_common() does this already?


On 9/3/13 12:52 PM, Roland Westrelin wrote:
> http://cr.openjdk.java.net/~roland/8024067/webrev.00/
> For the replace_in_map() in GraphKit::maybe_cast_profiled_receiver() to propagate the type change, we must first call replace_in_map on the result of the null checks when it's ok to do so. In a subsequent change, I intend for safe_for_replace to be an argument to GraphKit::gen_instanceof() and allow the replace_in_map calls to be done for the Class.isInstance intrinsic but that requires some form of profiling there first.
> Roland.

More information about the hotspot-compiler-dev mailing list