RFR(T): 8214857: "bad trailing membar" assert failure at memnode.cpp:3220

Roland Westrelin rwestrel at redhat.com
Wed Dec 5 10:58:22 UTC 2018


http://cr.openjdk.java.net/~roland/8214857/webrev.00/

When going from a trailing membar to the corresponding leading membar,
if a dead end is hit (top/null control input), the logic must go back to
the last region it encountered and try its next input. Currently the
code pops all regions it encountered.

I consider this change trivial and will push it with a single review.

Roland.


More information about the hotspot-compiler-dev mailing list