-
Enhancement
-
Resolution: Fixed
-
P3
-
hs25
-
b56
Issue | Fix Version | Assignee | Priority | Status | Resolution | Resolved In Build |
---|---|---|---|---|---|---|
JDK-8027507 | 8 | Roland Westrelin | P3 | Resolved | Fixed | b114 |
We drop some profiling data because it's of no immediate use. Rather than dropping it, we should register it with the type system and let it flow in the graph. At points where the static type or the local profiling doesn't help, we could then query the type system for speculative type data and if there's some emit a guard so it becomes the current type.
- backported by
-
JDK-8027507 C2 needs some form of type speculation
- Resolved
- relates to
-
JDK-8027571 fatal error: meet not symmetric
- Resolved
-
JDK-8027632 assert(xtype->klass_is_exact()) failed: Should be exact at graphKit.cpp
- Resolved
-
JDK-8028536 Test cases to cover type system fixes pushed with 8024070
- Resolved
-
JDK-8162384 Performance regression: bimorphic inlining may be bypassed by type speculation
- Resolved
-
JDK-8321204 C2: assert(false) failed: node should be in igvn hash table
- Resolved
(1 relates to)