- Better reasoning about array properties (e.g. an Integer[] must be flat).
- I notice there are times when the strategy to do a type assert of the array fails and we have to go through the full series of type check.
- Better profiling, it should answer the questions: has a flat array been observed, has a non-flat array been observed, etc.
- relates to
-
JDK-8228355 [lworld] Optimize abstract class / interface (arrays) not implemented by inline types
-
- Open
-
-
JDK-8251971 [lworld] Improve performance of arraycopy intrinsics
-
- Open
-
-
JDK-8253160 [lworld] C2's superword optimization should vectorize flat inline type array accesses
-
- Open
-
-
JDK-8255938 [lworld] C2 should optimize inline array access when type is only known after parsing
-
- Open
-
-
JDK-8355382 [lworld] Optimization added by JDK-8228622 only applies to null-free, flat arrays
-
- Open
-
-
JDK-8368503 [lworld] Use profile information from (bimorphic) call site to optimize previous array load
-
- Open
-
-
JDK-8231346 [lworld] Further optimize loading/storing element of known type to object array
-
- Open
-
-
JDK-8252027 [lworld] C2 should optimize clone from non-escaping array with ZGC
-
- Open
-