RFR(S) 8031994: java/lang/Character/CheckProp test times out
igor.veresov at oracle.com
Mon Jun 2 08:36:41 UTC 2014
Looks good. I think it makes sense to emulate the interpreter here since C2 uses the fact the the class was unresolved as one of the profiling points (if it’s unresolved - it’s never been executed).
On Jun 1, 2014, at 8:30 PM, Rickard Bäckman <rickard.backman at oracle.com> wrote:
> can I please have a couple of review for this change?
> It starts out as a test that times out, the cause being that ciTypeFlow
> decides that one of the byte codes will always trap (a new
> java/io/FileReader) because the class has not been resolved in the
> constant pool and the compilation we are doing is an OSR compilation
> with entry far later than the trapping bytecode. This causes the OSR
> compilation to fail when trying to match blocks.
> The change makes C1 check if we need to update the constant pool for
> classes (running with TieredCompilation).
> Webrev: http://cr.openjdk.java.net/~rbackman/8031994/
> Bug: https://bugs.openjdk.java.net/browse/JDK-8031994
More information about the hotspot-compiler-dev