JDK-8072091 : [AARCH64] JSR292: LF sharing regression
  • Type: Bug
  • Component: hotspot
  • Sub-Component: compiler
  • Affected Version: port-stage-aarch64,9
  • Priority: P4
  • Status: Closed
  • Resolution: Duplicate
  • Submitted: 2015-02-02
  • Updated: 2015-02-26
  • Resolved: 2015-02-24
The Version table provides details related to the release that this issue/RFE will be addressed.

Unresolved : Release in which this issue/RFE will be addressed.
Resolved: Release in which this issue/RFE has been resolved.
Fixed : Release in which this issue/RFE has been fixed. The release containing this fix may be available for download as an Early Access Release or a General Availability Release.

To download the current JDK release, click here.
Other
port-stage-aarch64Resolved
Related Reports
Duplicate :  
JDK-8067250 - [mlvm] vm/mlvm/mixed/stress/regression/b6969574 fails and perf regression
Relates :  
JDK-8063135 - Enable full LF sharing by default
Relates :  
JDK-8061543 - JSR 292: Regression: MethodHandle.invokeWithArguments() became slower in jdk9 b31
Description
Test vm/mlvm/mixed/stress/regression/b6969574 fails in PIT (1.9.0-internal-fastdebug-20150121235241.vkozlov.jdk9_aarch64_merg-b00) on all Linux, Solaris and Windows with:

### TRACE 1:    MH.invokeWithArguments(), exact types:   927.4 ns (stddev:  72.4 =  7%) // 11.3 times slower than Reflection API Method.invoke()
### TRACE 1:    MH.invokeWithArguments() + type conv.:   937.6 ns (stddev:  60.9 =  6%) // 11.4 times slower than Reflection API Method.invoke()
### TRACE 1:                invokedynamic instruction:    54.6 ns (stddev:   0.5 =  0%) // 1.5 times FASTER than Reflection API Method.invoke()
### TRACE 1: 
======== Conclusions
### TRACE 1: Comparing invocation time orders
### TRACE 1: Reflection API Method.invoke() <= MH.invokeExact(): Good.
### TRACE 1: MH.invokeExact() <= Direct call: Good.
### TRACE 1: MH.invoke() <= Direct call: Good.
# ERROR: Test marked failed at vm.mlvm.mixed.stress.regression.b6969574.INDIFY_Test.verifyTimeOrder(INDIFY_Test.java:256):
# ERROR: MH.invokeWithArguments(), exact types invocation time order (927.4 ns) is greater than of MH.invokeExact()(78.6 ns)!
The following stacktrace is for Aurora. Used to create a RULE:
nsk.share.TestFailure: Test marked failed at vm.mlvm.mixed.stress.regression.b6969574.INDIFY_Test.verifyTimeOrder(INDIFY_Test.java:256): MH.invokeWithArguments(), exact types invocation time order (927.4 ns) is greater than of MH.invokeExact()(78.6 ns)!
	at nsk.share.Log.logExceptionForAurora(Log.java:410)
	at nsk.share.Log.complain(Log.java:379)
	at nsk.share.Log.complain(Log.java:392)
	at vm.mlvm.share.Env.complain(Env.java:150)
	at vm.mlvm.share.MlvmTest.markTestFailedImpl(MlvmTest.java:252)
	at vm.mlvm.share.MlvmTest.markTestFailed(MlvmTest.java:234)
	at vm.mlvm.mixed.stress.regression.b6969574.INDIFY_Test.verifyTimeOrder(INDIFY_Test.java:256)
	at vm.mlvm.mixed.stress.regression.b6969574.INDIFY_Test.run(INDIFY_Test.java:376)
	at vm.mlvm.share.MlvmTestExecutor.runMlvmTestInstance(MlvmTestExecutor.java:361)
	at vm.mlvm.share.MlvmTestExecutor.runMlvmTest(MlvmTestExecutor.java:305)
	at vm.mlvm.share.MlvmTestExecutor.launch(MlvmTestExecutor.java:222)
	at vm.mlvm.share.MlvmTestExecutor.launch(MlvmTestExecutor.java:164)
	at vm.mlvm.share.MlvmTestExecutor.launch(MlvmTestExecutor.java:135)
	at vm.mlvm.share.MlvmTest.launch(MlvmTest.java:312)
	at vm.mlvm.mixed.stress.regression.b6969574.INDIFY_Test.main(INDIFY_Test.java:400)
Comments
[~kvn] Vladimir, I agree.
04-02-2015
Thank you, Pavel. Do you agree that it is not stopper for Aarch64 open port?
04-02-2015
Looks like JDK-8067250 was reproduced.
04-02-2015
ILW=Perf regression LF sharing, hard to reproduce - closed in 9 as n ot reproducable, none=MLH=P4
04-02-2015
Can we prepare a standalone test for RH to reproduce the failure?
02-02-2015
Why the bug confidential? It is testing for open port.
02-02-2015
can be a reincarnation of JDK-8067250
02-02-2015