JDK-8251925 : C2: RenaissanceStressTest fails with assert(!had_error): bad dominance
  • Type: Bug
  • Component: hotspot
  • Sub-Component: compiler
  • Affected Version: 11.0.10-oracle,16
  • Priority: P3
  • Status: Resolved
  • Resolution: Fixed
  • Submitted: 2020-08-18
  • Updated: 2021-01-04
  • Resolved: 2020-11-23
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.
JDK 11 JDK 16
11.0.11-oracleFixed 16 b26Fixed
Related Reports
Relates :  
JDK-8249749 - modify a primitive array through a stream and a for cycle causes jre crash
Relates :  
JDK-8257561 - Some code is not vectorized after 8251925 and 8250607
Description
This appears to be a recent regression and unrelated to JDK-8249607 which is a fuzzer test failure detected some time ago. They both hit the same generic assert(!had_error).
Comments
Fix request (11u) I would like to downport this for parity with oracle. Applies clean. This is a follow up for JDK-8249749 which I want to downport to 11.0.11. I would like to push this to 11.0.11, too, as it is a follow up issue of that bug.
22-12-2020
Changeset: e4a32bea Author: Christian Hagedorn <chagedorn@openjdk.org> Date: 2020-11-23 08:37:09 +0000 URL: https://github.com/openjdk/jdk/commit/e4a32bea
23-11-2020
We need to backport this fix into 11u-oracle because JDK-8249749 was backported there.
30-10-2020
[~chagedorn] Please share your reproducer command-line. I'd like to use it as a replay test case.
29-10-2020
If you run with -Djdk.internal.lambda.dumpProxyClasses=DUMP_CLASS_FILES -Djava.lang.invoke.MethodHandle.DUMP_CLASS_FILES=true do you get the .class files you need for replay? [~vlivanov] also has a patch that might help.
28-10-2020
[~dlong] Thanks for the tips. Unfortunately, this does not generate the class file I need. It seems that the required class file is generated dynamically by the benchmark itself. However, I was finally able to reproduce it locally and attach a debugger to it by directly running the benchmark (without going over the RenaissanceStressTest). This should simplify the analysis.
28-10-2020
I think we need to accelerate our Compilation Replay improvement project.
27-10-2020
I attached the latest test failure output lca_output.txt. Note, JDK-8251994 was fixed in jdk-16+22-1282 but latest failure happened in jdk-16+22-1291.
27-10-2020
So the issue is why early node for Bool node: early(n): 2928 IfTrue === 2927 [[ 2929 323 2908 3349 ]] #1 !orig=[3305] !jvms: GenericMutableRow::isNullAt @ bci:2 GeneratedClass$SpecificUnsafeProjection::apply @ bci:397 is the node after loop exit ?: n->out(0): 2899 CountedLoopEnd === 2898 2874 [[ 2900 2923 ]] [lt] P=0.500000, C=13008.000000 !orig=[2386],[1641] !jvms: BaseGenericInternalRow$class::getInt @ bci:5 GenericMutableRow::getInt @ bci:2 GeneratedClass$SpecificUnsafeProjection::apply @ bci:387 2900 IfTrue === 2899 [[ 2893 ]] #1 !orig=[1306] !jvms: BaseGenericInternalRow$class::getInt @ bci:5 GenericMutableRow::getInt @ bci:2 GeneratedClass$SpecificUnsafeProjection::apply @ bci:387 2923 IfFalse === 2899 [[ 2927 ]] #0 !jvms: GenericMutableRow::isNullAt @ bci:2 GeneratedClass$SpecificUnsafeProjection::apply @ bci:397 Looks like domination information for Bool node was not updated when pre-loop was cloned.
27-08-2020
Control domination path: idom[4] 2893 CountedLoop === 2893 320 2900 [[ 2890 2891 2893 2894 2895 ]] inner stride: 1 pre of N2391 !orig=[2391],[2344],[978] !jvms: Integer::intValue @ bci:1 BoxesRunTime::unboxToInt @ bci:12 BaseGenericInternalRow$class::getInt @ bci:5 GenericMutableRow::getInt @ bci:2 GeneratedClass$SpecificUnsafeProjection::apply @ bci:387 idom[3] 2895 If === 2893 1879 [[ 2896 2904 ]] P=0.999999, C=-1.000000 !orig=[1639] !jvms: BoxesRunTime::unboxToInt @ bci:12 BaseGenericInternalRow$class::getInt @ bci:5 GenericMutableRow::getInt @ bci:2 GeneratedClass$SpecificUnsafeProjection::apply @ bci:387 idom[2] 2896 IfTrue === 2895 [[ 2892 2897 ]] #1 !orig=[1304] !jvms: BoxesRunTime::unboxToInt @ bci:12 BaseGenericInternalRow$class::getInt @ bci:5 GenericMutableRow::getInt @ bci:2 GeneratedClass$SpecificUnsafeProjection::apply @ bci:387 idom[1] 2897 If === 2896 2877 [[ 2898 2906 ]] P=0.999999, C=-1.000000 !orig=[977] !jvms: BoxesRunTime::unboxToInt @ bci:15 BaseGenericInternalRow$class::getInt @ bci:5 GenericMutableRow::getInt @ bci:2 GeneratedClass$SpecificUnsafeProjection::apply @ bci:387 idom[0] 2898 IfTrue === 2897 [[ 2876 2881 2886 2899 ]] #1 !orig=[557],1880 !jvms: BoxesRunTime::unboxToInt @ bci:15 BaseGenericInternalRow$class::getInt @ bci:5 GenericMutableRow::getInt @ bci:2 GeneratedClass$SpecificUnsafeProjection::apply @ bci:387 2899 CountedLoopEnd === 2898 2874 [[ 2900 2923 ]] [lt] P=0.500000, C=13008.000000 !orig=[2386],[1641] !jvms: BaseGenericInternalRow$class::getInt @ bci:5 GenericMutableRow::getInt @ bci:2 GeneratedClass$SpecificUnsafeProjection::apply @ bci:387
27-08-2020
Note it is pre-loop: 2893 CountedLoop inner stride: 1 pre of N2391 No vectors should be here.
27-08-2020
From logs: Bad graph detected in compute_lca_of_uses n: 2874 Bool === _ 2875 [[ 2899 ]] [lt] !orig=[2385],[1881] !jvms: BoxesRunTime::unboxToInt @ bci:9 BaseGenericInternalRow$class::getInt @ bci:5 GenericMutableRow::getInt @ bci:2 GeneratedClass$SpecificUnsafeProjection::apply @ bci:387 early(n): 2928 IfTrue === 2927 [[ 2929 323 2908 3349 ]] #1 !orig=[3305] !jvms: GenericMutableRow::isNullAt @ bci:2 GeneratedClass$SpecificUnsafeProjection::apply @ bci:397 n->in(1): 2875 CmpI === _ 2885 3303 [[ 2874 ]] !orig=[2384],[2045] !jvms: BaseGenericInternalRow$class::getInt @ bci:5 GenericMutableRow::getInt @ bci:2 GeneratedClass$SpecificUnsafeProjection::apply @ bci:387 early(n->in(1)): 2928 IfTrue === 2927 [[ 2929 323 2908 3349 ]] #1 !orig=[3305] !jvms: GenericMutableRow::isNullAt @ bci:2 GeneratedClass$SpecificUnsafeProjection::apply @ bci:397 n->in(1)->in(1): 2885 AddI === _ 2894 593 [[ 2875 2894 2914 2925 2929 ]] !orig=1144,... !jvms: BoxesRunTime::unboxToInt @ bci:12 BaseGenericInternalRow$class::getInt @ bci:5 GenericMutableRow::getInt @ bci:2 GeneratedClass$SpecificUnsafeProjection::apply @ bci:387 early(n->in(1)->in(1)): 2893 CountedLoop === 2893 320 2900 [[ 2890 2891 2893 2894 2895 ]] inner stride: 1 pre of N2391 !orig=[2391],[2344],[978] !jvms: Integer::intValue @ bci:1 BoxesRunTime::unboxToInt @ bci:12 BaseGenericInternalRow$class::getInt @ bci:5 GenericMutableRow::getInt @ bci:2 GeneratedClass$SpecificUnsafeProjection::apply @ bci:387 n->in(1)->in(2): 3303 MinI === _ 584 3302 [[ 2875 ]] !orig=[2935] !jvms: GeneratedClass$SpecificUnsafeProjection::apply @ bci:429 early(n->in(1)->in(2)): 2928 IfTrue === 2927 [[ 2929 323 2908 3349 ]] #1 !orig=[3305] !jvms: GenericMutableRow::isNullAt @ bci:2 GeneratedClass$SpecificUnsafeProjection::apply @ bci:397 LCA(n): 2898 IfTrue === 2897 [[ 2876 2881 2886 2899 ]] #1 !orig=[557],1880 !jvms: BoxesRunTime::unboxToInt @ bci:15 BaseGenericInternalRow$class::getInt @ bci:5 GenericMutableRow::getInt @ bci:2 GeneratedClass$SpecificUnsafeProjection::apply @ bci:387 n->out(0): 2899 CountedLoopEnd === 2898 2874 [[ 2900 2923 ]] [lt] P=0.500000, C=13008.000000 !orig=[2386],[1641] !jvms: BaseGenericInternalRow$class::getInt @ bci:5 GenericMutableRow::getInt @ bci:2 GeneratedClass$SpecificUnsafeProjection::apply @ bci:387 n->out(0)->out(0): 2900 IfTrue === 2899 [[ 2893 ]] #1 !orig=[1306] !jvms: BaseGenericInternalRow$class::getInt @ bci:5 GenericMutableRow::getInt @ bci:2 GeneratedClass$SpecificUnsafeProjection::apply @ bci:387 n->out(0)->out(1): 2923 IfFalse === 2899 [[ 2927 ]] #0 !jvms: GenericMutableRow::isNullAt @ bci:2 GeneratedClass$SpecificUnsafeProjection::apply @ bci:397 *** Use 2898 isn't dominated by def 2874 ***
27-08-2020