JDK-8058564 : Tiered compilation performance drop in PIT
  • Type: Bug
  • Component: hotspot
  • Sub-Component: compiler
  • Affected Version: 8u40
  • Priority: P1
  • Status: Closed
  • Resolution: Fixed
  • Submitted: 2014-09-16
  • Updated: 2017-07-26
  • Resolved: 2014-09-18
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 8 JDK 9
8u40Fixed 9 b34Fixed
Related Reports
Blocks :  
JDK-7115356 - assert(!m->was_never_executed()) failed
Duplicate :  
JDK-7115356 - assert(!m->was_never_executed()) failed
Duplicate :  
JDK-8043061 - Tiered: MethodCounters should be present for methods enqueued for compilation
Relates :  
JDK-8058184 - Move _highest_comp_level and _highest_osr_comp_level from MethodData to MethodCounters
Description
In latest PIT of hs25.40 several jtreg tests failed due to exceeded timeout.
Performed analysis showed that execution time of following tests increased significantly comparing to jdk8u40-b04:

compiler/5091921/Test6905845.java
compiler/5091921/Test6992759.java
compiler/5091921/Test7020614.java
compiler/7088419/CRCTest.java
java/util/regex/RegExTest.java

For example, java/util/regex/RegExTest.java take ~ 2minutes to finish with jdk8u40-b04 and ~ 15 minutes with PIT build.

Issue could be reproduced with Xcomp and +TieredCompilation. It seems like only compilation on tier 3 is affected: with -TieredCompilation and -XX:TieredStopAtLevel=2 execution time is the same as with 8u40-b04.
Comments
verified in PIT
26-07-2017
ILW=Perf regression in 8u40, PIT, none=HMH=P1
17-09-2014
PIT build spend a lot of time in counter_overflow stub (from 8u40_pit.er): Excl. Incl. Name User CPU User CPU sec. sec. 770.199 770.199 <Total> 102.252 590.283 Runtime1::counter_overflow(JavaThread*,int,Method*) 75.123 620.804 <Unknown> 74.272 74.272 RegisterMap::RegisterMap(JavaThread*,bool) 66.687 80.086 Thread::is_in_stack(unsigned char*)const 49.655 245.942 SimpleThresholdPolicy::event(methodHandle,methodHandle,int,int,CompLevel,nmethod*,JavaThread*) 39.718 632.202 RuntimeStub - counter_overflow Runtime1 stub 34.144 177.264 AdvancedThresholdPolicy::method_invocation_event(methodHandle,methodHandle,CompLevel,nmethod*,JavaThread*) 32.443 32.443 GenericGrowableArray::check_nesting() 24.687 91.354 AdvancedThresholdPolicy::common(bool(AdvancedThresholdPolicy::*)(int,int,CompLevel),Method*,CompLevel,bool) 22.726 46.633 CodeCache::find_blob(void*) 20.965 55.769 frame::sender(RegisterMap*)const 19.944 19.944 CodeHeap::find_start(void*)const 17.022 18.683 AdvancedThresholdPolicy::loop_predicate(int,int,CompLevel) 15.941 17.532 AdvancedThresholdPolicy::call_predicate(int,int,CompLevel) 14.290 14.290 IntervalArray::operator[](int)const 13.640 16.552 nmethod::get_deopt_original_pc(const frame*) 13.399 13.399 os::current_stack_pointer() 11.668 11.668 CompilationPolicy::is_compilation_enabled() 9.256 9.256 Method::is_accessor()const 8.406 8.406 Method::invocation_count() 8.166 11.558 CompileBroker::compilation_is_in_queue(methodHandle,int) 7.405 7.405 Method::backedge_count() 5.884 5.884 <static>@0x2681d38 5.394 5.394 Method::check_code()const 5.344 5.344 CodeBlob::is_nmethod()const
16-09-2014
It looks like performance dropped somewhere around JDK-8023461.
16-09-2014