RFR (M): 8210462: Fix remaining mentions of initial mark

Thomas Schatzl thomas.schatzl at oracle.com
Fri Jul 3 18:06:33 UTC 2020


Hi Leo,

   thanks for your review.

On 02.07.20 23:43, Leo Korinth wrote:
> 
> in: src/hotspot/share/gc/g1/g1Policy.cpp
> After initializing the variable "this_pause", maybe use it instead of 
> young_gc_pause_kind() two statements below?
> 
> in: src/hotspot/share/gc/g1/g1Policy.hpp
> #include "gc/g1/g1HeapRegionAttr.hpp"
> #include "gc/g1/g1ConcurrentStartToMixedTimeTracker.hpp"
> #include "gc/g1/g1MMUTracker.hpp"
> (no longer sorted after renaming)
> 
> in: src/hotspot/share/gc/g1/g1InitialMarkToMixedTimeTracker.hpp
> #include "utilities/globalDefinitions.hpp"
> #include "utilities/debug.hpp"
> (not sorted, pre-existing)
> 

All fixed.

> Some grepping (egrep -Ri "initial.?mark") shows more potential changes in:

Your query string has obviously been better than mine...

> open/src/jdk.hotspot.agent/share/classes/sun/jvm/hotspot/gc/shared/GCCause.java:  
> _cms_initial_mark ("CMS Initial Mark"),
> (oops, my fault, I should have removed it with CMS)
> 

Let's do this separately.

> open/src/hotspot/share/gc/g1/g1CollectedHeap.inline.hpp:    res |= 
> G1EvacuationFailureALotDuringInitialMark;
> open/src/hotspot/share/gc/g1/g1_globals.hpp:  develop(bool, 
> G1EvacuationFailureALotDuringInitialMark, true,             \
> (maybe not rename the flag, unsure about the amount of work when it is a 
> developer flag, just want to mention it)
> 

Renaming is easy because it is a development flag. Done.

> open/src/jdk.hotspot.agent/share/classes/sun/jvm/hotspot/gc/shared/G1YCType.java:  
> InitialMark ("Initial Mark"),
> open/src/hotspot/share/gc/g1/g1Policy.cpp: 
> assert(!is_concurrent_start_pause(this_pause), "The young GC before 
> mixed is not allowed to be an initial mark GC");
> open/src/hotspot/share/gc/g1/g1CollectedHeap.cpp:    // Try to schedule 
> an initial-mark evacuation pause that will
> open/src/hotspot/share/gc/g1/g1CollectedHeap.cpp:      // collection an 
> initial-mark won't be honored.  If we don't check for
> open/src/hotspot/share/gc/g1/g1CollectedHeap.cpp:  // We do not allow 
> initial-mark to be piggy-backed on a mixed GC.

Fixed.

> open/test/hotspot/gtest/gc/shared/test_preservedMarks.cpp:  // Make sure 
> initial marks are correct.

This is correct, and I kept it.

> open/test/hotspot/jtreg/gc/g1/TestGCLogMessages.java:        new 
> TestGCLogMessages().testWithInitialMark();
> open/test/hotspot/jtreg/gc/g1/TestGCLogMessages.java:    private void 
> testWithInitialMark() throws Exception {
> open/test/hotspot/jtreg/gc/g1/TestGCLogMessages.java: 
> GCTestWithInitialMark.class.getName());
> open/test/hotspot/jtreg/gc/g1/TestGCLogMessages.java:    static class 
> GCTestWithInitialMark {
> open/test/hotspot/jtreg/gc/g1/TestHumongousAllocInitialMark.java: * 
> @test TestHumongousAllocInitialMark
> open/test/hotspot/jtreg/gc/g1/TestHumongousAllocInitialMark.java: * @run 
> driver gc.g1.TestHumongousAllocInitialMark
> open/test/hotspot/jtreg/gc/g1/TestHumongousAllocInitialMark.java:public 
> class TestHumongousAllocInitialMark {

Done.

Webrevs:

http://cr.openjdk.java.net/~tschatzl/8210462/webrev.0_to_1/ (diff)
http://cr.openjdk.java.net/~tschatzl/8210462/webrev.1/ (full)

Thanks,
   Thomas


More information about the hotspot-gc-dev mailing list