RVM
RVM-778

Fix synchronization and data structures of FinalizableProcessor

Details

Type: Bug
Status: Closed
Priority: Major
Resolution: Fixed
Affects Version/s: None
Fix Version/s: None
Component/s: MMTk
Labels:
None

Description

FinalizableProcessor uses an array that gets resized. The resizing involves a spinloop. The spinloop leads to crashes. Fix it so that it doesn't have to spin, potentially by changing the data structures involved.

Issue Links

is related to

RVM-761 Threads occasionally remain IN_JAVA_TO_BLOCK and never reach a safepoint during a GC stop-the-world request

Activity

Ascending order - Click to sort in descending order

Hide

Permalink

Filip Pizlo added a comment - 08/Feb/09 5:52 AM

Resolved in r15408.

Show

Filip Pizlo added a comment - 08/Feb/09 5:52 AM Resolved in r15408.

Hide

Permalink

Filip Pizlo added a comment - 09/Feb/09 5:08 AM

Repeated runs of xalan in 1-CPU mode, a benchmark/environment known to be particularly affected by this problem, show that stability has probably improved:

Number Of Failures: (out of 100 runs)
revision 15407 : 11
revision 15408 : 3
revision 15409 : 1
revision 15411 : 3

It's not definitive; according to Poisson statistics the error is +-10 on each of these failure counts. But it is indicative of improvement.

Show

Filip Pizlo added a comment - 09/Feb/09 5:08 AM Repeated runs of xalan in 1-CPU mode, a benchmark/environment known to be particularly affected by this problem, show that stability has probably improved: Number Of Failures: (out of 100 runs) revision 15407 : 11 revision 15408 : 3 revision 15409 : 1 revision 15411 : 3 It's not definitive; according to Poisson statistics the error is +-10 on each of these failure counts. But it is indicative of improvement.

People

Assignee:

Filip Pizlo

Reporter:

Filip Pizlo

Vote (0)

Watch (0)

Dates

Created:

08/Feb/09 1:58 AM

Updated:

11/May/09 8:18 AM

Resolved:

08/Feb/09 5:52 AM