RFR (XXS) JDK-8164124,The fix for JDK-8071770 incorrectly uses os::elapsed_counter()

Joseph Provino joseph.provino at oracle.com
Tue Aug 16 19:01:29 UTC 2016


Yes, it seems "now" should also used elapsedTime().

Thanks.

joe


On 8/16/2016 2:35 PM, Jon Masamitsu wrote:
> Joe,
>
> "last" seems to get its value from elapsedTime() (through 
> last_known_gc_end_time_sec()
> from _prev_collection_pause_end_ms in G1Analytics).
> 2479   double last = analytics->last_known_gc_end_time_sec();
> for example
>
>> void G1DefaultPolicy::record_full_collection_end() {
>>   // Consider this like a collection pause for the purposes of allocation
>>   // since last pause.
>>   double end_sec = os::elapsedTime();
>>   double full_gc_time_sec = end_sec - _full_collection_start_sec;
>>   double full_gc_time_ms = full_gc_time_sec * 1000.0;
>>
>>   _analytics->update_recent_gc_times(end_sec, full_gc_time_ms);
>
>
> Should "now" use elapsedTime()?
>
> Jon
>
>
> On 8/16/2016 11:20 AM, Joseph Provino wrote:
>>
>> Please review this one line change:
>>
>> jlong G1CollectedHeap::millis_since_last_gc() {
>>
>> This: jlong now = os::elapsed_counter() / NANOSECS_PER_MILLISEC;
>>
>> Should be: jlong now = os::javaTimeNanos() / NANOSECS_PER_MILLISEC;
>>
>> CR: JDK-8164124 <https://bugs.openjdk.java.net/browse/JDK-8164124> 
>> The fix for JDK-8071770 incorrectly uses os::elapsed_counter()
>>
>> Webrev: http://cr.openjdk.java.net/~jprovino/8164124/webrev.00
>>
>

-------------- next part --------------
An HTML attachment was scrubbed...
URL: <https://mail.openjdk.java.net/pipermail/hotspot-gc-dev/attachments/20160816/d1e02b38/attachment.htm>


More information about the hotspot-gc-dev mailing list