RFR: 8232594: Make the output of the JFR command duration more user friendly

Chihiro Ito chiroito107 at gmail.com
Sat Nov 16 16:49:58 UTC 2019


Hi Erik,

Could you please review this?

JBS : https://bugs.openjdk.java.net/browse/JDK-8232594
Webrev : http://cr.openjdk.java.net/~cito/JDK-8232594/webrev.01/

I improved this based on your idea and test case.

The test was passed test-tier1 and jdk/jdk/jfr.

Regards,
Chihiro


2019年11月11日(月) 23:57 Chihiro Ito <chiroito107 at gmail.com>:

> Hi Erik,
>
> Thank you for giving the idea and implementing unit tests.
> This is good for me.
>
> I will improve according to this idea.
>
> Regards,
> Chihiro
>
> 2019年11月8日(金) 2:39 Erik Gahlin <erik.gahlin at oracle.com>:
>
>> Hi Chihiro,
>>
>> Markus and I discussed what would be a suitable format and this is what
>> we came up with:
>>
>> - A duration of 0 is displayed as "0 s"
>> - A duration less than 1 second is displayed in ms with 3 significant
>> digits.
>> - A duration between 1 second and 60 seconds is displayed in s with three
>> significant units
>> - A duration of 60 second or more uses a combination of two units, for
>> example "1 m 30 s" or "3 h 30 m"
>> - The maxmim unit is a day.
>>
>> I wrote a unit tests that you can use. Hopefully it is correct.
>>
>> http://cr.openjdk.java.net/~egahlin/8232594/
>>
>> Thanks
>> Erik
>>
>> On 2019-10-31 15:47, Chihiro Ito wrote:
>>
>> Hi Erik,
>>
>> I can't imagine one event lasting more than a day.
>> Therefore, I think the number of days is unnecessary.
>> I also think we don't need duration as detailed as ns.
>>
>> So how about using 0.001 ms to hours?
>>
>> Also, it would be difficult to see all the units displayed.
>> About rounding, how about using up to 3 units?
>>
>> For example:
>> 27 h 34 m 56 s
>> 34 m 56 s 789 ms
>> 56 s 789.012 ms
>> 789.012 ms
>>
>> Regards,
>> Chihiro
>>
>>
>> 2019年10月30日(水) 2:57 Erik Gahlin <erik.gahlin at oracle.com>:
>>
>>> Hi Chihiro,
>>>
>>> I don't think 1 day is a good representation of for example 27 hours.
>>> The precision is too low.  I also think numbers should be rounded up
>>> where applicable.
>>>
>>> When a duration is short, it is often convenient to use ms (with some
>>> fractions) instead of microseconds. It makes it easier to compare
>>> values. This is especially true if you are looking at multiple values
>>> and want to compare them, for example GC pause time.
>>>
>>> It is probably a good idea to first decide how values should be
>>> formatted before attempting an implementation.
>>>
>>> Erik
>>>
>>> > Hi,
>>> >
>>> > I tried to implement this to output an accurate time duration.
>>> > Could you review this, please?
>>> >
>>> > Is a rounded duration better than an exact duration?
>>> >
>>> > JBS https://bugs.openjdk.java.net/browse/JDK-8232594
>>> > webrev http://cr.openjdk.java.net/~cito/JDK-8232594/webrev.00/
>>> >
>>> > Regards,
>>> > Chihiro
>>>
>>>


More information about the hotspot-jfr-dev mailing list