river-dev mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From Peter Firmstone <j...@zeus.net.au>
Subject Re: TaskManager progress
Date Wed, 21 Jul 2010 02:35:29 GMT
Thanks Patricia, looking good, will take some time to digest it further.

We don't have a set of coding conventions, unless someone wants to write 
a tool, there used to be one in com.sun.jini.tool, as evidenced by one 
of the jtreg tests

trunk/qa/jtreg/com/sun/jini/tool/CheckCodeStyle

Perhaps there are some widely available tool that we could settle on?

I like to follow Kent Beck's style in his book Implementation Patterns 
ISBN-10 0-321-41309-1, it's quite a small book and makes easy reading, 
but that's just my personal preference.

Cheers,

Peter.

Patricia Shanahan wrote:
> I've uploaded my current work-in-progress code as 
> http://www.patriciashanahan.com/apache/NewTaskManager.java
>
> Please send me any comments, questions, or suggestions for improvement.
>
> The change of name is temporary, to allow a smoother transition. I 
> plan to work through the callers, changing them one at a time to use 
> the new Task interface. When they have all been changed, and there are 
> no more TaskManager references, the name can be changed to TaskManager.
>
> I'll need to set up the correct formatting in Eclipse, but once I find 
> the rules that won't take long. Any other coding conventions I need to 
> watch out for?
>
> Meanwhile, I'm working on more testing and benchmarking. It definitely 
> improves performance when there are a lot of tasks or runAfter 
> dependencies, but I need to do more testing for short tasks in simple 
> cases, the case in which it is most likely to be worse than the 
> current code.
>
> Patricia
>
>
>
> On 7/20/2010 2:48 PM, Peter Firmstone wrote:
>> Looking forward to seeing some code.  SVN builds clean again.
>>
>>
>> Patricia Shanahan wrote:
>>> I did the first tests of my new TaskManager today. I can't benchmark
>>> very accurately because of a QA test running on the same computer, but
>>> it seems to be about the same without dependencies, and significantly
>>> faster with dependencies. Specifically, it removes the single task
>>> bottleneck.
>>>
>>> I'll next do more testing, benchmarking, and tuning in my own
>>> environment.
>>>
>>> Patricia
>>>
>>
>>
>
>


Mime
View raw message