MASON-INTEREST-L Archives

October 2015

MASON-INTEREST-L@LISTSERV.GMU.EDU
Options: Use Monospaced Font
Show HTML Part by Default
Show All Mail Headers

Message: [<< First] [< Prev] [Next >] [Last >>]
Topic: [<< First] [< Prev] [Next >] [Last >>]
Author: [<< First] [< Prev] [Next >] [Last >>]

Print Reply
Subject:
Re: -parallel
From:
Joey Harrison <[log in to unmask]>
Reply To:
MASON Multiagent Simulation Toolkit <[log in to unmask]>
Date:
Wed, 28 Oct 2015 10:26:20 -0400
Content-Type:
multipart/related
Parts/Attachments:
text/plain (22 kB) , text/html (5 kB) , MasonParallelHeatBugs.png (22 kB) 
In case anyone's curious, here are the results of my quick test:

$ time java -jar HeatBugs.jar -for 10000 -repeat 12 -parallel 1 -quiet

real 0m12.452s
user 0m16.332s
sys 0m2.002s
$ time java -jar HeatBugs.jar -for 10000 -repeat 6 -parallel 2 -quiet

real 0m6.744s
user 0m17.053s
sys 0m2.008s
$ time java -jar HeatBugs.jar -for 10000 -repeat 4 -parallel 3 -quiet

real 0m6.198s
user 0m22.361s
sys 0m2.993s
$ time java -jar HeatBugs.jar -for 10000 -repeat 3 -parallel 4 -quiet

real 0m5.228s
user 0m25.628s
sys 0m3.064s
$ time java -jar HeatBugs.jar -for 10000 -repeat 2 -parallel 6 -quiet

real 0m4.618s
user 0m27.471s
sys 0m2.144s


To summarize:
Jobs, Threads, Time
12, 1, 12.452s
12, 2, 6.744s
12, 3, 6.198s
12, 4, 5.228s
12, 6, 4.618s
[image: Inline image 1]

Not too shabby. I was a bit surprised to see a continued speed-up with 6
threads since I only have 4 cores. Maybe it's a hyper-threading thing or
something.

Joey

On Wed, Oct 28, 2015 at 9:49 AM, Joey Harrison <[log in to unmask]>
wrote:

> Sean,
>
> That certainly changes the situation. Amdahl, here I come!
>
> Joey
>
> On Wed, Oct 28, 2015 at 9:16 AM, <[log in to unmask]> wrote:
>
>> I'm pretty sure that's not it. I'm seeing the same thing when I run it on
>> one of the cluster machines.
>>
>>
>>
>> On Oct 28, 2015, at 6:10 AM, Ernesto Carrella <[log in to unmask]
>> <[log in to unmask]>> wrote:
>>
>> Is your RAM being the bottleneck by any chance?
>>
>> On Wed, Oct 28, 2015 at 5:16 AM Joey Harrison <[log in to unmask]>
>> wrote:
>>
>>> I'm experimenting with the -parallel flag and finding that using 2
>>> parallel threads is slower than a single thread, and using 3 is slower
>>> still. I have 4 cores on my machine, FYI.
>>>
>>> $ time java -jar HeatBugs.jar -for 10000 -repeat 4 -parallel 1 -quiet
>>>
>>> real 0m4.514s
>>> user 0m6.098s
>>> sys 0m0.749s
>>>
>>> $ time java -jar HeatBugs.jar -for 10000 -repeat 4 -parallel 2 -quiet
>>>
>>> real 0m4.799s
>>> user 0m12.019s
>>> sys 0m1.437s
>>>
>>> $ time java -jar HeatBugs.jar -for 10000 -repeat 4 -parallel 3 -quiet
>>>
>>> real 0m5.972s
>>> user 0m21.190s
>>> sys 0m2.933s
>>>
>>> What's the deal here?
>>>
>>> Joey
>>>
>>
>

ATOM RSS1 RSS2