Hi Dave,
Thanks for your reply. I did not check the lines put_rx and get_rx as
I used the X10_TRACE_RXTX. Now it is clear and everything is fine.
BR,
Long
On 3/5/13, David P Grove <gro...@us.ibm.com> wrote:
> Hi,
>
> I was wondering what version of X10 you are using? When I run your
> test program on some inputs I see
>
> [dgrove@linchen samples]$ ./a.out 5
> 5 is done
> Place: 0 msg_rx: 199/10 msg_tx: 193/8
> Place: 0 put_rx: 0(&0)/0 put_tx: 28(&20)/1
> Place: 0 get_rx: 0(&0)/0 get_tx: 0(&0)/0
> Place: 1 msg_rx: 201/9 msg_tx: 207/11
> Place: 1 put_rx: 28(&20)/1 put_tx: 0(&0)/0
> Place: 1 get_rx: 0(&0)/0 get_tx: 0(&0)/0
> [dgrove@linchen samples]$ ./a.out 500
> 500 is done
> Place: 0 msg_rx: 199/10 msg_tx: 193/8
> Place: 0 put_rx: 0(&0)/0 put_tx: 28(&2000)/1
> Place: 0 get_rx: 0(&0)/0 get_tx: 0(&0)/0
> Place: 1 msg_rx: 201/9 msg_tx: 207/11
> Place: 1 put_rx: 28(&2000)/1 put_tx: 0(&0)/0
> Place: 1 get_rx: 0(&0)/0 get_tx: 0(&0)/0
>
> The bytes moved via asyncCopy show up in the put_rx and get_rx lines, and
> do seem to be growing as expected as the input size grows.
>
> --dave
>
>
> Long Cheng <parach...@gmail.com> wrote on 03/04/2013 08:07:51 AM:
>>
>> I am trying to monitor the message send/receive between each place
>> with X10_TRACE_RXTX for my codes, but it seems there is something
>> wrong with the profiling results that: The RX bytes/TX bytes recorded
>> by the network card is total different from that recorded by
>> X10_TRACE_RXTX. In my test, the RX bytes/TX bytes recorded by the
>> network card of a node (testing with multi-nodes) is about 2GB, but
>> X10RT_RXTX shows only about 3MB per place (there are 12 places on that
>> node).
>>
>> Then I run a simple test on a single node with 2 places, the code and
>> the results are shown as below. It shows that the RX/TX of sending an
>> array with size 10,000 is the same as 100,000,000. Then I just doubt
>> that the X10_TRACE_RXTX can not catch the RX/TX of asyncCopy, is that
>> right? Is there any method that can record the RX/TX between each
>> place in this situation?
>>
>> Thanks in advance!
>>
>> Long
>>
>> public class test {
>> public static def main(args: Array[String]) {
>> val N:Int=Place.MAX_PLACES;
>>
>> val NUM:Int=Int.parseInt(args(0));
>> val r:Region=0..(N-1);
>> val d:Dist=Dist.makeBlock(r);
>>
>> val receive=DistArray.make[RemoteArray[Int]](d);
>> var sent:Array[Int]=new Array[int](NUM);
>> for(i in 0..(NUM-1)){
>> sent(i)=i;
>> }
>>
>> val p1=Place.place(1);
>> at(p1){
>> receive(here.id)= new RemoteArray(new Array
> [int](NUM));
>> }
>> Array.asyncCopy( sent, at (p1) receive(here.id));
>>
>> Console.OUT.println(NUM+" is done");
>> }
>> }
>>
>>
>> 10000 is done
>> Place: 0 rx: 167/6 tx: 161/4
>> Place: 1 rx: 189/5 tx: 167/6
>> 100000 is done
>> Place: 0 rx: 167/6 tx: 161/4
>> Place: 1 rx: 189/5 tx: 167/6
>> 1000000 is done
>> Place: 0 rx: 167/6 tx: 161/4
>> Place: 1 rx: 189/5 tx: 167/6
>> 10000000 is done
>> Place: 0 rx: 167/6 tx: 161/4
>> Place: 1 rx: 189/5 tx: 167/6
>> 100000000 is done
>> Place: 0 rx: 167/6 tx: 161/4
>> Place: 1 rx: 189/5 tx: 167/6
>>
>>
> ------------------------------------------------------------------------------
>
>> Everyone hates slow websites. So do we.
>> Make your web apps faster with AppDynamics
>> Download AppDynamics Lite for free today:
>> http://p.sf.net/sfu/appdyn_d2d_feb
>> _______________________________________________
>> X10-users mailing list
>> X10-users@lists.sourceforge.net
>> https://lists.sourceforge.net/lists/listinfo/x10-users
>>
------------------------------------------------------------------------------
Everyone hates slow websites. So do we.
Make your web apps faster with AppDynamics
Download AppDynamics Lite for free today:
http://p.sf.net/sfu/appdyn_d2d_feb
_______________________________________________
X10-users mailing list
X10-users@lists.sourceforge.net
https://lists.sourceforge.net/lists/listinfo/x10-users
