Re: COMS W4995: Breaking RSA

[Vijay Saraswat]

Which version of X10 did you use to get the Java backend numbers?

Certainly if u r scaling on Java you should be able to scale on the
native backend.

Your scaling numbers in this graph are excellent.

Send me your code and the scripts you ran to get the perf numbers below
(so I know how your env flags were set).

On 12/8/2010 7:07 PM, Albert Jimenez wrote:
> Hello Professor Saraswat,
>
> Thank you for you help today. Out of curiosity I tried running my
> program on a different cluster using the java backend. The results
> show that not only does the program have good load balancing but also
> very good parallelism. Attached is a graph of the execution time for 1
> to 16 asyncs. Each time is an average of 4 trials. For 1 async the
> program took 37.9 s, for 16 asyncs it took 3.2 s. So considering that
> these times each contain about 1 s of serial execution, the program
> was able to perfectly scale to 16 asyncs.
>
> We've spent 2 weeks trying to get reasonable results on slurm but now
> I am convinced that there was nothing wrong with our code. Perhaps the
> problem is that we are using 2.0.6? Is it possible that we can use
> this cluster instead to show the performance of our program?
>
> Thanks
> Albert