[parsec-users] Question about swaptions output

Christian Bienia cbienia at CS.Princeton.EDU
Mon May 3 11:52:51 EDT 2010


Hi,

Unless you want to do experiments for which the contents of the register
file and memory is important I'd suggest to simply keep using swaptions as
it is. If the contents does matter then you shouldn't use any form of
synthetic input anyway.

Best,
Chris


-----Original Message-----
From: parsec-users-bounces at lists.cs.princeton.edu
[mailto:parsec-users-bounces at lists.cs.princeton.edu] On Behalf Of Matt
Sinclair
Sent: Monday, May 03, 2010 11:11 AM
To: PARSEC Users
Subject: Re: [parsec-users] Question about swaptions output

Hi Chris,

Thanks for getting back to me!  Is there a fix I could put in
temporarily into 2.1?  Or when will the next release of PARSEC be
coming out?

Thanks,
Matt

On Mon, May 3, 2010 at 10:06 AM, Christian Bienia
<cbienia at cs.princeton.edu> wrote:
> Hi Matt,
>
> This is expected and has already been changed for the next version of
> PARSEC.
>
> Best,
> Chris
>
>
> -----Original Message-----
> From: parsec-users-bounces at lists.cs.princeton.edu
> [mailto:parsec-users-bounces at lists.cs.princeton.edu] On Behalf Of Matt
> Sinclair
> Sent: Monday, May 03, 2010 10:40 AM
> To: parsec-users at lists.cs.princeton.edu
> Subject: [parsec-users] Question about swaptions output
>
> Hello list,
>
> I have a question regarding the output of swaptions.  I have tried
> running swaptions on several different machines, for all of the
> different input sizes (test, simdev, ..., native), but regardless of
> input, I am finding that the answer for each swaption (0 --> 128 for
> native, 0 --> 3 for simdev, etc.) are the same.  By the same, I mean
> swaption[0] = swaption[1] = ... = swaption[N-1] for that particular
> test (not that the answers are the same for simdev and native).  I
> looked through the code to see why this might be occurring, and the
> conclusion I reached is that it must be because the worker() function
> uses a constant value as the seed for all threads:
>
>  for(int i=beg; i < end; i++) {
>        int iSuccess = HJM_Swaption_Blocking(pdSwaptionPrice,
>
> swaptions[i].dStrike,
>
> swaptions[i].dCompounding,
>
> swaptions[i].dMaturity,
>
> swaptions[i].dTenor,
>
> swaptions[i].dPaymentInterval,
>
> swaptions[i].iN,
>
> swaptions[i].iFactors,
>
> swaptions[i].dYears,
>
> swaptions[i].pdYield,
>
> swaptions[i].ppdFactors,
>                                                                 100, // **
> iRndSeed set to 100 for all threads
>                                                                
NUM_TRIALS,
>                                                                
BLOCK_SIZE,
>                                                                 0,
>                                                                 gpuPdZ,
>                                                                 gpuRandZ);
>      ...
>  }
>
> I looked through the archive and did not find this issue posted
> previously, so I wanted to ask if this is expected?  Thanks for any
> help you may be able to provide!
>
> Regards,
> Matt Sinclair
> University of Wisconsin-Madison
> College of Engineering, ECE Dept.
> Vertical Research Group
> _______________________________________________
> parsec-users mailing list
> parsec-users at lists.cs.princeton.edu
> https://lists.cs.princeton.edu/mailman/listinfo/parsec-users
>
> _______________________________________________
> parsec-users mailing list
> parsec-users at lists.cs.princeton.edu
> https://lists.cs.princeton.edu/mailman/listinfo/parsec-users
>
_______________________________________________
parsec-users mailing list
parsec-users at lists.cs.princeton.edu
https://lists.cs.princeton.edu/mailman/listinfo/parsec-users



More information about the parsec-users mailing list