Re: [sparta-users] [EXTERNAL] Particle count using transparent surface and issue with multiple core processors

[Vincent C. <vac...@ni...>]

Hello Mr. Moore!

Okay I understand. I'll submit the ticket on github now, thank you very
much for your response.


Sincerely,
Vincent

On Fri, Jun 2, 2023 at 6:50 AM Moore, Stan <st...@sa...> wrote:

> Hi Vincent, for some reason this email didn't seem to make it to the
> SPARTA mail list archive, and my email blocks *.zip files. Can you please
> submit a ticket on Github and attach the zip file?
> https://github.com/sparta/sparta/issues
>
>
> Thank you,
>
> Stan
> ------------------------------
> *From:* Vincent Cabalar <vac...@ni...>
> *Sent:* Wednesday, May 31, 2023 7:42 AM
> *To:* Moore, Stan <st...@sa...>
> *Cc:* spa...@li... <
> spa...@li...>
> *Subject:* Re: [EXTERNAL] [sparta-users] Particle count using transparent
> surface and issue with multiple core processors
>
> You don't often get email from vac...@ni.... Learn why this
> is important <https://aka.ms/LearnAboutSenderIdentification>
> +++ SNL email system blocked unsafe attachment +++
>
> This external incoming email included an attachment that is potentially
> unsafe and was removed from the message. This is a security feature of our
> email infrastructure to block files that may put your computer at risk for
> viruses and other threats. If the disallowed file was included in an
> archive or compressed file, then that entire archive will be removed.
> Please DO NOT carelessly contact the sender through reply or other means
> without first establishing authenticity of this incoming message.
>
> The removed file(s) are not recoverable. If you need this file to support
> a critical business need, please arrange for other means to transfer this
> file such as the Sandia Managed File Transfer service:
> https://snltransfer.sandia.gov/. Only internal SNL recipients will
> receive this notification. If you have any questions, please call CCHD at
> 845-2243.
>
> Common disallowed file types:
> http://technet.microsoft.com/en-us/library/cc179163.aspx
>
> Message Details:
> Date: 05/31/23
> Subject: Re: [EXTERNAL] [sparta-users] Particle count using transparent
> surface and issue with multiple core processors
> Sender: vac...@ni...
> Recipients: st...@sa...
> Message Body below:
> Good day Mr. Moore,
>
> Yes, I do get the expected results when using only a single processor. To
> help you check the issue that I have, I attached in this email a zipped
> folder containing the input script that I am using together with the
> necessary files to run the simulation. I also attached in the folder some
> sample results that show the issue that I have, including the dump txt
> files of the particle flux data from simulation run using 1 core, 2 cores,
> and 4 cores of the processor, and a spreadsheet with the plotted data for
> visualization. Thank you very much for attending to my concern and I look
> forward to resolving the issue for the better development of the SPARTA
> program.
>
>
> Sincerely,
> Vincent
>
> On Thu, May 18, 2023 at 5:11 AM Moore, Stan <st...@sa...> wrote:
>
> Hi Vincent,
>
> Do you get the expected behavior with a single processor? This sounds like
> a bug, but we would need a small-as-possible working example to debug.
>
> Stan
>
>
> ------------------------------
> *From:* Vincent Cabalar <vac...@ni...>
> *Sent:* Wednesday, May 17, 2023 3:01 PM
> *To:* spa...@li... <
> spa...@li...>
> *Subject:* [EXTERNAL] [sparta-users] Particle count using transparent
> surface and issue with multiple core processors
>
> Some people who received this message don't often get email from
> vac...@ni.... Learn why this is important
> <https://aka.ms/LearnAboutSenderIdentification>
> Good day SPARTA users,
>
> I just want to ask for help about some concern regarding the data that I
> generate from SPARTA. Basically, I introduce several particles in my
> simulation space through a small section of a surface and as it propagates
> horizontally, it expands and then passes through a transparent surface
> composed of 500 surface segments and perpendicular to the direction of the
> moving particles. What I want to do is to count the number of particles
> that pass through each of these line segments, and then generate a plot to
> show the distribution of the particles that passed through the whole
> transparent surface.
>
> If you try to imagine the scenario, a group of particles moving towards
> the middle of the transparent surface and passing through, we expect that
> the distribution plot should look like a cosine distribution, with the
> "peak" or thickest part in the middle of the transparent surface and then
> becomes thinner as it spreads to both ends. And this is where the problem
> happens. After creating a distribution plot of the data that I have, the
> plot shows two peaks in the surface, with the middle part having zero
> recorded particles. This is very far from what we expected a while ago
> which is the correct distribution to get from this.
>
> After doing some investigations, I found out that the number of peaks I
> generate is somewhat related to the number of core processors of the
> computer that I use in doing simulations. For additional context, I am
> doing my simulations by parallel processing. I was able to confirm this
> because the plot with two peaks is generated from my simulations done on my
> laptop with 2 core processors. Meanwhile, the data and plot that I
> generated from simulations done on my laboratory's 32-core desktop computer
> produces 32 peaks across the transparent surface, even if I use the exact
> same input script with the one I used on my 2-core laptop that produces
> data with "two peaks".
>
> Now, I am asking for your help on what I should do to be able to generate
> the correct data from my simulations. My adviser asked me to check on "how
> my output data are distributed on the processors, because it is possible
> that the processors are alternating in storing 500 surface segment data
> which produces the peaks". I am not sure how to solve this issue or where
> to even start, so your replies and suggestions are greatly appreciated.
>
> Thank you very much everyone!
>
> Sincerely,
> Vincent
>
>