How to use CUDA for AvlancheMC

Garfield++
1

Hello,

I have found that the current Garfield++supports the use of CUDA.

I want to use CUDA to accelerate running AvalancheMC, and I would like to know if any examples can help me.

2

To avoid creating an additional discussion, I have another question here：

image
image1090×684 27 KB

When using TrackSrim, I encountered an error message above.

3

Hi,

I’m afraid there is no GPU version of AvalancheMCat the moment.

Regarding the crash: can you provide a minimal working example that reproduces the issue?

4

Hello，

Thanks for your reply.

I can run this program in my original environment using the latest version of Garfield++and ROOT 6.36, but I found that the program executes very slowly, and even if I wait for a few hours, I cannot calculate the first cluster. So I wanted to run this program in another old version of the Garfield++environment, and then I encountered the error mentioned earlier.

I am using WSL, and I checked the total CPU usage in Windows while I was running the program:

1755588559617
1755588559617449×390 97.8 KB

It shows that the total CPU utilization is only 10%. I have allocated resources for my WSL, and my server has 64 logical processors. But I still have a slower running speed compared to others in this discussion.

https://root-forum.cern.ch/t/signal-question/63443/14

The code and parameters I used are consistent with those discussed in the link.

Here is the code and corresponding files. Thank you in advance for your help.

mre.zip (12.8 KB)

5

This issue is caused by an error in Garfield++calling ROOT.

I solved this problem by changing the environment variable configuration of ROOT.

But I still face the problem of slow AvalancheMC computation, even with dynamic OpenMP.