Seems GPU is not being used for calculation

[katsu...@gmail.com]

Hi,

I am trying to install and implement Accelerad.

What I did is following.

1. Installed Radiance
2. Checked GPU (NVIDIA Quadro K2000)
3. Copy the contents of the Accelerad bin folder into  Radiance bin folder.
4. In the Radiance bin folder, rename accelerad-rpict and accelerad-rtrace to rpict and rtrace, respectively, and replace the original program files. 
5. Copy the .ptx files from the Accelerad lib folder to the Radiance lib folder.

In these condition, I executed a simulation.

But checking at GPU monitor, GPU is not being used.(just one core of CPU was used )

Do you know why?

Thank you in advance.

[Nathaniel Jones]

Hi Katsuya,

The steps that you described should work, but there are a few things you should check. First, make sure that Accelerad is working by running test_accelerad_rpict.bat and test_accelerad_rtrace.bat with the accelerad-rtrace.exe in the location where it installed.

Second, check to see if you have any other installations of Radiance on your computer. In the command line, you can type echo %path% to see a list of all the places your computer will look for the Radiance programs. If you have multiple installations, your computer will run the first one it finds in that list, which is generally not a problem unless you have two different programs with the same name, as could be the case when you rename the Accelerad version to replace the regular version.

If all that checks out, then take a look at the call to Radiance to see what's happening. I'm not sure how you are calling rpict or rtrace, but if you can see it through the command line, please show me the command that is issued and the output that prints. Accelerad always starts by printing a list of available GPUs, unless you run it with the -g 0 option which forces it to use only the CPU.

What tool are you using to monitor the GPU? It may depend on the tool you use, but some GPUs may not check often enough to notice Accelerad running if the job is short.

I hope this helps,

Nathaniel

[Yafim Simanovsky]

Hi,

I've noticed the same issue in my setup (CPU running instead of GPU).

This is the output of the echo %path% command:

C:\Program Files\Microsoft MPI\Bin\;

C:\Program Files (x86)\Common Files\Intel\Shared Libraries\redist\intel64\compiler;

C:\ProgramData\Oracle\Java\javapath;.;

c:\daysim\lib\;

c:\daysim\bin\;

c:\program files (x86)\lenovo\fusionengine;

c:\program files (x86)\intel\icls client\;

c:\program files\intel\icls client\;

c:\windows\system32;

c:\windows;

c:\windows\system32\wbem;

c:\windows\system32\windowspowershell\v1.0\;

c:\program files\intel\intel(r) management engine components\dal;

c:\program files\intel\intel(r) management engine components\ipt;

c:\program files (x86)\intel\intel(r) management engine components\dal;

c:\program files (x86)\intel\intel(r) management engine components\ipt;

c:\program files (x86)\lenovo\motion control\;

c:\program files (x86)\common files\lenovo\easyplussdk\bin;

C:\Program Files\Intel\WiFi\bin\;

C:\Program Files\Common Files\Intel\WirelessCommon\;

C:\WINDOWS\system32;

C:\WINDOWS;

C:\WINDOWS\System32\Wbem;

C:\WINDOWS\System32\WindowsPowerShell\v1.0\;

C:\Program Files (x86)\NVIDIA Corporation\PhysX\Common;

C:\Program Files\Common Files\Autodesk Shared\;

C:\Program Files (x86)\Autodesk\Bac;

C:\Program Files\Microsoft SQL Server\130\Tools\Binn\;

C:\WINDOWS\system32;

C:\WINDOWS;

C:\WINDOWS\System32\Wbem;

C:\WINDOWS\System32\WindowsPowerShell\v1.0\;

C:\Program Files (x86)\Heroku\bin;

C:\Program Files (x86)\git\cmd;

C:\Program Files (x86)\Windows Live\Shared;

C:\Program Files\Microsoft SQL Server\120\Tools\Binn\;

C:\Python27\;

C:\Program Files\Git\cmd;

C:\Program Files (x86)\GtkSharp\2.12\bin;

C:\Qt;C:\Python27\Scripts;

C:\WINDOWS\System32\OpenSSH\;

C:\Boost\bin;

C:\Program Files\nodejs\;

C:\Program Files\Accelerad\bin;

C:\Users\yafim\AppData\Local\Programs\Python\Python37-32\Scripts\;

C:\Users\yafim\AppData\Local\Programs\Python\Python37-32\;.;

C:\DAYSIM\lib\;

C:\DAYSIM\bin\;

C:\Ruby25-x64\bin;

C:\Program Files\MongoDB\Server\3.2\bin;

C:\Users\yafim\AppData\Local\Microsoft\WindowsApps;

C:\Python27\;C:\Program Files\Heroku\bin;

C:\Users\yafim\AppData\Roaming\npm

Any idea what I should change in the Environment Variables in order to set this correctly? I have several overlaps probably between the User and System variable paths but I'm not sure what should be in each.

Thanks

[Nathaniel Jones]

Hi Yafim,

Are you using Honeybee? Honeybee ignores your environment, so this should not be relevant.

Your path variable is quite messy. I usually recommend to copy everything to the system path variable and delete the user path variable. Then remove all duplicate entries. The order of entries matters. If Daysim is listed before the folder that has the Accelerad programs, the applications that use the path to find programs will use Daysim versions.

Can you describe the problem you are having?

Nathaniel

[bimbase...@gmail.com]

The problem was that I didn't see my GH script running on the GPU. It was running on the CPU.

But I'm running a Honeybee script. I didn't know it ignored the environment. Does this mean that with Honeybee it will always run on the CPU?

Does this mean that Ladybug runs on the GPU?

I've deleted the path from the User variables and added the C:\Radiance\bin and C:\Radiance\lib to the System variables, above the Daysim variables in order.

I think this is the correct setup. Let me know if I'm mistaken.

Thank you

[Nathaniel Jones]

Because Ladybug/Honeybee ignores the environment variables that you set up, you can cause them to use Accelerad by copying the programs to C:\Radiance\bin, which is where Ladybug/Honeybee expect to find the Radiance programs.

In your earlier thread, it was apparent that you had successfully caused Honeybee to use the Accelerad programs. Unless you provide the -g- option, the ray tracing calculations will be performed on your GPU. What do you mean when you say that you didn't "see" your script running on the GPU? Only the ray tracing is performed on the GPU, and as this is very fast, you should not expect to see it in a GPU monitoring program.

Nathaniel

[bimbase...@gmail.com]

I was trying to see what happens on the task manager, just to see CPU vs. GPU activity. When I ran a Daylight Simulation I thought that I would be able to see how those are used.

I only saw activity in the CPU when running a simulation that ran for 3-5 minutes.

Is there a way I can "make sure" that Radiance is being run on the GPU?

Thanks.

[Nathaniel Jones]

If your output messages include a line like this:

rpict: OptiX x.x.x found display driver xxx.xx, CUDA driver x.x.x, and x GPU device(s):

Then the ray tracing is running on your GPU.

Nathaniel