Accelerad 0.7 + Ladybug Tools 1.2

156 views
Skip to first unread message

Maciej Nisztuk

unread,
Jun 12, 2021, 9:08:41 AM6/12/21
to Accelerad Users

Hey there,

I am trying to use Accelerad 0.7 + ladybug Tools 1.2 for calculating annual daylight (Rhino 6). I used the example file provided within Ladybug Tools installation files. I managed to run analysis with the latest Radiance release - Radiance 5.4a (2021-03-28). I also tried one which is presented in “compatibility matrix”, but had some errors like “Invalid exception: invalid resource folder”, described here Is it possible to get LBT 1.2.0 working with Accelerad? - #9 by chris.

I instaled Accelerad as described in their documentation (switched Radiance files for those provided by Accelerad) and everything seems to work ok, but I see in the logs that Honeybee/Radiance still uses CPU - it looks like the Accelerad reverts to basic Radiance CPU behaviour. Does anybody knows how to make it to use GPU?

My computer should work with the GPU support (RTX 2060). In addition, I used old Honeybee+ 0.0.06 release to do similar thing (used annual_daylight_gridbased example file, using the same Radiance/Accelerad installation) and the logs clearly showed the Accelerad GPU usage. The problem with Honeybee+ is it uses only single CPU for preparing the scene

Nathaniel Jones

unread,
Jun 12, 2021, 4:19:25 PM6/12/21
to Accelerad Users
Hi Maciej,

Could you share the log file that you are referring to, or a GH file that will reproduce this problem? In general, if you are calling an Accelerad program, it will run on the GPU unless you tell it not to with the -g option. Is it possible that the Honeybee workflow you are using does not make use of one of the programs replaced by Accelerad, or that there is another installation of Radiance on your computer that it is referring to?

If the log file you are looking at contains the line "OptiX x.x.x found display driver xxx.xx, CUDA driver x.x.x, and x GPU device(s)", then you can be assured the GPU is being used. However, as I'm not familiar with the scripts in Ladybug Tools 1.2, I can't guarantee that the standard output from Accelerad is passed through into the log.

Nathaniel

Maciej Nisztuk

unread,
Jun 13, 2021, 7:34:46 AM6/13/21
to Accelerad Users
Hi Nathaniel,

Thank you for the response. Just to be sure I am not missing something - I do the Annual Daylight analysis using this Ladybug Tools component:
5.PNG
I see the "OptiX x.x.x found display driver xxx.xx, CUDA driver x.x.x, and x GPU device(s)" info when using Honeybee + 0.0.06. The output from Ladybug Tools 1.2 does not have this info and  looks more less like this (just keep repeating as it uses 8 cpus):1.PNG2.PNG3.PNG4.PNG

Maciej.

Nathaniel Jones

unread,
Jun 13, 2021, 8:53:25 AM6/13/21
to Accelerad Users
Hi Maciej,

It appears that the Ladybug Tools 1.2 log is showing the output of a scheduler process, and it is unclear from what you have provided if it is possible to see the command-line output of Accelerad or not. Therefore, I cannot tell if Accelerad is being used or not. You could test this by providing an Accelerad setting that is not valid for CPU Radiance, such as the -ac parameter. If the simulation runs normally, then it is using Accelerad, but if it fails due to an unknown command, then it is using CPU Radiance.

I notice that Ladybug Tools 1.2 is splitting the ray tracing step between multiple processes. This is not helpful with Accelerad and will actually slow Accelerad down due to GPU context switching. A solution might be to limit Ladybug to a single worker process. However, there are also non-ray-tracing steps that are similarly being launched on worker processes, and it's beneficial to keep these in parallel. I'm not sure if the two parts can be controlled separately. You may wish to raise this issue on the Ladybug Tools discourse.

Nathaniel

Maciej Nisztuk

unread,
Jun 13, 2021, 9:01:09 AM6/13/21
to Accelerad Users
Hey Nathaniel,

Just as a curiosity - I've made the update of Ladybug Tools for the latest dev version and it started to show the Accelerad GPU usage in the logs. The computation times didn't change so I assume it's always used GPU, but v. 1.2.0 didn't show that info in logs. Problem solved :) Thank you.

4.PNG

Maciej
Reply all
Reply to author
Forward
0 new messages