Adding the option for using Accelerad for local runs

The notion happened upon me while wrestling with the Accelerad docs and the ladybug_tools install on my machine at another failed attempt to get Accelerad to work with the LBT install on my machine for the fourth time (human error likely).
The notion is that it would be really really really convenient for the user, if maybe ‘install with Accelerad’ was some sort of option for the Rhino and or Grasshopper PO installers for locally ran studies to allow for the gains of accelerad in lieu of the cloud where applicable.
I personally don’t have access to cloud studies so at least in my position the capability would be wonderful.
Understanding if I am not mistaken, and if so please correct me, AMD GPU’s don’t have CUDA cores which I believe Accelerad requires, which gave the idea of having some sort of Accelerad install be an optional feature.
All the best

Thanks @tfedyna ,

Someone more knowledgeable about Radiance might correct me (@mikkel or @mostapha ) but I think Accelerad might not have implemented all of the latest Radiance commands and options that we need for 100% of all Pollination recipes. Granted, you would definitely be able to run Point-in-Time studies (both Grid-based and View-based would be fully supported). And I imagine components like the HB Check Scene component would run a lot faster. But I’m not completely sure if all of the rcontrib functionality of Accelerad has been maintained enough to use with our latest annual recipes. So we may not easily be able add an option to just swap out the normal Radiance with Accelerad and know that 100% of everything works.

Also, there is this issue that you brought up:

I don’t know all of the details but I remember some things from back when I was at MIT while Nathaniel Jones was working on Accelerad for his PhD thesis. I know that the “SDK” he was using to build Accelerad was from NVIDIA. So, yes, I believe that Accelerad runs exclusively on NVIDIA GPUs and those people with AMD graphics cards may end up with an installation that does not work.

Granted, if you want an easy way to switch the version of Radiance used by your Ladybug Tools and Pollination installation, you can open a command prompt in Admin mode and use the following command:

"C:\Program Files\ladybug_tools\python\Scripts\honeybee-radiance.exe" set-config radiance-path [ACCELERAD_PATH]

Then, you can check what the Radiance path is set to using the following command

"C:\Program Files\ladybug_tools\python\Scripts\honeybee-radiance.exe" config

You can set your installation back to the default by running the set-config command again and not passing any arguments.

Let us know if you end up using this to test compatibility of Accelerad with the latest recipes.

Thankyou for the information @chriswmackey! and the easy way to switch between hb radiance and accelerad.
Now that I have the version of radiance switched properly I’ll definitely take a look at the radiance recipes, specifically the annual ones (but the other ones too) and report back here my findings.

1 Like

Hi @tfedyna,

All the recipes should work with Accelerad. Just be aware that some Radiance modifiers are not supported by Accelerad: Accelerad Documentation.

1 Like

Thank you, @tfedyna for posting this feature request. We have discussed this before internally, and I support adding an official extension for supporting Accelerad.

@mikkel already mentioned that some modifiers are missing. I imagine that should not be a major problem since most of the scenes don’t use the missing modifiers. The most common ones are probably mirror and BSDF that are missing.

The other challenge is that Pollination recipes are optimized for distributed computing which is not ideal for using Accelerad. Accelerad has its internal distributed computing using GPUs and expects all the sensors to be submitted as a single grid. You should ensure to use the recipe inputs to accommodate that. The easiest way to do that is to set the number of parallel CPUs to 1 and select a very large number for the number of sensors per grid.

Finally, Accelerad has its own unique command options (AKA parameters), that are different from the equivalent Radiance commands. You should keep that in mind when setting up the recipes.

1 Like

Thankyou for your explanation and instruction for using accelerad in studies!

@chriswmackey I followed your instructions and have ran into some issues.

All of the radiance components and even the HB Room From Solid compo calls the same error “1. Solution exception:C:\Program Files\Accelerad\bin is not a valid path to a Radiance installation.”
accelerad issues

Did I do something incorrectly here with the command to change the hb radiance path to that of the accelerad?

Hey @tfedyna ,

Maybe I’m just not familiar with how Accelerad is structured. Is there no bin folder within the Accelerad installation folder that contains all of the executable Radiance files?

You need to be setting the Honeybee-Radiance folder to something like that where the bin and lib folders are sub-folders of the Radiance installation folder.

@chriswmackey The solution to get accelerad to work (only tested PIT and Annual daylight so far but they are running) is to replace as per the docs:

the normal-rad exe’s with the A-rad ones, and to remove accelerad_ from the file names.
Next step is to pass the radiance param -g 0 to have accelerad behave like normalrad as per Mathias in this post

But once all that happens its looking like things are working so far!

Thanks, @tfedyna .

This is all good to know. I did not realize that, by default, the installer puts the word accelerad_ in front of all of the executable files. By this logic, my suggestion of using the CLI to change the Radiance installation could still work if you change the names of the executables in the Accelerad installation folder and copy over all of the other missing Radiance executables.

So this:


… would become this:

With that, you have a way to switch your system in between the “Normal” Radiance and an Accelerad version of Radiance. You are right that you would still need to pass the -g 0 option through the Radiance parameters to use the GPU with your Accelerad version of Radiance.