I’ve been working to get Honeybee+ working with accelerad and seem to have success with the annual Daylight recipe (required adding in a new variable -y which equals the quantity of calc points, and changing -c 1 (default = 1000). that all works but the calc still took 14min for 3400 points running on 2 GTX 1080s. seems slow…so i tried running a five phase and received this error in the report:
Found 31 opaque surfaces.
Found 12 fenestration surfaces.
Found 0 window-groups.
Writing recipe contents to: c:\ladybug\FOUNDRY_ASBUILT3\gridbased_fivephase
Runtime error (AssertionException): AssertionError
Traceback:
line 84, in write, “C:\Users\dan\AppData\Roaming\McNeel\Rhinoceros\6.0\scripts\honeybee\radiance\recipe\fivephase\gridbased.py”
line 64, in script
here’s the gh file: https://drive.google.com/file/d/1XeK37oCZqDCGuZiM17BZ7C8pXZdV5nkD/view?usp=sharing
Hi @daweissman, Welcome to the forum. As discussed on Accelerad mailing list I won’t be able to do much before our next deadline. Appreciate you posting it here and will get back to you after the deadline. Meanwhile do not use 5-Phase and stick to 2-Phase simulation.
1 Like
Hi @daweissman & @mostapha,
do you guys have any updates on that? I’m trying to do the same thing
Hi,@Diennymmm I have tested Accelerad 0.7 beta version.It can be coupled with HoneybeePlus.
You can try this version.
Here is the result of a sample file using 5-phase method.
@minggangyin Yeah, I have Accelerad 0.7 beta on my machine.
The thing I try to find out is how to connect my rhino model to Accelerad like shown in this video:
could you share that sample file?
Hi,@Diennymmm I just used the samples of HoneybeePlus.
Here is the link:
Oh cool! Did you use the there phase simulation?
Hi,@Diennymmm
I have tested the three-phased and five-phased samples. These two simulation worked fine on my workstation.
Here is the result of five-phased sample.
You can try to run HoneybeePlus following these tips:
https://nljones.github.io/Accelerad/documentation.html
1 Like
Hello, I would like to inquire about how to use accelerad, I installed accelerad according to the method of 使用Accelerad对HB Radiance进行GPU加速 - 知乎, but when the daylighting calculation of honeybee is performed, the program is not running, and my GPU is hardly working, I wonder if you have a good solution?
您好,我想咨询一下accelerad的使用方法,我根据 使用Accelerad对HB Radiance进行GPU加速 - 知乎的方法安装了accelerad,但在honeybee的采光计算时,该程序并没有运行,我的GPU也几乎没有在工作,不知您是否有好的解决方法?
Hi @JerrySun,
Can you provide more information about what is not working? For example, can you show the command-line output of your simulation to see what error messages, if any, are produced? Are you able to run the example files that come with Accelerad?
You should not expect your GPU to do much work. Accelerad’s use of GPU resources is very efficient and usually does not register on GPU monitors.
Nathaniel
Thank you for your reply. Figure 1 is my daylighting calculation module, and my purpose is to screen out the time when at least 60% of the room meets UDI needs throughout the year. My model calculates fine and can get results.
My problem is that the current Annual Daylight simulation is too slow, and even if I use a server with 100+ threads & a lot of memory, it still can’t effectively improve the speed of the calculation. I think it’s because parallel computing is not implemented, but I do set a reasonable number of workers in the HB Recipe Settings(The number of wokers is not set in the figure because the maximum number of CPUs minus 1 is chosen as the default). The Task Manager showed that only one or two CPUs were in high usage, which confirmed my suspicion.
Using Accerarad is a way I’ve seen accelerated computing. However, when I re-calculate in Honeybee after installation, the black task box that pops up does not show the characters of Accelerad, as far as I know, this does not happen when you successfully use the tool to calculate. Figure 2 and 3 are screenshots of my command prompt, I see the characters ‘GPU’ appear in Figure 3 but don’t understand if this is related to Accelerad.
My goal is just to speed up the calculations, if there are other effective methods please recommend them to me. I’m still a rookie to Grasshopper, and I can start a separate post about the above if necessary.
Thanks again.
Hi @JerrySun,
Your third figure shows the output of Accelerad, indicating that Accelerad is running and there are no errors reported. In other words, it appears that it is working as expected.
When you use Accelerad, you should use only one worker. This is because GPUs do not support having multiple workers, so trying to access the GPU by multiple workers slows the entire process down.
If you are comparing Accelerad on a single GPU to running 100+ CPU workers on a computer that has 100+ CPUs, I do not expect that Accelerad will give you a significant speedup in that case.
Thanks again for your reply, so
- Should I set the number of wokers to 1 and then go and see if there is an increase in running speed?
- Do you know how to use multiple CPUs in parallel to perform calculations for daylighting simulations? After all, I mentioned that I only had one or two CPU working when I did the simulation on a server with 100+ CPUs …
I reset the number of wokers to 1 and it turns out to make no difference. The speed is as usual. I guess I might have to try some other options.
Anyway, thank you so much for discussing this promblem with me!
@JerrySun
Hi,JerrSun
You are reading my article. Although I have already communicated with you in other communities, I still think it necessary to reply again and solve this problem.
Since you have modified the Radiance Parameter, when Recipe Type = 0
, it should correspond to the Point In Time Grid Base Recipe. This does not match the Annual Recipe currently in operation by you. Therefore, the computational efficiency you expected to obtain has not been achieved.
best.
Zhengrong