Maximum amount of points to calculate for HB[+]

Replies to This Discussion

Permalink Reply by Chris Mackey on August 30, 2017 at 3:53am

Peter,
I imagine that this is happening because you might be trying to load at least 50,000 x 8760 = 438,000,000 data points, which should take time but not 12 hours (it seems like there might be a bug at work here as well).
Does the simulation folder seems to contain all of the results for 50k points?
Also Mostapha and Sarith will know much more about this issue than I do.
-Chris

Permalink Reply by Peter Zaťko on August 30, 2017 at 4:11am

Hi Cris,

Yes, it has everything, I also ran it on Linux-CentOS, and tried to " load back in", but the same waiting time. I can´t upload the files anywhere cause it´s 45 GB.

Sorry to already pushing the boundaries of the tool :) 

Best,

Peter

Permalink Reply by Mostapha Sadeghipour Roudsari on August 30, 2017 at 8:40am

Hi Peter,

Which analysis recipe are you using? There are differences on how Honeybee[+] loads the results based on the recipe.

For most of the recipes Honeybee[+] loads 2 * number of hours * number of blind state for each test point. The 2 is there because it loads the results for total and direct contribution separately. If you don't have dynamic blinds use the Annual recipe which is a flavor of daylight coefficient recipe that doesn't load the results to memory and just calculates the annual metrics.

Let me know what is your goal and we can take it from there.

Mostapha

Permalink Reply by Peter Zaťko on August 30, 2017 at 9:11am

Hi Mostapha,

Since I´m not using any BSDF for this calc, and have just one state (unshaded)- cause of huge overhangs, I´m using Daylight coefitients. BUT, since I want to show the contribution of each facade + skylight I divided them into groups, that´s the reason of the size + compare the results with Daysim.

It´s not crucial, I´m just interested where are the limits

Best,

Peter

Attachments:

• PZ_DC_gridbased.gh, 1.1 MB

Permalink Reply by Mostapha Sadeghipour Roudsari on August 30, 2017 at 9:19am

We can and we will get to the place to handle large files like this. There are a couple of items that needs to be addressed before we get there.

For comparison studies to Daysim you should use the Annual Recipe. That is using the daylight coefficient recipe to run the analysis and similar to the Daysim workflow it doesn't load the results for each test point for calculating the annual metrics.

Here is the good news. If you copy your current files under the annual folder and change the recipe to annual and leave reuseMtx to True the component will read the results without re-running the analysis. That should give you the option to compare the values. No window group is supported though.

As I said there are solutions to be developed for specific cases. Keep us posted on your findings, needs and suggestions.

Mostapha

Permalink Reply by Peter Zaťko on August 31, 2017 at 1:48am

Thank you Mostapha,

I´ve got a wish/question, how it´s possible to do an annual simulation with different shading state each hour (light redirection). Do I need to write 8760 states, and then in postprocessing take them ?

Peter

Permalink Reply by Peter Zaťko on August 31, 2017 at 2:47am

Hi Mostpaha,

So I followed your advice and checked it with the Annual Recipe, but I´ve got this error at the post processing (see below), looks like just the first grid is read in, and the rest have some issues. Any suggestion?

Peter