DATA SAVING PROBLEM

Hello everyone,

I contact you because I have been struggling during last three days with a data saving problem and I hope that you have the sesame for this issue.

In fact,I would like to carry out a parametric study relative to a radiation analyse of a simple geometry. This parametric study includes 108000 simulations which are performed with the help of HoopSnake component. Each simulation lasts 1 second and ,at the end, I intend to have 108000 number stocked in a list ( panel). The problem is that at the beginning (between 1 and 2000 iterations) of the loop, each simulation lasts 1 second , as mentioned above. But as far as the iteration counter is getting higher, the simulation consumption time is getting greater (from 1 second to 2.5 second), and I ignore the reason of this artefact. I do believe this is related to stocking data in a list (panel) during the loop process and I have no idea how to save results with maintaining the simulation time constant (1 second).

I hope that I was limpid and clear and I will be very pleased with your contribution.

Thank you.

Replies to This Discussion

Permalink Reply by MARFOUQ MOHAMED AMINE on September 6, 2017 at 3:30am

I do believe it' s a memory allocation issue. If anyone has any trick to clear or refresh memory during the process, it will be very appreciable.

Thanks !

Permalink Reply by Joseph Oster on September 6, 2017 at 7:36am

Don't connect the panel until all the simulations are done?

Permalink Reply by MARFOUQ MOHAMED AMINE on September 7, 2017 at 5:48am

Thank you for your response.

I wish I could reach the simulations end. It started well at beginning, with 1 second simulation, but this time calculation is getting higher for simulation beyond 4000 iteration. All I need to save is only a value at each iteration, while the computer memory increases willy nilly dramatically for a reason that I ignore. I think that the program stock same breps for each iteration which is not necessary for my case and which increases the memory allowance part of calculation. And this, make the simulations slower for higher threshold iterations.

have you been before in a similar situation ?

Permalink Reply by Joseph Oster on September 7, 2017 at 6:11am

I have crashed Rhino/Grasshopper many times, to the point of having to reboot my computer because it's frozen so hard. Without any doubt, Rhino/Grasshopper is the most ill-behaved software I have ever used when it comes to bringing a computer to its knees. VERY ANNOYING!

We all know it's far too easy to graft the wrong inputs/outputs and cause Grasshopper to turn ten thousand object references into ten million (oops!). Looping (I use Anemone) can also consume more memory than expected when 'Record data' is enabled on 'Loop End'.

Does disconnecting the panel make any difference?

Any time I am dealing with potentially large numbers of geometry that will be SLOW to process, I reduce the "resolution" or loop count until I'm sure I've fixed/optimized the algorithm before letting it run on the whole data set.

Permalink Reply by MARFOUQ MOHAMED AMINE on September 7, 2017 at 6:24am

I disconnected the panel but unfortunately the problem persists.

I think that my program is optimized since one simulation lasts 1 second. I tried both Anemone and Hoopsnake. It appears that the program calculation has the same behavior with regard to both components. The thing is I want to perform 108000 simulations,and with this tools it will take an eternity.

Permalink Reply by Joseph Oster on September 7, 2017 at 6:54am

Did you try Anemone's 'Fast Loop' components? They have some quirks, like not displaying any output while running, but are MUCH faster. Hoopsnake never made any sense to me.

The only other idea that comes to mind is breaking the problem into pieces, saving results to files that can be combined later.

Permalink Reply by MARFOUQ MOHAMED AMINE on September 7, 2017 at 7:58am

Yes, I have just tried Anemone's Fast loop, but it is so quick that my function does consider only the last counter number... Burlesque.

Splitting the simulations into many pieces comes to my mind yesterday,which is so annoying as an approach. I hoped to find a better solution by sharing this problem here. Especially , an efficient way to clear memory during the process.

I thank you for your contribution.

RSS