Parallelize my Code

OVITO
1

Hello community members,

I have written a code (attached below) to compute interface heights for my melt-solid-melt simulation setup. I wish to parallelize it for an improved efficiency. I want to divide the loop in chunks of frames and analyze them in parallel. When I do that, i get an error - “cannot pickle object ovito.pipeline”. Can you help me fix this by suggesting how i can parallelize the for loop below.

Original Code:

function to compute interface heights

error_list = []
def compute_interface_heights(pipeline):
frame = 0
top_height = np.zeros((snapshot_count,bins))
bottom_height = np.zeros((snapshot_count,bins))

for frame in range(pipeline.source.num_frames-1):
    try:
        clear_output(wait=True)
        print(frame)
        data = pipeline.compute(frame)

        #compute the bottom interface height for each frame
        data_TopCopy = data.clone()
        data_BottomCopy = data.clone()

        #top interface
        data_TopCopy.apply(modifier0)
        data_TopCopy.apply(modifier1)
        data_TopCopy.apply(modifier3)
        top_height[frame,:] = data_TopCopy.tables['binning'].xy()[:,1]

        #bottom interface
        data_BottomCopy.apply(modifier0)
        data_BottomCopy.apply(modifier2)
        data_BottomCopy.apply(modifier3)
        bottom_height[frame,:] = data_BottomCopy.tables['binning'].xy()[:,1]

        #delete objects for reuse in next loop
        del data_TopCopy
        del data_BottomCopy

    except Exception as e:
        error_message = f"An error occurred at frame {frame}. Error: {e}"
        error_list.append(error_message)
        continue

return top_height, bottom_height
2

Hi! Thanks for the question.

Since this is a recurring question, I have written a new section for the OVITO documentation explaining how to use the multiprocessing module. It also discusses the existing limitations you need to be aware of.

https://ovito.org/docs/dev/python/introduction/advanced_topics.html#multiprocess-module-usage

3

Thank you for the response. It was helpful.