Improving performance of multi-site experiments

[juliomateoslangerak]

We are currently running a classic z-stack over multiple sites and I would like to improve the performance to be able to capture more stacks per cycle.

There is a significant delay between one site and the next one. Some 10sec. The computer is old and we are going to replace it soon, but I'd like to anticipate cause I have the feeling that it is not only this.

I believe there a number of things could be worked on on the multi-site experiment:

• For every position, there are a number of things that are 'unnecessarily' being recalculated, like the whole actions table. Could we wrap the calculation of the table in a function that could be cached? Would it be possible to split out the z positioner actions generation so that only that part is recalculated?
• The movement time to the next position seems slow, while the stage it self is moving in 150msec. I believe there might be some time lost waiting for the stage to answer and also the x and y movements are sequential...
• Saving images also seems slow. Is there something might help in that front?

Another alternative is to setup a single-site experiment doing the multiple sites. I believe that might provide more flexibility to optimize some of these things. Eg. with the NI-fpga executor, when I start the execution I can specify at which position of the table to start. It is possible to pre-compute the z-stack for the different positions and have no overhead in loading a new table for every position (as long as memory is not an issue).

This might not be necessary or not so pressing with a new PC. As soon as the PC is replaced I'm planning to run some timing tests to see which points might require more attention.