Improve runtime for MIRI emicorr step

[stscijgbot-jp]

Issue JP-3592 was created on JIRA by David Law:

The MIRI emicorr step seems to be working well, but is much slower than it needs to be, taking upwards of 25 minutes for some LRS TSO data.  From talking to Michael Regan it sounds like it should be straight forward to speed it up substantially, since the step does not need to search over all samples in order to get a good measurement of the EMI.

Michael Regan or Eddie Bergeron to provide further information as necessary for modification of the emicorr routine.

[stscijgbot-jp]

Comment by Tyler Pauly on JIRA:

Hi David, I've been assuming the "assigned to Tyler" step to mean that this is ready to be worked, but a dev looking at this might not know where to start. Could we get the further information attached to this ticket? Are there suggested bounds involved in the search, or something else?

[stscijgbot-jp]

Comment by David Law on JIRA:

Sorry, it should be ready to be worked within a few days, but to avoid confusion I should probably have assigned it to Michael Regan first until that information is added.  I'll reassign.

[stscijgbot-jp]

Comment by Eddie Bergeron on JIRA:

Maria and I did some work on this in the original emicorr ticket: https://jira.stsci.edu/browse/JP-3248

See comments on Dec 19, 2023. I don't know if any of these things were implemented, other than using nbins=500 for 10hz. I don't know if that is handled in the code itself or via optional parameters. however it does sound like we need to implement one of the more sophisticated speed-up algorithms that we talked about (or something else).

[stscijgbot-jp]

Comment by David Law on JIRA:

Michael Regan Just checking in to see if you think there's any additional information that would be helpful to include here before sending this ticket over to SCSB?

[stscijgbot-jp]

Comment by David Law on JIRA:

Adding an example file with very long runtime: jw01283001001_03101_00001_mirimage_uncal.fits currently takes 12 minutes to run through emicorr (100 groups, 10 ints)

[stscijgbot-jp]

Comment by Michael Regan on JIRA:

Determining the phase only takes around two cycles. So depending on the frame time. We can set an upper limit of the number of frames for the different emi sources.