[FEA] Split `experimental/row_operators.cuh`

[ttnghia]

Currently, experimental/row_operators.cuh file is huge: over 1200LOC and is growing. Including the entire file into a source file that only uses a small part of it is burdensome. I also feel dizzy while navigating it trying to find a struct/function of interest.

We should split it up (and re-organize files) into separate files like:

• experimental/row_operators/common.cuh
• experimental/row_operators/lexicographic.cuh
• experimental/row_operators/equality.cuh
• experimental/row_operators/hashing.cuh
• ...

[ttnghia]

Tag directly related devs: @devavret @jrhemstad @bdice.

[devavret]

The only reason it was one monolithic header is because it was a replacement of the single row_operators.cuh header. Splitting it is fine. We might need to keep equality and hashing in the same header because they use each other's pre-processed table

[ttnghia]

Then the hashing header can just include the equality header. Some source files just need equality comparison.

Wait: Using each other table? Then I'm not sure...

[bdice]

This idea seems fine to me. We'll want to make sure that the designs stay parallel (wherever possible) despite being in separate files.

[ttnghia]

Thanks all. Then I'll go ahead and do this split. Will have a PR up shortly.

[jrhemstad]

I'm fine with this, but let's wait on splitting them up until we are ready to move them out of "experimental". Then we can split them as we move their namespaces.

[github-actions]

This issue has been labeled inactive-30d due to no recent activity in the past 30 days. Please close this issue if no further response or action is needed. Otherwise, please respond with a comment indicating any updates or changes to the original issue and/or confirm this issue still needs to be addressed. This issue will be labeled inactive-90d if there is no activity in the next 60 days.