Robert (Bobby) Evans

Looking through the DRPC code, it occurred to me that for all of the authorizations for DRPC we probably also want to pass in the "func" along with the "execute",...

Having multiple threads helps, especially for larger devices that have enough memory and compute to do everything multiple things at once. I am not sure how deep you can get...

> If I understand correctly, a copy to/from an unpinned memory still requires CPU's involvement to move it between the "bounce buffer" and the real host memory? Yes that is...

This is not needed. It was a nice to have even when it was filed. feel free to close it.

We can remove it whenever. It is just a question of finding time to do the code cleanup.

We have some code that @ttnghia wrote. It will convert a range of tokens to a normalized string that matches what Spark wants. We did this for some Spark specific...

On the Spark side we use bounce buffers for UCX communication. Generally RMM does not play nicely with UCX so we allocate several bounce buffers using regular `cudaMalloc` that we...

Actually `pattern[0-9]{3,4}` is a string followed by 3 or 4 digits. I don't know if we want to write a custom kernel for that just yet. I don't know how...

I still don't see a lot of generality for `STATIC_STRING[RANGE]{X, Y}`.

build