In preparation for the upcoming talk, I've created a spreadsheet that contrasts several FFI and GPU options available to clojure (two separate sheets). Question, comments, and feedback appreciated. https://docs.google.com/spreadsheets/d/1ViLHNUgrO2osh2AH0h7MaCaXz8g0UpLbyWojY5f10kk/edit?usp=sharing
I appreciated the comparison, felt like I learned a thing or two. Liked the columns you chose, felt it was easy to read.
@veckon maybe you'll have some thoughts on the above spreadsheet, preparing for our https://clojureverse.org/t/data-recur-meeting-6-clojure-meets-metal-working-with-native-libraries-and-the-gpu/.
jank, even though in early stages, might be worth an honorable mention
oh you're right, and I guess I should mention the meetup on the #jank channel
I considered adding jank to the spreadsheet, but it's improving so quickly that I'm not sure there anything I could say that would be true for more than a month or two.
How dare you. 😄
I'm happy to fill in columns or provide write access if you want jank
Nah. Anything for jank there would be tentative. I think you made the right call by leaving jank out for now. Every other entry in there is available today.
cc @blueberry
I also enjoyed the comparison, thank you! Shouldn’t the performance of dtype-next be same as panama, even if it supports JNA as fallback?
There’s some mention of generating structs using clang in https://cnuernber.github.io/dtype-next/tech.v3.datatype.ffi.clang.html I haven’t tried it. Guess it doesn’t count as fully “automatic api generation”’.
Fair points! I just updated the sheet.