Maksim Levental and I recently started livestreaming regular discussions about MLIR and compilers. We’d like to invite anyone who wants to ask questions, talk design or bugs, or just chat to join the livestream chat or the video call itself (subject to moderation).
Newcomers to MLIR and undergraduate students will be given priority. We want to be welcoming and to help people who find MLIR intimidating to find a solid footing.
Our next scheduled livestream is Friday, March 29th at 3 PM - 4:30 PM US Pacific time. Our planned schedule is every two weeks at that time. You can subscribe to be notified when we start here.
In particular, if you think MLIR or compilers is/are interesting and you can’t get over the hump of “this is hugely complicated stuff” or “this must be very sophisticated technology/math/science, can’t be for me”, feel free to drop by and ask whatever questions might help you get oriented. Stuff like “who cares about SSA?”, “what’s tiling/vectorization?”, “what’s the difference between a GreedyPatternRewriter and a ConversionPatternRewriter?”. I personally don’t guarantee I’ll have all the answers (Jeremy might…) but I’ll certainly google furiously and try to find a decent one on the spot (or admit I’m out of my depth).
I went back and watched your previous streams. Thank you very much for recording them. I would encourage you to keep making them.
I am a beginner. MLIR is kind of my first introduction to inner working of compilers. I am struggling to create a mental model around it. For example, I may “understand” the concepts in the toy language tutorial. But, I really cannot start something from scratch yet.
So, I would like you to consider continuing with the idea of working your own projects in those streams meanwhile explaining your thought process. That way its a meaningful use of your time and helpful to people like me.