Post #1750297
2026-04-15 19:12 UTC
@dougmerritt @j2kun In my current research project, which involves implementing a stripped down LLVM IR, I implemented the parser first, which was a bad idea. I don't want both an AST and IR, so I'm iteratively refactoring my AST into an IR. I also already built an SMT IR with backends for Z3 and SMT-LIB, making this the opposite of my usual approach.
Replies (2)
-
@dougmerritt@mathstodon.xyz 2026-04-15 19:14
@thalia @j2kun Hmm, why was it worthwhile to do the effort, did it obscure things, or are you space or time limited? > SMT-LIB I keep hearing about that usage more and more and more.
-
@j2kun@mathstodon.xyz 2026-04-15 19:40
@thalia @dougmerritt have you heard of the SMT backend added to MLIR? (Or maybe you were involved in that?)