3 ms·

The example in the readme: http://ix.io/cdR http://ix.io/cdR Emitted llvm ir doesn't look much more compact either: http://ix.io/cdS http://ix.io/cdS Clearly

The example in the readme: http://ix.io/cdR http://ix.io/cdR Emitted llvm ir doesn't look much more compact either: http://ix.io/cdS http://ix.io/cdS Clearly there is still plenty of room for optimizing brainfuck compilers!

eridius 12y ago
Eh, compile with -O and let LLVM optimize it.
- ben0x539 12y ago
  Is `rustc --opt-level 3 --emit ir` not right?
  - eridius 12y ago
    The optimization level is given to LLVM. The IR itself should be the same for all optimization levels AFAIK.
    - dbaupp 12y ago
      Almost all LLVM optimisations map IR to IR, and --emit=ir prints the final IR just before codegen (i.e. after optimisations happen).
      - eridius 12y ago
        Oh? I was under the impression it was the IR that was given to LLVM to begin with. Interesting.
dbaupp 12y ago
Yeah, the implementation is literally just a token-by-token transliteration into Rust, just hoping that LLVM makes it go fast (and it does go pretty fast, e.g. it's faster than bff4[1], which Esolangs[2] touts as the fastest interpreter). The Rust output & optimised IR clearly show there's room for improvement, though. [1]: http://mazonka.com/brainf/ http://mazonka.com/brainf/ [2]: http://esolangs.org/wiki/Brainfuck_implementations#Optimizing_implementations http://esolangs.org/wiki/Brainfuck_implementations#Optimizin...