chrisdone / advent-2017-maze-rust-haskell Goto Github PK

They don’t make a difference and are source of criticism.

If they do, it should be raised as a compiler ticket perhaps.

My machine

$ uname -a
Linux chalcis 4.4.0-98-generic #121-Ubuntu SMP Tue Oct 10 14:24:03 UTC 2017 x86_64 x86_64 x86_64 GNU/Linux

Lower single-threaded perf than a 6700K or 7700K but otherwise good parallelism.

Rust

Haskell, native codegen

Haskell, LLVM backend

Seems broadly similar. Would need to setup a criterion harness to get a good sample.

Using LLVM made the answer twice as fast on my machine:

# Without LLVM
reading file into bytestring took 127.78 µs
parsing 1058 lines took 47.70 µs
25608482, is the answer, it took 85.76 ms

# With LLVM
reading file into bytestring took 134.40 µs
parsing 1058 lines took 31.28 µs
25608482, is the answer, it took 39.82 m

Just to keep comparisons fair, try running the Rust version with cargo run --release. The additional release flag will let cargo know to remove all the debug symbols from the build and do additional optimizations from the final build.

Please also update the Readme to include this.