Performance regressions on nom-8?

[hadronized]

Hello,

I’d rather post here because I’m not entirely sure whether the regression is genuine or if it’s just randomness. I have some data gathered via cargo bench and cargo benchcmp.

Let me know what you think

[marcdejonge]

Can you come with a clean example when the performance has a clear regression? That would help to figure out why there is any regression. Some simple example parser that we can compare v7 and v8 on would be helpful.

[Geal]

@marcdejonge I know @hadronized, we were already discussing off band. There's a benchmark output at https://paste.sr.ht/blob/03436c4cb74d9a9c058bb9cd5650ad9f27033222, and the benchmarking code is here https://git.sr.ht/~hadronized/glsl/tree/master/item/glsl/src/parse_benches.rs

I suspect nom is missing some inline tags in parts that would be easily inlined away by the compiler before