17:54 <FromGitter> <mattrberry> Heh, according to the Instruments profiler on mac, 2.3% of my application's performance is taken by Slice#check_writable

17:55 <FromGitter> <mattrberry> Instruments says that's the heaviest stack trace in the entire app lol

19:10 <FromGitter> <mattrberry> Is there a decent way to effectively template a method over constant parameters? Today, I have something like ⏎ ⏎ ```code paste, see link``` ⏎ ⏎ but I figure this would actually be worse for performance. While it saves computing the `second_instr` on every call, it'd require looking it up on the heap (if my assumption is correct) ... [https://gitter.im/cry

19:10 <FromGitter> ... stal-lang/crystal?at=635ecc0b0a8c6e22a1e8616c]

19:40 <FromGitter> <mattrberry> Having trouble completely trusting the Benchmark results haha. When running with only two reports I consistently see this ⏎ ⏎ ```code paste, see link``` [https://gitter.im/crystal-lang/crystal?at=635ed3195b0dbc04b8aa5246]

19:46 <FromGitter> <mattrberry> With a few more values on the heap / parsed at runtime, the closured approach seems to be increasingly bad and the macro approach pulls ahead a little ⏎ ⏎ ```code paste, see link``` [https://gitter.im/crystal-lang/crystal?at=635ed4aa27f328266d62cf29]

19:48 <FromGitter> <mattrberry> I guess the real takeaway here is that the macro approach is super ugly and that (as expected) checking bits at runtime is basically a nop

19:49 <FromGitter> <mattrberry> So my take is that macros aren't worth the fugliness for a sub-2% perf improvement even in a hot benchmark loop

19:51 <FromGitter> <mattrberry> This is a situation where it'd be nice to have some compile-time evaluation with the interpreter as opposed to the macro language :p

20:02 <yxhuvud>

20:02 <yxhuvud>

20:02 <yxhuvud> I don't think what you have told us is really enough to give any possibility for us to help you. In general I'd avoid an overreliance on procs as I wouldn't trust llvm to be able to optimize across proc boundaries.

20:30 <FromGitter> <mattrberry> What is confusing?