Conversation

Notices

1. Embed this notice
   Haelwenn /элвэн/ :triskell: (lanodan@queer.hacktivis.me)'s status on Sunday, 22-Sep-2024 20:28:35 JST Haelwenn /элвэн/ :triskell:

   • Michael Dexter
   @dexter Perl as the highest sounds pretty weird given how inefficient Python is in comparison, and pretty sure Lua has among the lowest impacts. Go and Java in the first part also seems kind of weird.
   Really wonder how they came up with those numbers, specially given how different all those languages are.
   • Embed this notice
     Haelwenn /элвэн/ :triskell: (lanodan@queer.hacktivis.me)'s status on Sunday, 22-Sep-2024 20:50:34 JST Haelwenn /элвэн/ :triskell:

     in reply to

     • Michael Dexter
     @dexter Plus kind of gives score-vibes instead of like, how efficient each language are for their tasks, things like networking, decoding/encoding, high-activity daemon, low-activity daemon, throwaway utility (like most unix commands), …
   • Embed this notice
     Guillaume Rossolini (guillaumerossolini@infosec.exchange)'s status on Monday, 23-Sep-2024 20:27:29 JST Guillaume Rossolini

     in reply to

     @lanodan yeah, same here on all counts

     Plus the Wikipedia page on this methodology does warn about shortcomings in this benchmark

     The developers themselves highlight the fact that those doing research should exercise caution when using such microbenchmarks

     https://en.m.wikipedia.org/wiki/The_Computer_Language_Benchmarks_Game

     Haelwenn /элвэн/ :triskell: likes this.
   • Embed this notice
     Haelwenn /элвэн/ :triskell: (lanodan@queer.hacktivis.me)'s status on Monday, 23-Sep-2024 20:52:13 JST Haelwenn /элвэн/ :triskell:

     in reply to

     • Guillaume Rossolini
     @GuillaumeRossolini Yeah, I think a better one would be to analyze (rather than score) the standard/extended-standard library where quite a lot of languages ship cryptographic algorithms, json/xml encoders/decoders, http client/server library, …
     And I guess measuring the CPU time + memory usage of runtimes for initial launch, threads, `fork()`, idle/sleep, …
     
     Which reminds me of http://bulk.fefe.de/scalability/ which analyses OS and libc runtime.
   • Embed this notice
     Guillaume Rossolini (guillaumerossolini@infosec.exchange)'s status on Monday, 23-Sep-2024 21:02:30 JST Guillaume Rossolini

     in reply to

     @lanodan yep, this looked like some interesting problems to determine the capabilities of a language, gauge how fit each is as a general purpose tool, identify possible shortcomings

     But actual applications in the wild don’t necessarily implement these edge cases often (or ever), and that’s perhaps what that warning was about

     If you’re going to benchmark, at least share your versions and settings so that they are readily available to readers. And if you write an article covering the research, do the same?

     That said, I’m no researcher

     Haelwenn /элвэн/ :triskell: likes this.
   • Embed this notice
     Guillaume Rossolini (guillaumerossolini@infosec.exchange)'s status on Monday, 23-Sep-2024 21:05:38 JST Guillaume Rossolini

     in reply to

     @lanodan even in the link you shared, they say at the start

     On all of the operating systems, I took the default settings

     Which it seems to me, fine, but that isn’t complete or representative

     There could be at least three categories:

     • default settings
     • I took a stab at optimization, don’t quote me
     • community reviewed best settings
     Haelwenn /элвэн/ :triskell: likes this.
   • Embed this notice
     Haelwenn /элвэн/ :triskell: (lanodan@queer.hacktivis.me)'s status on Monday, 23-Sep-2024 21:11:01 JST Haelwenn /элвэн/ :triskell:

     in reply to

     • Guillaume Rossolini
     @GuillaumeRossolini Well changing the settings would be appropriate for knowing how to tune the operating system, which is all too often non-existant or lacking in official documentation.
     
     > community reviewed best settings
     
     Well for libre systems (which is what fefe used) that quite ought to be the default one.