Newest at the top
| 2024-09-21 21:50:03 +0200 | <tomsmeding> | tuxpaint: https://paste.tomsmeding.com/2YnbokDI |
| 2024-09-21 21:49:59 +0200 | <EvanR> | it |
| 2024-09-21 21:49:56 +0200 | <EvanR> | probably because I have to tricked out with fancy font |
| 2024-09-21 21:49:21 +0200 | <EvanR> | in xterm there is noticable delay in haskell and C :( |
| 2024-09-21 21:48:45 +0200 | <tuxpaint> | terminal=st |
| 2024-09-21 21:48:43 +0200 | <EvanR> | the best term |
| 2024-09-21 21:48:35 +0200 | <tuxpaint> | i notice a difference in my zsh between ./pure and ./pureinc |
| 2024-09-21 21:48:34 +0200 | <EvanR> | I'll try xterm |
| 2024-09-21 21:48:20 +0200 | <tomsmeding> | I notice it less in fish, apparently fish prompt drawing takes time too :') |
| 2024-09-21 21:47:57 +0200 | <tomsmeding> | in bash, I do notice a difference between 'echo hi' and './pure' |
| 2024-09-21 21:47:24 +0200 | <EvanR> | imperceptibly different from the equivalent C program |
| 2024-09-21 21:47:11 +0200 | <tomsmeding> | that is true |
| 2024-09-21 21:47:07 +0200 | <dolio> | Sufficiently instantly for like 99.999% of use cases. |
| 2024-09-21 21:46:45 +0200 | <tomsmeding> | how instant is "instantly" :P |
| 2024-09-21 21:46:33 +0200 | <EvanR> | without the time command |
| 2024-09-21 21:46:20 +0200 | <EvanR> | it says real = 0.012s or 0.013s but when you press enter on ./pure, I am back to the prompt instantly |
| 2024-09-21 21:45:27 +0200 | <tuxpaint> | hmm. so it's waiting to exit or something? i noticed it was "slow" initially because of the time it spent to return me back to interactive |
| 2024-09-21 21:45:11 +0200 | synchromesh | (~john@2406:5a00:241a:5600:5db8:4c32:8611:a0fa) |
| 2024-09-21 21:44:58 +0200 | <tomsmeding> | how long does that take? |
| 2024-09-21 21:44:54 +0200 | <tomsmeding> | repeat that last command ~10 times |
| 2024-09-21 21:44:47 +0200 | <tomsmeding> | EvanR: `echo 'main = pure ()' >pure.hs; ghc pure.hs -o pure; time ./pure` |
| 2024-09-21 21:44:28 +0200 | <tomsmeding> | and the first thing after the wait is a futex wait |
| 2024-09-21 21:44:24 +0200 | <EvanR> | I have no noticed any startup time in haskell |
| 2024-09-21 21:43:52 +0200 | synchromesh | (~john@2406:5a00:241a:5600:5db8:4c32:8611:a0fa) (Read error: Connection reset by peer) |
| 2024-09-21 21:43:35 +0200 | <tomsmeding> | with `main = putStrLn "hi"`, and looking at strace output with nanosecond timestamps attached, a whole bunch of stuff happens, then the write, then 8ms of nothing, then a few more syscalls and exit |
| 2024-09-21 21:42:46 +0200 | <tomsmeding> | it seems the time is spent in shutdown, not in startup |
| 2024-09-21 21:42:01 +0200 | <tuxpaint> | huh, but if i set both above 20ms, it then caps at 22ms runtime again |
| 2024-09-21 21:41:25 +0200 | <tuxpaint> | so it seems it waits for either the first tick of the reschedule timer, or the clock timer or something? |
| 2024-09-21 21:41:05 +0200 | <tuxpaint> | yeah. if i set V=100ms, and i=5ms, i get 7ms |
| 2024-09-21 21:40:12 +0200 | <tomsmeding> | oops |
| 2024-09-21 21:40:09 +0200 | <tomsmeding> | s |
| 2024-09-21 21:39:07 +0200 | <tuxpaint> | i'm looking around for 20ms in the code. maybe it has something to do with the rescheduling timer? |
| 2024-09-21 21:39:00 +0200 | <tomsmeding> | case in point: I also got 12ms for the `main = pure ()` with default settings on my machine, but starting a full blown web server and immediately exiting with `exitSuccess` if there are no command line arguments (to ensure that nothing can be optimised away) takes 15ms |
| 2024-09-21 21:36:49 +0200 | <tuxpaint> | haskell is fun; it was my first language when learning to code many years ago. the IO monad was very nice to work with - didn't have a chance to do IO when i first touched haskell. |
| 2024-09-21 21:36:27 +0200 | <tuxpaint> | of course there's a balance between the user experience in how easy it is for me to write the cli (more features), and how fast it executes |
| 2024-09-21 21:36:20 +0200 | <tomsmeding> | same would hold for nodejs |
| 2024-09-21 21:36:06 +0200 | <tuxpaint> | i am, which is why it was very confusing that haskell was starting up in the same amount of time |
| 2024-09-21 21:35:51 +0200 | <tomsmeding> | tuxpaint: one thing that you may or may not be taking into account with perl and python is that they are interpreted, hence if you import a bunch of libraries, your startup time is going to suffer significantly |
| 2024-09-21 21:35:49 +0200 | merijn | (~merijn@204-220-045-062.dynamic.caiway.nl) (Ping timeout: 248 seconds) |
| 2024-09-21 21:33:36 +0200 | <tomsmeding> | strict functional language, funny syntax sometimes (especially data type declarations), somewhat fewer fancy type system features than haskell, first-class mutability |
| 2024-09-21 21:32:35 +0200 | <tuxpaint> | ocaml sounds fun, i may give it a try |
| 2024-09-21 21:32:11 +0200 | <tuxpaint> | yeah, so far i have perl, python, c, go, nim, rust, haskell, c# done |
| 2024-09-21 21:31:40 +0200 | <tomsmeding> | rust is a good candidate if you like haskell, otherwise (but I don't know anything about startup time here, so YMMV) ocaml |
| 2024-09-21 21:31:27 +0200 | <tuxpaint> | so i've been going through seeing how the basic experience of reading flags and files is, along with writing to stdout. |
| 2024-09-21 21:31:03 +0200 | merijn | (~merijn@204-220-045-062.dynamic.caiway.nl) |
| 2024-09-21 21:30:41 +0200 | <tuxpaint> | 12ms startup time in a cli is perceivable to a user. i'm exploring languages to write clis in. |
| 2024-09-21 21:30:01 +0200 | <tuxpaint> | interestingly, if i set it to a 8ms clock, runtime is 10ms. if i set it to 15ms clock, runtime is 17ms, up to 20ms, where anything there and above gives me ~22ms |
| 2024-09-21 21:29:50 +0200 | <tomsmeding> | tuxpaint: is this startup-time benchmark just curiosity, or do you have a real usecase for 1. wanting to use a high-level language like haskell, and 2. needing fast startup time? |
| 2024-09-21 21:28:43 +0200 | connrs | (~connrs@user/connrs) |
| 2024-09-21 21:28:11 +0200 | <tomsmeding> | ooh |