00:09 (quit) jonrafkind: Ping timeout: 255 seconds
00:11 (join) jonrafkind
00:19 (quit) jonrafkind: Ping timeout: 245 seconds
00:26 (join) jonrafkind
00:44 (quit) jonrafkind: Ping timeout: 256 seconds
00:57 (join) jonrafkind
01:29 (quit) jao: Ping timeout: 245 seconds
01:31 (quit) yoklov: Quit: Leaving.
01:38 (quit) dmac: Ping timeout: 260 seconds
01:39 (join) jao
03:06 (quit) jonrafkind: Ping timeout: 245 seconds
04:04 (quit) jao: Ping timeout: 260 seconds
04:45 (join) MayDaniel
05:09 (quit) MayDaniel: Read error: Connection reset by peer
05:10 (join) realitygrill_
05:11 (quit) realitygrill: Ping timeout: 245 seconds
05:11 (nick) realitygrill_ -> realitygrill
05:12 (join) MayDaniel
05:14 (join) Blkt
05:16 Blkt: good day eveyone
05:17 (quit) MayDaniel: Read error: Connection reset by peer
05:17 ohwow: hi
05:44 (join) Blkt`
05:46 (quit) Blkt: Ping timeout: 248 seconds
06:12 (join) neilv
06:20 (quit) neilv: Quit: Leaving
06:20 (join) masm
07:24 (quit) Blkt`: Ping timeout: 276 seconds
08:11 (join) mceier
08:49 (join) bluezenix
08:55 (join) yoklov
09:30 (quit) yoklov: Quit: Leaving.
09:37 (join) Lajla
10:04 (quit) bluezenix: Quit: Leaving.
10:25 (join) yoklov
10:34 (join) Blkt
11:29 (join) MayDaniel
11:31 (quit) MayDaniel: Read error: Connection reset by peer
11:40 (quit) yoklov: Quit: Leaving.
11:48 (join) bmp
11:49 (quit) bmp: Client Quit
12:02 (join) jonrafkind
12:31 (quit) jonrafkind: Ping timeout: 245 seconds
12:35 (join) anRch
12:36 (join) dmac
12:38 (join) jonrafkind
12:45 (quit) anRch: Quit: anRch
12:50 (join) anRch
12:52 (quit) anRch: Client Quit
12:56 (quit) mceier: Ping timeout: 260 seconds
12:57 (quit) chemuduguntar: Remote host closed the connection
12:57 (join) mceier
13:05 (join) anRch
13:07 (quit) dmac: Ping timeout: 276 seconds
13:10 (quit) jonrafkind: Ping timeout: 256 seconds
13:13 (join) dnolen
13:17 (quit) anRch: Quit: anRch
13:20 (join) jonrafkind
13:37 (join) yoklov
14:15 (join) jao
14:17 yoklov: if i'm already using flonums, is there anything i can do to improve the performance of math-intensive code?
14:17 yoklov: assuming the computation can't be like memoized
14:21 jonrafkind: use typed/racket
14:21 yoklov: yeah i had considered that, is it still able to work with things like racket/gui?
14:22 jonrafkind: well im not sure..
14:22 (quit) dnolen: Quit: dnolen
14:22 bremner: I guess if you split things into modules appropriately, it should be ok
14:22 bremner: emphasis on guess
14:22 yoklov: yeah
14:23 jonrafkind: how much faster do you need things to go?
14:23 jonrafkind: maybe its time to write stuff in C?
14:24 yoklov: haha, i mean its possible
14:24 yoklov: i'm implementing a (basic) 3d rendering engine myself doing pixel operations, and the framerate is about 20fps right now
14:24 yoklov: it was 3 to start out with though
14:25 yoklov: i could get better performance if i used draw-polygon, but i've done that before.
14:26 yoklov: and the point of it is that i'm not doing it in C or a faster language, because i've seen impressive results with racket before, but how they're achieved seems totally mysterious to me.
14:28 (quit) Blkt: Ping timeout: 256 seconds
14:29 (quit) jao: Ping timeout: 245 seconds
14:33 yoklov: but yeah i'm already rendering to a bytestring before drawing it so i could just do that in typed racket and see if that helps at all
14:37 (join) jao
14:43 (quit) jonrafkind: Ping timeout: 256 seconds
14:52 (nick) samth_away -> samth
14:53 samth: yoklov: are you using the unsafe operations?
14:55 yoklov: no
14:55 yoklov: but i was about to try them
14:56 (join) jonrafkind
14:56 yoklov: are they significantly unsafe like the reference implies, or are they just like c code unsafe
14:56 samth: those are the same thing, yoklov :)
14:57 samth: if you follow the contracts, they work correctly
14:57 samth: if you don't, your behavior will be unpredictable
14:57 samth: but they will allow you to get significantly faster performance
14:57 samth: jonrafkind: look here: http://shootout.alioth.debian.org/u32/benchmark.php?test=mandelbrot&lang=all
14:57 rudybot: http://tinyurl.com/3zhzebm
14:58 yoklov: yeah, i've seen that before
14:58 yoklov: very impressive
14:58 samth: note that racket #3 (the serial version) is faster than racket #4 (the parallel version)
14:58 samth: i'm mostly just answering jonrafkind's question from thursday
14:58 samth: he wasn't sure if they'd both be up there still
14:58 yoklov: hm, i think the last time i saw the mandelbrot function was a macro.
14:59 samth: it still is
14:59 jonrafkind: shouldnt the cpu secs be much larger than elapsed secs for the parallel version?
14:59 samth: look at `M'
14:59 jonrafkind: they lookt o be the same for all racket versions
14:59 yoklov: i'm looking at the wrong one then
14:59 samth: jonrafkind: that's on the single core machine
14:59 samth: jonrafkind: look here: http://shootout.alioth.debian.org/u32q/benchmark.php?test=mandelbrot&lang=all
14:59 rudybot: http://tinyurl.com/c4qfut
14:59 samth: that's on 4 cores
14:59 yoklov: ah
15:00 yoklov: yeah i didnt realize there were four
15:00 jonrafkind: on the quad core are all the languages parallel?
15:00 jonrafkind: i mean racket is close to ocaml, but is ocaml serial or parallel?
15:01 jonrafkind: oh i guess it is.. oops
15:01 jonrafkind: thats impressive that racket is at the same level as ocaml
15:03 yoklov: yeah
15:03 samth: jonrafkind: if you look at the ocaml impl, you'll see that it forks separate processes
15:03 samth: b/c ocaml doesn't have a parallel runtime
15:03 samth: also, you shouldn't be so surprised that racket is as fast as ocaml
15:04 samth: your advisor is pretty good at this stuff :)
15:04 jonrafkind: i thought it became competitive due to typed/racket optimizations
15:04 jonrafkind: no?
15:05 samth: jonrafkind: typed/racket doesn't do anything you can't do yourself (with a lot of effort)
15:06 jonrafkind: that was not the answer i was looking for
15:06 jonrafkind: is the mandelbrot benchmark using typed/racket with optimizations turned on?
15:06 samth: jonrafkind: racket is competetive w/ ocaml b/c the runtime and compiler and jit compiler are all well implemented
15:06 samth: jonrafkind: no
15:07 samth: currently, using typed racket incurs noticeable startup performance cost
15:07 samth: so using it for benchmarketing isn't a good idea
15:07 samth: the plan is that sub-modules will fix this
15:08 samth: at which point i hope to use TR in some shootout benchmarks
15:08 samth: also, TR can't prove the safety of some of the fixnum and bounds-check optimization being done manually in those programs
15:10 (quit) realitygrill: Ping timeout: 248 seconds
15:10 (join) realitygrill
15:13 (join) bluezenix
15:13 yoklov: well, i guess this is slightly faster
15:13 yoklov: it feels a lot smoother
15:14 yoklov: but my fps counter still says that it's about 2fps faster than it was now that i'm using the unsafe ops
15:14 yoklov: maybe i'm loosing my speed elsewhere
15:24 (quit) yoklov: Quit: Leaving.
15:26 (join) yoklov
15:27 yoklov: weird, switching to unsafe operations offered almost no speed increase on my laptop, but it's around 4x faster on my desktop
15:32 samth: yoklov: presumably the speed on your laptop is dominated by something else
15:32 (quit) mceier: Read error: Operation timed out
15:32 (join) mceier
15:34 yoklov: hm, maybe its that profiling is turned on on my laptop
15:36 samth: yoklov: are you running this under drracket?
15:36 yoklov: yeah, now it's about twice as fast on my laptop as it was
15:36 yoklov: yes
15:36 yoklov: should i not be?
15:36 samth: that's likely to introduce overhead that makes benchmarking less reliable
15:36 yoklov: yeah
15:36 samth: when i want to benchmark something, i just run it in plain racket at the shell
15:37 yoklov: that was far more significant of a performance boost than i expected
15:38 yoklov: 80fps -> 260fps
15:39 yoklov: do you guys still use drracket to edit code? or do you use emacs or something?
15:39 samth: i use drracket all the time
15:39 samth: except when i've broken it :)
15:40 samth: online check-syntax is so nice to have; now i'll definitely never go back to emacs
15:40 yoklov: hm, i wasn't really aware that was a thing
15:40 samth: yoklov: you need the nightly/git versions of drracket
15:40 samth: it'll be in the next release
15:41 yoklov: oh, i have the nightly version on my laptop
15:41 yoklov: because it runs mac and was incredibly buggy for a while
15:48 yoklov: still though, thanks for letting me know that running in racket creates overhead.
15:50 samth: yoklov: one of the interesting things about drracket is that it runs programs in-process
15:51 samth: racket has, over the past 15 years, developed a lot of interesting technology for safely running user programs in-process, whereas other systems (eclipse, chrome, firefox) are moving to running everything out-of-process
15:58 (nick) samth -> samth_away
16:35 (quit) bluezenix: Quit: Leaving.
16:40 (join) dnolen
17:04 (join) poet
17:30 (quit) adzuci: Ping timeout: 255 seconds
17:38 (quit) jao: Ping timeout: 248 seconds
17:43 (quit) dnolen: Quit: dnolen
17:44 (join) dnolen
17:56 (join) chturne
18:13 (quit) realitygrill: Read error: Connection reset by peer
18:15 (join) realitygrill
18:30 (quit) mceier: Quit: leaving
18:34 (join) adzuci
19:21 (join) jao
19:43 (join) neilv
20:10 (quit) neilv: Remote host closed the connection
20:15 (quit) chturne: Ping timeout: 256 seconds
20:25 (join) offby1`
20:26 (join) neilv
20:27 (quit) offby1: Ping timeout: 245 seconds
20:31 (quit) masm: Quit: Leaving.
20:36 (join) _danb_
20:47 (nick) offby1` -> offby1
21:24 (quit) offby1: Changing host
21:24 (join) offby1
21:39 (quit) jao: Ping timeout: 248 seconds
21:42 (join) ChibaPet
21:47 (quit) ChibaPet: Quit: Leaving.
23:32 (join) chemuduguntar