2016-06-20T00:05:46Z eschatologist joined #sbcl
2016-06-20T01:13:57Z stassats quit (Ping timeout: 276 seconds)
2016-06-20T01:19:59Z karswell quit (Remote host closed the connection)
2016-06-20T01:52:00Z ASau quit (Read error: Connection reset by peer)
2016-06-20T01:53:00Z ASau joined #sbcl
2016-06-20T01:56:31Z karswell joined #sbcl
2016-06-20T02:13:24Z PuercoPop is now known as Puercopop
2016-06-20T02:38:59Z loke joined #sbcl
2016-06-20T02:54:42Z DeadTrickster joined #sbcl
2016-06-20T02:56:32Z gema` joined #sbcl
2016-06-20T03:51:38Z DeadTrickster quit (Read error: Connection reset by peer)
2016-06-20T03:53:27Z DeadTrickster joined #sbcl
2016-06-20T04:24:38Z DougNYC quit
2016-06-20T05:03:50Z shka joined #sbcl
2016-06-20T05:08:17Z logrus_ quit (Quit: Leaving)
2016-06-20T05:18:28Z shka quit (Ping timeout: 244 seconds)
2016-06-20T05:28:55Z joshe joined #sbcl
2016-06-20T05:58:39Z DeadTrickster quit (Ping timeout: 276 seconds)
2016-06-20T06:09:21Z angavrilov joined #sbcl
2016-06-20T06:24:21Z jdz quit (Ping timeout: 258 seconds)
2016-06-20T06:29:09Z jdz joined #sbcl
2016-06-20T06:32:47Z gingerale joined #sbcl
2016-06-20T06:36:15Z attila_lendvai joined #sbcl
2016-06-20T06:36:15Z attila_lendvai quit (Changing host)
2016-06-20T06:36:15Z attila_lendvai joined #sbcl
2016-06-20T06:36:36Z gingerale- joined #sbcl
2016-06-20T06:37:09Z gingerale quit (Disconnected by services)
2016-06-20T06:37:18Z gingerale- is now known as gingerale
2016-06-20T06:56:13Z GrembleGremble joined #sbcl
2016-06-20T06:58:14Z GrembleGremble left #sbcl
2016-06-20T07:46:22Z flip214 quit (Remote host closed the connection)
2016-06-20T08:32:31Z DavidGu_Mobile joined #sbcl
2016-06-20T08:39:58Z guicho joined #sbcl
2016-06-20T08:54:53Z DavidGu_Mobile quit (Quit: Mutter: www.mutterirc.com)
2016-06-20T08:57:22Z karswell quit (Remote host closed the connection)
2016-06-20T08:57:37Z karswell joined #sbcl
2016-06-20T09:09:53Z DavidGu_Mobile joined #sbcl
2016-06-20T09:16:24Z guicho quit (Quit: さようなら)
2016-06-20T09:21:01Z barbone quit (Quit: KVIrc 4.2.0 Equilibrium http://www.kvirc.net/)
2016-06-20T09:27:23Z gingerale- joined #sbcl
2016-06-20T09:27:54Z gingerale quit (Disconnected by services)
2016-06-20T09:27:56Z gingerale- is now known as gingerale
2016-06-20T09:41:50Z DavidGu_Mobile_ joined #sbcl
2016-06-20T09:45:34Z DavidGu_Mobile quit (Ping timeout: 240 seconds)
2016-06-20T09:45:56Z ASau quit (Ping timeout: 244 seconds)
2016-06-20T09:53:18Z DavidGu_Mobile_ quit (Remote host closed the connection)
2016-06-20T10:16:32Z flip214 joined #sbcl
2016-06-20T11:01:11Z loke quit (Remote host closed the connection)
2016-06-20T11:20:18Z scymtym joined #sbcl
2016-06-20T12:05:53Z stassats joined #sbcl
2016-06-20T12:14:26Z angavrilov quit (Remote host closed the connection)
2016-06-20T12:14:39Z angavrilov joined #sbcl
2016-06-20T12:15:03Z gingerale quit (Ping timeout: 240 seconds)
2016-06-20T12:15:52Z gingerale joined #sbcl
2016-06-20T12:17:50Z flip214 quit (Ping timeout: 244 seconds)
2016-06-20T12:19:07Z flip214 joined #sbcl
2016-06-20T12:25:50Z ym joined #sbcl
2016-06-20T12:26:43Z stassats quit (Ping timeout: 252 seconds)
2016-06-20T12:27:58Z stassats joined #sbcl
2016-06-20T12:37:56Z eudoxia joined #sbcl
2016-06-20T13:11:14Z abruanese quit (Ping timeout: 244 seconds)
2016-06-20T13:17:50Z salva quit (Ping timeout: 272 seconds)
2016-06-20T13:25:55Z foom joined #sbcl
2016-06-20T13:29:52Z dougk_ joined #sbcl
2016-06-20T13:32:52Z salva joined #sbcl
2016-06-20T13:33:59Z salva quit (Max SendQ exceeded)
2016-06-20T13:34:18Z stassats quit (Ping timeout: 272 seconds)
2016-06-20T13:35:41Z salva joined #sbcl
2016-06-20T13:35:48Z stassats joined #sbcl
2016-06-20T13:36:32Z abruanese joined #sbcl
2016-06-20T13:37:33Z salva quit (Max SendQ exceeded)
2016-06-20T13:47:40Z salva joined #sbcl
2016-06-20T13:49:01Z stassats quit (Ping timeout: 258 seconds)
2016-06-20T13:51:22Z stassats joined #sbcl
2016-06-20T13:52:43Z salva quit (Ping timeout: 250 seconds)
2016-06-20T13:55:37Z salva joined #sbcl
2016-06-20T13:58:01Z stassats quit (Ping timeout: 252 seconds)
2016-06-20T13:58:51Z stassats joined #sbcl
2016-06-20T14:00:16Z nzambe quit (Quit: http://www.kiwiirc.com/ - A hand crafted IRC client)
2016-06-20T14:03:14Z salva quit (Ping timeout: 244 seconds)
2016-06-20T14:04:16Z nzambe joined #sbcl
2016-06-20T14:05:16Z barbone joined #sbcl
2016-06-20T14:05:52Z stassats quit (Ping timeout: 264 seconds)
2016-06-20T14:09:21Z stassats joined #sbcl
2016-06-20T14:13:03Z barbone quit (Ping timeout: 240 seconds)
2016-06-20T14:13:43Z stassats quit (Ping timeout: 244 seconds)
2016-06-20T14:15:22Z stassats joined #sbcl
2016-06-20T14:21:49Z salva joined #sbcl
2016-06-20T14:24:02Z abruanese quit (Read error: Connection reset by peer)
2016-06-20T14:25:51Z salva quit (Ping timeout: 240 seconds)
2016-06-20T14:25:55Z abruanese joined #sbcl
2016-06-20T14:30:40Z abruanese quit (Ping timeout: 272 seconds)
2016-06-20T14:36:12Z abruanese joined #sbcl
2016-06-20T14:40:38Z salva joined #sbcl
2016-06-20T14:48:25Z salva quit (Ping timeout: 260 seconds)
2016-06-20T14:58:41Z abruanese quit (Ping timeout: 244 seconds)
2016-06-20T15:08:20Z abruanese joined #sbcl
2016-06-20T15:26:49Z attila_lendvai quit (Ping timeout: 260 seconds)
2016-06-20T15:42:21Z abruanese quit (Quit: ZNC 1.6.3+deb1 - http://znc.in)
2016-06-20T15:43:09Z abruanese joined #sbcl
2016-06-20T15:55:15Z abruanese quit (Quit: ZNC 1.6.3+deb1 - http://znc.in)
2016-06-20T16:01:10Z abruanese joined #sbcl
2016-06-20T16:19:52Z DeadTrickster joined #sbcl
2016-06-20T16:28:15Z shka joined #sbcl
2016-06-20T16:33:35Z DeadTrickster quit (Ping timeout: 244 seconds)
2016-06-20T16:51:14Z eschatologist quit (Ping timeout: 250 seconds)
2016-06-20T16:57:03Z abruanese quit (Ping timeout: 240 seconds)
2016-06-20T16:57:39Z salva joined #sbcl
2016-06-20T17:03:43Z barbone joined #sbcl
2016-06-20T17:16:40Z salva quit (Ping timeout: 264 seconds)
2016-06-20T17:17:04Z salva joined #sbcl
2016-06-20T17:25:28Z salva quit (Ping timeout: 250 seconds)
2016-06-20T17:26:59Z salva joined #sbcl
2016-06-20T17:32:24Z salva quit (Ping timeout: 250 seconds)
2016-06-20T17:35:53Z DeadTrickster joined #sbcl
2016-06-20T17:46:30Z abruanese joined #sbcl
2016-06-20T17:47:26Z DavidGu_Mobile_ joined #sbcl
2016-06-20T17:52:33Z eschatologist joined #sbcl
2016-06-20T17:53:53Z eschatologist quit (Client Quit)
2016-06-20T17:56:18Z eschatologist joined #sbcl
2016-06-20T17:57:35Z eudoxia quit (Quit: Leaving)
2016-06-20T18:09:46Z DavidGu_Mobile_ quit (Remote host closed the connection)
2016-06-20T18:49:24Z eudoxia joined #sbcl
2016-06-20T18:59:48Z DGASAU quit (Ping timeout: 244 seconds)
2016-06-20T19:04:38Z DGASAU joined #sbcl
2016-06-20T19:50:24Z mordocai_ is now known as mordocai
2016-06-20T20:15:41Z jasom: any interest in replacing MT-19937 with something like xorshift1024*?  It is definitely faster, and considered higher quality.
2016-06-20T20:17:14Z eudoxia quit (Quit: Leaving)
2016-06-20T20:24:47Z angavrilov quit (Remote host closed the connection)
2016-06-20T21:04:36Z chris2 quit (Read error: No route to host)
2016-06-20T21:23:41Z ASau joined #sbcl
2016-06-20T21:23:55Z gingerale quit (Remote host closed the connection)
2016-06-20T21:25:33Z chris2 joined #sbcl
2016-06-20T21:26:36Z pkhuong_: more interest in figuring out how to make our PRNG strategy generic (enough)
2016-06-20T21:33:18Z jasom: ah
2016-06-20T21:33:39Z jasom: as in make random-chunk not hard-coded to 32-bits? or something else?
2016-06-20T21:34:16Z pkhuong_: I don't know what the best approach is
2016-06-20T21:34:27Z pkhuong_: do we want to support both float and bit-oriented PRNGs?
2016-06-20T21:34:48Z jasom: ah
2016-06-20T21:37:03Z scymtym quit (Ping timeout: 240 seconds)
2016-06-20T21:53:30Z fe[nl]ix: pkhuong_: do what /dev/urandom does
2016-06-20T21:55:04Z pkhuong_: fe[nl]ix: bit oriented
2016-06-20T21:55:33Z fe[nl]ix: yepp
2016-06-20T21:56:08Z fe[nl]ix: floats can be obtained from that
2016-06-20T22:09:59Z pkhuong_: yes. some PRNGs directly generate floats.
2016-06-20T22:10:06Z pkhuong_ is now known as pkhuong
2016-06-20T22:13:46Z stassats: all those test/test-not/key/from-end make things hard to optimize
2016-06-20T22:17:30Z fe[nl]ix: stassats: have you tried using beach's version of the sequence functions ?
2016-06-20T22:17:35Z stassats: no
2016-06-20T22:19:19Z fe[nl]ix: he said he evaluated the current state-of-the-art regarding algorithms and the ones in SICL are faster than SBCL's in some edge cases
2016-06-20T22:19:46Z stassats: i made things faster in some edge cases in the last couple of days
2016-06-20T22:21:31Z stassats: without using any new approaches, just simply doing less stuff
2016-06-20T22:21:46Z stassats: like not doing nthcdr on the same list twice
2016-06-20T22:22:11Z stassats: or copying remaining bits from end to length, but sharing them
2016-06-20T22:22:52Z stassats: short of generating a cross between all the options, i don't really see any way to make it significantly faster
2016-06-20T22:23:47Z stassats: (which some function already do, but for complicated stuff like remove-duplicates that may be too much)
2016-06-20T22:23:57Z stassats: size wise
2016-06-20T22:26:57Z stassats: generating only the test-checking bits does help somewhat, but that adds function call overhead
2016-06-20T22:33:32Z stassats: maybe with compiler support it can generate better test, instead of (lambda (...) (funcall test (funcall key))) the compiler would generate (lambda (...) (eql (car x) (car y)))
2016-06-20T22:33:49Z stassats: less call overhead, less code size explosion
2016-06-20T22:40:14Z shka quit (Ping timeout: 260 seconds)
2016-06-20T22:49:52Z stassats: that might actually be the best strategy to deal with test/test-not/key
2016-06-20T23:09:05Z eschatologist quit (Quit: Textual IRC Client: www.textualapp.com)
2016-06-20T23:13:44Z barbone quit (Quit: KVIrc 4.2.0 Equilibrium http://www.kvirc.net/)
2016-06-20T23:13:46Z jasom: Question: there doesn't seem to be any logic to generate e.g. (rol x 10) for (logior (ash x -54) (ash (ldb (byte 54 0) x) 10)); at what point in the compiler would it make most sense to try and add it?
2016-06-20T23:14:26Z stassats: it's not very good at these things
2016-06-20T23:14:52Z stassats: use sb-rotate-byte
2016-06-20T23:14:52Z jasom: does it have any equivalent of peepholes?
2016-06-20T23:15:14Z jasom: e.g. in a C compiler I'd look for "shift shift rotate" in the low-level IR
2016-06-20T23:15:24Z jasom: rather "shift shift or"
2016-06-20T23:16:07Z stassats: the c compiler doesn't have to deal with unknown types
2016-06-20T23:17:02Z jasom: fair enough.
2016-06-20T23:21:23Z jasom: sb-rotate-byte is in contrib, which is probably why I didn't find it at first
2016-06-20T23:26:35Z jasom: heh https://bugs.launchpad.net/sbcl/+bug/721463
2016-06-20T23:27:33Z stassats: working at the target instruction level is not a good idea
2016-06-20T23:27:41Z stassats: well, the whole compiler is not very good at this sort of thing
2016-06-20T23:29:27Z jasom: aren't types concrete at IR2?
2016-06-20T23:29:45Z stassats: do you consider T a concrete type?
2016-06-20T23:30:00Z jasom: I was thinking more unsigned-reg
2016-06-20T23:30:54Z stassats: you can't go from T to unsigned-reg
2016-06-20T23:33:06Z jasom: Just wondering if there is a pass that transforms IR2 to IR2 that can replace sequences of VOPs with a different sequence of VOPs
2016-06-20T23:33:44Z stassats: there is
2016-06-20T23:34:03Z jasom: that seems like it would  be a good place to do this, right?
2016-06-20T23:34:40Z stassats: well, any place is a good place, but how do you know which vops do what?
2016-06-20T23:38:35Z abruanese quit (Ping timeout: 258 seconds)
2016-06-20T23:42:59Z abruanese joined #sbcl
2016-06-20T23:43:31Z ASau quit (Ping timeout: 240 seconds)
2016-06-20T23:48:35Z jasom: fast-ash-c/unsigned->unsigned appears to have the same semantics on all archs, for example
2016-06-20T23:48:56Z ASau joined #sbcl
2016-06-20T23:49:03Z jasom: to make it arch independent, it would require that the vops with the same name on different archs have the same semantics.
2016-06-20T23:49:24Z jasom: but the only alternative to enforcing that would be to work at the assembly level
2016-06-20T23:52:37Z stassats: that's not going to fly
2016-06-20T23:55:01Z stassats: especially when for some patterns it will be too late to deal with at the VOP level
2016-06-20T23:56:13Z jasom: Personally, I don't care about "doesn't work for some patterns", I care about "will work for some real-world patterns"; if the latter is true, than it could be useful.  The former is true of essentially all possible optimizations.
2016-06-20T23:56:34Z stassats: for real world patterns
2016-06-20T23:56:45Z ASau quit (Ping timeout: 260 seconds)
2016-06-20T23:57:25Z jasom: but if the requirement that all vops with the same name have semantics that are arch independent is a non-starter, than I agree there's no good way to do this.
2016-06-20T23:57:40Z stassats: optimizing (logior (ash x -54) (ash (ldb (byte 54 0) x) 10)) isn't even that important