| --- Log opened Tue May 10 00:00:33 2016 |
| 00:20 | | * McMartin fiddles with syscalls in Linux through assembly code |
| 00:20 | <&McMartin> | ... this is a lot more like MS-DOS than I was expecting it to be |
| 00:21 | < [R]> | https://i.imgur.com/F30oHlS.png |
| 00:21 | < [R]> | Oh? Is that a good or a bad thing? |
| 00:22 | <&McMartin> | Hrm. I suppose at the end of the day it's not much of any kind of thing, actually |
| 00:22 | <&McMartin> | It's more comfortable *for me* |
| 00:22 | <&McMartin> | But it means syscalls work via triggering interrupts and passing data around in registers |
| 00:23 | <&McMartin> | Which I'm more comfortable with than messing around with alignment guarantees and passing stuff on the stack, as Windows and (I think?) BSD do. |
| 00:23 | <&McMartin> | On the minus side, the Windows approach means there's no real programmatic difference between making a syscall and calling something from a DLL |
| 00:23 | <&McMartin> | Which is cleaner[*] |
| 00:24 | <&McMartin> | [*] module ABIs across your DLLs, which are actually quite likely to be different after 20 years, but you're an assembly programmer, you take that shit in stride |
| 00:24 | <&McMartin> | *modulo |
| 00:26 | | Derakon[AFK] is now known as Derakon |
| 00:27 | | catadroid [catalyst@Nightstar-labuc5.dab.02.net] has joined #code |
| 00:29 | | catalyst [catalyst@Nightstar-bt5k4h.81.in-addr.arpa] has quit [[NS] Quit: Leaving] |
| 00:49 | <&McMartin> | catadroid: Hey, I had a point of C++ style to ask you about |
| 00:50 | <&McMartin> | If I am seized by the brainworms and actually make use of the offsetof() macro, how do I actually get a pointer out of base + offset? |
| 00:51 | <&McMartin> | Is the standard way really to reinterpret_cast base pointer to char *, add the offset, then reinterpret_cast back to member type? |
| 00:51 | <&McMartin> | *member type pointer |
| 00:51 | <&McMartin> | Or is there some less C-y way to rock this |
| 00:54 | < catadroid> | Rather too tired to think it through right now |
| 00:55 | <&McMartin> | OK, no problem |
| 00:56 | <&McMartin> | You're just the most authoritative person I know of on this. If there's a "here's a good place to find stuff like that" source too, I will also happily take pointers on that |
| 00:56 | <&McMartin> | But, uh, yeah, it's 1 AM for you now, isn't it |
| 00:56 | <&McMartin> | I suggest peppermint tea |
| 00:56 | < catadroid> | But I guess I would cast to uintptr_t or similar rather than char* |
| 00:57 | | catadroid` [catalyst@Nightstar-g4k01n.dab.02.net] has joined #code |
| 00:57 | < catadroid`> | Erk |
| 01:00 | | catadroid [catalyst@Nightstar-labuc5.dab.02.net] has quit [Ping timeout: 121 seconds] |
| 01:18 | | Thalass is now known as thalass|brb |
| 02:13 | <&Derakon> | ...should I be reading uintptr_t as anything other than "unsigned int pointer"? |
| 02:14 | < [R]> | No |
| 02:14 | < [R]> | I don't get why you'd use that over void * for something like that, but w/e |
| 03:02 | <@celticminstrel> | Isn't uintptr_t the int type guaranteed to be the same size as void*? |
| 03:02 | <@celticminstrel> | ie, not actually a pointer type, but an integer type that could store a pointer. |
| 03:07 | <&McMartin> | What CM said |
| 03:07 | <&McMartin> | And you don't use void * because void * cannot be serialized nor can it have math done to it |
| 03:33 | <@Reiv> | HEY VORN (and McMartin and other math dudes) |
| 03:34 | <@Reiv> | I have a fun math problem |
| 03:34 | <@Reiv> | Part of it being insane units, but here goes |
| 03:34 | <@Reiv> | Data set of 14000, 200 samples per week results in 15min of clashes as people accidentally take the same sample (it is psuedorandom, so obviously they're all picking from nearer the top of the list) |
| 03:35 | <~Vornicus> | wat |
| 03:35 | <~Vornicus> | "pseudorandom" should not give |
| 03:35 | <@Reiv> | They want to increase the sample rate to 400 pw, which will also start to eat into the data set by 200 pw (which is to say, after 10 weeks there should be 12000 left) |
| 03:36 | <~Vornicus> | I feel like there's even more missing context. |
| 03:36 | <@Reiv> | ... so, of course, after a year there will be a smaller data set and the increased data sampling, so the clashes are going to increase exponentially, yes? |
| 03:37 | <@Reiv> | "We have a huge pile of work. We're going to increase our rate we work on it so it actually goes down instead of staying steady. We currently waste time when two people pick the same data point simultaneously*. How much time is going to get wasted if we increase the sampling rate, which will also reduce the pile of work?" |
| 03:38 | <~Vornicus> | Okay. So you're pulling in 200 items of work a week. You're then "trying" to parcel out 200 items of work per week |
| 03:38 | <&Derakon> | Higher sampling rate leads to more frequently grabbing the same piece of work, aye? |
| 03:38 | <@Reiv> | Correct |
| 03:38 | <~Vornicus> | But people are colliding on the work selected. |
| 03:38 | <&Derakon> | Fix your goddamn sampling function~ |
| 03:38 | <@Reiv> | And I suspect it is quadratic at minimum |
| 03:38 | <@Reiv> | Derakon: Our sampling function is literally a list that's spat out each day |
| 03:39 | <@Reiv> | People are choosing 'at random' from it, but they are /people/ and there are no dice on hand |
| 03:39 | <@Reiv> | I also don't really know where to buy a d14000 |
| 03:39 | <&Derakon> | Oh. |
| 03:39 | <@Reiv> | So it's not exactly true-random |
| 03:39 | <&Derakon> | Replace your people with computers, problem solved~ |
| 03:39 | <@Reiv> | I am writing the justification for 'how much time will my proper, actually-sane solution save' |
| 03:40 | <@Reiv> | And ran into a lovely little math knot while I was there |
| 03:40 | <&Derakon> | This sounds like a dynamic systems / linear algebra problem, I think. |
| 03:40 | <@Reiv> | "So wait, we do 200/wk at the moment, and we waste maybe 15mins on clashes. Not too bad. But you want to increase to 400/wk, and this is going to start reducing the body of work by 200/wk accordingly, so ... shit, what the hell does that do to your collision rate" |
| 03:41 | <@Reiv> | "What does this look like in a years time, because then I can justify spending the money now" |
| 03:41 | <&Derakon> | Assuming perfect randomness, collision rate should be a function of the number of items and the amount of time it takes for a single person to process one item. |
| 03:41 | <&Derakon> | And the number of people doing processing. |
| 03:42 | <&Derakon> | Like, if you have 1 item, then you will have N collisions, N = number of people. |
| 03:42 | <@Reiv> | I was hoping for a ratio and/or multiplier, because I don't know how many actual collisions are involved, just the amount of time wasted accordingly |
| 03:42 | <@Reiv> | because administrative staff are lovely people, but they don't think in the raw beauty of mathematics like we do~ |
| 03:42 | <&Derakon> | I'm inclined to say "throw Monte Carlo at it". |
| 03:43 | <&Derakon> | I don't appreciate math's raw beauty either~ |
| 03:43 | <@Reiv> | ... heh |
| 03:43 | | * Vornicus thinks at it |
| 03:44 | <@Reiv> | If 200 samples from 14000 is 15min, what is 400 samples from 4000? |
| 03:44 | <~Vornicus> | Shrieking. |
| 03:44 | <@Reiv> | I'm pretty sure that's exponential or something equally silly. |
| 03:44 | <&Derakon> | Like I said, I think it's a linear algebra problem. |
| 03:44 | <&Derakon> | Those are not trivial to calculate. |
| 03:44 | <@Reiv> | Pity I don't know linear algebra |
| 03:44 | <@Reiv> | I note I am overthinking this of course |
| 03:44 | <~Vornicus> | Just shrieking. |
| 03:44 | <@Reiv> | Half my co-workers just write in "Value: 1500" or something and hope no-one notices |
| 03:45 | <@Reiv> | But nooo, I want to do maaaath don't I, because I am a craven soul who grew up valuing truth |
| 03:45 | <@Reiv> | I could just go a linear comparison. |
| 03:45 | <@Reiv> | It'll be much, much worse than it really is |
| 03:45 | <@Reiv> | But it might be right! |
| 03:45 | <@Reiv> | Or at least, it'll be a 'conservative estamite' |
| 03:45 | <@Reiv> | Which is probably for the best |
| 03:46 | <@Reiv> | Because who are we kidding, if it gets bad enough and the sample small enough they'll just stop using psuedorandom |
| 03:46 | <@Reiv> | So fuckit, I'll do that. |
| 03:46 | <&Derakon> | Like I said, throw Monte Carlo at it. |
| 03:46 | <@Reiv> | !14000/200*15 |
| 03:46 | <@Reiv> | ... aw, no dicebot |
| 03:46 | <@Reiv> | I'm not clever enough to do a monte carlo |
| 03:46 | <@Reiv> | And certainly not on work time |
| 03:47 | <@Reiv> | ... I am not clever enough to algebra either, apparently |
| 03:48 | <@Reiv> | If one in 70 is 15; one in 10 would be... 105? |
| 03:49 | <@Reiv> | Fuckit, it's 105. So then I take the average of these (60) and declare that at the accelerated pace, we have provided a solution to avoid an hour a weeks' wasted work. |
| 03:49 | <@Reiv> | This is of course hilariously wrong |
| 03:49 | <&Derakon> | Heh. |
| 03:50 | | thalass|brb is now known as Thalass|slep |
| 03:51 | <@Reiv> | But it has numbers and there are so many intangibles to the numbers, including whether or not 400/wk would really cut it down or if the rate of incoming increases so we stay steady [which reduces the value], or if they start hitting more collisions so change their method [which reduces the value] or if they increase the number of people working on it [which increases the value] that number-outta-a-hat might as well stay simple when it's impossibl |
| 03:54 | <&Derakon> | I maintain that you should just assign people their work items so there are no collisions. |
| 03:54 | | Thalass|slep [thalass@Nightstar-283.o7s.158.104.IP] has quit [Operation timed out] |
| 03:54 | <~Vornicus> | Yes, the difficulty here is convincing manglement at that |
| 03:59 | <@Reiv> | No, the trick is arranging a method so that people don't get collisions |
| 03:59 | <@Reiv> | We have a tool for that |
| 03:59 | <@Reiv> | Literally |
| 04:00 | <@Reiv> | I'm just justifying the effort in arranging it to be used, and justifying training the people in how to use it |
| 04:04 | <@Reiv> | It will literally take a list constructed in SQL (like the one they currently have) and loads it nightly, /checks against the previous lists/ to see if something is a duplicate or not. If so, leaves it on the list. If it's new, adds it. If it's gone, hides it. |
| 04:04 | <@Reiv> | This persistent list then has the ability to have notes saved against it. |
| 04:04 | <@Reiv> | And 'come back to it later', 'ignore it's done now' etc. |
| 04:05 | <@Reiv> | Little bit useful. |
| 04:05 | <@Reiv> | Takes a bit of config to get the SQL to spit out the right shaped columns. |
| 04:05 | <@Reiv> | So they go through us |
| 04:05 | <@Reiv> | And so now I have to pull a dollar value to my config time. ;_; |
| 04:11 | | Turaiel is now known as Turaiel[Offline] |
| 04:21 | | Crossfire [Z@Nightstar-r9lk5l.cust.comxnet.dk] has quit [Ping timeout: 121 seconds] |
| 05:04 | | catadroid` [catalyst@Nightstar-g4k01n.dab.02.net] has quit [The TLS connection was non-properly terminated.] |
| 05:04 | | catadroid [catalyst@Nightstar-g4k01n.dab.02.net] has joined #code |
| 05:11 | | Derakon is now known as Derakon[AFK] |
| 05:47 | | celticminstrel [celticminst@Nightstar-q0f7bb.dsl.bell.ca] has quit [[NS] Quit: And lo! The computer falls into a deep sleep, to awake again some other day!] |
| 06:09 | | Reiv [NSwebIRC@Nightstar-q8avec.kinect.net.nz] has quit [Ping timeout: 121 seconds] |
| 06:58 | | Turaiel[Offline] is now known as Turaiel |
| 06:59 | | Turaiel is now known as Turaiel[Offline] |
| 07:27 | | catadroid` [catalyst@Nightstar-huiopn.dab.02.net] has joined #code |
| 07:29 | | Turaiel[Offline] [Brandon@Nightstar-7mqsi0.mi.comcast.net] has quit [Operation timed out] |
| 07:31 | | catadroid [catalyst@Nightstar-g4k01n.dab.02.net] has quit [Ping timeout: 121 seconds] |
| 07:39 | | Turaiel[Offline] [Brandon@Nightstar-7mqsi0.mi.comcast.net] has joined #code |
| 07:49 | < catadroid`> | McMartin: I admit I think I'll need to do more thinking about offsetof before I have actual advice |
| 07:51 | <&McMartin> | I've been wandering the Internets for insight while you slept, as it happens |
| 07:51 | <&McMartin> | My tentative conclusion is "even if the standard says this is undefined, if it doesn't work, all hell breaks loose" |
| 07:51 | < catadroid`> | There's a lot of that |
| 07:52 | < catadroid`> | Our resource reading routines rely on it being common enough across platforms |
| 07:52 | < catadroid`> | For example |
| 07:52 | < catadroid`> | I think I came to the same conclusions about alignment |
| 07:53 | <&McMartin> | Well, so, (not to interrupt, but) the fun thing is that it's in question whether it *is* undefined; the standard apparently demands that standard-layout objects be in contiguous memory, which may mean that in fact *does* insist that seeking within it as if it were a byte array must be legal |
| 07:53 | < catadroid`> | The standard doesn't mandate the tightest packing scheme, but if a compiler doesn't it isn't competitive |
| 07:53 | <&McMartin> | Ah yes |
| 07:54 | <&McMartin> | That's off in #pragma land, isn't it |
| 07:54 | < catadroid`> | Sort of |
| 07:54 | < catadroid`> | In practice you can make simple assumptions |
| 07:54 | <&McMartin> | Yeah, my weird corner is that if my case doesn't work, offsetof() is part of the standard but it's undefined behavior to ever make use of the value as anything other than a charming curiosity |
| 07:55 | < catadroid`> | Whilst true, you can always set up a couple of unit tests as canaries and assume it's standard behaviour until they ever fire |
| 07:55 | < catadroid`> | (I doubt they will) |
| 07:55 | <&McMartin> | Yep |
| 07:55 | <&McMartin> | static_assert can be readily set up for this, even |
| 07:55 | < catadroid`> | Aye |
| 07:56 | < catadroid`> | I like that keyword |
| 07:56 | < catadroid`> | It's great fun in templates |
| 07:56 | <&McMartin> | And actually, I should have unit tests for it anyway if not straight-up in the raw production code: std::is_std_layout<> in the static_asserts don't cost anything at runtime and will scream if we make our value types too fancy |
| 07:56 | < catadroid`> | Mhm :) |
| 07:56 | <&McMartin> | "This unit test is: does the software compile" |
| 07:57 | < catadroid`> | \o/ |
| 07:57 | < catadroid`> | The best kind of test |
| 07:57 | <&McMartin> | Heh heh |
| 07:57 | <&McMartin> | I've been in ObjC mode lately though |
| 07:57 | <&McMartin> | Very different headspace, has been as trong challenge to my design sense |
| 07:57 | <&McMartin> | It takes me three or four tries to get an architecture I like out of it |
| 07:57 | <&McMartin> | But each attempt is so much better than the last that it feels pretty great |
| 08:14 | < catadroid`> | :) |
| 08:17 | <&McMartin> | (Porting some of that design back into C++ is where I ran into this offsetof hilarity; it's getting around C++'s lack of reflection) |
| 08:18 | <&McMartin> | (Or, um, hm. Introspection? The line is fuzzy in ObjC) |
| 08:42 | < abudhabi> | Python compiles to bytecode, right? |
| 08:43 | | Kindamoody[zZz] is now known as Kindamoody |
| 08:44 | <~Vornicus> | Yup. |
| 08:44 | < [R]> | In most circumstances, yes |
| 09:01 | | Kindamoody is now known as Kindamoody|afk |
| 09:57 | | Netsplit Deepthought.Nightstar.Net <-> Golgafrincham.Nightstar.Net quits: abudhabi, @Derakon[AFK], @Reiver, Xon, ion, @himi, pjdelport, @ErikMesoy, @Ogredude, @McMartin, (+10 more, use /NETSPLIT to show all of them) |
| 09:58 | | Netsplit over, joins: @Ogredude, &Reiver, @JustBob, @gnolam, @Alek, @Tamber, starkruzr, &Derakon[AFK], @Thalass|bedzZzzZzz, catadroid` (+9 more) |
| 09:58 | | jeroud [sid10043@Nightstar-h2b233.irccloud.com] has joined #code |
| 09:59 | | mode/#code [+ao jeroud jeroud] by ChanServ |
| 10:09 | | Vornicus [Vorn@ServerAdministrator.Nightstar.Net] has quit [Connection reset by peer] |
| 10:51 | < pjdelport> | abudhabi: Among the common implementations, CPython and PyPy compile to Python bytecode, and Jython compiles to either Python bytecode or Java class files (IIRC). |
| 10:53 | < pjdelport> | Cython is mainly used for its C bridging features, but it can compile most pure Python code to C (using the CPython runtime and C API, so it doesn't necessarily give you a magic speedup compared to bytecode) |
| 10:54 | < pjdelport> | Then there's probably at least dozen less common implementations that do various other things. |
| 11:10 | | VirusJTG_ [VirusJTG@Nightstar-6i5vf7.sta.comporium.net] has joined #code |
| 11:10 | | VirusJTG_ [VirusJTG@Nightstar-6i5vf7.sta.comporium.net] has quit [Connection closed] |
| 11:13 | | Thalass|slep [thalass@Nightstar-283.o7s.158.104.IP] has joined #code |
| 11:19 | | Thalass|slep is now known as Thalass |
| 11:20 | | mode/#code [+o Thalass] by ChanServ |
| 11:31 | | Thalass is now known as Thalatwerk |
| 11:35 | | Thalatwerk [thalass@Nightstar-283.o7s.158.104.IP] has quit [Ping timeout: 121 seconds] |
| 12:02 | | Emmy is now known as Emmy-Werk |
| 12:10 | | Crossfire [Z@Nightstar-r9lk5l.cust.comxnet.dk] has joined #code |
| 12:10 | | mode/#code [+o Crossfire] by ChanServ |
| 12:14 | | catadroid` [catalyst@Nightstar-huiopn.dab.02.net] has quit [Ping timeout: 121 seconds] |
| 12:30 | | Turaiel[Offline] is now known as Turaiel |
| 12:31 | | Turaiel is now known as Turaiel[Offline] |
| 12:54 | | Emmy [NSkiwiirc@Nightstar-41pbej.static.chello.nl] has joined #code |
| 12:59 | < Emmy> | Behold. the rare warking Emmy |
| 13:33 | < abudhabi> | How does a "web services" thingamajig work? |
| 13:33 | < abudhabi> | (Java.) |
| 13:34 | < abudhabi> | I'm told to write one of these things, but I haven't the faintest of what "WSDL" is and how one uses it. |
| 13:40 | | Emmy [NSkiwiirc@Nightstar-41pbej.static.chello.nl] has quit [NickServ (RECOVER command used by M-E)] |
| 13:40 | | emmy [M@Nightstar-9p7hb1.direct-adsl.nl] has joined #code |
| 13:43 | | Emmy-Werk [M@Nightstar-9p7hb1.direct-adsl.nl] has quit [Ping timeout: 121 seconds] |
| 13:43 | | M-E [NSkiwiirc@Nightstar-41pbej.static.chello.nl] has joined #code |
| 13:58 | | * M-E blarghs violently at access. |
| 14:00 | < M-E> | apparently it is utterly impossible to know or use the row number of an item on a form in table mode or continuous form mode. |
| 14:08 | | catadroid [catalyst@Nightstar-gi24sc.dab.02.net] has joined #code |
| 15:14 | | mac [macdjord@Nightstar-r9vt2h.mc.videotron.ca] has quit [Ping timeout: 121 seconds] |
| 15:18 | | mac [macdjord@Nightstar-r9vt2h.mc.videotron.ca] has joined #code |
| 15:18 | | mode/#code [+o mac] by ChanServ |
| 15:47 | | Vornicus [Vorn@ServerAdministrator.Nightstar.Net] has joined #code |
| 15:47 | | mode/#code [+qo Vornicus Vornicus] by ChanServ |
| 15:56 | | M-E [NSkiwiirc@Nightstar-41pbej.static.chello.nl] has quit [[NS] Quit: http://www.kiwiirc.com/ - A hand crafted IRC client] |
| 16:58 | | crystalclaw|AFK is now known as crystalclaw |
| 17:10 | | celticminstrel [celticminst@Nightstar-q0f7bb.dsl.bell.ca] has joined #code |
| 17:10 | | mode/#code [+o celticminstrel] by ChanServ |
| 17:47 | | catalyst [catalyst@Nightstar-bt5k4h.81.in-addr.arpa] has joined #code |
| 17:48 | | catadroid [catalyst@Nightstar-gi24sc.dab.02.net] has quit [[NS] Quit: Bye] |
| 18:55 | | ion [Owner@Nightstar-6grqph.vs.shawcable.net] has quit [[NS] Quit: broken shit] |
| 18:57 | | ion [Owner@Nightstar-6grqph.vs.shawcable.net] has joined #code |
| 19:19 | | ErikMesoy [Erik@Nightstar-hq72t5.customer.cdi.no] has quit [[NS] Quit: Leaving.] |
| 19:48 | | ErikMesoy [Erik@Nightstar-hq72t5.customer.cdi.no] has joined #code |
| 19:48 | | mode/#code [+o ErikMesoy] by ChanServ |
| 21:38 | | Kindamoody|afk is now known as Kindamoody |
| 22:11 | | Reiv [NSwebIRC@Nightstar-q8avec.kinect.net.nz] has joined #code |
| 22:11 | | mode/#code [+o Reiv] by ChanServ |
| 22:32 | | Thalatwerk [thalass@Nightstar-283.o7s.158.104.IP] has joined #code |
| 23:05 | | Turaiel[Offline] is now known as Turaiel |
| 23:24 | | emmy is now known as Emmy-zZz |
| 23:41 | | himi [fow035@Nightstar-v37cpe.internode.on.net] has quit [Ping timeout: 121 seconds] |
| 23:49 | | Kindamoody is now known as Kindamoody[zZz] |
| 23:51 | | Thalatwerk is now known as Thalass |
| 23:51 | | mode/#code [+o Thalass] by ChanServ |
| --- Log closed Wed May 11 00:00:48 2016 |