Logs: liberachat/#haskell
| 2021-07-18 19:32:11 | <tom__> | aha thanks |
| 2021-07-18 19:32:47 | <tom__> | by purely do you mean - not with IO? |
| 2021-07-18 19:32:48 | <davean> | Note there are plenty of monadic forms that aren't IO also, for prfs |
| 2021-07-18 19:33:22 | <geekosaur> | yes |
| 2021-07-18 19:33:27 | × | Atum_ quits (~IRC@user/atum/x-2392232) (Ping timeout: 250 seconds) |
| 2021-07-18 19:33:52 | <davean> | I'm not really a fan of the 1.2 changes :/ |
| 2021-07-18 19:34:21 | <davean> | tom__: if you want good sampling, see random-fu and such BTW |
| 2021-07-18 19:34:31 | <davean> | distributions is seperate from random numbers |
| 2021-07-18 19:34:57 | <geekosaur> | on the one hand it cleaned up my use of the mersenne twister a bit; on the other I found 1.1 much easier |
| 2021-07-18 19:35:15 | × | Sgeo quits (~Sgeo@user/sgeo) (Quit: Leaving) |
| 2021-07-18 19:35:17 | <geekosaur> | gripping hand I didn't have to fake normal distributions any more |
| 2021-07-18 19:35:31 | <davean> | geekosaur: I don't think the 1.2 design is valid :/ I think theres some actual contradictions in its design, but thats far too long a discussion for here and now |
| 2021-07-18 19:36:13 | <davean> | I'm actually rather upset about 1.2 |
| 2021-07-18 19:36:27 | <davean> | Not that 1.1 was great |
| 2021-07-18 19:36:34 | × | mdunnio quits (~textual@205.178.28.97) (Quit: Textual IRC Client: www.textualapp.com) |
| 2021-07-18 19:37:25 | → | wroathe joins (~wroathe@c-68-54-25-135.hsd1.mn.comcast.net) |
| 2021-07-18 19:37:50 | → | mdunnio joins (~textual@205.178.28.97) |
| 2021-07-18 19:38:13 | → | Sgeo joins (~Sgeo@user/sgeo) |
| 2021-07-18 19:38:50 | <tom__> | I'm using MWC-probability |
| 2021-07-18 19:39:05 | → | Erutuon joins (~Erutuon@user/erutuon) |
| 2021-07-18 19:39:53 | <mdunnio> | hello! is anyone aware of any up-to-date guides on how to use LLVM backend with GHC? I'm looking to utilize GHC 8.10.5's first class support for M1, but I'm not exactly sure where to get started. |
| 2021-07-18 19:40:14 | → | Atum_ joins (~IRC@user/atum/x-2392232) |
| 2021-07-18 19:41:18 | <geekosaur> | using it is just doing "ghc-options: -fllvm" in the cabal file. but I think 8.10.5 is known buggy |
| 2021-07-18 19:41:44 | <mdunnio> | oh interesting. ok |
| 2021-07-18 19:42:12 | <mdunnio> | I'll try that out |
| 2021-07-18 19:42:24 | × | wroathe quits (~wroathe@c-68-54-25-135.hsd1.mn.comcast.net) (Ping timeout: 258 seconds) |
| 2021-07-18 19:42:26 | <geekosaur> | and true first class (-fasm) support is supposed to come in 9.2.1 |
| 2021-07-18 19:42:45 | × | falafel quits (~falafel@pool-96-255-70-50.washdc.fios.verizon.net) (Ping timeout: 255 seconds) |
| 2021-07-18 19:42:55 | <geekosaur> | there's a prerelease available in ghcup but it has some known bugs |
| 2021-07-18 19:43:22 | <davean> | Yah, on one hand llvm is kinda first class, on the other, LLVM isn't a general system and we don't fit its preconception of being C |
| 2021-07-18 19:43:30 | <davean> | So it'll never be quite right to use LLVM |
| 2021-07-18 19:43:48 | <davean> | You'll always eat a penalty of LLVM not being a general system |
| 2021-07-18 19:44:22 | × | Erutuon quits (~Erutuon@user/erutuon) (Ping timeout: 265 seconds) |
| 2021-07-18 19:44:33 | × | chris_ quits (~chris@81.96.113.213) (Remote host closed the connection) |
| 2021-07-18 19:44:42 | <mdunnio> | penalty as in performance, or just support generally? |
| 2021-07-18 19:44:56 | <geekosaur> | both |
| 2021-07-18 19:44:57 | <davean> | More the former, but there CAN be a few issues with the later |
| 2021-07-18 19:45:07 | <davean> | LLVM is very much a C IR |
| 2021-07-18 19:45:20 | <davean> | It claims to be general, but frankly, thats wishful thinking |
| 2021-07-18 19:45:22 | <geekosaur> | llvm doesn't understand the cps-encoded code ghc emits and pessimizes it |
| 2021-07-18 19:45:26 | <davean> | it has a ton of assumptions baked in at hte bottom |
| 2021-07-18 19:45:37 | → | wroathe joins (~wroathe@c-68-54-25-135.hsd1.mn.comcast.net) |
| 2021-07-18 19:45:43 | <Rembane> | So every language that looks like C if you squint works for LLVM? |
| 2021-07-18 19:45:57 | <davean> | Rembane: Close, and LLVM works more generally than that, just not well |
| 2021-07-18 19:46:10 | <davean> | LLVM can't think some normal thoughts, and never will be able to |
| 2021-07-18 19:46:13 | <mdunnio> | I see |
| 2021-07-18 19:46:25 | <Rembane> | davean: Got it. Doesn't sound too bad if your needs are C et al |
| 2021-07-18 19:46:26 | <davean> | Theres ways to make them "work" because the turing tarpit, but that doesn't mean they work well |
| 2021-07-18 19:46:37 | <davean> | Rembane: Yah, its actually great for that |
| 2021-07-18 19:46:48 | <davean> | Rembane: I mean you make sometihng better for one thing, usually you make it worse for another |
| 2021-07-18 19:46:58 | <davean> | the only issue I see here is the claim that its general, which it just clearly isn't |
| 2021-07-18 19:47:16 | <davean> | but! It works pretty well for GHC for most thing |
| 2021-07-18 19:47:22 | <davean> | so, go use it |
| 2021-07-18 19:47:31 | <davean> | Its actually the best option for a few specific cases |
| 2021-07-18 19:47:38 | <davean> | and you can do funky stuff to use it only where it is the best ... |
| 2021-07-18 19:48:05 | <davean> | So like, its good to have, I'm not trying to bash it, just explain what one might feel about "first class ness" |
| 2021-07-18 19:48:32 | × | shailangsa quits (~shailangs@host86-186-196-229.range86-186.btcentralplus.com) (Ping timeout: 258 seconds) |
| 2021-07-18 19:48:36 | <mdunnio> | yeah, to be fair, I have a pretty broad interpretation of what "first class support" even means |
| 2021-07-18 19:48:58 | <Rembane> | davean: Totally, it's trade offs all the way down. |
| 2021-07-18 19:49:07 | <mdunnio> | at this point, I'm willing to take "does it work at all?" |
| 2021-07-18 19:49:14 | <davean> | mdunnio: ok! |
| 2021-07-18 19:49:16 | <davean> | so, it does work! |
| 2021-07-18 19:49:32 | × | wroathe quits (~wroathe@c-68-54-25-135.hsd1.mn.comcast.net) (Client Quit) |
| 2021-07-18 19:50:14 | <davean> | mdunnio: So my understanding is that the nix stuff is still the best option |
| 2021-07-18 19:50:39 | → | niemand joins (~niemand@p2e52f785.dip0.t-ipconnect.de) |
| 2021-07-18 19:50:43 | ← | niemand parts (~niemand@p2e52f785.dip0.t-ipconnect.de) () |
| 2021-07-18 19:50:57 | × | econo quits (uid147250@user/econo) (Quit: Connection closed for inactivity) |
| 2021-07-18 19:51:02 | <davean> | (I uh, keep an arms length from Apple, I just uh, make sure we have CI for it) |
| 2021-07-18 19:51:56 | <davean> | Talking of which, yah rossetta, uh, don't run a ton of x86 binaries. |
| 2021-07-18 19:53:05 | <mdunnio> | fair |
| 2021-07-18 19:53:10 | <mdunnio> | yeah, that's the plan |
| 2021-07-18 19:53:38 | <davean> | yah, its a problem (I don't know if you know, but it doesn't clear the cache, so you'll run out of disk space and be stuck unless you disable SIP) |
| 2021-07-18 19:53:46 | <mdunnio> | I haven't actually worked much with nix yet, might be time to experiment |
| 2021-07-18 19:54:01 | × | Maxdamantus quits (~Maxdamant@user/maxdamantus) (Ping timeout: 268 seconds) |
| 2021-07-18 19:54:19 | <davean> | mdunnio: So, I'm not sure they want me to refer you, but the bigest expert I know in this area is moritz angermann |
| 2021-07-18 19:54:31 | <davean> | I think we're due for a writeup on the state of M1 stuff |
| 2021-07-18 19:54:42 | <davean> | It would be a public service |
| 2021-07-18 19:55:26 | <mdunnio> | I recall moritz wrote up a bunch about cross-compiling for ARM. I did some rpi work back a year or so ago |
| 2021-07-18 19:56:01 | <davean> | I'll admit I was taking care of a lot of the mac stuff and then moritz stepped in and not being a mac user I disowned it as fast as possible |
| 2021-07-18 19:56:23 | <mdunnio> | seems like the appropriate time to pass the torch |
| 2021-07-18 19:56:51 | <mdunnio> | ok, I'm going to experiment a little bit and see if I can get somewhere. I appreciate the help |
| 2021-07-18 19:56:53 | × | P1RATEZ quits (piratez@user/p1ratez) (Remote host closed the connection) |
| 2021-07-18 19:57:02 | <davean> | Someone needed to do it, but better it be someone who loves it and has base skills for it |
| 2021-07-18 19:57:10 | × | mdunnio quits (~textual@205.178.28.97) (Quit: My MacBook has gone to sleep. ZZZzzz…) |
| 2021-07-18 19:58:41 | → | myShoggoth joins (~myShoggot@97-120-70-214.ptld.qwest.net) |
| 2021-07-18 19:58:44 | → | yauhsien_ joins (~yauhsien@61-231-35-149.dynamic-ip.hinet.net) |
| 2021-07-18 19:59:01 | <davean> | Hum, I failed to full say what I should have above - which is that the CI config is where that knowlege should be accessible |
| 2021-07-18 19:59:04 | <davean> | well I'm a bad helper |
| 2021-07-18 19:59:34 | × | yauhsien quits (~yauhsien@61-231-25-245.dynamic-ip.hinet.net) (Ping timeout: 268 seconds) |
| 2021-07-18 20:00:56 | × | peterhil quits (~peterhil@dsl-hkibng32-54f849-252.dhcp.inet.fi) (Ping timeout: 255 seconds) |
| 2021-07-18 20:05:07 | × | juhp quits (~juhp@128.106.188.66) (Ping timeout: 268 seconds) |
| 2021-07-18 20:05:31 | × | Lycurgus quits (~juan@cpe-45-46-140-49.buffalo.res.rr.com) (Quit: Exeunt) |
| 2021-07-18 20:06:54 | × | zmt00 quits (~zmt00@user/zmt00) (Quit: Gone.) |
| 2021-07-18 20:07:20 | → | juhp joins (~juhp@128.106.188.66) |
| 2021-07-18 20:09:14 | × | drd quits (~drd@93-39-151-19.ip76.fastwebnet.it) (Ping timeout: 258 seconds) |
| 2021-07-18 20:10:19 | × | eggplantade quits (~Eggplanta@108-201-191-115.lightspeed.sntcca.sbcglobal.net) (Remote host closed the connection) |
| 2021-07-18 20:11:25 | × | _ht quits (~quassel@82-169-194-8.biz.kpn.net) (Remote host closed the connection) |
| 2021-07-18 20:19:32 | → | shailangsa joins (~shailangs@host86-145-14-107.range86-145.btcentralplus.com) |
| 2021-07-18 20:19:56 | × | Atum_ quits (~IRC@user/atum/x-2392232) (Quit: Atum_) |
| 2021-07-18 20:19:59 | → | drd joins (~drd@93-39-151-19.ip76.fastwebnet.it) |
All times are in UTC.