Newest at the top
| 2025-05-08 22:44:02 +0200 | Frostillicus | (~Frostilli@pool-71-174-119-56.bstnma.fios.verizon.net) |
| 2025-05-08 22:35:49 +0200 | Frostillicus | (~Frostilli@pool-71-174-119-56.bstnma.fios.verizon.net) (Ping timeout: 245 seconds) |
| 2025-05-08 22:28:57 +0200 | dofsyl^ | (~dofsyl@50.168.231.214) |
| 2025-05-08 22:28:22 +0200 | Frostillicus | (~Frostilli@pool-71-174-119-56.bstnma.fios.verizon.net) |
| 2025-05-08 22:27:59 +0200 | dofsyl^ | (~dofsyl@50.168.231.214) (Ping timeout: 244 seconds) |
| 2025-05-08 22:27:57 +0200 | Frostillicus | (~Frostilli@pool-71-174-119-56.bstnma.fios.verizon.net) (Remote host closed the connection) |
| 2025-05-08 22:27:54 +0200 | peterbecich | (~Thunderbi@syn-047-229-123-186.res.spectrum.com) peterbecich |
| 2025-05-08 22:22:30 +0200 | dofsyl^ | (~dofsyl@50.168.231.214) |
| 2025-05-08 22:13:49 +0200 | ttybitnik | (~ttybitnik@user/wolper) ttybitnik |
| 2025-05-08 22:13:05 +0200 | takuan | (~takuan@d8D86B601.access.telenet.be) (Remote host closed the connection) |
| 2025-05-08 22:09:15 +0200 | Frostillicus | (~Frostilli@pool-71-174-119-56.bstnma.fios.verizon.net) |
| 2025-05-08 22:06:45 +0200 | Tuplanolla | (~Tuplanoll@91-159-69-59.elisa-laajakaista.fi) Tuplanolla |
| 2025-05-08 22:00:47 +0200 | rvalue- | rvalue |
| 2025-05-08 21:58:38 +0200 | <EvanR> | hell yeah |
| 2025-05-08 21:55:00 +0200 | <EvanR> | wow |
| 2025-05-08 21:54:55 +0200 | rvalue | (~rvalue@user/rvalue) (Ping timeout: 272 seconds) |
| 2025-05-08 21:53:53 +0200 | <[exa]> | I knew a guy who did type inference in the C code as the preprocessing step, can't find the thesis now tho |
| 2025-05-08 21:53:51 +0200 | rvalue- | (~rvalue@user/rvalue) rvalue |
| 2025-05-08 21:52:24 +0200 | <[exa]> | EvanR: for gcc, see the note below here https://gcc.gnu.org/onlinedocs/gcc/Preprocessor-Options.html |
| 2025-05-08 21:51:17 +0200 | acidjnk | (~acidjnk@p200300d6e71c4f49f4d0e34d1c0f8686.dip0.t-ipconnect.de) (Ping timeout: 248 seconds) |
| 2025-05-08 21:51:04 +0200 | acidjnk_new | (~acidjnk@p200300d6e71c4f4911c235e3ddcab2f1.dip0.t-ipconnect.de) acidjnk |
| 2025-05-08 21:46:42 +0200 | <tomsmeding> | https://hackage.haskell.org/package/record-dot-preprocessor ctrl-F for "use this magic" |
| 2025-05-08 21:46:30 +0200 | <EvanR> | cool |
| 2025-05-08 21:46:14 +0200 | <tomsmeding> | ghc can be |
| 2025-05-08 21:44:49 +0200 | <EvanR> | can gcc be told to use a different preprocessor |
| 2025-05-08 21:44:46 +0200 | manwithluck | (~manwithlu@2a09:bac1:5b80:20::49:f6) manwithluck |
| 2025-05-08 21:44:23 +0200 | wootehfoot | (~wootehfoo@user/wootehfoot) (Read error: Connection reset by peer) |
| 2025-05-08 21:44:10 +0200 | <tomsmeding> | so if I confuse them, GHC yells at me |
| 2025-05-08 21:44:10 +0200 | <EvanR> | the blog post related to datatype99 "What's the point of the C preprocessor, actually?" is kind of inspiring. And makes me think, what if you had a different preprocessor entirely |
| 2025-05-08 21:44:03 +0200 | <tomsmeding> | and reverse-mode in a highly typed way where adjoints (cotangents) and primal values have _distinct types_ |
| 2025-05-08 21:43:43 +0200 | <tomsmeding> | I write two kinds of AD code: dual-numbers forward mode where I just use "x" and "dx", and the code is simple enough anyway that it's all clear |
| 2025-05-08 21:43:21 +0200 | <tomsmeding> | ah, I think I see why I don't have problems with this |
| 2025-05-08 21:42:56 +0200 | <Athas> | No, it's a bigger problem for forward mode (where we use a _tan suffix). |
| 2025-05-08 21:42:48 +0200 | <tomsmeding> | with conventional meaning of "'", naming an adjoint of x as x' doesn't even make much sense |
| 2025-05-08 21:42:31 +0200 | <tomsmeding> | and furthermore, is an adjoint really a derivative of a thing? |
| 2025-05-08 21:42:15 +0200 | <tomsmeding> | the ' notation for derivatives is not helpful in the context of AD anyway, because it carries much too little information about with respect to what you're actually differentiating |
| 2025-05-08 21:41:42 +0200 | <tomsmeding> | funny, I've never even thought about this being a problem |
| 2025-05-08 21:41:10 +0200 | <Athas> | https://github.com/diku-dk/futhark/blob/master/src/Futhark/AD/Rev.hs#L5-L8 |
| 2025-05-08 21:41:03 +0200 | <tomsmeding> | I see -- yes, "dx" is what I use |
| 2025-05-08 21:40:33 +0200 | <Athas> | In Haskell it is common to use the name x' for some variable derived from x, but that doesn't mean x' is the mathematical derivative of x, but when implementing AD you _do_ need a naming convention for that. |
| 2025-05-08 21:40:03 +0200 | <tomsmeding> | just make it dx instead |
| 2025-05-08 21:39:14 +0200 | <sprout> | derivative naming? |
| 2025-05-08 21:38:17 +0200 | <tomsmeding> | how do variable naming conventions make implementing AD harder? |
| 2025-05-08 21:37:44 +0200 | tomsmeding | doesn't get |
| 2025-05-08 21:37:20 +0200 | <Athas> | tomsmeding: the most difficult part of implementing AD in Haskell is that the usual x' convention becomes confusing. |
| 2025-05-08 21:36:59 +0200 | <Athas> | Not sure. It may be something as trivial as most programmers don't like writing documentation, so they don't. |
| 2025-05-08 21:36:53 +0200 | <tomsmeding> | I have to prevent myself so often from writing "we can derive xyz equations" when talking about AD |
| 2025-05-08 21:35:57 +0200 | <[exa]> | ("differentiated" not as in AD now! :D ) |
| 2025-05-08 21:35:26 +0200 | <[exa]> | Athas: well the question then is what happened that differentiated these |
| 2025-05-08 21:31:22 +0200 | ljdarj1 | ljdarj |