| 31 Oct 2024 |
 K900 | That's really not the kind of thing that's going to help | 16:13:21 |
| K900 | The interpreter isn't slow because it's complex | 16:13:38 |
| K900 | It's slow because it was not written to be fast | 16:13:44 |
| K900 | If that makes any sense | 16:13:47 |
 .. | 😕 | 16:14:14 |
 @aloisw:kde.org | In reply to @k900:0upti.me
The interpreter isn't slow because it's complex I think they might mean "complexity" as in "algorithmic complexity". | 16:14:41 |
| K900 | Yes, I get that | 16:15:23 |
| .. | I'm looking for contributing some open-source projects to make my CV | 16:15:25 |
| K900 | And it's not going to help the interpeter | 16:15:27 |
| K900 | Because the interpreter is just ... not very designed to be fast | 16:15:40 |
| .. | K900: anyways, thanks for sharing your experience with us. | 16:17:43 |
 connor (burnt/out) (UTC-8) | If you want to do open source work and have a background in numerical analysis, check out https://herbie.uwplse.org/ | 16:19:04 |
| .. | not bad, thanks a lot | 16:20:38 |
| .. | connor (he/him) (UTC-7): I have just sent an email to them. many thanks. | 16:27:16 |
| .. | it seems they are not active on the project. their email simply doesn't work to get inquiries.
i have sent an email to tensorflow probability library to see how it goes. | 16:51:25 |
|  Adda joined the room. | 17:17:45 |
| connor (burnt/out) (UTC-8) | Added GC numbers to https://github.com/NixOS/nix/pull/11767; with the change it allocates more memory than before (which is odd, given that the implementation we currently have just mallocs and memcpy's like wild -- I'd expect at least some decrease, though maybe that's because I haven't figured out how to work transient yet). | 17:56:45 |
|  Philipp Jungkamp left the room. | 22:54:54 |
| Philipp Jungkamp joined the room. | 22:55:15 |
| 1 Nov 2024 |
 Mic92 | Bumped all nix versions and added some automation for updating nix: https://github.com/NixOS/nixpkgs/pull/352834 | 10:53:28 |
|  Ryan Yin joined the room. | 14:21:04 |
|  Samuel set a profile picture. | 21:25:16 |
| Samuel changed their display name from Samuel to Samuel (DrRed). | 21:41:39 |
| Samuel changed their display name from Samuel (DrRed) to Samuel. | 21:42:31 |
| 2 Nov 2024 |
|  error_pro joined the room. | 05:55:57 |
|  @icarus_dh:matrix.org joined the room. | 14:07:09 |
| connor (burnt/out) (UTC-8) | Does anyone have any recommended reading on ways to implement call-by-name (call-by-need?) in CPP? | 20:51:34 |
| connor (burnt/out) (UTC-8) | TBH getting a little frustrated with https://github.com/NixOS/nix/pull/11767 and concerned that to improve performance with data structures that have structural sharing we’ll need to move more data structures over to permit sharing (instead of re-allocating when interacting with STL containers) | 21:00:53 |
 emily | FWIW, even in Haskell it's sometimes faster to use the contiguous, copy-happy vector than fancier structures just because of constant factors – not to discourage you, just saying that I'm not sure the performance trade-off is going to unambiguously be in favour of the elegant functional data structures | 21:05:02 |
| emily | memcpy is a very cheap operation these days | 21:05:22 |
