| 29 Mar 2024 |
 Alex | tau you're in luck because I was planning on writing some instructions in the shared notes (see room description) this evening.
I'll let you know when the instructions are ready. | 18:10:19 |
 hive ⬡ | oh awesome | 18:20:00 |
| Alex | tau Steven Keuchel fgaz the shared notes now have instructions for how to get a working GHC.
(Disclaimer: It's been a while since I first booted GHC and I haven't yet had time to properly test the instructions.) | 19:39:30 |
| hive ⬡ | it's been building for like | 23:58:16 |
| hive ⬡ | since you sent that message | 23:58:22 |
| hive ⬡ | and it only got to ghc like, 30 minutes ago | 23:58:34 |
| hive ⬡ | and this is on my like build server lmao | 23:58:49 |
| hive ⬡ | i'm going to be pushing this to cachix so other people don't have to build it lol | 23:59:17 |
| Alex | I couldn't find a way of only patching cc-wrapper and binutils-wrapper for GHC, so you had to rebuild the entire stdenv. | 23:59:26 |
| hive ⬡ | womp | 23:59:38 |
| 30 Mar 2024 |
| Alex | Wait until you see how long native GHC takes to build on RISC-V hardware :)
(I doubt you'll be able to build much beyond GHC itself with the cross-compiled GHC.) | 00:01:37 |
| hive ⬡ | i don't intend to be doing much building on my riscv dev board lol | 00:02:22 |
| Alex | If you just want to cross-compile Haskell packages for RISC-V, you don't actually need my changes; Nixpkgs can already do build = host != target. | 00:04:31 |
| hive ⬡ | couldn't get that working | 00:04:48 |
| Alex | I'm afraid I can't help there, because I've spent all my time on getting working native compilation and no time on cross-compiling haskellPackages :D | 00:05:32 |
| hive ⬡ | ah ok | 00:05:42 |
| Alex | Btw I suspect your build server isn't particularly fast... My Ryzen 5 5600X desktop finished the GHC build over 90 minutes ago (I can't remember when I started it, but it was definitely not more than an hour before I sent that message). | 00:09:50 |
| hive ⬡ | including all of stdenv ? | 00:10:35 |
| Alex | Yes | 00:10:41 |
| hive ⬡ | wowee | 00:10:45 |
|  @jackleightcap:matrix.org joined the room. | 00:11:01 |
| Alex | Just a few months ago I noticed that my system was using the powersave CPU frequency scaling driver instead of amd-pstate. Changing that made things significantly faster. If you're on an AMD CPU, maybe you have the same problem? | 00:15:17 |
| hive ⬡ | we do have the server set to powersave mode, so it's probably something similar, but i don't think we could afford to run it without it | 00:15:53 |
| hive ⬡ | oh well | 00:16:01 |
| Alex | IIRC powersave means it will always use the min frequency | 00:16:35 |
| hive ⬡ | ye | 00:16:39 |
| hive ⬡ | tbh after thinking it over, it's probably costing more to have the server running this long, so i will probably turn power saver off after this | 00:17:47 |
| Alex | On the other hand lower cpufreq should be more power efficient (due to lower core temperatures).
I guess if you wanted to maximise power efficiency the only way of finding the sweet spot is a lot of benchmarking. | 00:20:04 |
| hive ⬡ | yeah | 00:26:32 |
|  @shalokshalom:kde.org left the room. | 00:27:37 |
