| 1 Jun 2026 |
 llakala | working on a speedup to lib.platforms generation using groupBy, and had the idea to put the most frequent systems first in the individual lists | 17:11:46 |
| llakala | so the elem lookups could find the correct system earlire | 17:14:16 |
| llakala | * so the elem lookups could find the correct system earlier | 17:14:19 |
 Sergei Zimmerman (xokdvium) | In reply to @llakala:matrix.org
judging by my quick look at the impl in noogle, I assume groupBy doesn't preserve order (where the values in the individual groups would be in the order of the original list) Why do you say that it doesn’t preserve order? | 17:15:41 |
| llakala | was just looking at the alreadySorted part | 17:15:57 |
| llakala | but apparently i'm wrong, since i just checked and order seems to be preserved | 17:16:08 |
| llakala | meaning we can take
[
"aarch64-linux"
"arc-linux"
"armv5tel-linux"
"armv6l-linux"
"armv7a-linux"
"armv7l-linux"
"i686-linux"
"loongarch64-linux"
"m68k-linux"
"sh4-linux"
"microblaze-linux"
"microblazeel-linux"
"mips-linux"
"mips64-linux"
"mips64el-linux"
"mipsel-linux"
"powerpc-linux"
"powerpc64-linux"
"powerpc64le-linux"
"riscv32-linux"
"riscv64-linux"
"s390-linux"
"s390x-linux"
"x86_64-linux"
]
and turn it into
[
"x86_64-linux"
"aarch64-linux"
"arc-linux"
"armv5tel-linux"
"armv6l-linux"
"armv7a-linux"
"armv7l-linux"
"i686-linux"
"loongarch64-linux"
"m68k-linux"
"sh4-linux"
"microblaze-linux"
"microblazeel-linux"
"mips-linux"
"mips64-linux"
"mips64el-linux"
"mipsel-linux"
"powerpc-linux"
"powerpc64-linux"
"powerpc64le-linux"
"riscv32-linux"
"riscv64-linux"
"s390-linux"
"s390x-linux"
]
| 17:16:32 |
| Sergei Zimmerman (xokdvium) | In reply to @llakala:matrix.org
was just looking at the alreadySorted part That’s the regular stuff to avoid sorting stuff that’s already sorted internally. | 17:17:43 |
| Sergei Zimmerman (xokdvium) | (And by sorting I mean the internal symbol table order which isn’t totally unobservable, but it’s not about the list orderings) | 17:18:23 |
| llakala | interesting, neat to know | 17:18:41 |
|  mikabo joined the room. | 17:54:52 |
| 2 Jun 2026 |
|  @m0jibake:matrix.org left the room. | 01:39:59 |
|  halloibims joined the room. | 05:23:36 |
|  Max joined the room. | 07:38:11 |
|  Sr. Galán joined the room. | 09:17:03 |
 toonn | The documentation on Floats is rather sparse. Is the following correct? Nix Floats are double-precision floating point numbers, excluding -0, subnormals, NaN and +/-∞. | 14:25:16 |
| toonn | And only decimal literals are available? | 14:26:21 |
| Sergei Zimmerman (xokdvium) | In reply to @toonn:matrix.org
The documentation on Floats is rather sparse. Is the following correct? Nix Floats are double-precision floating point numbers, excluding -0, subnormals, NaN and +/-∞. Internally everything is just a double. String formatting uses like 4 different formats and is totally busted. You can produce +-0, canonical NaN or +-inf with arithmetic | 14:40:08 |
| Sergei Zimmerman (xokdvium) | Subnormals is kinda meh, depends on the c++ stdlib when it comes to parsing. I seem to recall there were issues with TOML and libc++ | 14:40:56 |
| Sergei Zimmerman (xokdvium) | Basically it’s a flaming pile of nonreproducibility | 14:42:14 |
| Sergei Zimmerman (xokdvium) | And yeah everything by related to NaNs is platform dependent | 14:43:43 |
| Sergei Zimmerman (xokdvium) | * And yeah everything related to NaNs is platform dependent | 14:43:47 |
| Sergei Zimmerman (xokdvium) | Consult the closest ISA manual /s | 14:44:10 |
 piegames | If I could I'd rip out floats entirely, and then reintroduce them with a clean implementation later maybe | 14:44:24 |
| piegames | Maybe I will, but priorities | 14:44:30 |
| Sergei Zimmerman (xokdvium) | I was hoping to just plonk down softfloat and be done with it, native floats is garbage and compilers have a tendency to miscompile it still | 14:45:00 |
| toonn | Oh, * works yeah, couldn't make it happen with +. -1.0 * 0.0 -> -0, 1.e308 * 1.e308 -> inf (negatable) and 1.e308 * 1.e308 * 0.0 -> NaN (negatable). | 14:50:11 |
| toonn | Thank you, Sergei! | 14:50:36 |
| toonn | And of course the only one of those that roundtrips through to/fromJSON is -0.0, even though that's not a legal literal >.< | 14:55:35 |
|  Nuwa changed their profile picture. | 15:43:33 |
