!FBuJyWXTGcGtHTPphC:nixos.org

Nix Rust

671 Members
Rust154 Servers

Load older messages


SenderMessageTime
1 Aug 2025
@emilazy:matrix.orgemily(this is the entire link the reddit post goes to)12:06:28
@dramforever:matrix.orgdramforeverit's literally not the same instruction encodings12:07:41
@emilazy:matrix.orgemilyRISC-V feels like the kind of architecture I would have produced as a teenager who was monomaniacally fixated on orthogonality and minimality in all things12:07:57
@emilazy:matrix.orgemilyI guess this comparison is probably not using uhh RV32 or whatever though12:08:13
@emilazy:matrix.orgemilyor whatever you're meant to be using now. the real architecture12:08:20
@k900:0upti.meK900It's literally like that12:08:26
@emilazy:matrix.orgemilyyeah ok the reddit comments say as much12:08:34
@k900:0upti.meK900Their response to "why" is "just do more uop fusion lol"12:08:37
@emilazy:matrix.orgemilywell, no12:09:02
@emilazy:matrix.orgemilyhttps://godbolt.org/z/19jnvda7G12:09:03
@emilazy:matrix.orgemilyit looks like rva23u64 beats MIPS there12:09:20
@emilazy:matrix.orgemilyso it's just a dumb comparison12:09:27
@k900:0upti.meK900Well yeah I don't mean this specific case12:09:37
@k900:0upti.meK900I just mean in general12:09:39
@emilazy:matrix.orgemilyalthough the fragmentation of the RISC-V instruction sets in general is its own kind of dumb I suppose12:09:39
@emilazy:matrix.orgemilybut also the thing teenage me would have done12:09:45
@k900:0upti.meK900There's like12:09:46
@k900:0upti.meK900A List12:09:47
@k900:0upti.meK900Of like Officially Mandated Fusions You Must Do To Fast12:10:01
@k900:0upti.meK900That compilers also know12:10:07
@emilazy:matrix.orgemilyI think x86 also involves known uop fusions that compilers take into account12:10:08
@emilazy:matrix.orgemilyit's arguably better to write it down than not12:10:15
@k900:0upti.meK900Not really12:10:17
@k900:0upti.meK900(for x86)12:10:23
@k900:0upti.meK900x86 mostly just has scheduling fuckery because it's very wide12:10:37
@k900:0upti.meK900But then modern aarch64 is also very wide12:10:46
@k900:0upti.meK900So it's likely actually fast RISC-V will also be very wide12:10:55
@k900:0upti.meK900So you'd have to consider both scheduling AND uop fusion and this is when you get the real funni12:11:12
@emilazy:matrix.orgemilymy opinion as someone with no ISA design experience and no right to be making any judgement is that it feels like nobody has actually made anything better than AArch64 so far12:11:17
@dramforever:matrix.orgdramforeverx86 and arm64 absolutely fuse12:11:24

Show newer messages


Back to Room ListRoom Version: 6