Linux Kernel Performance Bottlenecks Spotted By Mold Developer

[rjzak]

If the Mold developer was involved with LLVM’s linker, why not have the performance improvements in LLVM instead of a new project?

[caligula]

lmao, dude is so good at making fast linkers, the kernel is now holding him down
this guy is super talented, i hope he gets better support

Like the total opposite of ESR who started whining when his shiny workstation class computer wasn't fast enough for that shitty python toy code.

[stormcrow]

If the Mold developer was involved with LLVM’s linker, why not have the performance improvements in LLVM instead of a new project?

Because the goal of MOLD is performance at all costs. That means you make sacrifices elsewhere others may not be willing to make, including instruction code quality, size, potentially even sacrifice security considerations.

[davidbepo]

Like the total opposite of ESR who started whining when his shiny workstation class computer wasn't fast enough for that shitty python toy code.

whos that?

[DumbFsck]

whos that?

Eric Raymond. A writer who had his writings inspiring much of the FOSS movement, especially at the beginning.


Sometimes he pretends to be a dev or to know literally anything about code... But he isn't afaict. EDIT: apoarebtly he was in fact a professional developer in the 80's. I don't know why but my (recent) memory is only of people saying he isn't a programmer let alone a good one, so take from that what you will - I'm no more familiar with his work than someone who only read cathedral and bazaar a couple of decades ago.


I also don't know about his struggles with workstation performance

caligula do you mind clueing us in?

[Volta]

Bottlenecks compared to what?

[coder]

Eric Raymond. A writer who had his writings inspiring much of the FOSS movement, especially at the beginning.

To hear him tell it, you'd think so. He was more of a self-appointed spokesman for FOSS than a real leader, IMO. The FOSS movement he described was more how he wished to see it, than a faithful reporting of it as it really was.

Back in the day, someone published web comic featuring leading personalities from the FOSS movement and tech industry, called Everybody Loves Eric Raymond (which was itself a parody of the TV series Everybody Loves Raymond, which I should disclose that I've never seen).

https://geekz.co.uk/lovesraymond/

TL;DR: a grizzly, old libertarian blow-hard. Yes, he did actually write some code at one point in time.

[braiam]

If only there was a way to figure out what the Mold dev was doing:

ext4 on a PCIe Gen.5 SSD, but I guess it probably doesn't matter much
because we observed similar results even on tmpfs (~1.75s vs. 2.45s
when linking clang).

But that won't matter, because he was using fallocate anyways, and still happens. So, something is inherent to freshly created files that makes them slower to write.

[Mark Rose]

For all those asking what filesystem he says both ext4 and tmpfs.

[rmfx]

It's nice to see these types of feedbacks.
For security, it's common to see loads of entire teams spending time on finding issues, but for performance, it's more up to the kernel and driver devs themselves to find the problems' source.
That could be nice to see perf bottleneck bounties more often.