Magik6k

Magik6k

@magik.net

434 followers648 following155 posts

Another one of small 15 minute hacks to improve my workflow now that local multi-modal AI is good enough. github.com/magik6k/scro... Trigger a hotkey, select screenshot area, ask about a thing, fully private and super snappy

So easy to fix things with the recent wave of agentic coding tools like OpenCode. I was debugging a fairly complex issue involving 10s of goroutine trees and over multiple hours Previously this would be a few jank bash scripts, now it's a nice tool I could opensource - github.com/magik6k/gscr...

github.comGitHub - magik6k/gscrape: A toolkit for scraping, indexing, and analyzing Go pprof goroutine dumps over time across services.A toolkit for scraping, indexing, and analyzing Go pprof goroutine dumps over time across services. - magik6k/gscrape

The brew guide from GLITCH comes with a grind size reference in a small bag (left). Seemingly wants me to set the grind setting on my Niche Zero to “a different grinder”

Maybe instead of fighting with RDMA and the annoyance that is closed source nvidia drivers on Arch, I can just pipe the 4 Display Port data lanes straight into a qsfp28 port? Both have 4 differential pairs, with 100ohm impedance, with roooughly similar signal levels, so maybe just maybe 🤔

Flashing random firmware and casually passing the "yeah don't worry the hardware id is different" flag will never feel nice

Magik6k@magik.net

Current boss: Convincing libfabric to see my NICs on Arch Linux

Let’s see how smooth setting up RDMA between two non-server AMD machines goes

Magik6k@magik.net

NVIDIA DIGITS couldn't be announced at a better time. I have a dark fiber that I could use for remote display with Looking Glass over RDMA with lgproxy (4k 120hz hdr, ~60Gbps), and this cute little thing solves all of my client side problems with its hopefully 2x 100Gbps NICs

View full thread

I have borderline zero idea what all those random names mean/do, but the thing compiled now. No idea if it works tho. Pretty fun to enter a new technology and feel absolutely clueless. Next step is to get LookingGlass setup running, shouldn't take long, but I'm also tired enough today.

Magik6k reposted

Replying to @magik.net

I have borderline zero idea what all those random names mean/do, but the thing compiled now. No idea if it works tho. Pretty fun to enter a new technology and feel absolutely clueless. Next step is to get LookingGlass setup running, shouldn't take long, but I'm also tired enough today.

Oh yeah, I run Arch (btw), and of course nothing can compile with the most latestest libraries. I guess we contribute some to open source tomorrow ¯\_(ツ)_/¯

Magik6k@magik.net

Added a cable between computers, setup OpenSM and now it links up! Benchmarked at ~60Gbps with ib_read_bw. The bottleneck is explainable by.... PCIe on the non-threadripper end of this setup LnkSta: Speed 8GT/s (downgraded), Width x8 (downgraded)

Let’s see how smooth setting up RDMA between two non-server AMD machines goes

Magik6k@magik.net

NVIDIA DIGITS couldn't be announced at a better time. I have a dark fiber that I could use for remote display with Looking Glass over RDMA with lgproxy (4k 120hz hdr, ~60Gbps), and this cute little thing solves all of my client side problems with its hopefully 2x 100Gbps NICs

View full thread

Added a cable between computers, setup OpenSM and now it links up! Benchmarked at ~60Gbps with ib_read_bw. The bottleneck is explainable by.... PCIe on the non-threadripper end of this setup LnkSta: Speed 8GT/s (downgraded), Width x8 (downgraded)

Magik6k reposted

Replying to @magik.net

It's ALIVE! Now just need to put a super jank fan in the other computer and do the same setup there.

Let’s see how smooth setting up RDMA between two non-server AMD machines goes

Magik6k@magik.net

NVIDIA DIGITS couldn't be announced at a better time. I have a dark fiber that I could use for remote display with Looking Glass over RDMA with lgproxy (4k 120hz hdr, ~60Gbps), and this cute little thing solves all of my client side problems with its hopefully 2x 100Gbps NICs

View full thread

Let’s see how smooth setting up RDMA between two non-server AMD machines goes

Magik6k@magik.net

NVIDIA DIGITS couldn't be announced at a better time. I have a dark fiber that I could use for remote display with Looking Glass over RDMA with lgproxy (4k 120hz hdr, ~60Gbps), and this cute little thing solves all of my client side problems with its hopefully 2x 100Gbps NICs

* The card seems to have Infiniband disabled * To enable I need mellanox drivers * It also seems to see PCIe errors (?) not sure if that's a big issue :shrug: * It's also getting high temperature errors, probably due to being put into a peasant desktop case instead of a real machine

The IB disabled issue wouldn't be half bad, but both of those machines are on Arch Linux, and best packages are either Debs or RPMs. At least there are also from-source versions, but I don't think I have the patience for that I guess let's reboot and see if there are any bios knobs. Also add a fan

NVIDIA DIGITS couldn't be announced at a better time. I have a dark fiber that I could use for remote display with Looking Glass over RDMA with lgproxy (4k 120hz hdr, ~60Gbps), and this cute little thing solves all of my client side problems with its hopefully 2x 100Gbps NICs

Magik6k reposted

Just throwing this out there, but back in Ye Old Internet Times the links you had already visited would be purple and links you hadn't been to would be blue and I'm wondering if that sort of paradigm might be useful for helping people spot social media accounts engaged in impersonation.

Magik6k reposted

"Just use Postgres" isn't solely about technology, it's about organizational simplicity. For every later you add to your data stack you're adding essentially adding a team. Just Postgres really does work, and you could see team sizes 1/10 to 1/100 of other organizations.

My kernel was doing some really weird stuff recently, where something (in the kernel?) is allocating a ton of memory that isn't really accounted for anywhere (not in slab, not anywhere userspace (processes add up to 4GiB total), just kinda there Rolling back to older LTS kernel seems to fix things

View full thread

Likely related is that the burst of allocated memory in the first post happened after a large download, and also my network card is throwing a ton of AER errors for some reason (probably need to re-seat the CPU, tried other things) - but those errors were there for over a year now

For now this is where I'm at, I could just try to fix my hardware, but I'd like to figure out what kernel patch is causing this issue first - that way is just more fun!

My kernel was doing some really weird stuff recently, where something (in the kernel?) is allocating a ton of memory that isn't really accounted for anywhere (not in slab, not anywhere userspace (processes add up to 4GiB total), just kinda there Rolling back to older LTS kernel seems to fix things

This has started recently, and looking at my kernel update history, 6.10.10 is fine, 6.11.6 was probably fine too, but 6.11.9 was definitely not. 6.11.8 iirc had this problem too. Now, the main two drives that allocate user pages that I suspect are NVIDIA and i40e (inter networking)

Magik6k reposted

This is not a criticism of Bluesky, just three features that would be useful for me: - bookmarks, ideally with folders and even more ideally, with annotations (ie, you can attach a text note to a bookmarked post) - a native thread composer - drafts post and draft threads, synced between devices

Pushed some updates that made it both much better and slightly worse Should be much less buggy, and waay smoother on chrome, but sadly the new chunked renderer currently doesn't flip the board when playing as black :( (and I've run out of o1-preview message limit + need to do some real work)

Magik6k@magik.net

Here’s another chess site: chess280.com It’s like any other chess site, but this one only has one board