Kevin Milczewski

I write code and build BattleBots.

[Kevin Milczewski @KevinMilczewski]

@immanuel_vibe >kernel never spends a cycle

Except the driver runs in kernel...

eBPF is great, but a lot of these solutions are just workarounds for having a monolithic kernel vs micro kernel. We run most of these same features directly from userspace with much lower overhead than eBPF.

[Kevin Milczewski @KevinMilczewski]

@_dylanga And this is new how?

[Kevin Milczewski @KevinMilczewski]

@andersonc0d3 You should rewrite in your own voice or teach your agent a better voice. Way too many LLM-isms in your article.

[Kevin Milczewski @KevinMilczewski]

@jhleath It doesn't take much to want to ignore almost anything DHH says

[Kevin Milczewski @KevinMilczewski]

@0xCodez The irony of this post being full of easily identifiable claude text patterns is hilarious.

[Kevin Milczewski @KevinMilczewski]

@iximiuz Very nice explanation of the startup internals. Another fun fact, main is just a linker symbol, at least with 10 year old compilers it doesn't even need to be in a code segment. https://t.co/pMNo88oTUC

[Kevin Milczewski @KevinMilczewski]

@techneo @TrisH0x2A Its actually slow. The multiply alone is the same or more cycle cost as tzcnt, then there is a memory lookup which can be ~500x slower if not in cache.

Unless you mean the complier replacing it with 1 instruction, then yes that's awesome.

[Kevin Milczewski @KevinMilczewski]

@CalebChamberla6 Do you have the list of supported materials?

[Kevin Milczewski @KevinMilczewski]

@CalebChamberla6 Had my first order ship from there a week ago or so. Somehow the cut quality is better than the last time I got the part. Amazing job having the cuts dialed from day 1.

[Kevin Milczewski @KevinMilczewski]

@CalebChamberla6 @repkord Are those bottles the gas for the lasers? Are those like carriers are you move groups of them around?

[Kevin Milczewski @KevinMilczewski]

@BenjDicken Off by ~5x in some cases. That is likely lightly loaded latency on nvme. Under load reads that hit the nand on it are way worse than 60us, and your kernel + pci transaction time on gen3 is close to 60us. Gen5 this is close. Also EBS is a lot faster than that :)

[Kevin Milczewski @KevinMilczewski]

@jhleath >ebs drives are like the disks that are attached to your laptop. they don't have any logic inside of them to allow multiple machines to coordinate on the same set of data.

https://t.co/oY8RwUZgts

[Kevin Milczewski @KevinMilczewski]

@Geenimetsuri @straceX Yeah. They compile to the same thing. Clang generates a branchless vectorized loop for both.

https://t.co/qvzzxXCfIW

[Kevin Milczewski @KevinMilczewski]

@Sirupsen The 1-2us "SSD" speeds look like block cache times. I'm pretty darn sure you're not getting through PCI + interrupt stack in 2 us. Hell, if your machine has deep sleep power modes on, the interrupt wake latency is over 100us.

[Kevin Milczewski @KevinMilczewski]

@CalebChamberla6 Near norwalk? :P

[Kevin Milczewski @KevinMilczewski]

@max0x7ba @ChShersh I feel like the talk probably about the lock free queue itself where you already have data in L1d and want to shove a pointer through a queue. The cache coherency there is usually high if there is contention with traditional CAS based approaches.

I want to find the talk!

[Kevin Milczewski @KevinMilczewski]

@ChShersh @thedcoffman What CPU is that running on? Sub 10ns implies Shared L2 unless it's some 5ghz monster? And maybe doing it without locked instructions for the write? So no CAS and opportunistic reads?

[Kevin Milczewski @KevinMilczewski]

@chrisgpt So it recreated something that already existed and was highly text based? Not really surprising. Still cool.

[Kevin Milczewski @KevinMilczewski]

@Adriksh "no memcpy" -> writes a bytewise memory copy manually and stores the whole data in the ring.

The way to do this right is to pair it with an object pool and pass the pointers through the ring. That is zero copy. Of course your object pool is likely to use the same structure.

[Kevin Milczewski @KevinMilczewski]

@CalebChamberla6 Xometry went to crap when it became more B2B focused (among other things). These dinosaur companies are the heart of the problem you've fixed with innovation. Don't let them drag you down into their game of waste an inefficiency.