Zach Bjornson (@[email protected])

Microbiology and immunology Ph.D. working in bioinformatics for @CellCarta on @CellEngine. @[email protected]

[Zach Bjornson (@[email protected]) @Zbjorn]

Super excited to officially launch CellEngine for #flowcytometry and #cytof analysis. We've been developing it for five years and have tested it with over 3.5 million FCS files. It's super fast, reliable and has powerful visualization capabilities. 🧵

[CellCarta @CellCarta]

NEWS: we’re launching CellEngine – a cloud-based cytometry analysis #software for fast analysis and easy collaboration.
Learn more: https://t.co/K1wJ3iFdGM

#CellCarta #CROlaboratory #precisionmedicine #biomarkers #cytometry #bioinformatics #datascience #flowcytometry

[Zach Bjornson (@[email protected]) @Zbjorn]

@metadaddy Glad you liked it! The numbers are still pretty accurate today.

🙄 Thanks for finding that copy. Oh well.

[CellEngine @CellEngine]

@CellEngine isn't just for cytometry data! Our poster at #CYTO2024 analyzes #CITEseq data by comparing and combining hierarchical gating with unsupervised clustering.

Check out the details on our blog: https://t.co/6rYRgpJd4I

[CellEngine @CellEngine]

Next up in our tutorial series, we cover advanced gating features - automatic gate adjustment, percentile gating, combination gates, and more. If you don’t have a #CellEngine account yet, sign up for a trial at https://t.co/NWGTHEpi8c and follow along.

https://t.co/TDCYnjrwwc

[Zach Bjornson (@[email protected]) @Zbjorn]

@tom_forsyth @WillPBach No... See 3.1 of https://t.co/RP25t21SOH. Lactase works optimally around pH 6.0, which is close to the pH of cow milk. It's done at a controlled temperature in batch processing.

[Zach Bjornson (@[email protected]) @Zbjorn]

@WillPBach @tom_forsyth Yeah, the lactase breaks it down immediately to <0.1 g lactose/L.

[Zach Bjornson (@[email protected]) @Zbjorn]

@jarredsumner Coming to SpiderMonkey now: https://t.co/cXaah3v7zQ

[Zach Bjornson (@[email protected]) @Zbjorn]

@jarredsumner I *think* Node makes no guarantees here -- that fs.write will write whatever's in the passed Buffer when it reaches the kernel. Is Bun making stronger guarantees?

[Marius Millea @cosmic_mar]

One of the greatest mysteries in our current understanding of the universe is why the Hubble constant appears to be different if you use CUDA v11.4 vs v11.5

[Zach Bjornson (@[email protected]) @Zbjorn]

@joseph_h_garvin I tried to make a PEG grammar for it but there are inconsistencies in Intel's material that makes it hard to finish. https://t.co/46mG6SqjP1 has the draft of it and shows how to get the XML.

[Zach Bjornson (@[email protected]) @Zbjorn]

@joseph_h_garvin For instruction and timing, https://t.co/CUGpW9IEbD has XML files: https://t.co/zgqqjPdkCR and Intlatx64 has fixed-width tables https://t.co/RuKyw1iO57. For intrinsics and pseudocode, Intel's intrinsic reference page is powered by an XML that you can easily download. 1/

[CellEngine @CellEngine]

CellEngine is now live on Twitter/X! Follow us to get the latest #CellEngine updates and #flowcytometry tips.

[Zach Bjornson (@[email protected]) @Zbjorn]

@JonasBadalic @bunjavascript Third and fourth commits in the link below. Close to the smallest useful function: get one arg, check its type, get its data (it's a TypedArray), return void. The OP benchmarks were with a length-1 array to focus on call overhead.

https://t.co/3Ifpz1RKpV

[Zach Bjornson (@[email protected]) @Zbjorn]

. @bunjavascript's C/C++ interface overhead is so much lower than Node.js'. Pretty awesome to be able to use NAPI without a performance penalty.

[Zach Bjornson (@[email protected]) @Zbjorn]

@jarredsumner @Constellation I worked out quite a few of the other conversions for x86 in https://t.co/HCNpS2Axrh, in case it's useful

[Zach Bjornson (@[email protected]) @Zbjorn]

@jarredsumner Maybe consider investing resources in a continuous fuzzing suite instead, plus a bug bounty program. Given how quickly bun is changing, I think you'd get more benefit than periodic pen testing.

[Zach Bjornson (@[email protected]) @Zbjorn]

@andrey_akinshin Great post as always. I think the fourth image is the wrong one though (efficiency for uniform distribution).

[Zach Bjornson (@[email protected]) @Zbjorn]

@jasnell A spec should be stored somewhere central, be continuously updated by multiple people as the feature evolves, document research and design decisions, etc. A PR isn't that.

[Zach Bjornson (@[email protected]) @Zbjorn]

@jasnell I think an actual spec is critical, and it's one of my biggest frustrations with Node's APIs. Without the spec, who/what's to say a behavior is intended and not to be changed? Unit tests can help, but rarely have commentary. Node Streams keep having subtle breaking changes e.g.

[Zach Bjornson (@[email protected]) @Zbjorn]

@tomjcleveland @zainy Yes, constant time equals in the crypto sense.

[Zach Bjornson (@[email protected]) @Zbjorn]

@Kruithne @eddyw_dev @jarredsumner Hum, I was mistaken, deleted previous tweet to not spread wrong info. It looks like V8 does do an allocation of a FixedArray for holey SMI elements, up to ~512 or 1024 MB.