
🇨🇦 • leukemia instructor @sloan_kettering • @SwiftLang language steering group member
New York, USA
Joined April 2007
- Tweets77
- Following173
- Followers255
- Likes631
And now for something different…
Is it possible to write an implementation of 128-bit decimal floating-point arithmetic in Swift, from scratch, with comparable performance to major reference implementations in C/C++?
Yes! github.com/xwu/decimalannex
1
1
45
I wrote up some of the challenges and learnings along the way.
forums.swift.org/t/native-12…
19
For years, lab lore was that the workhorse Scl-CreERT transgenic driver is integrated on mouse chr4, thanks to yeoman’s work tracking genetic crosses by @sanchezvelamd. In 2024, we decided to find its exact locus of integration by inspection of aligned Illumina short-read WGS. /1
Indeed, we found the transgene on chr4 just upstream of Echdc2. We’ve since been using this knowledge to streamline our mouse colony. But how best to share such a simple finding with the field? That we still haven’t figured out. So here’s the preprint. /2
biorxiv.org/content/10.64898…
Thanks belatedly to @TrowbridgeLab for the encouragement that it’s interesting enough to share, and of course @rosslevinemd for indulging this side quest!
1
2
378
Xiaodi Wu, MD, PhD retweeted
“some AIs were found to be picking up on the text font that certain hospitals used to label the scans. As a result, fonts from hospitals with more serious caseloads became predictors of covid risk” 😬 technologyreview.com/2021/07…
Xiaodi Wu, MD, PhD retweeted
The next time you’re in a quiet ED, I encourage you to proclaim loudly that we now have RCT evidence that uttering the word “quiet” does not, in fact, influence work flow.
See where that gets you.
bmj.com/content/367/bmj.l644…
They’ve *finally* fixed the alignment of the send button next to the outgoing iMessage. Thank you, Tim Apple.
Xiaodi Wu, MD, PhD retweeted
Teachers! Parents!
It is legal and perfectly welcome to email scientists asking for a PDF of their paper.
Until all papers are free, please RT to make this more widely known. And fyi, scientists don't get the $$ from sales of papers so all it costs us is 1 minute of time!
This quoted post is unavailable.
let (x: b, y: a) = (x: 1, y: 2) // OK
let (y: a, x: b) = (x: 1, y: 2) // shuffle, OK
let (y: a, x: b) = (1, 2) // warn: extraneous label 'y'
WRT tuples, another solution without touching shuffling: warn when labels are newly added during destructuring and immediately discarded.
1
That'd be problematic: I've had reason to name a member `T` (yes, uppercase)--didn't shadow a generic type T, but easily could have.
1
You don't need to deal with tuple shuffle to avoid this: warn when any variable name shadows a type name, allow silencing with backticks
It's why I keep saying SE-25 is so deeply problematic; in this example, IMO, S : R should be rejected for lack of actual public default impl
1
Yeah that's the part that's nonconformant with SE-25; it was decided that access modifier in front of extension is default *and* max allowed
Extensions have different rules as of SE-25: not apparently consistently enforced, but internal extension can't have public member
still trying out twitter! nitter.cf/xwu/status/43674782
3








