/dev/kelas

Hi Jordan, This is a known issue. The primary reason Pragmata is offered in monospace is to provide a fallback for such cases: 

obviously, most of this is well-known by now, and is generally or academic interest. **except maybe one thing.** ``` 8801e6ca interrupts-extended 8801e6fc interrupt-parent ``` this is not meant to be,...

> Thanks for the PR! not at all — and I also totally appreciate why nobody's in a hurry to merge this little piece of public heresy :)

> Did you use ghidra to generate the pseudocode? If so, do you have a link to your project anywhere? yes, a combination of the excellent r2 work from this...

> Macros tend to harm readability and maintainability as you know. depends on your choice of macros :) this particular style indeed doesn't match the neighbourhood, it's a lazy copypaste...

yup from the looks of it simdjson version would get auto-unrolled like a boss. blue skies. but yeah - i think it is the first time i've hit a major...

> time 30% higher it actually seems to fluctuate between 10% and 25%, i guess depending on the cost of `lea` instruction on a given microarch. > then you have...

> purely aesthetic reasons. thanks for digging that out, alex! i second that, i originally de-unrolled both loops in the C port just because i didn't want to have zzzzzzzz...

> If you really want unrolling, you better do it yourself in the code. this little hiccup at hand provides ample evidence that it just ain't so. a human getting...

> haswell ok > evidence lacking here's the bare-minimum example of the exact `lea` condition on `haswell` microarch, which manifests in hand-unrolled version, and doesn't when unrolling is left to...