fuz

Not many news here.

sync

thanks

bömp

bump

Nope, I have not been able to.

interesting

Merry Christmas to you too!

Interesting, thanks for looking this up!

Perhaps if you had an 8086 / 80286, you could only access register-register operations and the register-memory operations came later.

There was at least thisthing: x.com/david_schor/status/1205565182992429056

btw, do you know anything about the Weitek 1067/1167 FPUs?

interesting

^ this manual says that video modes 0bh and 0ch are internal to the EGA BIOS. Do you know what they did?

techhelpmanual.com/114-video_modes.html

I noticed that the other sites don't really like questions about outdated tools.

Retrocomputing is probably better.

The curator's assistant recommends to come on any day but Tuesday (the one day in the week the museum is officially open) as there is a “danger of visitors being present” when you come during the museums opening hours. Instead, send them an email and agree on coming some other day.

Got to visit the Stuttgart Computer Museum on Tuesday. I highly recommend a visit!

github.com/rodneyknaap/atx-286at-v1-mainboard

^ do you know the guy who built this one?

I hope it'll be worth the money

finally getting a new laptop with AVX-512

Fascinating. I had wondered how this worked.

A bit smaller than last time I was there, but neverthless nice

thanks, it was great

Whatever works for you!

Looking forwards to this year's VCFB!

brömpf

At least that's what the audio engineer told me.

They'll fix it in post production. It'll be fine on the final video.

^ here's the talk I gave at EuroBSDcon 2024.

youtube.com/live/OzX38cWdivc?si=Y0AsQeFaguCbSXlr&t=22345

thanks

Not sure if there could be any impact from adding new sections like that

e.g. they write their own linker scripts to pull it in

People do weird things with libc

However, we decided not to do so to avoid extra complexity.

I agree and I had considered this; it would be trivial to implement.

I spent a lot of time trying to work out a solution without swapping and no duplicate code, but didn't manage to come up with anything.

I wrote both and compared them in a microbenchmark. The branchless variant is significantly slower, by like 10%. This is because swapping the various registers using conditional move instructions turns out to be a lot slower than one would naively believe.

Google paid for a student to port my work to AArch64. He's great and we'll do a joint talk at EuroBSDcon 2024 on our work. Another student attempted a port to RISC-V, but he only managed to complete 6 functions.

Here's some current work complementing what I talked about in the video: github.com/soppelmann/freebsd-src/commits/exprunSIMDports

As for @10:34, I'd say you are most likely right. This is an inaccuracy on my part.

I don't really care about AI, so I didn't even think of mentioning it. The talk was a bit of a mess as I didn't spend enough time preparing for it. Got drunk the night before and then forgot that a time zone change stole another hour of preparation time from me.

F5

bümp

Intel has an all new microarchitecture with Lunar Lake that departs significantly from the stagnant old stuff

I'm mainly interested due to the AVX-512 improvements.

The performance is very promising