@joseph_garvin I voted for option 0 but there are a lot of unstated assumptions ("you are allowed to write to the top byte", "bottom 24-bits of the already stored value are (apparently) zero", etc).

TBH, even while understanding a lot of the microarchitecture of many x86 chips, the best answer to questions like these is very implementation dependent but as a rule of thumb on modern processors: avoid partial register usage and minimize number of stores.

@reidrac this is likely fixable; I’ve replaced many non-user serviceable batteries, but the challenge is finding the replacement battery. AliExpress used to be excellent for this before the tariffs.

@didoesdigital @wezm Alas the effect is completely temporary and it turns into a tedious game of whack-a-mole

This is a reminder that April is Procrastination Awareness Month.

@kenshirriff Which makes me wonder what the reliability was of these transistors when stored and operated in ideal conditions.

@kenshirriff @arafel IIRC, they really like to operate with strange negative voltages too.

It might be fun to implement something new with IBM style circuits.

@wren6991 flags are extra pain for high performance cores. I’d actually prefer multiple output instructions instead, which would also allow load-pairs.

@bitsplusatoms @interlisp @amoroso @masinter OMG, didn’t realize how huge the box was until I saw the picture with a tiny human next to it. My favorite esthetic is still the Daylight and I’d love a 3D pronted scaled replica.

@timonsku Before that had direct flights I often went through Frankfurt but I tried to avoid it because people were smoking everywhere. It’s one of my least favorite airports.

@ids1024 I'm using Ubuntu on both ends (different versions though) and the server FS is ZFS. I *think* I also didn't have problems back when I was serving from FreeNAS.

I'd be curious to understand the root cause of your issue as in my humble opinion, NFSv4 is so much better than NFSv3.

(My file server went down from a really stupid mistake of mine and it took a few *weeks* before I fixed it. My workstation patiently waited and resumed without missing a beat when the server came back.)

@ids1024 I use Cargo over NFSv4 all day long, but it's Linux server, Linux client. The only issue I have had to deal with is that, unlike 20 years ago, software is ill prepared for multiple hosts on the same file system. I had to move .cargo (and .rustup) to .cargo.$ARCH and set the appropriate env vars.

Apple however goes completely fubar if you try to share $HOME across hosts so I have given up on that.

@jornane @juandesant Yeah it usually goes "there are two hard issues in CS: naming, caching, and off-by-one errors".

In my humble experience, at least the first two are indeed very common.

My contractor is working with https://wastefree23.org
I thought it looked pretty interesting so I'm sharing it here. Boosts welcome.

#wastefree #recycle #sustainability

@markd @kenshirriff Ha, very similar: my first ever PC (circa 1990) was the cheapest laptop that could run Emacs under Linux; I got a 4 MiB monochrome 640x480 Toshiba T1800, sporting a 16 MHz (?) 386Sx.

I want a new job.

Anyone need an embedded engineer?

I've done work on embedded Linux, RTOSs, bare metall you name it.

I've also worked on the security side of embedded, TPM and HAB and what have you. And I'm especially good on weird platforms. I mean... weird machines. And I dig deep on the weird debugging issues.

I'm good at skunkworks type projects. You got a new project, I can work with the team to create something improbable.

Edit: US based, remote.

#getfedihired
#FediHire
#fedihired

@dngrs rust_discussions are a mediocre bot? May be, but I got at least one reaction :)

I’ve raised similarly concerns, though I phrased it as the crate system encourages bloat. To my surprise I found nobody agreed. I have deliberately avoided even common crates at work because all dependencies are vendored and the review process is real. That’s a good policy with some obvious downsides.

@rust_discussions they scare me too

@robstewartUK Nice work but why isn’t there a direct comparison to Reduceron? The frequency looks slower than what Reduceron could hit in the same tech, but you maybe have a per-clock advantage. Maybe.

@jonathan @cstross

Public Service info from the volunteer chief archivist at datamuseum.dk:

NEVER use flash-based storage for archival purposes.

Not CF, not SD, not USB-sticks, not SSD, not NVME, nor any other flash-technology based storage media.

Flash is simply not a stable storage mechanism: Just like Dynamic RAM it will eventually forget your bits.

Unfortunately there are no good choices for long term data storage, but flash is a particular bad choice.

Because I have an odd sense of humor, I implemented Chacha20 in BLS Pascal on the Nascom2, running in my browser (https://tommythorn.github.io/virtual-nascom/virtual-nascom.html)

Turns out adding 2 32-bit values is really painful this way. Achieved ~ 10s for a 64 byte block, but written in assembler it would be a lot faster.
#Z80 #Nascom2 #retrocomputing