Jarkko Sakkinen
Posts
4943Following
327Followers
492OpenPGP: 3AB05486C7752FE1
Jarkko Sakkinen
jarkko
@wamserma JavaScript has this crippled number presentation where number is either 32-bit signed int or a double depending on operation applied.
So they added https://developer.arm.com/documentation/dui0801/h/A64-Floating-point-Instructions/FJCVTZS
:-)
So they added https://developer.arm.com/documentation/dui0801/h/A64-Floating-point-Instructions/FJCVTZS
:-)
1
0
0
@polarity @roger_booth i prefer this over phoscyon and abl303 because it does not have e.g. sequencer, yet another delay and stuff like that which is not that useful...
1
0
0
Jarkko Sakkinen
jarkko
@roger_booth @polarity i could support midilab but do not see any link to make a donation.
1
0
0
Jarkko Sakkinen
jarkko
Edited 1 year ago
To this day even tho I've contributed #Intel #SGX support to the kernel, I don't know how to check the chain as an end user.
E.g. Signal claims to use Intel SGX. How do I verify that for my benefit? There really should be some sort of universal standard for attestation of SGX/TDX/SNP workloads.
I mean the workload itself can be with a proprietary technology but attestation should be standardised. With that we could perhaps have something like certification chain that goes from data center up to the web browser.
I think confidential computing today is broken because of this and for most somewhat useless, expect in the white papers speaking about military grade security and all that :-) The hardware is expensive, attestation is broken and even the terminology is broken. In normal crypto-terminology confidentiality does not guarantee integrity. Better name would be thus trusted computing and somewhat easier to put into your mouth too. I've hated that term since I first heard about it.
E.g. Signal claims to use Intel SGX. How do I verify that for my benefit? There really should be some sort of universal standard for attestation of SGX/TDX/SNP workloads.
I mean the workload itself can be with a proprietary technology but attestation should be standardised. With that we could perhaps have something like certification chain that goes from data center up to the web browser.
I think confidential computing today is broken because of this and for most somewhat useless, expect in the white papers speaking about military grade security and all that :-) The hardware is expensive, attestation is broken and even the terminology is broken. In normal crypto-terminology confidentiality does not guarantee integrity. Better name would be thus trusted computing and somewhat easier to put into your mouth too. I've hated that term since I first heard about it.
2
0
0
Jarkko Sakkinen
jarkko
Edited 1 year ago
Another thing learned from local #rustlang meet up. There is one really useful sounding application for #Rust #embedded #USB stack: #firmware updates from web browser.
Some audio hardware uses #WebMIDI for this but this would widen the scope.
I'm not sure tho how the access would be provided to the device if the USB stack was compiled in wasm.
Some audio hardware uses #WebMIDI for this but this would widen the scope.
I'm not sure tho how the access would be provided to the device if the USB stack was compiled in wasm.
0
0
1
I was actually surprised how crappy the thing is in the standard ATM but obviously it will get fixed at some point.
Another thing I learned is that ARMv8 has special opcodes for JavaScript :D Fixing a horrible programming language with hardware features is pretty interesting.
Another thing I learned is that ARMv8 has special opcodes for JavaScript :D Fixing a horrible programming language with hardware features is pretty interesting.
1
0
0
@troglobit it took me about a week even to find a form to write anything to the customer service :-) well hidden feature.
1
0
0
@troglobit amazon goes kafka way with the algorithm...
1
0
1
Jarkko Sakkinen
repeated
repeated
The Last Psion | Alex
thelastpsion@bitbang.socialSide note: The article in that last toot is by the Castle Game Engine project, a game engine entirely in #ObjectPascal.
At first glance, their docs look to be very clearly written and approachable.
Also, they're written in #AsciiDoc, stored in a Git repo and turned into a static site. And you all know I love me a bit of AsciiDoc.
Thumbs up from me!
0
1
1
Jarkko Sakkinen
jarkko
I learned over the weekend (I was at a local #rustlang meetup) that interrupt controller caps #riscv for competing with #ARM on real-time tasks. I also learned about #CLIC. Just wondering are there any other extensions making waves or is CLIC where the world is converging.
More information:
https://ieeexplore.ieee.org/document/9660345
More information:
https://ieeexplore.ieee.org/document/9660345
2
0
1
Jarkko Sakkinen
jarkko
@troglobit thanks for the compassion, i'm not even pissed for some reason because the situation is so ridiculous :D cannot honestly believe this
1
0
0
Jarkko Sakkinen
repeated
repeated
Ludovic :Firefox: :FreeBSD:
usul@piaille.frComing to fosdem this year? you use #gnupg? Want to sign your key? Good news, I'm organizing a key signing party (https://en.wikipedia.org/wiki/Key_signing_party).
details are at https://ludovic.hirlimann.net/2024/01/key-signing-party-at-fosdem-2024.html
please boost or share, so people come and attend.
#privacy #email #security #gpg
0
2
1
Jarkko Sakkinen
repeated
repeated
K. Ryabitsev 🍁
monsieuricon
Would be nice if all package managers understood/aliased all common English verbs meaning "remove this package", because I can never remember if it's "remove", "erase", or "uninstall".
5
5
21
Jarkko Sakkinen
jarkko
Edited 1 year ago
Right, someone reminded me yesterday that if you want rigid memory allocation scheme with heap, you can implement that with heapless crate. It will deliver allocation errors from pre-allocated heap per-allocation. For stuff that I might be doing with #Rust in user space that is good enough.
The de-facto response from Rust application for OOM is panic.
#rustlang
The de-facto response from Rust application for OOM is panic.
#rustlang
0
0
2
@alerque I'm mostly looking at things from operating systems and alike side and I have total buy-in uapi and drivers side because merits are really something I count (i.e. preventing subcategories of exploits thanks to compiler) but the "core" side of the equation is totally alien to me. Sorry, totally off-topic but just writing these down merely as remarks.
0
0
0
@alerque And also sort of might be a bit harsh and controversial to say this but when there's big noise about a new language feature (which in the end means zero value at least at its epoch) I just think "when are these language features going to end". I mean on C side in kernel we were before like C89 + gnu extensions and now I suppose we are like C11 standard('ish).
This rapidness might work for user space but especially for core kernel features feels a bit brutal to say the least...
I'm actually just opening my views because in LKML I see "yes Rust" and "please no Rust" so trying to address "in-between-but-doubtful-of-Rust" camp's concerns :-)
This rapidness might work for user space but especially for core kernel features feels a bit brutal to say the least...
I'm actually just opening my views because in LKML I see "yes Rust" and "please no Rust" so trying to address "in-between-but-doubtful-of-Rust" camp's concerns :-)
1
0
0