Posts
4417
Following
315
Followers
471
Linux kernel hacker and maintainer etc.

OpenPGP: 3AB05486C7752FE1
@lkundrak @ljs SUN employees doing pair programming
0
0
1
I still hate reading GAS syntax after all these years. I learned assembly with Turbo Assembler and Intel syntax, which I still think is so much more readable.
0
0
1
Edited 11 months ago

Just so that you know llvm-ir is the language tag for #LLVM #IR when formatting it for #Markdown. I did not know this but found it by trial and error.

0
0
1
@oleksandr @gromit @vbabka I made my own independent fork that I use from now on ;-) mbsync should almost never change to anything, and it is only 15 KSLOC program.

I have some ideas how to modernize it too a bit (to make it easier to maintain) so might work on that too. Got familiar with the code base thanks this bug.
1
0
0
@vbabka Thanks! I applied like wide array of places because you know economy is not doing too well, and ended up to this situation.
0
0
0
What does this mean?

```
install transitive failure
└─ install main transitive failure
└─ zig build-exe main Debug native 1 errors
```

I tried --verbose-cc and --verbose-link but neither made this more verbose.

#zig
0
0
0
@gromit @adamw [it hurts my head that the project is called "isync" the artifact is called "mbsync"]
0
0
0
@montar Not really. Please tell me if you hear. And please do not hit me, I'm just a random end user ;-)

EvenConsoleCowboys is wondering it in some episode of his Radare2 tut's WTF is this shit why there are two versions of this.

Fedora has Rizin and Cutter nicely packaged so I've randomly ended up using those.
1
0
1
@gromit @adamw

Sorry I was too busy having a weekend but here's a screen cast that proves that -fno-lto addresses the issue :-)

I definitely going to check also later on https://gitlab.archlinux.org/archlinux/packaging/packages/isync/-/issues/1 but at sight I don't really believe this is the same issue, it fails elsewhere. Obviously this can be considered most as guess because I don't know how the internals of mbsync interact with each other.
1
0
0
@gimulnautti I have this big idea that would like to see how much you can pull juices out of P-166 Mhz MMX version of Pentium, which is sort of nice architecture, two ALU pipeline and concurrent MMX/FPU pipeline. With modern dev tools doing heavy recalculation there is a lot of resources in that type of machine.
1
0
1
@gimulnautti nice work!

I looked over my holidays how people target retro / old-school stuff, or at least majority of people, to MS-DOS. It was across the board OpenWatcom or even Turbo-C.

So I quickly put together a build pipeline, with custom linker scripts defining binary layout and no crt0 linked to the final binary. So now I can build .COM files with the code written in latest GCC or clang (LLVM) :-)

I've written the currently used arch x86 code in Linux that trampolines to real-mode and back when going into and away of suspend in 2012, so with those learnings of doing this type of custom binaries I also wrote shenanigans to access full 4 gigabytes of flat address space.

And just to refresh a bit I wrote a bit code that initializes so called "Mode-X" i.e. 320x240 planar squarepixel mode. I was able to run my stuff in 86box, which felt good ;-) I loose the C standard libraries as payback I have like the 2024 level development environment.

Might put it fully protected mode later on but BIOS access is nice for now. Protected mode is better than real-mode for performance, especially for data heavy real-time stuff like DSP and texture mapping because 32-bit memory accesses in tweaked unreal mode add a prefix to the opcode, which consumes icache ;-)

Not sure what I'm doing with this yet, probably just add random stuff to it without a plan. Also, MZ EXE bin generation is a target that I pursue.
1
0
1
Edited 11 months ago
@ljs I'm waiting for the book, I'm really into memory management stuff so buying it is a no-brainer for me. I learned quite a bit of internals when working SGX so looking forward to build on top of that :-) Especially this VMA merge stuff is interesting given that I have had my fights with that side of mm.
0
0
0
At this point I can say there's no return to Dropbox.

I have a simple "cache hierarchy":

1. Data in transit @ iCloud
2. Data at rest for foreseeable future use @ Storj
3. Data at rest for archiving @ NAS.

And since local NAS and Storj speak standard protocols it is dead easy to automate everything and never loose any data again by mistake.

I've now deeply seen how Dropbox style of even based synchronization can be harmful, when you **** ** your files. Now that I have more like "commit and sync" cloud workflow, the collateral damage of that is isolated to the source :-)
0
0
0
I've been using #Storj as my only cloud storage since March, except 50GB iCloud for being "L2 cache" for photos, which go to my NAS.

Just S3 without extra clutter is just soo convenient, and there is #CyberDuck, #s3fs and #rclone to make using it super smooth.
1
0
3
@mekkaokereke Might be too idealistic but I see it actually only viable commercial threat to the existing "Oligarch status quo" of IT ;-) Not that different how Internet was in let's say 1995.

It is good to denote also. that:

1. Mastodon is a self-governing social network, which does not have keep up any target rate of visitors in order to kept funded because it is not a company. We can have it as long as we want it. It is up to us.
2. Self-governance is sealed and stamped by Affero GPL, which prevents also running modified instances of the code without publishing the changes on request, in addition to the usual copyleft stuff.

Affero GPL in modern AI fuelled Internet is probably the most important license FSF has ever created, not from fairness perspective, but from confidentiality perspective. It is in. a way same as using something like Intel SGX but instead of CPU features and attestation to the Intel CA, legal enforcement is used to seal the confidentiality.

E.g. Signal uses AGPL because of this, as a commercial company.
0
0
0
@mekkaokereke For instance, if I put my own startup (any possible startup), I would probably create isolated ActivityHub network to play the role of intranet instead 365 or Google Cloud shenanigans. It would open so many options to pivot in the future how to interact with outside world and best of class integration to the best possible network to hire best of their breed developers and hackers :-)
1
0
0
@mekkaokereke I has potential to build also commercial services but more like within "scope of instance", i.e. offer healthy commercial environment. I.e. just like in town you know when you enter a grocery store, or movie rental service, and when you leave from there.

Normally **** drives Internet evolution at least when it comes to bandwidth but in the case of core services it is developers and I could imagine services that I might even use for some dev stuff based on ActivityHub. I've replaced Github gists with Mastodon for instance with the basic stuff. If something like that ever succeed, that could start to stretch to something bigger.

It is like moving from monarchy to democracy in the virtual space.
1
0
1
Edited 11 months ago
this is really super smooth workflow-wise kick: https://audija.com/kickdrum

Ignoring that it so much cleaner (except colors suck) than #KICK2, it is has one benefit over: the EQ multiplies the source signal with the EQ curve like per Fourier coefficient (i.e. like local EQ in #Harmor and the EQ in Phase Plant's wave editor). This cancels any possible phasing errors.

You cannot really have that otherwise, except other kick drum generators with similar built-in EQ (of which I'm not aware of).

#audija #MusicProduction
0
0
0
@ljs Just interested on topic because we tried to solve bunch of bottlenecks with SGX and mm locally in that stack. Like for instance we fought a lot with open device files, reserved memory which is also shared memory and do_exit() flow.
1
0
0
Show older