JATth

nohz_full confusingly also helps with power usage… if the cpu doesn’t have anything to run, no point waking it up with a scheduler-tick IPI… but also no point trying to run the scheduler if a core is peaking with a single task… With nohz the kernel overheard basically ceases to exist for a task while the it is running. (Thought the overhead just moves to non-nohz cpu cores)

To be fair, there should be some heuristics to boost priority of anything that has received input from the hardware. (a button click e.g.) The no-care-latency jobs can be delayed indefinitely.

Why can Windows do it when Linux can’t?

Windows lies to you. The only way they don’t get this problem is that they are reserving some CPU bandwidth for the UI beforehand. Which explains the 1-2% y-cruncher worse results on windows.

I agree that UI should always take priority. I shouldn’t have to do anything to guarantee this.

I have HZ_1000, tickless kernel with nohz_full set up. This all has a throughput/bandwidth cost (about 2%) in exchange for better responsiveness by default.

But this is not enough, because the short burst UI tasks need near-zero wake-up latency… By the time the task scheduler has done its re-balancing the UI task is already sleeping/halted again, and this cycle repeats. So the nice/priorities don’t work very well for UI tasks. Only way a UI task can run immediately is if it can preempt something or if the system has a somewhat idle CPU to put it on.

The kernel doesn’t know any better which tasks are like this. The on-going EEVDF, sched_ext scheduler projects attempt to improve the situation. (EEVDF should allow specifying the desired latency, while sched_ext will likely allow tuning the latency automatically)

No, I definitely want it to use as many resources it can get.

taskset -c 0 nice -n+5 bash -c 'while :; do :; done' &
taskset -c 0 nice -n+0 bash -c 'while :; do :; done'

Observe the cpu usage of nice +5 job: it’s ~1/10 of the nice +0 job. End one of the tasks and the remaining jumps back to 100%.

Nice’ing doesn’t limit the max allowed cpu bandwidth of a task; it only matters when there is contention for that bandwidth, like running two tasks on the same CPU thread. To me, this sounds exactly what you want: run at full tilt when there is no contention.

“The kernel runs out of time to solve the NP-complete scheduling problem in time.”

More responsiveness requires more context-switching, which then subtracts from the available total CPU bandwidth. There is a point where the task scheduler and CPUs get so overloaded that a non-RT kernel can no longer guarantee timed events.

So, web browsing is basically poison for the task scheduler under high load. Unless you reserve some CPU bandwidth (with cgroups, etc.) beforehand for the foreground task.

Since SMT threads also aren’t real cores (about ~0.4 - 0.7 of an actual core), putting 16 tasks on a 16/8 machine is only going to slow down the execution of all other tasks on the shared cores. I usually leave one CPU thread for “housekeeping” if I need to do something else. If I don’t, some random task is going to be very pleased by not having to share a core. That “spare” CPU thread will be running literally everything else, so it may get saturated by the kernel tasks alone.

nice +5 is more of a suggestion to “please run this task with a worse latency on a contended CPU.”.

(I think I should benchmark make -j15 vs. make -j16 to see what the difference is)

BTW. With clang lto’d kernel 6.9. When non-Arch get the buggy updates, We have already moved on.

cries silently So no sleep until I crash. Got it.

I once helped a person with their computer. They complained the they cant save the their photos. Well, their onedrive was filled to brim with crap, while the local 1Tb disk was empty because they had zero idea how storage and folders work. I had to explain her there is literally 1000x more fast disk space available, so please dont save into onedrive.

Wayland has much more up-to-date graphics technologies behind it to put it simply.

Mint was the distro that converted me. After 8-10 years, I’m still using the cinnamon DE, but just on top of Arch. Next hope is the devs port it over wayland so I can also ditch xorg. (There is a demo/alpha available)

I’m actually bit sad that I had to move onto a ISP which has zero IPv6 support, as I previously did have IPv6. The last thing I did on that connection was to debug the hell out of my IPv6 code I had developed.

What’s happening in the area, I.e. a view of the overall scenery. This vast perspective is different, and it’s like looking at a live painting. When I’m on the move, such a observer perspective shrinks to few tens of meters, which kind of makes sense.

I don’t think this is anything most people don’t do however. I do remember places quite vividly though, and I practically never get lost. People however in the scenery, I forget in about a minute.

I once said that the current “AI” is just a excel spread sheet with a few billion rows, from what all of the answer gets interpolated from…

But it’s the spaghetti cabling that makes it work and highly robust.

Put that 20 billions first into fusion research. Like yesterday. How on earth would we power and cool that 100 km of superconductors otherwise? Unless the 100 km FCC is required by the fusion research, then we have a pickle.

So, now they are slowly (or immediately and forever, I don’t know the time span) injecting propaganda into their clone of wikipedia and they are simultaneously thus admitting they are doing it. (to further brainwash the russian citizens)

So lettme repeat: FUCK PUTIN, and stuff your rubber clones in your ass. (which there are many of)

The in-rush of endorphins when the modded LCD thing actually worked probably knocked you out?

A bit of a side story: I disassemble probably 1–5 panels in a week. (For recycling, it takes about 20-45min per panel.) The flat cables alone are so flimsy, I would say just assembling a display again from known, matched and functional parts would take days. I would triple or quadruple that to assemble a display from random parts. Considering this, that modded panel is quite amazing to me.

In standalone panels, the PSU has a chance of killing you: the main capacitors can hold multiple joules worth of nope, charged to about 400v. So, if the main caps are not allowed to discharge (if they discharge at all that is), there is a possibility of death when disassembling a panel with an integrated PSU. Waiting “5min” is bad; a PSU needs to be unplugged for a day or two at least before the charge drains out.

I put too way too much effort in this reply… Yes… it’s nerve racking, especially if you are resorting to BIOS flashback to boot the CPU on an older (new) board.

Can’t get visuals (except maybe leds/indicators on the motherboard itself) when your CPU is incapable of accessing the ram or the devices yet. All external devices normally communicate through the RAM. (And by external, I mean not on the CPU package) Yet, the CPU has to solve out this chicken-and-egg problem of how to progress from the cold-boot without knowing what external RAM is installed. There are plethora of timing/clock-cycle/voltage settings for one stick of ram, which are tested on POST. Establishing sane DDR5/4 parameters is non-trivial. (I think it is order of +20!, twenty factorial: 2432902008176640000, if there were no starting point of XMP, JEDEC etc.)

I use hand tuned settings for DDR4, and on cold boot, the BIOS adjust the settings which I didn’t forbid it to do. Unless I unplug the PSU from the wall, the BIOS won’t retrain the memory again. I suspect my settings still aren’t 100% stable. (over period of years) Non-cold-boot assumes the ram works 100% same on each power up. If some OC setting drifts past a threshold once the system is heat soaked or receives more EMI interference, this could provoke a crash/BSOD etc. in absurd theory having a busy wifi router next the ram could cause the bios to select more robust/conservative settings to counter the EMI interference. Would be fun to know, if this would be true.

I just did: “rm -rf xz”

pacman -Syu
find / -name "*xz*"  | sort | grep -e '\.xz$' | xargs -o -n1 rm -i 
pacman -Qqn | pacman -S -

(and please, absolutely don’t run above as root. Just don’t.) I carefully answered to retain any root owned files and my backups, despite knowing the backdoor wasn’t included in the culprit package. This system has now “un-trusted” status, meaning I’ll clean re-install the OS, once the full analysis of the backdoor payload is available.

Edit: I also booted the “untrusted” system without physical access to the web, no gui, and installed the fixed package transferred to it locally. (that system is also going to be dd if=/dev/zero'd)