https://make-linux-fast-again.com/
Since others may be curious, here is the discussion on those specific config changes: https://news.ycombinator.com/item?id=22830330
Spectre mitigations murder userspace performance
101–110 of 294 posts
Re: Spectre mitigations murder userspace performance
#102I would love to see new chips that take all of the transistors used for speculation and use them for more cores. I asked electrical engineers once how many of the transistors on a chip serve speculation, and they said, "To a first approximation, 100%." That means we could add a lot more cores. Add enough, and the OS could easily pin basically all processes/threads to their own cores. And that, I believe, might win so…
> Add enough, and the OS could easily pin basically all processes/threads to their own cores.
But usually (when you care about performance) you don't have 20 processes requiring 5% CPU load each, but 1-2 processes requiring everything. Making those slower won't accomplish much.
And if your problem is actually embarrassingly parallel [0], you are looking to run your code on a GPU (as another commenter notes).
Re: Spectre mitigations murder userspace performance
#103Earlier quoted context omitted.
“Close browser tab” - immediately exits a Google Meet. Closing a Zoom/Webex meeting, who knows since it’s still running in the background. I also like meetings sandboxed in a browser so weird things like “automatically take control of your screen and maximize window” doesn’t happen when someone in a Zoom meeting starts sharing their screen. Even at the expense of more CPU.
That meme with the actor taking to a bloodied Jesus comes to mind while reading you guys comparing google with zoom. You guys are so lucky. I work on Skype for Business over a Citrix Workspace connection. While Skype is an unmitigated disaster that can’t do simple stuff like copying text there is Citrix that requires a wizard installer with admin rights that deploys 3 background services and requires an audio plugin…
Since you are on Skype for Business I'm going to assume you are not using Teams currently. Teams is actually a lot worse in almost every way than SfB when it comes to the functions both systems share.
> While Skype is an unmitigated disaster that can’t do simple stuff like copying text
Do you mean from shared contents or from the chat? The latter works for me, but since you also mention using Citrix Workspace, which sounds like a remote desktop/application tool, it seems likely to me that this is actually the fault of Citrix, not Skype. Remote clipboards seem to be rather unreliable, I'm using DCV 2017 and the clipboard breaks basically every five minutes, necessitating a reconnect.
Re: Spectre mitigations murder userspace performance
#104Earlier quoted context omitted.
How long would you have to run JavaScript on your desktop to leak sensitive information and has there been any known exploits in the wild ?
https://leaky.page/ I don't know about the wild, but if this were tuned (i.e. this requires a lot of work for the first byte, the rest are easy) for a HVT you wouldn't know.
The problem is that PoC is extracting data which the PoC itself created specifically to facilitate said extraction. AFAIK no one has created a PoC which can extract specific data which hasn't been constructed to facilitate the PoC.
Re: Spectre mitigations murder userspace performance
#105I would love to see new chips that take all of the transistors used for speculation and use them for more cores. I asked electrical engineers once how many of the transistors on a chip serve speculation, and they said, "To a first approximation, 100%." That means we could add a lot more cores. Add enough, and the OS could easily pin basically all processes/threads to their own cores. And that, I believe, might win so…
> Add enough, and the OS could easily pin basically all processes/threads to their own cores. But usually (when you care about performance) you don't have 20 processes requiring 5% CPU load each, but 1-2 processes requiring everything. Making those slower won't accomplish much. And if your problem is actually embarrassingly parallel [0], you are looking to run your code on a GPU (as another commenter notes). [0]: htt…
I think you need to read my full blog post. :)
Re: Spectre mitigations murder userspace performance
#106`mitigations=off` in your GRUB_CMDLINE_LINUX_DEFAULT to disable Spectre/Meltdown mitigations, in case anyone is wondering. This page has a pretty decent write-up on it: https://leochavez.org/index.php/2020/11/16/disabling-intel-a... .
has there been any research on what hardware changes would be necessary to reclaim performance? I've noticed extreme stutter events on desktop OSX for the last 2 years. Would be curious if its related.
Seems like there’s a lot of people with stuttering macOS https://apple.stackexchange.com/questions/245429/cursor-free...
USB-C and USB 3.0 devices are one common cause, Bluetooth mouse interference by USB another, and external monitors after sleep/wake another.
Re: Spectre mitigations murder userspace performance
#107Earlier quoted context omitted.
I don't want to go into this topic because frontend developers are very defensive of their capability to Javascript on people, they often cite javascript adoption numbers as proof that it's fine to make it mandatory or cite complex web applications as a reason for it to be mandatory for all sites, which I personally find to be a false dichotomy. I tend to agree with what you're saying but the ship has sailed very muc…
Entirely? Use NoScript to temporarily enable only those portions of the single site that you need. No affiliation, just a happy user.
I use a browser called Qutebrowser which doesn't have a noscript addon; but I can disable javascript loading on a domain level.
However, overall I can tell you for absolute certain: if you have JS partially disabled things break in non-obvious ways and I find myself playing whack-a-mole with allowing various domains to load javascript to get the page working.
I'm pretty certain you do also, because it's basically impossible to tell why certain damned sites are broken and the most obvious thing to do is just enable JS temporarily to see if it works at all.
This is especially annoying on some part of a site such as checkout- where reloading the page causes a form resubmission.
Re: Spectre mitigations murder userspace performance
#108Earlier quoted context omitted.
I love that this is a toggle like this, having control of my system is why I love Linux. But I must caution desktop users against doing this for performance, it's _much_ better to have some kind of build server somewhere else with this kernel flag than to run it on your desktop. Why? because your desktop executes untrusted and rather arbitrary code pretty often, not just in the form of Javascript but that's the large…
> I love that this is a toggle like this, having control of my system is why I love Linux. You can disable mitigations on Windows too.
Yeah but the next update will probably turn it back on silently...
Re: Spectre mitigations murder userspace performance
#109Who is ultimately at risk because of Spectre and Meltdown? What additional risks am I taking on by disabling mitigations?
Those exploits enable any program to read essentially all memory.
This means nothing you did between turning on your computer and running an application with potentially malicious code (e.g. looking at a website with javascript enabled) is secret.
Re: Spectre mitigations murder userspace performance
#110… in the presence of frequent syscalls.
This crucial qualification in the original title is missing from the HN title.