=== himcesjf_ is now known as him-cesjf | ||
=== Elimin8r is now known as Elimin8er | ||
alkisg | Hi, https://wiki.ubuntu.com/Security/Features#Built_as_PIE says that it has 5-10% performance penalty | 09:00 |
---|---|---|
alkisg | Does that mean that if I run `echo 0 > /proc/sys/kernel/randomize_va_space` on a pentium 4, it'll run Ubuntu 16.04 faster? | 09:00 |
alkisg | Or compiling with PIE enabled will result in performance loss anyway? | 09:05 |
apw | compiling as PIE will alter the emitted code | 10:21 |
alkisg | apw: I mean, does PIE mean a syscall that can be disabled in the kernel at runtime, avoiding the extra overhead, or is it a compiler time thing only, which cannot be disabled after compilation? | 10:22 |
apw | pie is a compilation option which changes the code to not assume its location, which has a cost associated with it | 10:23 |
alkisg | apw, thank you, so the only way to measure that overhead would be to recompile, which isn't really an option if one wants to use a recent Ubuntu version :) | 10:24 |
apw | right, you'd need to turn it off and rebuild the kernel | 10:24 |
alkisg | The kernel or the apps? | 10:24 |
alkisg | I thought that was about the apps... | 10:25 |
apw | both indeed, depends what you are trying to measure | 10:25 |
alkisg | Well in general how much slower a "desktop feels", maybe that would be 80% userspace and 20% kernel... | 10:25 |
apw | hard to say indeed, and humans are immensly sensitive to that kind of thing | 10:26 |
alkisg | E.g. a few years ago it was possible to watch youtube with a p4 1.7 ghz, now it's not even with p4 at 2.4 ghz | 10:26 |
apw | a few % and you can tell often | 10:26 |
apw | are the videos at the same resolution i wonder | 10:26 |
alkisg | Yes, but I'm not sure about the players | 10:26 |
apw | but it doesn't supprise me entierly, old h/w does get left behind over time | 10:27 |
apw | i wonder how old a pentium-4 mbased mchine could be at its youngest now | 10:27 |
alkisg | I used the same video back then and now, and tried with flash now as well as then when html5 didn't exist, but I don't know if flash or youtube added unrelated code that made them slower | 10:27 |
apw | browsers are more tightly controlled now days too | 10:29 |
apw | and video playback is one of those things where your computer is either fast enough or not | 10:29 |
alkisg | And there's compositing even in the legacy desktops, adding to delays... https too... | 10:29 |
alkisg | Ah, playing with vlc is fast now too | 10:29 |
apw | it might have been 1% faster than needed before and be 2% slower now and ... you lose | 10:29 |
alkisg | The browsers make it like 4 times slower | 10:30 |
apw | so it is ok in vlc ? | 10:30 |
alkisg | Yeah | 10:30 |
apw | so not inherantly unplayable then | 10:30 |
alkisg | Ubuntu 16.04 i386 is still fast enough to play an e.g. 1024x768 full screen movie without dropped frames | 10:30 |
alkisg | Indeed, but there's no way to convince youtube/firefox etc etc to make them more efficient for older machines, so I was looking to gain that 10% performance in case it makes things barely more functional | 10:31 |
alkisg | apw: so with regards to aslr/pie, we're not expecting any performance differences between e.g. 12.04 and 18.04, since all "major" programs like the kernel or firefox were already using them, correct? | 10:33 |
apw | that'd be my feeling | 10:33 |
alkisg | Thank you, I was worried that I would have to hold those machines with 16.04 | 10:34 |
alkisg | Let them work with 18.04 as long as they have a bit of life in them | 10:34 |
apw | alkisg, as with all things, testing it out is a good plan | 10:40 |
alkisg | Indeed... hmm I'll actually do a larger comparison, 10.04/12.04/18.04 live cds, and watching the same youtube video... | 10:41 |
apw | sounds like "fun" | 10:42 |
=== Dmitrii-Sh is now known as Dmitrii-Sh-PTO | ||
=== _ruben_ is now known as _ruben |
Generated by irclog2html.py 2.7 by Marius Gedminas - find it at mg.pov.lt!