adept is incredibly bloaty

[Andreas Mohr]

Hello,

the subject line could be considered to be rather inflammatory, but this is somewhat intentional given that a simply checkmark toggle (install/don't install application) in adept takes longer than 5 seconds (!) on a P3/450 with 256MB.
This is, ahem, friggin' insane. Please stop this bloat.
Startup of adept and general appearance is similarly bloaty.

This is on a very, very optimized setup (hdparm etc.) with an especially fast -ck kernel (swap prefetching, better scheduler, ...), so it's doubly painful to have adept still be that slow.

Its memory requirements are also far from modest:

# ps aux|grep adept
yun 4661 1.9 4.7 29844 12288 ? S 22:14 0:39 adept_notifier
root 7040 18.8 25.0 88612 64280 ? S 22:28 3:45 adept_installer -icon adept_installer -miniicon adept_installer -caption Add/Remove Programs

I'd suggest treating this with a fair dose of oprofile, gprof and/or kcachegrind to stop the worst excesses.

Despite all this criticism I want to add that I'm very happy that there's finally a rather simple program installer available that's usable by mere lazy mortals, too -- keep up this great work!

Thanks,

Andreas Mohr

[Phil Bull (philbull)]

Thanks for the report.

Could you provide some more information, please? For example, what version of adept are you using? Do you get this issue on any other systems? Do you have the same issues on a standard kernel? Has the application always been this slow?

[mornfall (mornfall)]

Patches welcome.

[Asraniel (asraniel)]

yes, adept is realy becoming slow for startup, sometimes im even thinking that it crashed because there is no progressbar and nothing.

[Phil Bull (philbull)]

Could you (or someone else running Kubuntu, I'm not) maybe run a suitable profiler over adept and find out what's taking the time? Is there any terminal output? What processor usage does adept show during load? How long do certain steps take?

It could be something simple, and a small patch would really speed things up. But we'll never know unless someone does some useful profiling...

BTW, this is very, very unlikely to get fixed for Dapper.

Thanks

[Andreas Mohr (andi)]

I would do a profile run, problem is that I stupidly forgot to fetch the power supply for this notebook, so it will take up to 2 weeks to get it back into working condition, and I don't have any way to test (install) it easily anywhere else (no DSL at my current home).

Possibly adept is doing relatively complicated package dependency checks (via a backend?) when toggling a checkmark? In that case a 6 second delay would be more understandable, although still far from "joyful".

Doing a profiling run is not too hard, just install oprofile (see e.g. linux/Documentation/basic-profiling.txt) and check events such as unhalted clock, cache misses, branch mispredictions.

Meanwhile I'll keep concentrating on kernel optimizations ;)
(I don't think *I* should be the one to delve into adept optimization, since I'm very far from hacking on it on a daily basis)

Oh, and if it doesn't get fixed for Dapper, then bad luck, but it'd be nice to see it fixed at all, and then if at all possible still in Dapper.

Anyway, adept seems to be a new and exciting app, so it's just normal that people will oncentrate on the basic things (read: bare functionality) first before actually making it truly usable.

[mornfall (mornfall)]

Oh come on. Of course i know where the problem is. The original reporter assumed i was too stupid to know where my performance problems are so i let him find out. When i get a semi-polite request i may do something about it. However, a fix will probably cost functionality and i am not sure it's worth it. Most people do not complain about this specific problem, so i don't think it's too pressing. It will be fixed in adept 3.0 for sure though.

[Phil Bull (philbull)]

Thanks for the reply, I'm leaving the bug as rejected.