How to benchmark gentoo?

Message

C1REX · Post by **C1REX** » Tue Feb 18, 2020 12:26 am

Sooo...

What tools should I use to compare my gentoo build vs other distros I'm testing?
Also how would you benchmark your own gentoo to test effects of changes in make.conf?

Tony0945 · Post by **Tony0945** » Tue Feb 18, 2020 1:07 am

phoronix-test-suite is in portage. Warning: It's a real slow build and takes a lot of disk space.
But it would give you a chance to compare your system against published benchmarks at www.phoronix.com

C1REX · Post by **C1REX** » Tue Feb 18, 2020 1:10 am

That's an amazing tip. Everything ready in portage :) Nice.

erm67 · Post by **erm67** » Tue Feb 18, 2020 6:54 am

C1REX wrote:That's an amazing tip. Everything ready in portage Nice.

Don't waste your time, there is no conspiracy like some fool thinks, without strong CFLAGS and disabling portage CFLAGS filter your regular gentoo system IS slower than clearLinux . Maybe benchmark some of the gentooLTO CFLAGS

C1REX · Post by **C1REX** » Tue Feb 18, 2020 8:05 am

If gentoo is slower than clear Linux then I want to know how much.
If it's faster than than Ubuntu and Arch I want to test see the difference.

If GentooLTO is any quicker - I want to see if it's worth it.

Muso · Post by **Muso** » Tue Feb 18, 2020 8:24 am

C1REX wrote:If gentoo is slower than clear Linux then I want to know how much.
If it's faster than than Ubuntu and Arch I want to test see the difference.

Is GentooLTO any quicker - I want to see if it's worth it.

For me, it's a matter of control. I like what portage provides and how I can tune my system exactly to my needs. Yes, it's fast... but even better, it's mine.

I'm not obsessed with pure speed, as I do like having the ability to debug as needed (so I no longer use the -fomit-frame-pointer flag). With modern hardware, you're looking at nano seconds (unless it's completely bloated crap *cough Ubuntu*). Distro choice, from my perspective, is a matter of which tools you need, and how comfortable you are with the package manager and architecture of the system.

Gentoo has a strange elegance that I love for desktop use. For forensics and pentesting, I prefer Kali... but Gentoo provides an excellent solution via stefantalpalaru's overlay (though I just locally mirror the ebuilds from his overlay instead of using the full thing). His vmplayer ebuild allows me to run a virtual Kali system.

Everything I need or want on a desktop system is there at my fingertips with Gentoo. Not so with other distros. While I can use SystemD, I don't care for it. I like OpenRC. I don't like GUI login managers, and use i3-gaps with a selected suite of tools I've found comfortable for my needs over the years.

The speed of Gentoo is nice, but it really shines in that it is a metadistro. You design your own system for your own needs... and can choose whether or not you want to take the poettering-pill.

fpemud · Post by **fpemud** » Tue Feb 18, 2020 12:18 pm

I think speed and control are equally important.
Why not set the throttle to its maximum when Gentoo has the ability?

But I'd like the optimization process be automated.
There should be a way to let users submit their optimization best practice, which are some kind of rules I think, perhaps hardware related.
Others choose and apply these rules to their system, by one command, or one button press.
This can even be part of the world update if extreme simplicity is pursued.

C1REX · Post by **C1REX** » Tue Feb 18, 2020 12:22 pm

If people start benchmarking we can start finding the best solutions and some flags might be recommended or even put in ebuilds.

This could help to move gentoo forward and for many people to have fun experimenting :)

Post by **NeddySeagoon** » Tue Feb 18, 2020 12:34 pm

C1REX,

There are lots of open source benchmarks around.
To get people to run them and post results, you only need to find or write ebuilds.

roylongbottom has a good selection which he has used to benchmark Raspberry Pis.
e.g. 64 Bit Raspberry Pi 4B Benchmarks

I can't find the post at the moment but one of his posts compares a Pi to a Cray supercomputer and the Pi wins. :)

-- edit --

I gave in. The best overall Geomean of 285 MFLOPS, for one core, is nearly 24 times faster than the Cray 1 supercomputer, that cost $7 million (in 1978).