Wanting to Try A BSD variant- Which is Best For Me? Slacker, Debian, Arch Veteran

sevendogsbsd · 05-22-2019, 01:57 PM

I agree, I am talking about extreme cases where the CPU is pegged out or the whole system is just sluggish because of ram use. I haven't noticed any difference between performance on Linux compared to FreeBSD but my use is very different on FreeBSD: I do not use a DE at all now and used to use either Gnome or KDE on Linux. KDE was good with performance on Linux, Gnome was horrible, both running on a machine with essentially server sized hardware.

I just meant that to me bloat is referring to performance and not size. Just my opinion though.

cynwulf · 05-23-2019, 03:28 AM

Bloat is indeed a performance issue, but the definition of bloat does not simply equate to using more RAM or CPU.

A quick ddg for "code bloat" revealed some fairly useful results, but the wikipedia article gets straight to the point and sums it up in a paragraph: https://en.wikipedia.org/wiki/Code_bloat

Quote:

In computer programming, code bloat is the production of program code (source code or machine code) that is perceived as unnecessarily long, slow, or otherwise wasteful of resources. Code bloat can be caused by inadequacies in the programming language in which the code is written, the compiler used to compile it, or the programmer writing it. Thus, while code bloat generally refers to source code size (as produced by the programmer), it can be used to refer instead to the generated code size or even the binary file size.

A quick read up on the OpenBSD project will show that it's goals are pretty much at odds with that. More specifically, any submitted "bloated" code is very unlikely to be merged.

cynwulf · 05-23-2019, 03:43 AM

Quote:

Originally Posted by hitest

I guess. OpenBSD uses more RAM to run XFCE than Linux does, but, typically the BSDs use more RAM than Linux.

I've not encountered this - truth be told, never really thought about it, do you have any actual data to support this?

For example, have you looked at RAM usage on e.g. OpenBSD, NetBSD, FreeBSD compared to a Linux system with similar configuration and the Xfce desktop?

jggimi · 05-23-2019, 04:55 AM

Quote:

Originally Posted by cynwulf

A quick read up on the OpenBSD project will show that it's goals are pretty much at odds with that.

Except this week.

The amdgpu drm code just added to -current (but not yet in the build) adds 1,543,049 lines of code. The src/sys/dev/pci/drm directory grew from 16M to 126M.

Source: conversation on Mastodon between @stsp and @brynet.

Commit log entry:

Code:

----------------------------
revision 1.1
date: 2019/05/21 22:40:06;  author: jsg;  state: Exp;  commitid: bbJhylodHLWDncc6;
add amdgpu from linux 4.19.44 for recent AMD Radeon parts

committing now so this can be worked on in tree

Thanks to the OpenBSD Foundation for sponsoring this work and kettenis@
for helping.
=============================================================================

cynwulf · 05-23-2019, 05:47 AM

Heh... that's one of the hazards of porting Linux bloat code - and what is effectively a vendor driver focused primarily on 3D rendering / acceleration - my understanding of the ported Linux drm code is that it was the best solution when compared with the old and deprecated drivers in userspace solution.

hitest · 05-23-2019, 09:46 AM

Quote:

Originally Posted by cynwulf

For example, have you looked at RAM usage on e.g. OpenBSD, NetBSD, FreeBSD compared to a Linux system with similar configuration and the Xfce desktop?

I have my experience and a bit of data. At the moment I'm dual booting Slackware64-current and OpenBSD 6.5 on my Dell Optiplex 990. The unit has an i7 CPU @ 3.4 GHz with 8 GB RAM. Both installations are running XFCE. Here are two screenshots taken of top after the unit boots up in OpenBSD and in Slackware. There are some differences in memory usage with XFCE for OpenBSD and Slackware.

cynwulf · 05-23-2019, 11:25 AM

Due to the differences between GNU top(8) and OpenBSD's top(8), a comparison between ps(1) output for both OS might be more useful, complete and readable but will still be an apples and oranges comparison.

You have to consider a different compiler (llvm/clang) and differences in memory management, etc.

Many ports are compiled with pledge(2) and unveil(2) - I don't know enough about these to know if they - or similar - add any "overhead"?

jggimi · 05-23-2019, 11:58 AM

Quote:

Originally Posted by cynwulf

Many ports are compiled with pledge(2) and unveil(2) - I don't know enough about these to know if they - or similar - add any "overhead"?

Excepting Chromium, which added a module to support --enable-unveil and --disable-unveil, the additional code is simply making syscalls at the appropriate points in program flow. As few as one syscall, sometimes several more, depending on application complexity. I added 9 pledge(2) syscalls to archivers/p7zip, when I was its maintainer. It's quite complex ... but the pledges are not.

At this time, there are 94 pledged ports out of 8968 in the tree, which is just over 1%.