SMP Linux - What to choose?

[Nothinman]

But I have seen that X, ion, vim and a lot other programs use less CPU time and less memery

X I might believe, but vim?

I can understand that you can't see the power in making a world if you use the programs as the come in binary packages

There's no additional power unless you have some local patches you apply first. If I want I can use 'apt-get source' or 'apt-build' to compile 99% of the programs on my box (I have some closed source things like VMWare and Acroread installed) but there's just no compelling reason to do it. The speed advantage is long moot and I have the addition of the Debian package maintainer's QA. He normally knows the packages much better than I do and I would rather he work out the major bugs before I even have a chance to install it and run into them myself.

 

[n0cmonkey]

Originally posted by: bromer

No, we dont. But Im an OpenBSD user so my definition is correct.

Ahh. I see.. well. very sorry then

Anyhow, I prefer building from source most of the time, but I get lazy, impatient, or I need something right away so I will install binaries on occassion. I dont notice a difference for most applications, but if you do, great. I know I dont need absolutely the best performance I can get, if I did I would be hacking at the source a lot more than I do.

 

[bromer]

X I might believe, but vim?

Yes... I can't quite figure that one out myself. I might think that I also upgraded it and that did some trics.

 

[bromer]

Anyhow, I prefer building from source most of the time, but I get lazy, impatient, or I need something right away so I will install binaries on occassion. I dont notice a difference for most applications, but if you do, great. I know I dont need absolutely the best performance I can get, if I did I would be hacking at the source a lot more than I do.

Okay.. well I have a Celeron 700MHz, so I want the most out of the programs.. and I like to hacke source, recompile and that stuff

 

[manly]

Premature optimization is the root of all evil.

God aka Donald Knuth

FWIW, recompiling EVERYTHING falls under "premature optimization" in my book.

 

[bromer]

Originally posted by: manly

Premature optimization is the root of all evil.

God aka Donald Knuth

FWIW, recompiling EVERYTHING falls under "premature optimization" in my book.

God dosn't make mistakes.. Knuth however does. And talking about recompiling everything without actualy having done it is just premature. There are lot of shotcuts when you make a new world, so you dosn't need to recompile ALL the sourcecode. I takes me about 10 hours to make a world. That's not enough time to recompile all programs, just the world

 

[cleverhandle]

Originally posted by: Nothinman
Specifying the CPU makes little different, 95% of the asm generated by GCC is compatible across all generations of a CPU. And I don't believe GCC supports SSE or SSE2 the main things that might help on x86 processors and even then I doubt the speed increase would be too great.

For the record, GCC 3.x does support mmx, sse, and 3dnow. Not sure about sse2 off the top of my head. But I agree with the general opinion that the speed boosts are not a big deal in most cases. A happy medium would be to recompile the few big programs like X and Mozilla, and leave the rest.

Of course, if you want to compile everything just for the sake of puttering, like I do, that's a different story...

 

[n0cmonkey]

Originally posted by: bromer

Originally posted by: manly

Premature optimization is the root of all evil.

God aka Donald Knuth

FWIW, recompiling EVERYTHING falls under "premature optimization" in my book.

God dosn't make mistakes.. Knuth however does. And talking about recompiling everything without actualy having done it is just premature. There are lot of shotcuts when you make a new world, so you dosn't need to recompile ALL the sourcecode. I takes me about 10 hours to make a world. That's not enough time to recompile all programs, just the world

Dude, its time for an upgrade. I do the OpenBSD equiv of a make world and it takes me no where near that time.

 

[n0cmonkey]

Originally posted by: cleverhandle

Originally posted by: Nothinman
Specifying the CPU makes little different, 95% of the asm generated by GCC is compatible across all generations of a CPU. And I don't believe GCC supports SSE or SSE2 the main things that might help on x86 processors and even then I doubt the speed increase would be too great.

For the record, GCC 3.x does support mmx, sse, and 3dnow. Not sure about sse2 off the top of my head. But I agree with the general opinion that the speed boosts are not a big deal in most cases. A happy medium would be to recompile the few big programs like X and Mozilla, and leave the rest.

Of course, if you want to compile everything just for the sake of puttering, like I do, that's a different story...

From what I have read on the OpenBSD mailing list, gcc 3 isnt quite ready. And some of the posts make it sound like it wont ever be ready...

 

[N11]

These threads are interesting.

 

[bromer]

Originally posted by: N11
These threads are interesting.

What do you mean by that?

 

[Bremen]

Perhaps because its easier finding out about gccs capabilities here than digging through the documentation (tried looking at gcc.gnu.org once for info on gcc, never again!)

 

[N11]

Originally posted by: bromer

Originally posted by: N11
These threads are interesting.

What do you mean by that?

I mean that any thread which attempts to ascertain some "greater" distribution turns into an interesting discussion where individuals side with their preferences and make elaborate arguments to support their views -- wherein someone can come read and get a better understanding of the pros and cons of certain things where they otherwise wouldn't have given most of these issues a second thought.

People are at their best when they are talking about something they believe in and work with. Even though I personally wouldn't use debian in any production environment I am affiliated with, I still find comments that people such as Nothinman make to be interesting reads because I don't use Debian, he does, he's the one with the experience so to disagree without experiencing or listening is stubborn on my or anyone else's behalf.

 

[n0cmonkey]

Originally posted by: N11

Originally posted by: bromer

Originally posted by: N11
These threads are interesting.

What do you mean by that?

I mean that any thread which attempts to ascertain some "greater" distribution turns into an interesting discussion where individuals side with their preferences and make elaborate arguments to support their views -- wherein someone can come read and get a better understanding of the pros and cons of certain things where they otherwise wouldn't have given most of these issues a second thought.

People are at their best when they are talking about something they believe in and work with. Even though I personally wouldn't use debian in any production environment I am affiliated with, I still find comments that people such as Nothinman make to be interesting reads because I don't use Debian, he does, he's the one with the experience so to disagree without experiencing or listening is stubborn on my or anyone else's behalf.

Does that mean I should start pimping OpenBSD more?

 

[geekassault]

I'm a debian and openbsd user myself, and having been running both of them for a while. To me debian still is the best. Only because it's so easy to use (after a while of course) and very easy to upgrade. My debian box is for every day use, so I really can't afford building world on a system and crashing half way, the next morning I do want to get some work done. So my OpenBSD isn't upgraded, except after the usual stable change every six months.
I like to hack a bit on OpenBSD, primitive ports and stuff like that. But for the debian box I do prefer to have someone on the other side of the binary that knows the upstream authors and has read through the code. Looks like common sense to me.
As for SMP .. is that allready a long time in FreeBSD ? Stable ?? Or is it OpenBSD that is still lacking support ??

 

[n0cmonkey]

Originally posted by: geekassault
I'm a debian and openbsd user myself, and having been running both of them for a while. To me debian still is the best. Only because it's so easy to use (after a while of course) and very easy to upgrade. My debian box is for every day use, so I really can't afford building world on a system and crashing half way, the next morning I do want to get some work done. So my OpenBSD isn't upgraded, except after the usual stable change every six months.
I like to hack a bit on OpenBSD, primitive ports and stuff like that. But for the debian box I do prefer to have someone on the other side of the binary that knows the upstream authors and has read through the code. Looks like common sense to me.
As for SMP .. is that allready a long time in FreeBSD ? Stable ?? Or is it OpenBSD that is still lacking support ??

You do apply OpenBSD patches right?

OpenBSD does not support SMP. FreeBSD 2.2(?)-4.whatever supports some early SMP, but 5.0's SMP is supposed to be on par with Linux 2.4 or so.

 

[Bremen]

Does that mean I should start pimping OpenBSD more?

I bought the OpenBSD 3.2 CDs because of you, so I guess you're doing something right :0)

 

[geekassault]

You do apply OpenBSD patches right?

Yeah, of course, after all, running openbsd means running a secure box, right

 

[MGMorden]

Boy this thread went off on a tangent. I'm a compile from source type guy myself, but only for key applications (KDE, Xfree86 . . . . other important stuff ) and things that I install myself after the fact. I'm not gonna sit and wait for anything to compile (I can start the process before going out or something, or if neccessary just use my Windows machine), so that time it takes to compile the software isn't important to me (any automated tasks the machine needs to perform are gonna be daemons and won't be affected by compiling). To each his own though. Some people would rather compile and some would rather download binaries. The important thing is that both groups can choose to do whichever we prefer.