assigning processes to 1 proc?

assigning processes to 1 proc?

Post by Tulse X. Lupe » Tue, 12 Aug 2003 21:21:04



Hello, I'm wondering if Linux allows assigning specific processes to a
particular processor in a dual system. Thanks in advance; email copies of
responses preferred.


           "who is in your mind now tell me who you are thinking of"
---------------------------- Tulse Luper: 92 pernambucating pygmy-owls in VUE.

 
 
 

assigning processes to 1 proc?

Post by Douglas Bollinge » Fri, 22 Aug 2003 07:00:03



> Hello, I'm wondering if Linux allows assigning specific processes to a
> particular processor in a dual system. Thanks in advance; email copies of
> responses preferred.

No, the 2.4.x kernel series does not support processor affinity for
specific processes, at least in the vanilla kernel source.  Supposedly this
feature was to be added to the 2.6.x series, but I haven't checked-up on
that lately to see if it made it in the source.

--
Who goeth a-borrowing goeth a-sorrowing.
                -- Thomas Tusser

 
 
 

1. PROPOSAL: /proc standards (was dot-proc interface [was: /proc

Just think about it for a minute.

There are three ways to address "/proc":
 - 100% binary interface
  * human interaction doesn't belong in the kernel - period.
 - optimally formated text
  * not designed for humans, but in human format ("text")
 - human readable
  * thus the entire OS is reduced to "cat" and "echo"

Providing more than one interface/format means code duplication.  It doesn't
matter how much is actually compiled.  Someone has to write it.  Others have
to maintain it.  Suddenly a change in one place becomes a change in dozens
of places.

Personally, I vote for a 100% binary interface. (no surprise there.)  It
makes things in kernel land so much cleaner, faster, and smaller.  Yes,
it increases the demands on userland to some degree.  However, printf/scanf
is one hell of a lot more wasteful than a simple mov.

For my worst case scenerio, answer this:
  How do you tell how many processors are in a Linux box?

The kernel already knows this, but it isn't exposed to userland.  So, one
must resort to ass-backward, stupid shit like counting entries in
/proc/cpuinfo.  And to make things even worse, the format is different for
every arch! (I've been bitching about this for four (4) years.)

And for those misguided people who think processing text is faster than
binary, you're idiots.  The values start out as binary, get converted to
text, copied to the user, and then converted back to binary.  How the hell
is that faster than copying the original binary value? (Answer: it isn't.)

And those who *will* complain that binary structures are hard to work with,
(you're idiots too :-)) a struct is far easier to deal with than text
processing, esp. for anyone who knows what they are doing.  Yes, changes
to the struct do tend to break applications, but the same thing happens
to text based inputs as well.  Perhaps some of you will remember the stink
that arose when the layout of /proc/meminfo changed (and broke, basically,
everything.)

--Ricky

-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in

More majordomo info at  http://vger.kernel.org/majordomo-info.html
Please read the FAQ at  http://www.tux.org/lkml/

2. Help! scsi timeout???

3. PROPOSAL: /proc standards (was dot-proc interface [was: /proc stuff])

4. Cost of Sun Disks

5. SCSI ID Assigned But No Drive Letters Assigned *** SOLVED ***

6. Getting wait output within a script

7. Assigning a process to a CPU

8. Powerstorm on XP1000

9. Assign login process for in.telnetd

10. Assigning a process to a CPU

11. How can I get the process name from proc[NPROC] ?

12. How do I get a (struct proc *) for the current user process (in NEXTSTEP 3.3)?

13. How to creat a sound process in the /proc directory?