Tech Insider					   Technology and Trends


			   USENET Archives


Electronic mail:			      WorldWideWeb:
   tech-insider@outlook.com		         http://tech-insider.org/

From: rgo...@atnf.csiro.au (Richard Gooch)
Subject: Interesting scheduling times
Date: 1998/09/16
Message-ID: <199809161443.AAA00304@workaholix.atnf.CSIRO.AU>
X-Deja-AN: 391751418
Approved: g...@greenie.muc.de
Sender: muc.de!l-linux-kernel-owner
Newsgroups: muc.lists.linux-kernel

  Hi, all. I've been playing around with measuring Linux context
switch times, and I noticed something curious: a Pentium/MMX 200 is
doing much better than a PPro 180. Furthermore, a PPro 180 isn't doing
heaps better than a Pentium 100.

CPU			process switch	thread switch	Kernel version
Pentium 100		12		12		2.1.109
PPro 180		8		4		2.1.122-pre2
Pentium/MMX 200		4		2		2.1.104

all times in microseconds for UP machines.

Do these times seem a little odd to people?
FYI: I've appended my testcode.

				Regards,

					Richard....
===============================================================================
Code

-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majord...@vger.rutgers.edu
Please read the FAQ at http://www.tux.org/lkml/

From: ander...@inconnect.com (Erik Andersen)
Subject: Re: Interesting scheduling times
Date: 1998/09/16
Message-ID: <Pine.GSO.4.02A.9809161407400.25101-100000@ultra1>#1/1
X-Deja-AN: 391891454
Approved: g...@greenie.muc.de
Sender: muc.de!l-linux-kernel-owner
References: <199809161443.AAA00304@workaholix.atnf.CSIRO.AU>
Newsgroups: muc.lists.linux-kernel

On Thu, 17 Sep 1998, Richard Gooch wrote:

>   Hi, all. I've been playing around with measuring Linux context
> switch times, and I noticed something curious: a Pentium/MMX 200 is
> doing much better than a PPro 180. Furthermore, a PPro 180 isn't doing
> heaps better than a Pentium 100.
> 
> CPU			process switch	thread switch	Kernel version
> Pentium 100		12		12		2.1.109
> PPro 180		8		4		2.1.122-pre2
> Pentium/MMX 200	4		2		2.1.104
> 
> all times in microseconds for UP machines.
> 
> Do these times seem a little odd to people?

Maybe you should run them all using the same kernel, so we see an
"apples-to-apples" comparison.  I lot has happened during the last
20 kernel releases...

 -Erik

--
Erik B. Andersen   Web:    http://www.inconnect.com/~andersen/ 
                   email:  ander...@debian.org
--This message was written using 73% post-consumer electrons--



-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majord...@vger.rutgers.edu
Please read the FAQ at http://www.tux.org/lkml/

From: rgo...@atnf.csiro.au (Richard Gooch)
Subject: Re: Interesting scheduling times
Date: 1998/09/16
Message-ID: <199809162351.JAA17376@vindaloo.atnf.CSIRO.AU>#1/1
X-Deja-AN: 391915302
Approved: g...@greenie.muc.de
Sender: muc.de!l-linux-kernel-owner
References: <Pine.GSO.4.02A.9809161407400.25101-100000@ultra1>
Newsgroups: muc.lists.linux-kernel

Erik Andersen writes:
> On Thu, 17 Sep 1998, Richard Gooch wrote:
> 
> >   Hi, all. I've been playing around with measuring Linux context
> > switch times, and I noticed something curious: a Pentium/MMX 200 is
> > doing much better than a PPro 180. Furthermore, a PPro 180 isn't doing
> > heaps better than a Pentium 100.
> > 
> > all times in microseconds for UP machines.
> > 
> > Do these times seem a little odd to people?
> 
> Maybe you should run them all using the same kernel, so we see an
> "apples-to-apples" comparison.  I lot has happened during the last
> 20 kernel releases...

OK:

CPU			process switch	thread switch	Kernel version
Pentium 100		12		12		2.1.109
Pentium 100		19		9		2.1.122-pre2
PPro 180		8		4		2.1.122-pre2
Pentium/MMX 200		6		2		2.1.122-pre2
Pentium/MMX 200		4		2		2.1.104

Even more interesting: 2.1.122-pre2 is slower for process switching
and is faster for thread switching than 2.1.104/2.1.109.

				Regards,

					Richard....

-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majord...@vger.rutgers.edu
Please read the FAQ at http://www.tux.org/lkml/

From: lin...@z.ml.org (Gregory Maxwell)
Subject: Re: Interesting scheduling times
Date: 1998/09/17
Message-ID: <Pine.LNX.3.96.980917000946.21056A-100000@z.ml.org>#1/1
X-Deja-AN: 391951245
Approved: g...@greenie.muc.de
Sender: muc.de!l-linux-kernel-owner
References: <199809162351.JAA17376@vindaloo.atnf.CSIRO.AU>
Newsgroups: muc.lists.linux-kernel

On Thu, 17 Sep 1998, Richard Gooch wrote:

[snip]
> 
> OK:
> 
> CPU			process switch	thread switch	Kernel version
> Pentium 100		12		12		2.1.109
> Pentium 100		19		9		2.1.122-pre2
L1 is 2 way. 16k

> PPro 180		8		4		2.1.122-pre2
L1 is 2way/4way 16k

> Pentium/MMX 200		6		2		2.1.122-pre2
L1 is 4way 32k
> Pentium/MMX 200		4		2		2.1.104
[snip]

I'm sure if you test a PII it will be alot more like the PMMX.


-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majord...@vger.rutgers.edu
Please read the FAQ at http://www.tux.org/lkml/

From: rgo...@atnf.csiro.au (Richard Gooch)
Subject: Re: Interesting scheduling times
Date: 1998/09/17
Message-ID: <199809170431.OAA20382@vindaloo.atnf.CSIRO.AU>#1/1
X-Deja-AN: 391947277
Approved: g...@greenie.muc.de
Sender: muc.de!l-linux-kernel-owner
References: <Pine.LNX.3.96.980917000946.21056A-100000@z.ml.org>
Newsgroups: muc.lists.linux-kernel

Gregory Maxwell writes:
> On Thu, 17 Sep 1998, Richard Gooch wrote:
> 
> [snip]
> > 
> > OK:
> > 
> > CPU			process switch	thread switch	Kernel version
> > Pentium 100		12		12		2.1.109
> > Pentium 100		19		9		2.1.122-pre2
> L1 is 2 way. 16k
> 
> > PPro 180		8		4		2.1.122-pre2
> L1 is 2way/4way 16k
> 
> > Pentium/MMX 200		6		2		2.1.122-pre2
> L1 is 4way 32k
> > Pentium/MMX 200		4		2		2.1.104
> [snip]

Yeah, I know about the better caches. I'm just surprised at the
differences.
More importantly, I'm surprised by the slowdown from 2.1.104 to
2.1.122-pre2. 

				Regards,

					Richard....

-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majord...@vger.rutgers.edu
Please read the FAQ at http://www.tux.org/lkml/

From: pet...@varel.bg (Petko Manolov)
Subject: Re: Interesting scheduling times
Date: 1998/09/17
Message-ID: <36011D0C.1028BAF8@varel.bg>#1/1
X-Deja-AN: 392021034
Approved: g...@greenie.muc.de
Sender: muc.de!l-linux-kernel-owner
References: <199809161443.AAA00304@workaholix.atnf.CSIRO.AU>
Organization: Varel Ltd.
Newsgroups: muc.lists.linux-kernel

Richard Gooch wrote:
> 
> CPU                     process switch  thread switch   Kernel version
> Pentium 100             12              12              2.1.109
> PPro 180                8               4               2.1.122-pre2
> Pentium/MMX 200         4               2               2.1.104

I did a little different test on my pentium 166 MMX:

kernel 2.0.35		5us
kernel 2.1.121		6us
kernel 2.1.122		6us - 7us
kernel 2.1.122		9us (under X Windows???)

Petkan

-- 
Petko Manolov - pet...@varel.bg
http://www.varel.bg/~petkan

-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majord...@vger.rutgers.edu
Please read the FAQ at http://www.tux.org/lkml/

From: rgo...@atnf.csiro.au (Richard Gooch)
Subject: Re: Interesting scheduling times
Date: 1998/09/17
Message-ID: <199809171358.XAA25476@vindaloo.atnf.CSIRO.AU>#1/1
X-Deja-AN: 392071993
Approved: g...@greenie.muc.de
Sender: muc.de!l-linux-kernel-owner
References: <36011D0C.1028BAF8@varel.bg>
Newsgroups: muc.lists.linux-kernel

Petko Manolov writes:
> Richard Gooch wrote:
> > 
> > CPU                     process switch  thread switch   Kernel version
> > Pentium 100             12              12              2.1.109
> > PPro 180                8               4               2.1.122-pre2
> > Pentium/MMX 200         4               2               2.1.104
> 
> I did a little different test on my pentium 166 MMX:
> 
> kernel 2.0.35		5us
> kernel 2.1.121		6us
> kernel 2.1.122		6us - 7us
> kernel 2.1.122		9us (under X Windows???)

Did you run it as root? If you don't, the programme competes with X
for time. Running it as root allows it to grab all the CPU it wants,
and thus gives a more accurate test.

				Regards,

					Richard....

-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majord...@vger.rutgers.edu
Please read the FAQ at http://www.tux.org/lkml/

From: pet...@varel.bg (Petko Manolov)
Subject: Re: Interesting scheduling times
Date: 1998/09/17
Message-ID: <3601441B.25E31CA5@varel.bg>#1/1
X-Deja-AN: 392116202
Approved: g...@greenie.muc.de
Sender: muc.de!l-linux-kernel-owner
References: <199809161443.AAA00304@workaholix.atnf.CSIRO.AU> 
Organization: Varel Ltd.
Newsgroups: muc.lists.linux-kernel

Richard Gooch wrote:
> 
> Petko Manolov writes:
> > I did a little different test on my pentium 166 MMX:
> >
> > kernel 2.0.35         5us
> > kernel 2.1.121                6us
> > kernel 2.1.122                6us - 7us
> > kernel 2.1.122                9us (under X Windows???)
> 
> Did you run it as root? If you don't, the programme competes with X
> for time. Running it as root allows it to grab all the CPU it wants,
> and thus gives a more accurate test.

Yes, i was root during all tests.
But i thought that software switch is faster than
hardware. Other point is if this times can be correctly
computed from user level?

	Petkan

-- 
Petko Manolov - pet...@varel.bg
http://www.varel.bg/~petkan

-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majord...@vger.rutgers.edu
Please read the FAQ at http://www.tux.org/lkml/

From: torva...@transmeta.com (Linus Torvalds)
Subject: Re: Interesting scheduling times
Date: 1998/09/17
Message-ID: <6trfc3$io7$1@palladium.transmeta.com>#1/1
X-Deja-AN: 392174868
Approved: g...@greenie.muc.de
Sender: muc.de!l-linux-kernel-owner
References: <199809162351.JAA17376@vindaloo.atnf.CSIRO.AU> 
Organization: Transmeta Corporation, Santa Clara, CA
Newsgroups: muc.lists.linux-kernel

In article <199809170431.OAA20...@vindaloo.atnf.CSIRO.AU>,
Richard Gooch  <rgo...@atnf.csiro.au> wrote:
>
>More importantly, I'm surprised by the slowdown from 2.1.104 to
>2.1.122-pre2. 

What happened is that the very latest 2.1.x kernels are using a software
context switch, rather than just a jump through a task-gate and using
the hardware context switch code. 

Intel documents hardware context switching as slow, and various people
have at times complained to be about using it.  They can now see _why_ I
used it. 

The reason I had to make the context switch be done in software rather
than hardware was that I couldn't fix an oops any other way.  What
happens is that in order to streamline various other parts, I really
cannot guarantee that all user-level segment registers are always
up-to-date - when processes like Wine and DOSEMU change the LDT, the
process context may no longer be valid in another thread, and with the
hardware context switching I could force an oops in the context switch
that I had no way to recover from. 

In contrast, with the slightly slower software context switch I can
recover gracefully from bad segments. 

Oh, well.  2.0.x had this same problem too, but it's harder to trigger
because threads don't share the LDT in 2.0.x.  In addition, in 2.0.x the
context switch isn't protected by a spinlock, so the oops was less
damaging: in 2.1.x the oops would result in a completely dead system due
to the scheduler spinlock being held stale. 

The new software context switch routine might be slightly optimizable,
so we may be able to make it faster, but on the whole it's always easier
to make buggy code faster than correct code.  And while I am a
performance freak, fixing bugs takes precedence.. 

Oh, another change is that in the later kernels the scheduler also
correctly does the "tq_scheduler" bottom half processing, which earlier
2.1.x kernels didn't do because I couldn't handle the kernel lock
correctly.  That may or may not make a difference. 

		Linus

-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majord...@vger.rutgers.edu
Please read the FAQ at http://www.tux.org/lkml/

From: torva...@transmeta.com (Linus Torvalds)
Subject: Re: Interesting scheduling times
Date: 1998/09/18
Message-ID: <6ts8hp$uim$1@palladium.transmeta.com>#1/1
X-Deja-AN: 392305442
Approved: g...@greenie.muc.de
Sender: muc.de!l-linux-kernel-owner
References: <199809162351.JAA17376@vindaloo.atnf.CSIRO.AU> 
Organization: Transmeta Corporation, Santa Clara, CA
Newsgroups: muc.lists.linux-kernel

In article <6trfc3$io...@palladium.transmeta.com>,
Linus Torvalds <torva...@transmeta.com> wrote:
>In article <199809170431.OAA20...@vindaloo.atnf.CSIRO.AU>,
>Richard Gooch  <rgo...@atnf.csiro.au> wrote:
>>
>>More importantly, I'm surprised by the slowdown from 2.1.104 to
>>2.1.122-pre2. 
>
>What happened is that the very latest 2.1.x kernels are using a software
>context switch, rather than just a jump through a task-gate and using
>the hardware context switch code. 

Happily I checked some more, just to make sure, and profiled the kernel. 

Yes, it happened during the switch-over to a software task-switch
routine, but the reason the numbers got worse for some people is simply
that I screwed up the floating point save code, and it saved _every_
time through instead of doing the lazy save it was meant to do. 

The effect of this is not huge, but it's certainly noticeable.  With
that fixed, the software context switch is pretty comparable to the
hardware one. 

2.1.123 should have this context switch slowdown fixed.

		Linus

-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majord...@vger.rutgers.edu
Please read the FAQ at http://www.tux.org/lkml/

From: rgo...@atnf.csiro.au (Richard Gooch)
Subject: Re: Interesting scheduling times
Date: 1998/09/18
Message-ID: <199809180023.KAA29236@vindaloo.atnf.CSIRO.AU>#1/1
X-Deja-AN: 392305441
Approved: g...@greenie.muc.de
Sender: muc.de!l-linux-kernel-owner
References: <3601441B.25E31CA5@varel.bg>
Newsgroups: muc.lists.linux-kernel

Petko Manolov writes:
> Richard Gooch wrote:
> > 
> > Petko Manolov writes:
> > > I did a little different test on my pentium 166 MMX:
> > >
> > > kernel 2.0.35         5us
> > > kernel 2.1.121                6us
> > > kernel 2.1.122                6us - 7us
> > > kernel 2.1.122                9us (under X Windows???)
> > 
> > Did you run it as root? If you don't, the programme competes with X
> > for time. Running it as root allows it to grab all the CPU it wants,
> > and thus gives a more accurate test.
> 
> Yes, i was root during all tests.

OK, I think what this means is that under X you had more processes on
the run queue. Since the scheduler has to scan the run queue, the
longer it is the more time the scheduler takes.
You can see the effect of this by using the <num_running> option: this
will add the specified number of (low priority) processes to the run
queue. Increasing this number will slow the scheduler down.

In fact, this indicates that Linux only maintains a single run
queue. It could be argued that for RT performance, we'd be better off
with two run queues: one for SCHED_OTHER and the other for RT
scheduling classes. This way scheduling times for RT processes would
not be affected by the number of normal processes on the run
queue. However, it could be counter-argued that there are not usually
many processes on the run queue anyway. This is likely true for
embedded systems, but isn't necessarily the case for a large system
controlling an instrument as well as supporting users [1].

Another possible advantage of separating the run queue would be that
it would then become cheap to *always* call schedule() (or perhaps
schedule_rt()) upon return from interrupt/syscall. In most cases there
won't be any RT processes on the run queue, so this would only add a
few cycles overhead to the scheduler.
The advantage of doing this is that RT processes which are woken up
by a driver will preempt whatever is currently running, and hence will
actually start running immediately, rather than waiting N
jiffies. This would make Linux more RT-friendly.

Linus: what do you think of this idea? A valid project for 2.3?
I have to say I'm impressed with the soft-RT performance of Linux. In
my view the main limitation is the jiffies delay between when an RT
process is unblocked and when it starts running.

> But i thought that software switch is faster than hardware.

Not always. See Linus' message about this.

> Other point is if this times can be correctly computed from user
> level?

We have a pretty accurate gettimeofday(2) syscall, which makes doing
this easy. Particularly on the Pentium and above, where we use the TSC
for greater accuracy. I'm not sure how good it is on the 486 and
lower. In any case, in my code I'm careful to amortise the cost of
getting the time and reducing the inaccuracies of the timestamp by
running through the scheduler in a loop.

				Regards,

					Richard....

-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majord...@vger.rutgers.edu
Please read the FAQ at http://www.tux.org/lkml/

From: torva...@transmeta.com (Linus Torvalds)
Subject: Re: Interesting scheduling times
Date: 1998/09/18
Message-ID: <Pine.LNX.3.96.980917174717.495A-100000@penguin.transmeta.com>#1/1
X-Deja-AN: 392309321
Approved: g...@greenie.muc.de
Sender: muc.de!l-linux-kernel-owner
References: <199809180023.KAA29236@vindaloo.atnf.CSIRO.AU>
Reply-To: Linus Torvalds <torva...@transmeta.com>
Newsgroups: muc.lists.linux-kernel



On Fri, 18 Sep 1998, Richard Gooch wrote:
> 
> Linus: what do you think of this idea? A valid project for 2.3?
> I have to say I'm impressed with the soft-RT performance of Linux. In
> my view the main limitation is the jiffies delay between when an RT
> process is unblocked and when it starts running.

A single run-queue is almost always better than multiple run-queues, and
I'm very unlikely to change that.

The reason for a single run-queue is that it's about 10 times simpler than
any of the alternatives, and it's never slower in real life. Yes, we may
end up walking a few more entries, but the simplicity more than pays back
the cost of that walk.

Even under heavy load, the runqueue is seldom more than a few entries
deep.  More than 10 entries on the run-queue is already very rare, and
when it does happen the scheduling overhead is very small compared to what
else the machine is doing: having that many entries implies that the
scheduler isn't your biggest bottle-neck anyway.

That said, the idea of just having two run-queues, one with real-time
processes and one without is so far the best multi-runqueue idea I've
heard. So yes, I could imagine doing something like that, but I still
don't actually believe that the run-queue is the major bottle-neck. 

		Linus

PS. Here's the patch to make 2.1.122 perform as it should wrt scheduling,
and not save the FP register state all the time. Embarrassing.

diff -u --recursive --new-file v2.1.122/linux/arch/i386/kernel/process.c linux/arch/i386/kernel/process.c
--- v2.1.122/linux/arch/i386/kernel/process.c	Thu Sep 17 17:53:34 1998
+++ linux/arch/i386/kernel/process.c	Thu Sep 17 17:41:51 1998
@@ -540,10 +540,10 @@
 static inline void unlazy_fpu(struct task_struct *tsk)
 {
 	if (tsk->flags & PF_USEDFPU) {
-		tsk->flags &= ~PF_USEDFPU;
 		__asm__("fnsave %0":"=m" (tsk->tss.i387));
-		stts();
 		asm volatile("fwait");
+		tsk->flags &= ~PF_USEDFPU;
+		stts();
 	}
 }
 



-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majord...@vger.rutgers.edu
Please read the FAQ at http://www.tux.org/lkml/

From: pet...@varel.bg (Petko Manolov)
Subject: Re: Interesting scheduling times
Date: 1998/09/18
Message-ID: <36023855.4BE148E8@varel.bg>#1/1
X-Deja-AN: 392359527
Approved: g...@greenie.muc.de
Sender: muc.de!l-linux-kernel-owner
References: <Pine.LNX.3.96.980917174717.495A-100000@penguin.transmeta.com>
Organization: Varel Ltd.
Newsgroups: muc.lists.linux-kernel

Linus Torvalds wrote:
> 
> PS. Here's the patch to make 2.1.122 perform as it should wrt scheduling,
> and not save the FP register state all the time. Embarrassing.

OK, i've applied your patch and got strange results:
first time i run Richards prog i got 4us. All next 
values were 6us. I reboot the machine and the result
was the same!
I'm afraid i don't understand what exactly the patch
does. At first glance only the sequence of lines is
changed - so what? Or there is some relatione between
"fnsave" and setting task switch bit?

	Petkan

P.S.	I'll try to do some profile in kernel mode

-- 
Petko Manolov - pet...@varel.bg
http://www.varel.bg/~petkan

-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majord...@vger.rutgers.edu
Please read the FAQ at http://www.tux.org/lkml/

From: rgo...@atnf.csiro.au (Richard Gooch)
Subject: Re: Interesting scheduling times
Date: 1998/09/18
Message-ID: <199809180751.RAA02219@vindaloo.atnf.CSIRO.AU>#1/1
X-Deja-AN: 392352642
Approved: g...@greenie.muc.de
Sender: muc.de!l-linux-kernel-owner
References: <36023855.4BE148E8@varel.bg>
Newsgroups: muc.lists.linux-kernel

Petko Manolov writes:
> Linus Torvalds wrote:
> > 
> > PS. Here's the patch to make 2.1.122 perform as it should wrt scheduling,
> > and not save the FP register state all the time. Embarrassing.
> 
> OK, i've applied your patch and got strange results:
> first time i run Richards prog i got 4us. All next 
> values were 6us. I reboot the machine and the result
> was the same!
> I'm afraid i don't understand what exactly the patch
> does. At first glance only the sequence of lines is
> changed - so what? Or there is some relatione between
> "fnsave" and setting task switch bit?

Actually, after some further testing, I get the same result! The first
time it's faster, then subsequent runs are slower (when using RT
processes). For non-RT processes, the times are stable. Strange.

				Regards,

					Richard....

-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majord...@vger.rutgers.edu
Please read the FAQ at http://www.tux.org/lkml/

From: rgo...@atnf.csiro.au (Richard Gooch)
Subject: Re: Interesting scheduling times
Date: 1998/09/18
Message-ID: <199809180637.QAA00514@vindaloo.atnf.CSIRO.AU>#1/1
X-Deja-AN: 392362734
Approved: g...@greenie.muc.de
Sender: muc.de!l-linux-kernel-owner
References: <6ts8hp$uim$1@palladium.transmeta.com>
Newsgroups: muc.lists.linux-kernel

Linus Torvalds writes:
> Yes, it happened during the switch-over to a software task-switch
> routine, but the reason the numbers got worse for some people is simply
> that I screwed up the floating point save code, and it saved _every_
> time through instead of doing the lazy save it was meant to do. 
> 
> The effect of this is not huge, but it's certainly noticeable.  With
> that fixed, the software context switch is pretty comparable to the
> hardware one. 
> 
> 2.1.123 should have this context switch slowdown fixed.

I tried the patch you sent, and it makes no noticable difference on my
PPro 180 (less than 1 us).

				Regards,

					Richard....

-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majord...@vger.rutgers.edu
Please read the FAQ at http://www.tux.org/lkml/

From: torva...@transmeta.com (Linus Torvalds)
Subject: Re: Interesting scheduling times
Date: 1998/09/18
Message-ID: <Pine.LNX.3.96.980918110555.7095B-100000@penguin.transmeta.com>#1/1
X-Deja-AN: 392564799
Approved: g...@greenie.muc.de
Sender: muc.de!l-linux-kernel-owner
References: <199809180751.RAA02219@vindaloo.atnf.CSIRO.AU>
Newsgroups: muc.lists.linux-kernel



On Fri, 18 Sep 1998, Richard Gooch wrote:
> 
> Actually, after some further testing, I get the same result! The first
> time it's faster, then subsequent runs are slower (when using RT
> processes). For non-RT processes, the times are stable. Strange.

I haven't even looked at the benchmark you seem to be talking about - I
use "lmbench" myself which I trust to be reasonably realistic. It
certainly showed an effect of my FPU screwup, although it's not all that
large on any reasonable system (it's probably horrible on a i386/i387
combination where FP operations are slower).

lmbench uses a set of pipes and passes a token around to force scheduling,
and that should work fine. I'd be nervous about any other kind of
scheduling benchmark.

		Linus


-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majord...@vger.rutgers.edu
Please read the FAQ at http://www.tux.org/lkml/

			   USENET Archives


Notice
******

The materials and information included in this website may only be used
for purposes such as criticism, review, private study, scholarship, or 
research.


Electronic mail:			      WorldWideWeb:
   tech-insider@outlook.com		         http://tech-insider.org/