RE: Peculiar(?) slowdown with -CURRENT as of 21 March
On 23-Mar-2002 Jeff Roberson wrote: >> >> I saw some similar weirdness in my test machines last night where a dual >> processor DS20 (Alpha 21264 500x2) beat out a PII Xeon 450x4. Normally > the >> quad xeon beats the DS20. The quad xeon was using -j16 but was about 74% >> idle. >> The DS20 had used -j8. I didn't get a chacne to run top to see how it was >> doing during hte world since I didn't notice the weirdness until last > night >> after the DS20 had finsihed but the quad xeon was still chugging along. >> > > Are you both running with WITNESS and INVARIANTS? UMA is slightly slower > with these options on than the original malloc & vm_zone code. I'm not > sure why it would be even worse for SMP machines though. So maybe it > isn't UMA at all but it's worth looking into. Yes. -- John Baldwin <[EMAIL PROTECTED]> <>< http://www.FreeBSD.org/~jhb/ "Power Users Use the Power to Serve!" - http://www.FreeBSD.org/ To Unsubscribe: send mail to [EMAIL PROTECTED] with "unsubscribe freebsd-current" in the body of the message
Re: Peculiar(?) slowdown with -CURRENT as of 21 March
On Fri, Mar 22, 2002 at 06:34:00PM -0800, David Wolfskill wrote: > >Date: Fri, 22 Mar 2002 20:43:03 -0500 (EST) > >From: Jeff Roberson <[EMAIL PROTECTED]> > > >Are you both running with WITNESS and INVARIANTS? UMA is slightly slower > >with these options on than the original malloc & vm_zone code. I'm not > >sure why it would be even worse for SMP machines though. So maybe it > >isn't UMA at all but it's worth looking into. > > I don't speak (or write) for John, but yes, I have both WITNESS & > INVARIANTS in my -CURRENT kernels. (The link I posted refers to a page > that has copies of each kernel config, among other bits of trivia.) As more locks are added to the kernel the overhead of WITNESS becomes greater. 21 March was around the time of Jeff's UMA commit. Kris msg36453/pgp0.pgp Description: PGP signature
Re: Peculiar(?) slowdown with -CURRENT as of 21 March
>Date: Fri, 22 Mar 2002 20:43:03 -0500 (EST) >From: Jeff Roberson <[EMAIL PROTECTED]> >Are you both running with WITNESS and INVARIANTS? UMA is slightly slower >with these options on than the original malloc & vm_zone code. I'm not >sure why it would be even worse for SMP machines though. So maybe it >isn't UMA at all but it's worth looking into. I don't speak (or write) for John, but yes, I have both WITNESS & INVARIANTS in my -CURRENT kernels. (The link I posted refers to a page that has copies of each kernel config, among other bits of trivia.) Reason I posted about it is because the result was at such variance with my previous experience and expectations, after all :-) And one of the reasons for my surprise is that the (SMP) build machine has usually been quite a bit faster... and the kernel configurations are about as similar to each other as makes any sort of sense, given the differences in the hardware. If anything, I'd expect the laptop to be a bit more sluggish if only because it has the usual UI stuff, like sound and an X server (with less memory)... while the only access to the build machine is either SSH or the serial console; it doesn't have a keyboard or monitor. No speakers, either. :-} Cheers, david (links to my resume at http://www.catwhisker.org/~david) -- David H. Wolfskill [EMAIL PROTECTED] I believe it would be irresponsible (and thus, unethical) for me to advise, recommend, or support the use of any product that is or depends on any Microsoft product for any purpose other than personal amusement. To Unsubscribe: send mail to [EMAIL PROTECTED] with "unsubscribe freebsd-current" in the body of the message
Peculiar(?) slowdown with -CURRENT as of 21 March
> > I saw some similar weirdness in my test machines last night where a dual > processor DS20 (Alpha 21264 500x2) beat out a PII Xeon 450x4. Normally the > quad xeon beats the DS20. The quad xeon was using -j16 but was about 74% > idle. > The DS20 had used -j8. I didn't get a chacne to run top to see how it was > doing during hte world since I didn't notice the weirdness until last night > after the DS20 had finsihed but the quad xeon was still chugging along. > Are you both running with WITNESS and INVARIANTS? UMA is slightly slower with these options on than the original malloc & vm_zone code. I'm not sure why it would be even worse for SMP machines though. So maybe it isn't UMA at all but it's worth looking into. Thanks, Jeff To Unsubscribe: send mail to [EMAIL PROTECTED] with "unsubscribe freebsd-current" in the body of the message
Re: Peculiar(?) slowdown with -CURRENT as of 21 March
Robert Watson wrote: > > On a 0 - 10 "weirdness" scale, this ranks about a 4, perhaps, so it's > > hardly earth-shattering. But it's odd enough that I thought that a > > small reality check might be in order, in case the effect(s) in question > > were not expected. (And yes, I understand that a degree of uncertainty > > with respect to -CURRENT's performance is to be expected, even at the > > best of times. This is not a complaint.) > > Could be the removal of __P :-) Speed difference between K&R and ANSI compilers? 8-) 8-). -- Terry To Unsubscribe: send mail to [EMAIL PROTECTED] with "unsubscribe freebsd-current" in the body of the message
Re: Peculiar(?) slowdown with -CURRENT as of 21 March
On Fri, 22 Mar 2002, David Wolfskill wrote: > On a 0 - 10 "weirdness" scale, this ranks about a 4, perhaps, so it's > hardly earth-shattering. But it's odd enough that I thought that a > small reality check might be in order, in case the effect(s) in question > were not expected. (And yes, I understand that a degree of uncertainty > with respect to -CURRENT's performance is to be expected, even at the > best of times. This is not a complaint.) Could be the removal of __P :-) Robert N M Watson FreeBSD Core Team, TrustedBSD Project [EMAIL PROTECTED] NAI Labs, Safeport Network Services To Unsubscribe: send mail to [EMAIL PROTECTED] with "unsubscribe freebsd-current" in the body of the message
RE: Peculiar(?) slowdown with -CURRENT as of 21 March
On 22-Mar-2002 David Wolfskill wrote: > On a 0 - 10 "weirdness" scale, this ranks about a 4, perhaps, so it's > hardly earth-shattering. But it's odd enough that I thought that a > small reality check might be in order, in case the effect(s) in question > were not expected. (And yes, I understand that a degree of uncertainty > with respect to -CURRENT's performance is to be expected, even at the > best of times. This is not a complaint.) > > Briefly, my SMP "build machine" built today's -CURRENT (while running > yesterday's -- hence the Subject:) much more slowly than my laptop did. > > To illustrate, here are the timestamp messages from the respective > kernel builds: > Kernel build for FREEBEAST started on Fri Mar 22 07:59:58 PST 2002 Kernel build for FREEBEAST completed on Fri Mar 22 08:24:19 PST 2002 > Kernel build for LAPTOP_30W started on Fri Mar 22 08:09:25 PST 2002 Kernel build for LAPTOP_30W completed on Fri Mar 22 08:26:31 PST 2002 > > So that's not a huge difference in and of itself -- 24:21 vs. 17:06 -- > but what really stands out is that the laptop actually finished the whole > morning's processing before the build machine did. And that's somewhat > remarkable, given that: > > * build machine got a bit of a head start (though it did have a little > bit more work to do in one respect). > > * build machine is 2x866 MHz P3s; laptop is a 750 MHz P3. > > * laptop disk is 4500 RPM; build machine's disk is probably 5400 -- > certainly no slower than that. > > I don't really want to spam the entire list with the details of the > processes used, so I cobbled up a page at > http://www.catwhisker.org/~david/FreeBSD/speed.html that should > have enough excruciating details for anyone sufficiently interested. I saw some similar weirdness in my test machines last night where a dual processor DS20 (Alpha 21264 500x2) beat out a PII Xeon 450x4. Normally the quad xeon beats the DS20. The quad xeon was using -j16 but was about 74% idle. The DS20 had used -j8. I didn't get a chacne to run top to see how it was doing during hte world since I didn't notice the weirdness until last night after the DS20 had finsihed but the quad xeon was still chugging along. -- John Baldwin <[EMAIL PROTECTED]> <>< http://www.FreeBSD.org/~jhb/ "Power Users Use the Power to Serve!" - http://www.FreeBSD.org/ To Unsubscribe: send mail to [EMAIL PROTECTED] with "unsubscribe freebsd-current" in the body of the message
Peculiar(?) slowdown with -CURRENT as of 21 March
On a 0 - 10 "weirdness" scale, this ranks about a 4, perhaps, so it's hardly earth-shattering. But it's odd enough that I thought that a small reality check might be in order, in case the effect(s) in question were not expected. (And yes, I understand that a degree of uncertainty with respect to -CURRENT's performance is to be expected, even at the best of times. This is not a complaint.) Briefly, my SMP "build machine" built today's -CURRENT (while running yesterday's -- hence the Subject:) much more slowly than my laptop did. To illustrate, here are the timestamp messages from the respective kernel builds: >>> Kernel build for FREEBEAST started on Fri Mar 22 07:59:58 PST 2002 >>> Kernel build for FREEBEAST completed on Fri Mar 22 08:24:19 PST 2002 >>> Kernel build for LAPTOP_30W started on Fri Mar 22 08:09:25 PST 2002 >>> Kernel build for LAPTOP_30W completed on Fri Mar 22 08:26:31 PST 2002 So that's not a huge difference in and of itself -- 24:21 vs. 17:06 -- but what really stands out is that the laptop actually finished the whole morning's processing before the build machine did. And that's somewhat remarkable, given that: * build machine got a bit of a head start (though it did have a little bit more work to do in one respect). * build machine is 2x866 MHz P3s; laptop is a 750 MHz P3. * laptop disk is 4500 RPM; build machine's disk is probably 5400 -- certainly no slower than that. I don't really want to spam the entire list with the details of the processes used, so I cobbled up a page at http://www.catwhisker.org/~david/FreeBSD/speed.html that should have enough excruciating details for anyone sufficiently interested. Thanks, david -- David H. Wolfskill [EMAIL PROTECTED] I believe it would be irresponsible (and thus, unethical) for me to advise, recommend, or support the use of any product that is or depends on any Microsoft product for any purpose other than personal amusement. To Unsubscribe: send mail to [EMAIL PROTECTED] with "unsubscribe freebsd-current" in the body of the message