Hello. After a year in stasis, I've completely rebuilt my kernel
patch that implements page coloring. Improvements include:
- Page coloring is now hardwired into the kernel. The hash
queues now use bootmem, and page coloring is always on. The
patch still creates /proc/page_color for statistics, but that
will go away in time.
- Automatic detection of external cache size on many architectures.
I have no idea if any of this code works, since I don't have any
of the target machines. The preferred way to initialize the coloring
is by passing "page_color=<external cache size in kB>" as a boot
argument.
- NUMA-safe, discontig-safe
Right now the actual page coloring algorithm is the same as in previous
patches, and performs the same. In the next few weeks I'll be trying new
ideas that will hopefully reduce fragmentation and increase performance.
This is an early attempt to get some feedback on mistakes I may have made.
lmbench shows no real gains or losses compared to an unpatched kernel;
some of the page fault and protection fault times are slightly slower, but
it's close to the rounding error over five lmbench runs.
Here are all the performance results I have for the patch:
1. Compile of 2.4.20 kernel with gcc 3.1.1 on 466MHz DS10 Alphaserver with
2MB cache: repeatable 1% speedup (573 sec vs. 579 sec)
2. 1000x1000 matrix multiply: 10% speedup on Athlon II with 512kB cache
(Dieter N?tzel)
3. Without page coloring, the alpha gets 80% of max theoretical bandwidth
for working sets at most 1/8 the size of its L2 cache. For larger working
sets than that the achieved bandwidth is only 30%-50% of max. With page
coloring, the 80% figure applies to the entire L2 cache.
4. FFTW (alpha): 30% speedup for 64k-point FFTs, 20% speedup for 1M-point FFTs
Patch is available at
http://www.boo.net/~jasonp/page_color-2.4.20-20030104.patch
Thanks in advance for any feedback.
jasonp
On Sat, Jan 04, 2003 at 11:31:11PM -0500, Jason Papadopoulos wrote:
> Hello. After a year in stasis, I've completely rebuilt my kernel
> patch that implements page coloring. Improvements include:
> - Page coloring is now hardwired into the kernel. The hash
> queues now use bootmem, and page coloring is always on. The
> patch still creates /proc/page_color for statistics, but that
> will go away in time.
> - Automatic detection of external cache size on many architectures.
> I have no idea if any of this code works, since I don't have any
> of the target machines. The preferred way to initialize the coloring
> is by passing "page_color=<external cache size in kB>" as a boot
> argument.
> - NUMA-safe, discontig-safe
> Right now the actual page coloring algorithm is the same as in previous
> patches, and performs the same. In the next few weeks I'll be trying new
> ideas that will hopefully reduce fragmentation and increase performance.
> This is an early attempt to get some feedback on mistakes I may have made.
Any chance for a 2.5.x-mm port? This is a bit feature-ish for 2.4.x.
Thanks,
Bill
> > This is an early attempt to get some feedback on mistakes I may have made.
>
> Any chance for a 2.5.x-mm port? This is a bit feature-ish for 2.4.x.
I know. The problem is that 2.5.53 cannot finish booting on the Alpha I have
here (IDE issues). While I can port the patch over, I'm not comfortable being
unable to test it at all.
jasonp
---------------------------------------------
This message was sent using Endymion MailMan.
http://www.endymion.com/products/mailman/
At some point in the past, jasonp wrote:
>> Any chance for a 2.5.x-mm port? This is a bit feature-ish for 2.4.x.
On Sun, Jan 05, 2003 at 04:03:33PM +0000, [email protected] wrote:
> I know. The problem is that 2.5.53 cannot finish booting on the Alpha I have
> here (IDE issues). While I can port the patch over, I'm not comfortable being
> unable to test it at all.
What kind of Alpha? Got an oops/backtrace?
I probably can't reproduce it directly since my Alpha's diskless.
Bill
At 11:34 AM 1/5/03 -0800, you wrote:
>At some point in the past, jasonp wrote:
>>> Any chance for a 2.5.x-mm port? This is a bit feature-ish for 2.4.x.
>
>On Sun, Jan 05, 2003 at 04:03:33PM +0000, [email protected] wrote:
>> I know. The problem is that 2.5.53 cannot finish booting on the Alpha I have
>> here (IDE issues). While I can port the patch over, I'm not comfortable being
>> unable to test it at all.
>
>What kind of Alpha? Got an oops/backtrace?
>
>I probably can't reproduce it directly since my Alpha's diskless.
The machine in question is a DS10 Alphaserver (DP264 type chipset) using an
ALI M5229 rev c1 IDE controller (uses the ALI 15x3 driver). The stock 2.5.53
kernel panics at boot time because it can't find the root partition; when I
first reported the problem, one of the maintainers passed on a patch that
resolved some 2.5 IDE issues. With the patch in place, the boot process gets
farther along but occaisionally there will be a printout that hda lost an
interrupt. I don't know how far it really gets, because I gave up after other
subsystems started reporting errors.
I haven't tried 2.5.54, either. I will shortly.
Is 2.4 really in bug-fix mode now? 2.4.19 and 2.4.20 were huge patches.
Thanks,
jasonp
At 11:34 AM 1/5/03 -0800, you wrote:
>> What kind of Alpha? Got an oops/backtrace?
>> I probably can't reproduce it directly since my Alpha's diskless.
On Sun, Jan 05, 2003 at 03:04:05PM -0500, Jason Papadopoulos wrote:
> The machine in question is a DS10 Alphaserver (DP264 type chipset) using an
> ALI M5229 rev c1 IDE controller (uses the ALI 15x3 driver). The stock 2.5.53
> kernel panics at boot time because it can't find the root partition; when I
> first reported the problem, one of the maintainers passed on a patch that
> resolved some 2.5 IDE issues. With the patch in place, the boot process gets
> farther along but occaisionally there will be a printout that hda lost an
> interrupt. I don't know how far it really gets, because I gave up after other
> subsystems started reporting errors.
> I haven't tried 2.5.54, either. I will shortly.
> Is 2.4 really in bug-fix mode now? 2.4.19 and 2.4.20 were huge patches.
Sounds like a job for the IDE crew, who appear to know there are
pending ali + Alpha issues.
Bill
Version 2 of the page coloring patch is ready. This version includes
support for non-power-of-two cache sizes, fixes the ia64 cache detection
code (thanks due to Alex Williamson), and fixes a small initialization bug.
New patch available at
http://www.boo.net/~jasonp/page_color-2.4.20-20030114.patch
Thanks in advance,
jasonp