)
From: Ingo Molnar <mingo@elte.hu>
This is the latest version of the scheduler cache-hot-auto-tune patch.
The first problem was that detection time scaled with O(N^2), which is
unacceptable on larger SMP and NUMA systems. To solve this:
- I've added a 'domain distance' function, which is used to cache
measurement results. Each distance is only measured once. This means
that e.g. on NUMA distances of 0, 1 and 2 might be measured, on HT
distances 0 and 1, and on SMP distance 0 is measured. The code walks
the domain tree to determine the distance, so it automatically follows
whatever hierarchy an architecture sets up. This cuts down on the boot
time significantly and removes the O(N^2) limit. The only assumption
is that migration costs can be expressed as a function of domain
distance - this covers the overwhelming majority of existing systems,
and is a good guess even for more assymetric systems.
[ People hacking systems that have assymetries that break this
assumption (e.g. different CPU speeds) should experiment a bit with
the cpu_distance() function. Adding a ->migration_distance factor to
the domain structure would be one possible solution - but lets first
see the problem systems, if they exist at all. Lets not overdesign. ]
Another problem was that only a single cache-size was used for measuring
the cost of migration, and most architectures didnt set that variable
up. Furthermore, a single cache-size does not fit NUMA hierarchies with
L3 caches and does not fit HT setups, where different CPUs will often
have different 'effective cache sizes'. To solve this problem:
- Instead of relying on a single cache-size provided by the platform and
sticking to it, the code now auto-detects the 'effective migration
cost' between two measured CPUs, via iterating through a wide range of
cachesizes. The code searches for the maximum migration cost, which
occurs when the working set of the test-workload falls just below the
'effective cache size'. I.e. real-life optimized search is done for
the maximum migration cost, between two real CPUs.
This, amongst other things, has the positive effect hat if e.g. two
CPUs share a L2/L3 cache, a different (and accurate) migration cost
will be found than between two CPUs on the same system that dont share
any caches.
(The reliable measurement of migration costs is tricky - see the source
for details.)
Furthermore i've added various boot-time options to override/tune
migration behavior.
Firstly, there's a blanket override for autodetection:
migration_cost=1000,2000,3000
will override the depth 0/1/2 values with 1msec/2msec/3msec values.
Secondly, there's a global factor that can be used to increase (or
decrease) the autodetected values:
migration_factor=120
will increase the autodetected values by 20%. This option is useful to
tune things in a workload-dependent way - e.g. if a workload is
cache-insensitive then CPU utilization can be maximized by specifying
migration_factor=0.
I've tested the autodetection code quite extensively on x86, on 3
P3/Xeon/2MB, and the autodetected values look pretty good:
Dual Celeron (128K L2 cache):
---------------------
migration cost matrix (max_cache_size: 131072, cpu: 467 MHz):
---------------------
[00] [01]
[00]: - 1.7(1)
[01]: 1.7(1) -
---------------------
cacheflush times [2]: 0.0 (0) 1.7 (1784008)
---------------------
Here the slow memory subsystem dominates system performance, and even
though caches are small, the migration cost is 1.7 msecs.
Dual HT P4 (512K L2 cache):
---------------------
migration cost matrix (max_cache_size: 524288, cpu: 2379 MHz):
---------------------
[00] [01] [02] [03]
[00]: - 0.4(1) 0.0(0) 0.4(1)
[01]: 0.4(1) - 0.4(1) 0.0(0)
[02]: 0.0(0) 0.4(1) - 0.4(1)
[03]: 0.4(1) 0.0(0) 0.4(1) -
---------------------
cacheflush times [2]: 0.0 (33900) 0.4 (448514)
---------------------
Here it can be seen that there is no migration cost between two HT
siblings (CPU#0/2 and CPU#1/3 are separate physical CPUs). A fast memory
system makes inter-physical-CPU migration pretty cheap: 0.4 msecs.
8-way P3/Xeon [2MB L2 cache]:
---------------------
migration cost matrix (max_cache_size: 2097152, cpu: 700 MHz):
---------------------
[00] [01] [02] [03] [04] [05] [06] [07]
[00]: - 19.2(1) 19.2(1) 19.2(1) 19.2(1) 19.2(1) 19.2(1) 19.2(1)
[01]: 19.2(1) - 19.2(1) 19.2(1) 19.2(1) 19.2(1) 19.2(1) 19.2(1)
[02]: 19.2(1) 19.2(1) - 19.2(1) 19.2(1) 19.2(1) 19.2(1) 19.2(1)
[03]: 19.2(1) 19.2(1) 19.2(1) - 19.2(1) 19.2(1) 19.2(1) 19.2(1)
[04]: 19.2(1) 19.2(1) 19.2(1) 19.2(1) - 19.2(1) 19.2(1) 19.2(1)
[05]: 19.2(1) 19.2(1) 19.2(1) 19.2(1) 19.2(1) - 19.2(1) 19.2(1)
[06]: 19.2(1) 19.2(1) 19.2(1) 19.2(1) 19.2(1) 19.2(1) - 19.2(1)
[07]: 19.2(1) 19.2(1) 19.2(1) 19.2(1) 19.2(1) 19.2(1) 19.2(1) -
---------------------
cacheflush times [2]: 0.0 (0) 19.2 (19281756)
---------------------
This one has huge caches and a relatively slow memory subsystem - so the
migration cost is 19 msecs.
Signed-off-by: Ingo Molnar <mingo@elte.hu>
Signed-off-by: Ashok Raj <ashok.raj@intel.com>
Signed-off-by: Ken Chen <kenneth.w.chen@intel.com>
Cc: <wilder@us.ibm.com>
Signed-off-by: John Hawkes <hawkes@sgi.com>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
include/asm-ppc/ had #ifdef __KERNEL__ in all header files that
are not meant for use by user space, include/asm-powerpc does
not have this yet.
This patch gets us a lot closer there. There are a few cases
where I was not sure, so I left them out. I have verified
that no CONFIG_* symbols are used outside of __KERNEL__
any more and that there are no obvious compile errors when
including any of the headers in user space libraries.
Signed-off-by: Arnd Bergmann <arnd@arndb.de>
Signed-off-by: Paul Mackerras <paulus@samba.org>
We used to print a NUMA cpu summary at boot before the hotplug cpu code
was added. This has been useful for catching machine configuration as
well as firmware bugs in the past.
This patch restores that functionality. An example of the output is:
Node 0 CPUs: 0-7
Node 1 CPUs: 8-15
Signed-off-by: Anton Blanchard <anton@samba.org>
Signed-off-by: Paul Mackerras <paulus@samba.org>
The fix to topology.h (5cfccd7f13) seems to have
a typeo, struct sched_domain has an idle_idx member but not an idle_id
member. I assume this is the fix.
Signed-off-by: Michael Ellerman <michael@ellerman.id.au>
Signed-off-by: Paul Mackerras <paulus@samba.org>
PowerPC's NUMA domain doesn't currently set up some of the newer
sched-domains parameters.
Brian Twichell <tbrian@us.ibm.com> discovered and diagnosed a 1.5% OLTP
database regression on a 4 core POWER5 system that was due to the use of
NUMA scheduling on ppc64.
This patch applies some saneish values to the parameters, in line with
other architectures. This solves the regression.
Signed-off-by: Nick Piggin <npiggin@suse.de>
Signed-off-by: Paul Mackerras <paulus@samba.org>
Convert to sparsemem and remove all the discontigmem code in the
process. This has a few advantages:
- The old numa_memory_lookup_table can go away
- All the arch specific discontigmem magic can go away
We also remove the triple pass of memory properties and instead create a
list of per node extents that we iterate through. A final cleanup would
be to change our lmb code to store extents per node, then we can reuse
that information in the numa code.
Signed-off-by: Anton Blanchard <anton@samba.org>
Signed-off-by: Paul Mackerras <paulus@samba.org>
Remove ppc64 specific version of nr_cpus_node and use the generic one
provided.
Signed-off-by: Anton Blanchard <anton@samba.org>
Signed-off-by: Paul Mackerras <paulus@samba.org>
Merge a few asm-ppc and asm-ppc64 header files.
Note: the merge of setup.h intentionally does not carry
forward the m68k cruft. That means this patch continues
to break the already broken amiga on the ppc32.
Signed-off-by: Jon Loeliger <jdl@freescale.com>
Signed-off-by: Kumar Gala <kumar.gala@freescale.com>
Signed-off-by: Paul Mackerras <paulus@samba.org>
Dont include asm-generic/topology.h unconditionally, we end up overriding
all the ppc64 specific functions when NUMA is on.
Signed-off-by: Anton Blanchard <anton@samba.org>
Acked-by: Paul Mackerras <paulus@samba.org>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
asm-generic/topology.h must also be included if CONFIG_NUMA is set in order to
provide the fall back pcibus_to_node function.
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
Initial git repository build. I'm not bothering with the full history,
even though we have it. We can create a separate "historical" git
archive of that later if we want to, and in the meantime it's about
3.2GB when imported into git - space that would just make the early
git days unnecessarily complicated, when we don't have a lot of good
infrastructure for it.
Let it rip!
akpm@osdl.org