When it was introduced, zone_reclaim_mode made sense as NUMA distances
punished and workloads were generally partitioned to fit into a NUMA
node. NUMA machines are now common but few of the workloads are
NUMA-aware and it's routine to see major performance degradation due to
zone_reclaim_mode being enabled but relatively few can identify the
problem.
Those that require zone_reclaim_mode are likely to be able to detect
when it needs to be enabled and tune appropriately so lets have a
sensible default for the bulk of users.
This patch (of 2):
zone_reclaim_mode causes processes to prefer reclaiming memory from
local node instead of spilling over to other nodes. This made sense
initially when NUMA machines were almost exclusively HPC and the
workload was partitioned into nodes. The NUMA penalties were
sufficiently high to justify reclaiming the memory. On current machines
and workloads it is often the case that zone_reclaim_mode destroys
performance but not all users know how to detect this. Favour the
common case and disable it by default. Users that are sophisticated
enough to know they need zone_reclaim_mode will detect it.
Signed-off-by: Mel Gorman <mgorman@suse.de>
Acked-by: Johannes Weiner <hannes@cmpxchg.org>
Reviewed-by: Zhang Yanfei <zhangyanfei@cn.fujitsu.com>
Acked-by: Michal Hocko <mhocko@suse.cz>
Reviewed-by: Christoph Lameter <cl@linux.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
We replace the old way to configure the scheduler topology with a new method
which enables a platform to declare additionnal level (if needed).
We still have a default topology table definition that can be used by platform
that don't want more level than the SMT, MC, CPU and NUMA ones. This table can
be overwritten by an arch which either wants to add new level where a load
balance make sense like BOOK or powergating level or wants to change the flags
configuration of some levels.
For each level, we need a function pointer that returns cpumask for each cpu,
a function pointer that returns the flags for the level and a name. Only flags
that describe topology, can be set by an architecture. The current topology
flags are:
SD_SHARE_CPUPOWER
SD_SHARE_PKG_RESOURCES
SD_NUMA
SD_ASYM_PACKING
Then, each level must be a subset on the next one. The build sequence of the
sched_domain will take care of removing useless levels like those with 1 CPU
and those with the same CPU span and no more relevant information for
load balancing than its children.
Signed-off-by: Vincent Guittot <vincent.guittot@linaro.org>
Tested-by: Dietmar Eggemann <dietmar.eggemann@arm.com>
Reviewed-by: Preeti U Murthy <preeti@linux.vnet.ibm.com>
Reviewed-by: Dietmar Eggemann <dietmar.eggemann@arm.com>
Signed-off-by: Peter Zijlstra <peterz@infradead.org>
Cc: Andrew Morton <akpm@linux-foundation.org>
Cc: Benjamin Herrenschmidt <benh@kernel.crashing.org>
Cc: Bjorn Helgaas <bhelgaas@google.com>
Cc: Chris Metcalf <cmetcalf@tilera.com>
Cc: Christoph Lameter <cl@linux.com>
Cc: David S. Miller <davem@davemloft.net>
Cc: Fenghua Yu <fenghua.yu@intel.com>
Cc: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
Cc: Hanjun Guo <hanjun.guo@linaro.org>
Cc: Heiko Carstens <heiko.carstens@de.ibm.com>
Cc: Jason Low <jason.low2@hp.com>
Cc: Linus Torvalds <torvalds@linux-foundation.org>
Cc: Martin Schwidefsky <schwidefsky@de.ibm.com>
Cc: Tony Luck <tony.luck@intel.com>
Cc: linux390@de.ibm.com
Cc: linux-ia64@vger.kernel.org
Cc: linux-s390@vger.kernel.org
Link: http://lkml.kernel.org/r/1397209481-28542-2-git-send-email-vincent.guittot@linaro.org
Signed-off-by: Ingo Molnar <mingo@kernel.org>
With the preempt checking logic for __this_cpu_ops we will get false
positives from locations in the code that use numa_node_id.
Before the __this_cpu ops where introduced there were no checks for
preemption present either. smp_raw_processor_id() was used. See
http://www.spinics.net/lists/linux-numa/msg00641.html
Therefore we need to use raw_cpu_read here to avoid false postives.
Note that this issue has been discussed in prior years. If the process
changes nodes after retrieving the current numa node then that is
acceptable since most uses of numa_node etc are for optimization and not
for correctness.
There were suggestions to implement a raw_numa_node_id in order to do
preempt checks for numa_node_id as well. But I think we better defer
that to another patch since that would mean investigating how
numa_node_id() is used throughout the kernel which would increase the
scope of this patchset significantly. After all preemption was never
checked before when numa_node_id() was used.
Some sample traces:
__this_cpu_read operation in preemptible [00000000] code: login/1456
caller is __this_cpu_preempt_check+0x2b/0x2d
CPU: 0 PID: 1456 Comm: login Not tainted 3.12.0-rc4-cl-00062-g2fe80d3-dirty #185
Call Trace:
dump_stack+0x4e/0x82
check_preemption_disabled+0xc5/0xe0
__this_cpu_preempt_check+0x2b/0x2d
get_task_policy+0x1d/0x49
get_vma_policy+0x14/0x76
alloc_pages_vma+0x35/0xff
handle_mm_fault+0x290/0x73b
__do_page_fault+0x3fe/0x44d
do_page_fault+0x9/0xc
page_fault+0x22/0x30
generic_file_aio_read+0x38e/0x624
do_sync_read+0x54/0x73
vfs_read+0x9d/0x12a
SyS_read+0x47/0x7e
cstar_dispatch+0x7/0x23
caller is __this_cpu_preempt_check+0x2b/0x2d
CPU: 0 PID: 1456 Comm: login Not tainted 3.12.0-rc4-cl-00062-g2fe80d3-dirty #185
Call Trace:
dump_stack+0x4e/0x82
check_preemption_disabled+0xc5/0xe0
__this_cpu_preempt_check+0x2b/0x2d
alloc_pages_current+0x8f/0xbc
__page_cache_alloc+0xb/0xd
__do_page_cache_readahead+0xf4/0x219
ra_submit+0x1c/0x20
ondemand_readahead+0x28c/0x2b4
page_cache_sync_readahead+0x38/0x3a
generic_file_aio_read+0x261/0x624
do_sync_read+0x54/0x73
vfs_read+0x9d/0x12a
SyS_read+0x47/0x7e
cstar_dispatch+0x7/0x23
Signed-off-by: Christoph Lameter <cl@linux.com>
Acked-by: Ingo Molnar <mingo@kernel.org>
Cc: Alex Shi <alex.shi@intel.com>
Cc: Tejun Heo <tj@kernel.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
This patch builds on patch 2 and periodically decays that max value to
do idle balancing per sched domain by approximately 1% per second. Also
decay the rq's max_idle_balance_cost value.
Signed-off-by: Jason Low <jason.low2@hp.com>
Signed-off-by: Peter Zijlstra <peterz@infradead.org>
Link: http://lkml.kernel.org/r/1379096813-3032-4-git-send-email-jason.low2@hp.com
Signed-off-by: Ingo Molnar <mingo@kernel.org>
In this patch, we keep track of the max cost we spend doing idle load balancing
for each sched domain. If the avg time the CPU remains idle is less then the
time we have already spent on idle balancing + the max cost of idle balancing
in the sched domain, then we don't continue to attempt the balance. We also
keep a per rq variable, max_idle_balance_cost, which keeps track of the max
time spent on newidle load balances throughout all its domains so that we can
determine the avg_idle's max value.
By using the max, we avoid overrunning the average. This further reduces the
chance we attempt balancing when the CPU is not idle for longer than the cost
to balance.
Signed-off-by: Jason Low <jason.low2@hp.com>
Signed-off-by: Peter Zijlstra <peterz@infradead.org>
Link: http://lkml.kernel.org/r/1379096813-3032-3-git-send-email-jason.low2@hp.com
Signed-off-by: Ingo Molnar <mingo@kernel.org>
Since power saving code was removed from sched now, the implement
code is out of service in this function, and even pollute other logical.
like, 'want_sd' never has chance to be set '0', that remove the effect
of SD_WAKE_AFFINE here.
So, clean up the obsolete code, includes SD_PREFER_LOCAL.
Signed-off-by: Alex Shi <alex.shi@intel.com>
Signed-off-by: Peter Zijlstra <a.p.zijlstra@chello.nl>
Link: http://lkml.kernel.org/r/5028F431.6000306@intel.com
Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
Commit 8e7fbcbc22 ("sched: Remove stale power aware scheduling remnants
and dysfunctional knobs") removed SD_PERFER_SIBLING from the CPU domain.
On NUMA machines this causes that load_balance() doesn't perfer LCPU in
same physical CPU package.
It causes some actual performance regressions on our NUMA machines from
Core2 to NHM and SNB.
Adding this domain flag again recovers the performance drop.
This change doesn't have any bad impact on any of my benchmarks:
specjbb, kbuild, fio, hackbench .. etc, on all my machines.
Signed-off-by: Alex Shi <alex.shi@intel.com>
Signed-off-by: Peter Zijlstra <a.p.zijlstra@chello.nl>
Link: http://lkml.kernel.org/r/1342765190-21540-1-git-send-email-alex.shi@intel.com
Signed-off-by: Ingo Molnar <mingo@kernel.org>
Pull scheduler changes from Ingo Molnar:
"The biggest change is the cleanup/simplification of the load-balancer:
instead of the current practice of architectures twiddling scheduler
internal data structures and providing the scheduler domains in
colorfully inconsistent ways, we now have generic scheduler code in
kernel/sched/core.c:sched_init_numa() that looks at the architecture's
node_distance() parameters and (while not fully trusting it) deducts a
NUMA topology from it.
This inevitably changes balancing behavior - hopefully for the better.
There are various smaller optimizations, cleanups and fixlets as well"
* 'sched-core-for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/tip/tip:
sched: Taint kernel with TAINT_WARN after sleep-in-atomic bug
sched: Remove stale power aware scheduling remnants and dysfunctional knobs
sched/debug: Fix printing large integers on 32-bit platforms
sched/fair: Improve the ->group_imb logic
sched/nohz: Fix rq->cpu_load[] calculations
sched/numa: Don't scale the imbalance
sched/fair: Revert sched-domain iteration breakage
sched/x86: Rewrite set_cpu_sibling_map()
sched/numa: Fix the new NUMA topology bits
sched/numa: Rewrite the CONFIG_NUMA sched domain support
sched/fair: Propagate 'struct lb_env' usage into find_busiest_group
sched/fair: Add some serialization to the sched_domain load-balance walk
sched/fair: Let minimally loaded cpu balance the group
sched: Change rq->nr_running to unsigned int
x86/numa: Check for nonsensical topologies on real hw as well
x86/numa: Hard partition cpu topology masks on node boundaries
x86/numa: Allow specifying node_distance() for numa=fake
x86/sched: Make mwait_usable() heed to "idle=" kernel parameters properly
sched: Update documentation and comments
sched_rt: Avoid unnecessary dequeue and enqueue of pushable tasks in set_cpus_allowed_rt()
It's been broken forever (i.e. it's not scheduling in a power
aware fashion), as reported by Suresh and others sending
patches, and nobody cares enough to fix it properly ...
so remove it to make space free for something better.
There's various problems with the code as it stands today, first
and foremost the user interface which is bound to topology
levels and has multiple values per level. This results in a
state explosion which the administrator or distro needs to
master and almost nobody does.
Furthermore large configuration state spaces aren't good, it
means the thing doesn't just work right because it's either
under so many impossibe to meet constraints, or even if
there's an achievable state workloads have to be aware of
it precisely and can never meet it for dynamic workloads.
So pushing this kind of decision to user-space was a bad idea
even with a single knob - it's exponentially worse with knobs
on every node of the topology.
There is a proposal to replace the user interface with a single
3 state knob:
sched_balance_policy := { performance, power, auto }
where 'auto' would be the preferred default which looks at things
like Battery/AC mode and possible cpufreq state or whatever the hw
exposes to show us power use expectations - but there's been no
progress on it in the past many months.
Aside from that, the actual implementation of the various knobs
is known to be broken. There have been sporadic attempts at
fixing things but these always stop short of reaching a mergable
state.
Therefore this wholesale removal with the hopes of spurring
people who care to come forward once again and work on a
coherent replacement.
Signed-off-by: Peter Zijlstra <a.p.zijlstra@chello.nl>
Cc: Suresh Siddha <suresh.b.siddha@intel.com>
Cc: Arjan van de Ven <arjan@linux.intel.com>
Cc: Vincent Guittot <vincent.guittot@linaro.org>
Cc: Vaidyanathan Srinivasan <svaidy@linux.vnet.ibm.com>
Cc: Linus Torvalds <torvalds@linux-foundation.org>
Cc: Andrew Morton <akpm@linux-foundation.org>
Link: http://lkml.kernel.org/r/1326104915.2442.53.camel@twins
Signed-off-by: Ingo Molnar <mingo@kernel.org>
Since percpu_xxx() serial functions are duplicated with this_cpu_xxx().
Removing percpu_xxx() definition and replacing them by this_cpu_xxx()
in code. There is no function change in this patch, just preparation for
later percpu_xxx serial function removing.
On x86 machine the this_cpu_xxx() serial functions are same as
__this_cpu_xxx() without no unnecessary premmpt enable/disable.
Thanks for Stephen Rothwell, he found and fixed a i386 build error in
the patch.
Also thanks for Andrew Morton, he kept updating the patchset in Linus'
tree.
Signed-off-by: Alex Shi <alex.shi@intel.com>
Acked-by: Christoph Lameter <cl@gentwo.org>
Acked-by: Tejun Heo <tj@kernel.org>
Acked-by: "H. Peter Anvin" <hpa@zytor.com>
Cc: Ingo Molnar <mingo@elte.hu>
Cc: Thomas Gleixner <tglx@linutronix.de>
Signed-off-by: Stephen Rothwell <sfr@canb.auug.org.au>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Tejun Heo <tj@kernel.org>
The current code groups up to 16 nodes in a level and then puts an
ALLNODES domain spanning the entire tree on top of that. This doesn't
reflect the numa topology and esp for the smaller not-fully-connected
machines out there today this might make a difference.
Therefore, build a proper numa topology based on node_distance().
Since there's no fixed numa layers anymore, the static SD_NODE_INIT
and SD_ALLNODES_INIT aren't usable anymore, the new code tries to
construct something similar and scales some values either on the
number of cpus in the domain and/or the node_distance() ratio.
Signed-off-by: Peter Zijlstra <a.p.zijlstra@chello.nl>
Cc: Anton Blanchard <anton@samba.org>
Cc: Benjamin Herrenschmidt <benh@kernel.crashing.org>
Cc: Chris Metcalf <cmetcalf@tilera.com>
Cc: David Howells <dhowells@redhat.com>
Cc: "David S. Miller" <davem@davemloft.net>
Cc: Fenghua Yu <fenghua.yu@intel.com>
Cc: "H. Peter Anvin" <hpa@zytor.com>
Cc: Ivan Kokshaysky <ink@jurassic.park.msu.ru>
Cc: linux-alpha@vger.kernel.org
Cc: linux-ia64@vger.kernel.org
Cc: linux-kernel@vger.kernel.org
Cc: linux-mips@linux-mips.org
Cc: linuxppc-dev@lists.ozlabs.org
Cc: linux-sh@vger.kernel.org
Cc: Matt Turner <mattst88@gmail.com>
Cc: Paul Mackerras <paulus@samba.org>
Cc: Paul Mundt <lethal@linux-sh.org>
Cc: Ralf Baechle <ralf@linux-mips.org>
Cc: Richard Henderson <rth@twiddle.net>
Cc: sparclinux@vger.kernel.org
Cc: Tony Luck <tony.luck@intel.com>
Cc: x86@kernel.org
Cc: Dimitri Sivanich <sivanich@sgi.com>
Cc: Greg Pearson <greg.pearson@hp.com>
Cc: KAMEZAWA Hiroyuki <kamezawa.hiroyu@jp.fujitsu.com>
Cc: bob.picco@oracle.com
Cc: chris.mason@oracle.com
Cc: Linus Torvalds <torvalds@linux-foundation.org>
Cc: Andrew Morton <akpm@linux-foundation.org>
Link: http://lkml.kernel.org/n/tip-r74n3n8hhuc2ynbrnp3vt954@git.kernel.org
Signed-off-by: Ingo Molnar <mingo@kernel.org>
We want to override the default value of SD_NODES_PER_DOMAIN on ppc64,
so move it into linux/topology.h.
Signed-off-by: Anton Blanchard <anton@samba.org>
Acked-by: Peter Zijlstra <a.p.zijlstra@chello.nl>
Signed-off-by: Benjamin Herrenschmidt <benh@kernel.crashing.org>
Recently, Robert Mueller reported (http://lkml.org/lkml/2010/9/12/236)
that zone_reclaim_mode doesn't work properly on his new NUMA server (Dual
Xeon E5520 + Intel S5520UR MB). He is using Cyrus IMAPd and it's built on
a very traditional single-process model.
* a master process which reads config files and manages the other
process
* multiple imapd processes, one per connection
* multiple pop3d processes, one per connection
* multiple lmtpd processes, one per connection
* periodical "cleanup" processes.
There are thousands of independent processes. The problem is, recent
Intel motherboard turn on zone_reclaim_mode by default and traditional
prefork model software don't work well on it. Unfortunatelly, such models
are still typical even in the 21st century. We can't ignore them.
This patch raises the zone_reclaim_mode threshold to 30. 30 doesn't have
any specific meaning. but 20 means that one-hop QPI/Hypertransport and
such relatively cheap 2-4 socket machine are often used for traditional
servers as above. The intention is that these machines don't use
zone_reclaim_mode.
Note: ia64 and Power have arch specific RECLAIM_DISTANCE definitions.
This patch doesn't change such high-end NUMA machine behavior.
Dave Hansen said:
: I know specifically of pieces of x86 hardware that set the information
: in the BIOS to '21' *specifically* so they'll get the zone_reclaim_mode
: behavior which that implies.
:
: They've done performance testing and run very large and scary benchmarks
: to make sure that they _want_ this turned on. What this means for them
: is that they'll probably be de-optimized, at least on newer versions of
: the kernel.
:
: If you want to do this for particular systems, maybe _that_'s what we
: should do. Have a list of specific configurations that need the
: defaults overridden either because they're buggy, or they have an
: unusual hardware configuration not really reflected in the distance
: table.
And later said:
: The original change in the hardware tables was for the benefit of a
: benchmark. Said benchmark isn't going to get run on mainline until the
: next batch of enterprise distros drops, at which point the hardware where
: this was done will be irrelevant for the benchmark. I'm sure any new
: hardware will just set this distance to another yet arbitrary value to
: make the kernel do what it wants. :)
:
: Also, when the hardware got _set_ to this initially, I complained. So, I
: guess I'm getting my way now, with this patch. I'm cool with it.
Reported-by: Robert Mueller <robm@fastmail.fm>
Signed-off-by: KOSAKI Motohiro <kosaki.motohiro@jp.fujitsu.com>
Acked-by: Christoph Lameter <cl@linux.com>
Acked-by: David Rientjes <rientjes@google.com>
Reviewed-by: KAMEZAWA Hiroyuki <kamezawa.hiroyu@jp.fujitsu.com>
Cc: Benjamin Herrenschmidt <benh@kernel.crashing.org>
Cc: "Luck, Tony" <tony.luck@intel.com>
Acked-by: Dave Hansen <dave@linux.vnet.ibm.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
On top of the SMT and MC scheduling domains this adds the BOOK scheduling
domain. This is useful for NUMA like machines which do not have an interface
which tells which piece of memory is attached to which node or where the
hardware performs striping.
Signed-off-by: Heiko Carstens <heiko.carstens@de.ibm.com>
Signed-off-by: Peter Zijlstra <a.p.zijlstra@chello.nl>
LKML-Reference: <20100831082844.253053798@de.ibm.com>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
Define stubs for the numa_*_id() generic percpu related functions for
non-NUMA configurations in <asm-generic/topology.h> where the other
non-numa stubs live.
Fixes ia64 !NUMA build breakage -- e.g., tiger_defconfig
Back out now unneeded '#ifndef CONFIG_NUMA' guards from ia64 smpboot.c
Signed-off-by: Lee Schermerhorn <lee.schermerhorn@hp.com>
Tested-by: Tony Luck <tony.luck@intel.com>
Acked-by: Tony Luck <tony.luck@intel.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
No logic changes, only spelling.
Signed-off-by: Michael Neuling <mikey@neuling.org>
Cc: linuxppc-dev@ozlabs.org
Cc: David Howells <dhowells@redhat.com>
Cc: Peter Zijlstra <peterz@infradead.org>
LKML-Reference: <15249.1277776921@neuling.org>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
Check to see if the group is packed in a sched doman.
This is primarily intended to used at the sibling level. Some cores
like POWER7 prefer to use lower numbered SMT threads. In the case of
POWER7, it can move to lower SMT modes only when higher threads are
idle. When in lower SMT modes, the threads will perform better since
they share less core resources. Hence when we have idle threads, we
want them to be the higher ones.
This adds a hook into f_b_g() called check_asym_packing() to check the
packing. This packing function is run on idle threads. It checks to
see if the busiest CPU in this domain (core in the P7 case) has a
higher CPU number than what where the packing function is being run
on. If it is, calculate the imbalance and return the higher busier
thread as the busiest group to f_b_g(). Here we are assuming a lower
CPU number will be equivalent to a lower SMT thread number.
It also creates a new SD_ASYM_PACKING flag to enable this feature at
any scheduler domain level.
It also creates an arch hook to enable this feature at the sibling
level. The default function doesn't enable this feature.
Based heavily on patch from Peter Zijlstra.
Fixes from Srivatsa Vaddagiri.
Signed-off-by: Michael Neuling <mikey@neuling.org>
Signed-off-by: Srivatsa Vaddagiri <vatsa@linux.vnet.ibm.com>
Signed-off-by: Peter Zijlstra <a.p.zijlstra@chello.nl>
Cc: Arjan van de Ven <arjan@linux.intel.com>
Cc: "H. Peter Anvin" <hpa@zytor.com>
Cc: Thomas Gleixner <tglx@linutronix.de>
LKML-Reference: <20100608045702.2936CCC897@localhost.localdomain>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
Introduce numa_mem_id(), based on generic percpu variable infrastructure
to track "nearest node with memory" for archs that support memoryless
nodes.
Define API in <linux/topology.h> when CONFIG_HAVE_MEMORYLESS_NODES
defined, else stubs. Architectures will define HAVE_MEMORYLESS_NODES
if/when they support them.
Archs can override definitions of:
numa_mem_id() - returns node number of "local memory" node
set_numa_mem() - initialize [this cpus'] per cpu variable 'numa_mem'
cpu_to_mem() - return numa_mem for specified cpu; may be used as lvalue
Generic initialization of 'numa_mem' occurs in __build_all_zonelists().
This will initialize the boot cpu at boot time, and all cpus on change of
numa_zonelist_order, or when node or memory hot-plug requires zonelist
rebuild. Archs that support memoryless nodes will need to initialize
'numa_mem' for secondary cpus as they're brought on-line.
[akpm@linux-foundation.org: fix build]
Signed-off-by: Lee Schermerhorn <lee.schermerhorn@hp.com>
Signed-off-by: Christoph Lameter <cl@linux-foundation.org>
Cc: Tejun Heo <tj@kernel.org>
Cc: Mel Gorman <mel@csn.ul.ie>
Cc: Christoph Lameter <cl@linux-foundation.org>
Cc: Nick Piggin <npiggin@suse.de>
Cc: David Rientjes <rientjes@google.com>
Cc: Eric Whitney <eric.whitney@hp.com>
Cc: KAMEZAWA Hiroyuki <kamezawa.hiroyu@jp.fujitsu.com>
Cc: Ingo Molnar <mingo@elte.hu>
Cc: Thomas Gleixner <tglx@linutronix.de>
Cc: "H. Peter Anvin" <hpa@zytor.com>
Cc: "Luck, Tony" <tony.luck@intel.com>
Cc: Pekka Enberg <penberg@cs.helsinki.fi>
Cc: <linux-arch@vger.kernel.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
Rework the generic version of the numa_node_id() function to use the new
generic percpu variable infrastructure.
Guard the new implementation with a new config option:
CONFIG_USE_PERCPU_NUMA_NODE_ID.
Archs which support this new implemention will default this option to 'y'
when NUMA is configured. This config option could be removed if/when all
archs switch over to the generic percpu implementation of numa_node_id().
Arch support involves:
1) converting any existing per cpu variable implementations to use
this implementation. x86_64 is an instance of such an arch.
2) archs that don't use a per cpu variable for numa_node_id() will
need to initialize the new per cpu variable "numa_node" as cpus
are brought on-line. ia64 is an example.
3) Defining USE_PERCPU_NUMA_NODE_ID in arch dependent Kconfig--e.g.,
when NUMA is configured. This is required because I have
retained the old implementation by default to allow archs to
be modified incrementally, as desired.
Subsequent patches will convert x86_64 and ia64 to use this implemenation.
Signed-off-by: Lee Schermerhorn <lee.schermerhorn@hp.com>
Cc: Tejun Heo <tj@kernel.org>
Cc: Mel Gorman <mel@csn.ul.ie>
Reviewed-by: Christoph Lameter <cl@linux-foundation.org>
Cc: Nick Piggin <npiggin@suse.de>
Cc: David Rientjes <rientjes@google.com>
Cc: Eric Whitney <eric.whitney@hp.com>
Cc: KAMEZAWA Hiroyuki <kamezawa.hiroyu@jp.fujitsu.com>
Cc: Ingo Molnar <mingo@elte.hu>
Cc: Thomas Gleixner <tglx@linutronix.de>
Cc: "H. Peter Anvin" <hpa@zytor.com>
Cc: "Luck, Tony" <tony.luck@intel.com>
Cc: Pekka Enberg <penberg@cs.helsinki.fi>
Cc: <linux-arch@vger.kernel.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
SD_PREFER_SIBLING is set at the CPU domain level if power saving isn't
enabled, leading to many cache misses on large machines as we traverse
looking for an idle shared cache to wake to. Change the enabler of
select_idle_sibling() to SD_SHARE_PKG_RESOURCES, and enable same at the
sibling domain level.
Reported-by: Lin Ming <ming.m.lin@intel.com>
Signed-off-by: Mike Galbraith <efault@gmx.de>
Signed-off-by: Peter Zijlstra <a.p.zijlstra@chello.nl>
LKML-Reference: <1262612696.15495.15.camel@marge.simson.net>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
Yanmin reported that both tbench and hackbench were significantly
hurt by trying to keep tasks local on these domains, esp on small
cache machines.
So disable it in order to promote spreading outside of the cache
domains.
Reported-by: "Zhang, Yanmin" <yanmin_zhang@linux.intel.com>
Signed-off-by: Peter Zijlstra <a.p.zijlstra@chello.nl>
CC: Mike Galbraith <efault@gmx.de>
LKML-Reference: <1255083400.8802.15.camel@laptop>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
Sysbench thinks SD_BALANCE_WAKE is too agressive and kbuild doesn't
really mind too much, SD_BALANCE_NEWIDLE picks up most of the
slack.
On a dual socket, quad core, dual thread nehalem system:
sysbench (--num_threads=16):
SD_BALANCE_WAKE-: 13982 tx/s
SD_BALANCE_WAKE+: 15688 tx/s
kbuild (-j16):
SD_BALANCE_WAKE-: 47.648295846 seconds time elapsed ( +- 0.312% )
SD_BALANCE_WAKE+: 47.608607360 seconds time elapsed ( +- 0.026% )
(same within noise)
Signed-off-by: Peter Zijlstra <a.p.zijlstra@chello.nl>
LKML-Reference: <new-submission>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
And turn it on for NUMA and MC domains. This improves
locality in balancing decisions by keeping up to
capacity amount of tasks local before looking for idle
CPUs. (and twice the capacity if SD_POWERSAVINGS_BALANCE
is set.)
Signed-off-by: Peter Zijlstra <a.p.zijlstra@chello.nl>
LKML-Reference: <new-submission>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
If we're looking to place a new task, we might as well find the
idlest position _now_, not 1 tick ago.
Signed-off-by: Peter Zijlstra <a.p.zijlstra@chello.nl>
LKML-Reference: <new-submission>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
Make the idle balancer more agressive, to improve a
x264 encoding workload provided by Jason Garrett-Glaser:
NEXT_BUDDY NO_LB_BIAS
encoded 600 frames, 252.82 fps, 22096.60 kb/s
encoded 600 frames, 250.69 fps, 22096.60 kb/s
encoded 600 frames, 245.76 fps, 22096.60 kb/s
NO_NEXT_BUDDY LB_BIAS
encoded 600 frames, 344.44 fps, 22096.60 kb/s
encoded 600 frames, 346.66 fps, 22096.60 kb/s
encoded 600 frames, 352.59 fps, 22096.60 kb/s
NO_NEXT_BUDDY NO_LB_BIAS
encoded 600 frames, 425.75 fps, 22096.60 kb/s
encoded 600 frames, 425.45 fps, 22096.60 kb/s
encoded 600 frames, 422.49 fps, 22096.60 kb/s
Peter pointed out that this is better done via newidle_idx,
not via LB_BIAS, newidle balancing should look for where
there is load _now_, not where there was load 2 ticks ago.
Worst-case latencies are improved as well as no buddies
means less vruntime spread. (as per prior lkml discussions)
This change improves kbuild-peak parallelism as well.
Reported-by: Jason Garrett-Glaser <darkshikari@gmail.com>
Signed-off-by: Mike Galbraith <efault@gmx.de>
Signed-off-by: Peter Zijlstra <a.p.zijlstra@chello.nl>
LKML-Reference: <1253011667.9128.16.camel@marge.simson.net>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
CPU level should have WAKE_AFFINE, whereas ALLNODES is dubious.
Signed-off-by: Peter Zijlstra <a.p.zijlstra@chello.nl>
LKML-Reference: <new-submission>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
When merging select_task_rq_fair() and sched_balance_self() we lost
the use of wake_idx, restore that and set them to 0 to make wake
balancing more aggressive.
Signed-off-by: Peter Zijlstra <a.p.zijlstra@chello.nl>
LKML-Reference: <new-submission>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
The problem with wake_idle() is that is doesn't respect things like
cpu_power, which means it doesn't deal well with SMT nor the recent
RT interaction.
To cure this, it needs to do what sched_balance_self() does, which
leads to the possibility of merging select_task_rq_fair() and
sched_balance_self().
Modify sched_balance_self() to:
- update_shares() when walking up the domain tree,
(it only called it for the top domain, but it should
have done this anyway), which allows us to remove
this ugly bit from try_to_wake_up().
- do wake_affine() on the smallest domain that contains
both this (the waking) and the prev (the wakee) cpu for
WAKE invocations.
Then use the top-down balance steps it had to replace wake_idle().
This leads to the dissapearance of SD_WAKE_BALANCE and
SD_WAKE_IDLE_FAR, with SD_WAKE_IDLE replaced with SD_BALANCE_WAKE.
SD_WAKE_AFFINE needs SD_BALANCE_WAKE to be effective.
Touch all topology bits to replace the old with new SD flags --
platforms might need re-tuning, enabling SD_BALANCE_WAKE
conditionally on a NUMA distance seems like a good additional
feature, magny-core and small nehalem systems would want this
enabled, systems with slow interconnects would not.
Signed-off-by: Peter Zijlstra <a.p.zijlstra@chello.nl>
LKML-Reference: <new-submission>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
Now that SD_WAKE_IDLE doesn't make pipe-test suck anymore,
enable it by default for MC, CPU and NUMA domains.
Signed-off-by: Peter Zijlstra <a.p.zijlstra@chello.nl>
LKML-Reference: <new-submission>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
Re-organize the flag settings so that it's visible at a glance
which sched-domains flags are set and which not.
With the new balancer code we'll need to re-tune these details
anyway, so make it cleaner to make fewer mistakes down the
road ;-)
Cc: Peter Zijlstra <a.p.zijlstra@chello.nl>
Cc: Andreas Herrmann <andreas.herrmann3@amd.com>
Cc: Andreas Herrmann <andreas.herrmann3@amd.com>
Cc: Gautham R Shenoy <ego@in.ibm.com>
Cc: Balbir Singh <balbir@in.ibm.com>
LKML-Reference: <new-submission>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
The idea is that multi-threading a core yields more work
capacity than a single thread, provide a way to express a
static gain for threads.
Signed-off-by: Peter Zijlstra <a.p.zijlstra@chello.nl>
Tested-by: Andreas Herrmann <andreas.herrmann3@amd.com>
Acked-by: Andreas Herrmann <andreas.herrmann3@amd.com>
Acked-by: Gautham R Shenoy <ego@in.ibm.com>
Cc: Balbir Singh <balbir@in.ibm.com>
LKML-Reference: <20090901083826.073345955@chello.nl>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
Impact: cleanup, potential bugfix
Not sure what changed to expose this, but clearly that numa_node_id()
doesn't belong in mmzone.h (the inline in gfp.h is probably overkill, too).
In file included from include/linux/topology.h:34,
from arch/x86/mm/numa.c:2:
/home/rusty/patches-cpumask/linux-2.6/arch/x86/include/asm/topology.h:64:1: warning: "numa_node_id" redefined
In file included from include/linux/topology.h:32,
from arch/x86/mm/numa.c:2:
include/linux/mmzone.h:770:1: warning: this is the location of the previous definition
Signed-off-by: Rusty Russell <rusty@rustcorp.com.au>
Cc: Mike Travis <travis@sgi.com>
LKML-Reference: <200903132343.37661.rusty@rustcorp.com.au>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
Impact: cleanup
node_to_cpumask (and the blecherous node_to_cpumask_ptr which
contained a declaration) are replaced now everyone implements
cpumask_of_node.
Signed-off-by: Rusty Russell <rusty@rustcorp.com.au>
Impact: reduce stack usage, use new cpumask API.
This actually uses topology_core_cpumask() and
topology_thread_cpumask(), removing the only users of
topology_core_siblings() and topology_thread_siblings()
Signed-off-by: Rusty Russell <rusty@rustcorp.com.au>
Signed-off-by: Mike Travis <travis@sgi.com>
Cc: linux-net-drivers@solarflare.com
Impact: change task balancing to save power more agressively
Add SD_BALANCE_NEWIDLE flag at MC level and CPU level
if sched_mc is set. This helps power savings and
will not affect performance when sched_mc=0
Ingo and Mike Galbraith have optimised the SD flags by
removing SD_BALANCE_NEWIDLE at MC and CPU level. This
helps performance but hurts power savings since this
slows down task consolidation by reducing the number
of times load_balance is run.
sched: fine-tune SD_MC_INIT
commit 1480098470
Author: Mike Galbraith <efault@gmx.de>
Date: Fri Nov 7 15:26:50 2008 +0100
sched: re-tune balancing -- revert
commit 9fcd18c9e6
Author: Ingo Molnar <mingo@elte.hu>
Date: Wed Nov 5 16:52:08 2008 +0100
This patch selectively enables SD_BALANCE_NEWIDLE flag
only when sched_mc is set to 1 or 2. This helps power savings
by task consolidation and also does not hurt performance at
sched_mc=0 where all power saving optimisations are turned off.
Signed-off-by: Vaidyanathan Srinivasan <svaidy@linux.vnet.ibm.com>
Acked-by: Balbir Singh <balbir@linux.vnet.ibm.com>
Acked-by: Peter Zijlstra <a.p.zijlstra@chello.nl>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
Impact: cleanup
BALANCE_FOR_MC_POWER and similar macros defined in sched.h are
not constants and have various condition checks and significant
amount of code that is not suitable to be contain in a macro.
Also there could be side effects on the expressions passed to
some of them like test_sd_parent().
This patch converts all complex macros related to power savings
balance to inline functions.
Signed-off-by: Vaidyanathan Srinivasan <svaidy@linux.vnet.ibm.com>
Acked-by: Balbir Singh <balbir@linux.vnet.ibm.com>
Acked-by: Peter Zijlstra <a.p.zijlstra@chello.nl>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
Change arch_update_cpu_topology so it returns 1 if the cpu topology changed
and 0 if it didn't change. This will be useful for the next patch which adds
a call to this function in partition_sched_domains.
Signed-off-by: Heiko Carstens <heiko.carstens@de.ibm.com>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
Tune SD_MC_INIT the same way as SD_CPU_INIT:
unset SD_BALANCE_NEWIDLE, and set SD_WAKE_BALANCE.
This improves vmark by 5%:
vmark 132102 125968 125497 messages/sec avg 127855.66 .984
vmark 139404 131719 131272 messages/sec avg 134131.66 1.033
Signed-off-by: Mike Galbraith <efault@gmx.de>
Acked-by: Peter Zijlstra <a.p.zijlstra@chello.nl>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
# *DOCUMENTATION*
Impact: improve wakeup affinity on NUMA systems, tweak SMP systems
Given the fixes+tweaks to the wakeup-buddy code, re-tweak the domain
balancing defaults on NUMA and SMP systems.
Turn on SD_WAKE_AFFINE which was off on x86 NUMA - there's no reason
why we would not want to have wakeup affinity across nodes as well.
(we already do this in the standard NUMA template.)
lat_ctx on a NUMA box is particularly happy about this change:
before:
| phoenix:~/l> ./lat_ctx -s 0 2
| "size=0k ovr=2.60
| 2 5.70
after:
| phoenix:~/l> ./lat_ctx -s 0 2
| "size=0k ovr=2.65
| 2 2.07
a 2.75x speedup.
pipe-test is similarly happy about it too:
| phoenix:~/sched-tests> ./pipe-test
| 18.26 usecs/loop.
| 14.70 usecs/loop.
| 14.38 usecs/loop.
| 10.55 usecs/loop. # +WAKE_AFFINE on domain0+domain1
| 8.63 usecs/loop.
| 8.59 usecs/loop.
| 9.03 usecs/loop.
| 8.94 usecs/loop.
| 8.96 usecs/loop.
| 8.63 usecs/loop.
Also:
- disable SD_BALANCE_NEWIDLE on NUMA and SMP domains (keep it for siblings)
- enable SD_WAKE_BALANCE on SMP domains
Sysbench+postgresql improves all around the board, quite significantly:
.28-rc3-11474e2c .28-rc3-11474e2c-tune
-------------------------------------------------
1: 571 688 +17.08%
2: 1236 1206 -2.55%
4: 2381 2642 +9.89%
8: 4958 5164 +3.99%
16: 9580 9574 -0.07%
32: 7128 8118 +12.20%
64: 7342 8266 +11.18%
128: 7342 8064 +8.95%
256: 7519 7884 +4.62%
512: 7350 7731 +4.93%
-------------------------------------------------
SUM: 55412 59341 +6.62%
So it's a win both for the runup portion, the peak area and the tail.
Signed-off-by: Ingo Molnar <mingo@elte.hu>
This can result in an empty topology directory in sysfs, and requires
in-kernel users to protect all uses with #ifdef - see
<http://marc.info/?l=linux-netdev&m=120639033904472&w=2>.
The documentation of CPU topology specifies what the defaults should be if
only partial information is available from the hardware. So we can
provide these defaults as a fallback.
This patch:
- Adds default definitions of the 4 topology macros to <linux/topology.h>
- Changes drivers/base/topology.c to use the topology macros unconditionally
and to cope with definitions that aren't lvalues
- Updates documentation accordingly
[ From: Andrew Morton <akpm@linux-foundation.org>
- fold now-duplicated code
- fix layout
]
Signed-off-by: Ben Hutchings <bhutchings@solarflare.com>
Cc: Vegard Nossum <vegard.nossum@gmail.com>
Cc: Nick Piggin <nickpiggin@yahoo.com.au>
Cc: Chandra Seetharaman <sekharan@us.ibm.com>
Cc: Suresh Siddha <suresh.b.siddha@intel.com>
Cc: Mike Travis <travis@sgi.com>
Cc: Christoph Lameter <clameter@sgi.com>
Cc: John Hawkes <hawkes@sgi.com>
Cc: Zhang, Yanmin <yanmin.zhang@intel.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
improve the sysbench ramp-up phase and its peak throughput on
a 16way NUMA box, by turning on WAKE_AFFINE:
tip/sched tip/sched+wake-affine
-------------------------------------------------
1: 700 830 +15.65%
2: 1465 1391 -5.28%
4: 3017 3105 +2.81%
8: 5100 6021 +15.30%
16: 10725 10745 +0.19%
32: 10135 10150 +0.16%
64: 9338 9240 -1.06%
128: 8599 8252 -4.21%
256: 8475 8144 -4.07%
-------------------------------------------------
SUM: 57558 57882 +0.56%
this change also improves lat_ctx from 6.69 usecs to 1.11 usec:
$ ./lat_ctx -s 0 2
"size=0k ovr=1.19
2 1.11
$ ./lat_ctx -s 0 2
"size=0k ovr=1.22
2 6.69
in sysbench it's an overall win with some weakness at the lots-of-clients
side. That happens because we now under-balance this workload
a bit. To counter that effect, turn on NEWIDLE:
wake-idle wake-idle+newidle
-------------------------------------------------
1: 830 834 +0.43%
2: 1391 1401 +0.65%
4: 3105 3091 -0.43%
8: 6021 6046 +0.42%
16: 10745 10736 -0.08%
32: 10150 10206 +0.55%
64: 9240 9533 +3.08%
128: 8252 8355 +1.24%
256: 8144 8384 +2.87%
-------------------------------------------------
SUM: 57882 58591 +1.21%
as a bonus this not only improves the many-clients case but
also improves the (more important) rampup phase.
sysbench is a workload that quickly breaks down if the
scheduler over-balances, so since it showed an improvement
under NEWIDLE this change is definitely good.
* Remove empty cpumask_t (and all non-zero/non-null) variables
in SD_*_INIT macros. Use memset(0) to clear. Also, don't
inline the initializer functions to save on stack space in
build_sched_domains().
* Merge change to include/linux/topology.h that uses the new
node_to_cpumask_ptr function in the nr_cpus_node macro into
this patch.
Depends on:
[mm-patch]: asm-generic-add-node_to_cpumask_ptr-macro.patch
[sched-devel]: sched: add new set_cpus_allowed_ptr function
Cc: H. Peter Anvin <hpa@zytor.com>
Signed-off-by: Mike Travis <travis@sgi.com>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
Will be called each time the scheduling domains are rebuild.
Needed for architectures that don't have a static cpu topology.
Signed-off-by: Heiko Carstens <heiko.carstens@de.ibm.com>
Signed-off-by: Martin Schwidefsky <schwidefsky@de.ibm.com>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
WAKE_IDLE is too agressive on multi-core CPUs with the new
wake-affine code, keep it on for SMT/HT balancing alone
(where there's no cache affinity at all between logical CPUs).
Signed-off-by: Ingo Molnar <mingo@elte.hu>
most multicore CPUs today have shared L2 caches, so tune things so
that the spreading amongst cores is more aggressive.
Signed-off-by: Ingo Molnar <mingo@elte.hu>
Mel Gorman
Vincent Guittot
Christoph Lameter
Jason Low
Alex Shi
Linus Torvalds
Peter Zijlstra
Peter Zijlstra
Anton Blanchard
KOSAKI Motohiro
Heiko Carstens
Lee Schermerhorn
Michael Neuling
Mike Galbraith
Rusty Russell
Ingo Molnar
Vaidyanathan Srinivasan
Ben Hutchings
Mike Travis