From: Keith Busch <keith.busch@intel.com>
To: linux-kernel@vger.kernel.org, linux-acpi@vger.kernel.org,
linux-mm@kvack.org
Cc: Greg Kroah-Hartman <gregkh@linuxfoundation.org>,
Rafael Wysocki <rafael@kernel.org>,
Dave Hansen <dave.hansen@intel.com>,
Dan Williams <dan.j.williams@intel.com>,
Keith Busch <keith.busch@intel.com>
Subject: [PATCHv2 12/12] doc/mm: New documentation for memory performance
Date: Mon, 10 Dec 2018 18:03:10 -0700 [thread overview]
Message-ID: <20181211010310.8551-13-keith.busch@intel.com> (raw)
In-Reply-To: <20181211010310.8551-1-keith.busch@intel.com>
Platforms may provide system memory where some physical address ranges
perform differently than others, or is side cached by the system.
Add documentation describing a high level overview of such systems and the
performance and caching attributes the kernel provides for applications
wishing to query this information.
Signed-off-by: Keith Busch <keith.busch@intel.com>
---
Documentation/admin-guide/mm/numaperf.rst | 171 ++++++++++++++++++++++++++++++
1 file changed, 171 insertions(+)
create mode 100644 Documentation/admin-guide/mm/numaperf.rst
diff --git a/Documentation/admin-guide/mm/numaperf.rst b/Documentation/admin-guide/mm/numaperf.rst
new file mode 100644
index 000000000000..846b3f991e7f
--- /dev/null
+++ b/Documentation/admin-guide/mm/numaperf.rst
@@ -0,0 +1,171 @@
+.. _numaperf:
+
+=============
+NUMA Locality
+=============
+
+Some platforms may have multiple types of memory attached to a single
+CPU. These disparate memory ranges share some characteristics, such as
+CPU cache coherence, but may have different performance. For example,
+different media types and buses affect bandwidth and latency.
+
+A system supporting such heterogeneous memory by grouping each memory
+type under different "nodes" based on similar CPU locality and performance
+characteristics. Some memory may share the same node as a CPU, and others
+are provided as memory only nodes. While memory only nodes do not provide
+CPUs, they may still be directly accessible, or local, to one or more
+compute nodes. The following diagram shows one such example of two compute
+noes with local memory and a memory only node for each of compute node:
+
+ +------------------+ +------------------+
+ | Compute Node 0 +-----+ Compute Node 1 |
+ | Local Node0 Mem | | Local Node1 Mem |
+ +--------+---------+ +--------+---------+
+ | |
+ +--------+---------+ +--------+---------+
+ | Slower Node2 Mem | | Slower Node3 Mem |
+ +------------------+ +--------+---------+
+
+A "memory initiator" is a node containing one or more devices such as
+CPUs or separate memory I/O devices that can initiate memory requests. A
+"memory target" is a node containing one or more accessible physical
+address ranges from one or more memory initiators.
+
+When multiple memory initiators exist, they may not all have the same
+performance when accessing a given memory target. The highest performing
+initiator to a given target is considered to be one of that target's
+local initiators. Any given target may have one or more local initiators,
+and any given initiator may have multiple local memory targets.
+
+To aid applications matching memory targets with their initiators,
+the kernel provide symlinks to each other like the following example::
+
+ # ls -l /sys/devices/system/node/nodeX/local_target*
+ /sys/devices/system/node/nodeX/local_targetY -> ../nodeY
+
+ # ls -l /sys/devices/system/node/nodeY/local_initiator*
+ /sys/devices/system/node/nodeY/local_initiatorX -> ../nodeX
+
+The linked nodes will also have their node number set in the local_mem
+and local_cpu node list and maps.
+
+An example showing how this may be used to run a particular task on CPUs
+and memory that are both local to a particular PCI device can be done
+using existing 'numactl' as follows::
+
+ # NODE=$(cat /sys/devices/pci:0000:00/.../numa_node)
+ # numactl --membind=$(cat /sys/devices/node/node${NODE}/local_mem_nodelist) \
+ --cpunodebind=$(cat /sys/devices/node/node${NODE}/local_cpu_nodelist) \
+ -- <some-program-to-execute>
+
+================
+NUMA Performance
+================
+
+Applications may wish to consider which node they want their memory to
+be allocated from based on the node's performance characteristics. If the
+system provides these attributes, the kernel exports them under the node
+sysfs hierarchy by appending the local_initiator_access directory under
+the memory node as follows::
+
+ /sys/devices/system/node/nodeY/local_initiator_access/
+
+The kernel does not provide performance attributes for non-local memory
+initiators. These attributes apply only to the memory initiator nodes that
+have a local_initiatorX link, or are set in the local_cpu_nodelist. A
+memory initiator node is considered local to itself if it also is
+a memory target and will be set it its node list and map, but won't
+contain a symlink to itself.
+
+The performance characteristics the kernel provides for the local initiators
+are exported are as follows::
+
+ # tree /sys/devices/system/node/nodeY/local_initiator_access
+ /sys/devices/system/node/nodeY/local_initiator_access
+ |-- read_bandwidth
+ |-- read_latency
+ |-- write_bandwidth
+ `-- write_latency
+
+The bandwidth attributes are provided in MiB/second.
+
+The latency attributes are provided in nanoseconds.
+
+==========
+NUMA Cache
+==========
+
+System memory may be constructed in a hierarchy of elements with various
+performance characteristics in order to provide large address space
+of slower performing memory side-cached by a smaller higher performing
+memory. The system physical addresses that initiators are aware of is
+provided by the last memory level in the hierarchy, while the system uses
+higher performing memory to transparently cache access to progressively
+slower levels.
+
+The term "far memory" is used to denote the last level memory in the
+hierarchy. Each increasing cache level provides higher performing
+initiator access, and the term "near memory" represents the fastest
+cache provided by the system.
+
+This numbering is different than CPU caches where the cache level (ex:
+L1, L2, L3) uses a CPU centric view with each increased level is lower
+performing. In contrast, the memory cache level is centric to the last
+level memory, so the higher numbered cache level denotes memory nearer
+to the CPU, and further from far memory.
+
+The memory side caches are not directly addressable by software. When
+software accesses a system address, the system will return it from the
+near memory cache if it is present. If it is not present, the system
+accesses the next level of memory until there is either a hit in that
+cache level, or it reaches far memory.
+
+An application does not need to know about caching attributes in order
+to use the system, software may optionally query the memory cache
+attributes in order to maximize the performance out of such a setup.
+If the system provides a way for the kernel to discover this information,
+for example with ACPI HMAT (Heterogeneous Memory Attribute Table),
+the kernel will append these attributes to the NUMA node memory target.
+
+When the kernel first registers a memory cache with a node, the kernel
+will create the following directory::
+
+ /sys/devices/system/node/nodeX/side_cache/
+
+If that directory is not present, the system either does not not provide
+a memory side cache, or that information is not accessible to the kernel.
+
+The attributes for each level of cache is provided under its cache
+level index::
+
+ /sys/devices/system/node/nodeX/side_cache/indexA/
+ /sys/devices/system/node/nodeX/side_cache/indexB/
+ /sys/devices/system/node/nodeX/side_cache/indexC/
+
+Each cache level's directory provides its attributes. For example,
+the following is a single cache level and the attributes available for
+software to query::
+
+ # tree sys/devices/system/node/node0/side_cache/
+ /sys/devices/system/node/node0/side_cache/
+ |-- index1
+ | |-- associativity
+ | |-- level
+ | |-- line_size
+ | |-- size
+ | `-- write_policy
+
+The "associativity" will be 0 if it is a direct-mapped cache, and non-zero
+for any other indexed based, multi-way associativity.
+
+The "level" is the distance from the far memory, and matches the number
+appended to its "index" directory.
+
+The "line_size" is the number of bytes accessed on a cache miss.
+
+The "size" is the number of bytes provided by this cache level.
+
+The "write_policy" will be 0 for write-back, and non-zero for
+write-through caching.
+
+See also: https://www.uefi.org/sites/default/files/resources/ACPI_6_2.pdf
--
2.14.4
next prev parent reply other threads:[~2018-12-11 1:06 UTC|newest]
Thread overview: 29+ messages / expand[flat|nested] mbox.gz Atom feed top
2018-12-11 1:02 [PATCHv2 00/12] Heterogeneous memory node attributes Keith Busch
2018-12-11 1:02 ` [PATCHv2 01/12] acpi: Create subtable parsing infrastructure Keith Busch
2018-12-11 9:44 ` Rafael J. Wysocki
2018-12-19 23:19 ` Schmauss, Erik
2018-12-19 23:59 ` Dan Williams
2018-12-20 1:15 ` Schmauss, Erik
2018-12-20 8:57 ` Rafael J. Wysocki
2018-12-20 19:00 ` Schmauss, Erik
2018-12-13 9:05 ` kbuild test robot
2018-12-19 22:10 ` kbuild test robot
2018-12-11 1:03 ` [PATCHv2 02/12] acpi/hmat: Parse and report heterogeneous memory Keith Busch
2018-12-11 6:03 ` Dan Williams
2018-12-11 16:55 ` Keith Busch
2018-12-11 20:29 ` Dan Williams
2018-12-11 20:44 ` Keith Busch
2018-12-11 22:50 ` Dan Williams
2018-12-11 1:03 ` [PATCHv2 03/12] node: Link memory nodes to their compute nodes Keith Busch
2018-12-11 1:03 ` [PATCHv2 04/12] Documentation/ABI: Add new node sysfs attributes Keith Busch
2018-12-11 1:03 ` [PATCHv2 05/12] acpi/hmat: Register processor domain to its memory Keith Busch
2018-12-11 1:03 ` [PATCHv2 06/12] node: Add heterogenous memory performance Keith Busch
2018-12-11 1:03 ` [PATCHv2 07/12] Documentation/ABI: Add node performance attributes Keith Busch
2018-12-11 1:03 ` [PATCHv2 08/12] acpi/hmat: Register " Keith Busch
2018-12-11 1:03 ` [PATCHv2 09/12] node: Add memory caching attributes Keith Busch
2018-12-11 1:03 ` [PATCHv2 10/12] Documentation/ABI: Add node cache attributes Keith Busch
2018-12-11 1:03 ` [PATCHv2 11/12] acpi/hmat: Register memory side " Keith Busch
2018-12-11 1:03 ` Keith Busch [this message]
2018-12-11 6:45 ` [PATCHv2 12/12] doc/mm: New documentation for memory performance Mike Rapoport
2018-12-12 4:53 ` Aneesh Kumar K.V
2018-12-12 14:45 ` Keith Busch
Reply instructions:
You may reply publicly to this message via plain-text email
using any one of the following methods:
* Save the following mbox file, import it into your mail client,
and reply-to-all from there: mbox
Avoid top-posting and favor interleaved quoting:
https://en.wikipedia.org/wiki/Posting_style#Interleaved_style
* Reply using the --to, --cc, and --in-reply-to
switches of git-send-email(1):
git send-email \
--in-reply-to=20181211010310.8551-13-keith.busch@intel.com \
--to=keith.busch@intel.com \
--cc=dan.j.williams@intel.com \
--cc=dave.hansen@intel.com \
--cc=gregkh@linuxfoundation.org \
--cc=linux-acpi@vger.kernel.org \
--cc=linux-kernel@vger.kernel.org \
--cc=linux-mm@kvack.org \
--cc=rafael@kernel.org \
/path/to/YOUR_REPLY
https://kernel.org/pub/software/scm/git/docs/git-send-email.html
* If your mail client supports setting the In-Reply-To header
via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line
before the message body.
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox;
as well as URLs for NNTP newsgroup(s).