
On 6/24/20 10:49 AM, Michal Privoznik wrote:
To cite ACPI specification:
Heterogeneous Memory Attribute Table describes the memory attributes, such as memory side cache attributes and bandwidth and latency details, related to the System Physical Address (SPA) Memory Ranges. The software is expected to use this information as hint for optimization.
According to our upstream discussion [1] this is exposed under <numa/> as <cache/> under NUMA <cell/> and <latency> or <bandwidth/> under numa/latencies.
1: https://www.redhat.com/archives/libvir-list/2020-January/msg00422.html
Signed-off-by: Michal Privoznik <mprivozn@redhat.com> --- docs/formatdomain.html.in | 88 ++++++ docs/schemas/cputypes.rng | 110 ++++++- src/conf/numa_conf.c | 350 ++++++++++++++++++++- src/conf/numa_conf.h | 33 ++ src/libvirt_private.syms | 6 + tests/qemuxml2argvdata/numatune-hmat.xml | 52 +++ tests/qemuxml2xmloutdata/numatune-hmat.xml | 1 + tests/qemuxml2xmltest.c | 1 + 8 files changed, 625 insertions(+), 16 deletions(-) create mode 100644 tests/qemuxml2argvdata/numatune-hmat.xml create mode 120000 tests/qemuxml2xmloutdata/numatune-hmat.xml
diff --git a/docs/formatdomain.html.in b/docs/formatdomain.html.in index 07dcca57f5..78b2d2828d 100644 --- a/docs/formatdomain.html.in +++ b/docs/formatdomain.html.in @@ -1931,6 +1931,94 @@ using 10 for local and 20 for remote distances. </p>
+ <h4><a id="hmat">Heterogeneous Memory Attribute Table</a></h4> + +<pre> +... +<cpu> + ... + <numa> + <cell id='0' cpus='0-3' memory='512000' unit='KiB' discard='yes'/> + <cell id='1' cpus='4-7' memory='512000' unit='KiB' memAccess='shared'/> + <cell id='3' cpus='0-3' memory='2097152' unit='KiB'> + <cache level='1' associativity='direct' policy='writeback'> + <size value='10' unit='KiB'/> + <line value='8' unit='B'/> + </cache> + </cell> + <latencies> + <latency initiator='0' target='0' type='access' value='5'/> + <latency initiator='0' target='0' cache='1' type='access' value='10'/> + <bandwidth initiator='0' target='0' type='access' value='204800' unit='KiB'/> + </latencies> + </numa> + ... +</cpu> +...</pre> + + <p> + <span class='since'>Since 6.5.0</span> the <code>cell</code> element can + have <code>cache</code> child element which describes memory side cache
"have a <code>cache</code> child ...."
+ for memory proximity domains. The <code>cache</code> element has
"<code>cache</code> element has a"
+ <code>level</code> attribute describing the cache level and thus the + element can be repeated multiple times to describe different levels of + the cache. + </p> + + <p> + The <code>cache</code> element then has <code>associativity</code> + attribute (accepted values are <code>none</code>, <code>direct</code> and + <code>full</code>) describing the cache associativity, and + <code>policy</code> attribute (accepted values are <code>none</code>, + <code>writeback</code> and <code>writethrough</code>) describing cache + write associativity. The element has two mandatory child elements then: + <code>size</code> and <code>line</code> which describe cache size and + cache line size. Both elements accept two attributes: <code>value</code> + and <code>unit</code> which set the value of corresponding cache + attribute. + </p>
This second paragraph is about the same <cache> element that you started to describe in the previous paragraph. Might as well make a big paragraph talking just about the cache element. And there's a lot of different elements and attributes at different levels to unpack at once here. I think you can use a description list (<dl>) here, like you did below with the <latencies> element, or any other hierarchical HTML struct for that matter, to make it easier to display all the info you're providing here.
+ + <p> + The NUMA description has optional <code>latencies</code> element that
"The NUMA description has an optional ..." Everything else LGTM and these HTML changes are somewhat cosmetic, so: Reviewed-by: Daniel Henrique Barboza <danielhb413@gmail.com>