This third version is based on review comments and suggestion
from David.
changes from v2:
- patch 3 from v2: removed
- patch 2:
* added David's R-b
- patch 3 (former 4 in v2):
* added Greg and David R-bs
* added G_STATIC_ASSERT() before memcpy()
- patch 4 (former 5 in v2):
* clarified what 'node 0' means in the commit msg
* rewrote a bit to clarify what the logic does
* the translation of user distance to PAPR distance logic,
previously presented in the former patch 3 in v2, was folded
into spapr_numa_get_numa_level()
* removed needless 'if' when assigning associativities in
spapr_numa_define_associativity_domains()
- patch 5 (former 6 in v2):
* moved the section describing the current logic up
* created a new 'legacy' section describing the pre-5.2
behavior
* added a 'known limitations' section documentating that
we don't support asymmetrical topologies and we do a poor
job approximating 'non-transitive' topologies
v2 link: https://lists.gnu.org/archive/html/qemu-devel/2020-09/msg09073.html
Daniel Henrique Barboza (5):
spapr: add spapr_machine_using_legacy_numa() helper
spapr_numa: forbid asymmetrical NUMA setups
spapr_numa: change reference-points and maxdomain settings
spapr_numa: consider user input when defining associativity
specs/ppc-spapr-numa: update with new NUMA support
docs/specs/ppc-spapr-numa.rst | 206 +++++++++++++++++++++++++++++++++-
hw/ppc/spapr.c | 12 ++
hw/ppc/spapr_numa.c | 195 ++++++++++++++++++++++++++++++--
include/hw/ppc/spapr.h | 2 +
4 files changed, 406 insertions(+), 9 deletions(-)
--
2.26.2
