Different platforms have very different cluster / numa topologies. Therefore we have to add a generic option which accepts any core order or sequence for lockhammer to refer.
We add -o option with comma separated string as core order input. Then add another lstopo parser to test_lockhammer.py which can automatically generate the default order.
In future, we can easily extend test_lockhammer.py and add more complex algorithms to deduce the "best" core pinning order for each platform without modifying lockhammer.c
Different platforms have very different cluster / numa topologies. Therefore we have to add a generic option which accepts any core order or sequence for lockhammer to refer.
We add -o option with comma separated string as core order input. Then add another lstopo parser to test_lockhammer.py which can automatically generate the default order.
In future, we can easily extend test_lockhammer.py and add more complex algorithms to deduce the "best" core pinning order for each platform without modifying lockhammer.c