CPU load balancing in Heterogeneous Multicore Processing (HMP) architecture

Hi everyone,

I am looking for information regarding CPU task load balancing policy in Jetson devices.
Jetson devices have multiple clusters of CPU cores with different performance/power characteristics. I was wondering what are the criteria for mapping tasks to little and big cores, and when the migration happens. (pointer to kernel source would be super helpful)
Global Task Scheduling (GTS) seems to be the state-of-the-art technique, however, I couldn’t find relevant kernel configs in L4T’s kernel.
So, my question is what is the task-to-core mapping and migration policy implemented in Jetson devices?


For TX2, there is a special rule:
Cannot enable denver cores for TX2 (Jetpack 4.4 DP) - #40 by DaneLLL
You would need to manually schedule tasks to Denver cores.

For other Jetson platforms( including non-Denver cores in TX2 ), it is default policy in kernel and we don’s have private rule for it. Please check
Linux Scheduler — The Linux Kernel documentation

Thank you very much! That answered my question and it was very helpful.