Some Optimizations Inbound In The Linux Kernel Space For x86/x86_64
Of the locking/misc x86 work for this next version of the kernel there are lockdep scalability improvements and micro-optimizations, spin lock micro-optimizations, better fairness for qspinlocks, and support for relative references in jump tables to optimize jump labels.
That pull also has updates to the Linux Kernel Memory Consistency Model that was merged earlier this year plus other low-level x86 kernel work.
Meanwhile in x86/mm for the memory management work there is work postponed from last cycle for making the lazy TLB mode even lazier as well as to always use this mode. The x86 memory management work also has CPA optimizations, support for kdump'ing encrypted memory with AMD SME enabled, and other enhancements.
It will be interesting to see if any of these low-level kernel improvements in the x86 space translate to real-world performance improvements. Of course, once the merge window is over on the 4th of November, I'll begin with my usual benchmarking dance.
5 Comments
