Josh Poimboeuf from Red Hat has published a patch for the Linux kernel that speeds up the per_thread_ops test by 2.6%, which evaluates the number of operations that can be performed in a single thread. The patch uses pointer masking to reduce the number of slow calls to barrier_nospec() in the 64-bit copy_from_user() function used for copying data into the kernel from user space. Linus Torvalds has already merged this optimization into the kernel branch 6.12.
Source: opennet.ru