https://bugzilla.redhat.com/show_bug.cgi?id=1988151
--- Comment #7 from Ben Beasley <[email protected]> --- I’ll go ahead and create the tracker bugs for the missing architectures. It might still be possible to resolve them with some care, and perhaps consultation with upstream. I think I now basically understand the requirements for a minimal “empty” spin_loop_pause(). The fallback for ARMs that don’t support the “yield” instruction is a good example. It’s a “nop” instruction plus a memory clobber, to create a “compiler barrier” that keeps the compiler from eliding, hoisting, or otherwise subverting the pause function. This is exactly what the example you gave for PowerPC is doing. “or 1 1 1” is an idiomatic nop for PowerPC, or’ing a register with itself to waste time with no effect. I’m not so sure about the s390x example from Linux. As far as I can understand from the limited documentation I’ve found, it’s a conditional branch where the condition is zero so the branch is not taken. However, it seems the first argument somehow causes the BCR instruction to have a synchronization function as well (https://github.com/golang/go/issues/42479), although I haven’t found good documentation on the exact effects. I assume this is intentional, and I wonder why it is needed. -- You are receiving this mail because: You are on the CC list for the bug. You are always notified about changes to this product and component _______________________________________________ package-review mailing list -- [email protected] To unsubscribe send an email to [email protected] Fedora Code of Conduct: https://docs.fedoraproject.org/en-US/project/code-of-conduct/ List Guidelines: https://fedoraproject.org/wiki/Mailing_list_guidelines List Archives: https://lists.fedoraproject.org/archives/list/[email protected] Do not reply to spam on the list, report it: https://pagure.io/fedora-infrastructure
