https://bugs.kde.org/show_bug.cgi?id=415136
--- Comment #2 from Julian Seward <jsew...@acm.org> --- Thanks for the patch. I looked at it in some detail. An administrative request: please could you attach the test changes as a separate patch? With it all in one patch, I have to scroll through the entire giant patch to check there aren't some extra implementation bits somewhere in it. I am going to need some context to make sense of this. In particular, why is ARM64in_CAS emitted with a loop, where it did not have a loop before? It is true that a correct usage of CAS in general requires a loop to retry in case of failure, but that loop must exist in the guest code surrounding the incoming CAS instruction. The role of the IR representation and its host-side translation back into arm64 code is merely to attempt the CAS and return a success/failure indication to the guest state. Also, what is the reason for changing the hardwired register numbers? They now don't match the ARM64in_CAS case in getRegUsage_ARM64Instr. -- You are receiving this mail because: You are watching all bug changes.
