CSharperMantle wrote: > [...] on LoongArch, the kernel page size is typically 16K. If we still use 4K > here, could that introduce unnecessary store instructions during stack > probing?
Yes, but I think taking the min() here is safer. LoongArch kernel supports 4k paging, and AFAIC some translation tools like [Box64](https://github.com/ptitSeb/box64) requires it. Distros like AOSC ships 4k kernel by default. If 16k stride is used, this could be a problem on these systems. Moreover, this knob is adjustable by user code. In cases where this might be a performance bottleneck, user code can specify a larger value (yet, risking the compatibility w/ 4k kernels). https://github.com/llvm/llvm-project/pull/195595 _______________________________________________ cfe-commits mailing list [email protected] https://lists.llvm.org/cgi-bin/mailman/listinfo/cfe-commits
