On Mon, 2024-04-29 at 19:33 -0300, Jason Gunthorpe wrote: > On Mon, Apr 29, 2024 at 04:11:03PM -0600, Alex Williamson wrote: > > > This isn't very performance optimal already, we take a lock on > > > every > > > iteration, so there isn't much point in inlining multiple copies > > > of > > > everything to save an branch. > > > > These macros are to reduce duplicate code blocks and the errors > > that typically come from such duplication, > > But there is still quite a bit of repetition here.. I appears like duplications, I agree - but the vfio_pci_core_ioreadX and vfio_pci_core_iowriteX accessors are exported as such, or might be reused by way of vfio_pci_iordwrX in vfio_pci_core_do_io_rw for arbitrarily sized read/writes, too. > > as well as to provide type safe functions in the spirit of the > > ioread# and iowrite# helpers. > > But it never really takes any advantage of type safety? It is making > a memcpy.. At first, I was overwhelmed by the macro definitions, too. But after a while I started to like the strict typing once the value came out of memcpy or until it is memcpy'd. > > Jason >
