Quoting Mika Kuoppala (2020-02-27 10:51:46)
> Daniele Ceraolo Spurio <daniele.ceraolospu...@intel.com> writes:
> > static void gen8_set_pte(void __iomem *addr, gen8_pte_t pte)
> > {
> > writeq(pte, addr);
> > @@ -172,7 +179,7 @@ static void gen8_ggtt_insert_page(struct
> > i915_address_space *vm,
> > gen8_pte_t __iomem *pte =
> > (gen8_pte_t __iomem *)ggtt->gsm + offset / I915_GTT_PAGE_SIZE;
> >
> > - gen8_set_pte(pte, gen8_pte_encode(addr, level, 0));
> > + gen8_set_pte(pte, gen8_ggtt_pte_encode(addr, level, 0));
>
> Make me ponder why we don't use the vm->pte_encode all the way
> as we have it :P
Don't use vfuncs if the target routine is known.
Apparently it's even cheaper to do a bsearch on a page of target
routines than it is to use retpolines.
Anyway, the era of vfuncs is over.
-Chris
_______________________________________________
Intel-gfx mailing list
Intel-gfx@lists.freedesktop.org
https://lists.freedesktop.org/mailman/listinfo/intel-gfx
