5 ms·

I wonder if a CPU ISA change couldn't make virtualization more efficient? Dedicated instructions for IO for example instead of using Load/Store.. D But I don't

• xxpor 6y ago
  I guess the question is, what would the new instructions do differently? Plus, everything out there supports DMA already. Would you have to buy different HW to support this new scheme? An IOMMU + DMA is already pretty good, a lot of the bottleneck at this point is really in the IO devices themselves these days (network cards and SSDs are only so fast).
  • fithisux 6y ago
    The problem is that not all device drivers take advantage of IOMMU, to run device drivers as user level programs.
    • xxpor 6y ago
      Maybe I'm just ignorant, but my impression was that the device drivers don't have a choice if they're running in a VM. The hypervisor sets it up so that all memory access was forced through the IOMMU.
• usr1106 6y ago
  > I wonder if a CPU ISA change couldn't make virtualization more efficient Without knowing the details I am convinced the architecture has an impact. Until some 15 years ago Intel archtecture had the reputation of being impossible to virtualize. I don't remember whether KVM which requires HW support or Xen which doesn't but requires support by the guest (paravirtualization) was first. I guess it was the latter, but either way they became feasible in rather quick succession. On the other side IBM /370 architecture was designed for virtualization from the beginning. Already in the early 1980s virtualization was standard (possibly even earlier). I'm quite sure difficult virtualization does not come for free efficiency-wise either.