<html>
  <head>
    <meta content="text/html; charset=utf-8" http-equiv="Content-Type">
  </head>
  <body bgcolor="#FFFFFF" text="#000000">
    <p>What part of the article are you referring to and how exactly
      does it improve performance for you?</p>
    <p>- Nicolas<br>
    </p>
    <div class="moz-cite-prefix">On 2016-05-21 11:42, Colin Godsey
      wrote:<br>
    </div>
    <blockquote
cite="mid:CAExLznd9Gi_c3zKDccECN2mFJe=AYiQ-+z_soWyQJ=UKt=6xLA@mail.gmail.com"
      type="cite">
      <meta http-equiv="Content-Type" content="text/html; charset=utf-8">
      <div dir="ltr">Windows 10 forcing X2APIC: <a
          moz-do-not-send="true"
          href="https://support.microsoft.com/en-us/kb/2303458"><a class="moz-txt-link-freetext" href="https://support.microsoft.com/en-us/kb/2303458">https://support.microsoft.com/en-us/kb/2303458</a></a>
        <div><br>
        </div>
        <div>This helped a bunch for me at least on 4.4- I think there
          was some confusion on the correct way to signal the guest on
          what APIC to use, forcing this gave me much better %sys times.
          AFAIK it can also still be used with hyper-v APIC because
          that’s just the EOI para-virtualization. </div>
      </div>
      <br>
      <div class="gmail_quote">
        <div dir="ltr">On Fri, May 20, 2016 at 2:28 PM Alex Williamson
          <<a moz-do-not-send="true"
            href="mailto:alex.l.williamson@gmail.com">alex.l.williamson@gmail.com</a>>
          wrote:<br>
        </div>
        <blockquote class="gmail_quote" style="margin:0 0 0
          .8ex;border-left:1px #ccc solid;padding-left:1ex">
          <div dir="ltr">
            <div class="gmail_extra">
              <div class="gmail_quote">On Fri, May 20, 2016 at 11:38 AM,
                Nicolas Roy-Renaud <span dir="ltr"><<a
                    moz-do-not-send="true"
                    href="mailto:nicolas.roy-renaud.1@ens.etsmtl.ca"
                    target="_blank"><a class="moz-txt-link-abbreviated" href="mailto:nicolas.roy-renaud.1@ens.etsmtl.ca">nicolas.roy-renaud.1@ens.etsmtl.ca</a></a>></span>
                wrote:<br>
                <blockquote class="gmail_quote" style="margin:0 0 0
                  .8ex;border-left:1px #ccc solid;padding-left:1ex">
                  <div bgcolor="#FFFFFF" text="#000000">
                    <p>Did you have any issues with efifb or some other
                      driver grabbing the gard before vfio-pci? What are
                      your IOMMU groups like, which chard is your
                      boot_vga, what script or method do you use to get
                      vfio-pci to bind with your card, what's your
                      kernel command line and what modules are included
                      in your initramfs?</p>
                    <p>I've struggled with getting my primary GPU to
                      work correctly for a passthrough like this, and I
                      ended up switching it up with the secondary
                      because I simply couldn't get it to work properly
                      (I'd keep getting something about "<code>Invalid
                        ROM contents</code>") and I couldn't see my boot
                      logs until the host driver finished loading. It's
                      really inconvenient now becuase my motherboard is
                      designed so that having a full-size GPU in the
                      second PCIe slot blocks 4 SATA ports out of 6, so
                      I need to have L-shaped cables in a bunch of
                      places to make everything work properly.</p>
                  </div>
                </blockquote>
                <div><br>
                </div>
              </div>
            </div>
          </div>
          <div dir="ltr">
            <div class="gmail_extra">
              <div class="gmail_quote">
                <div>The boot VGA ROM is handled differently from other
                  PCI ROMs in Linux, when you read it you're actually
                  reading a shadow copy of it placed at 0xc0000, which
                  has its roots back in the original IBM PC history. 
                  Execution of the VGA ROM is allowed to modify this
                  copy of the ROM image in memory.  Thus you're really
                  not getting a pristine copy of the device ROM and it
                  can often lead to these invalid ROM contents
                  messages.  I wish we provided raw access to the PCI
                  option ROM in these cases as well.  For now, if you do
                  have these sorts of invalid ROM messages, I would
                  suggest that you at least boot with the card as
                  secondary, dump the ROM, then use that collected image
                  with the romfile= option rather than use the shadow
                  copy.  I had to add some code to vfio to fix device
                  IDs and checksums for IGD assignment because of this,
                  I don't know if there's any standard fixup we can do
                  for other vendors.  Thanks,</div>
                <div><br>
                </div>
                <div>Alex</div>
              </div>
            </div>
          </div>
          _______________________________________________<br>
          vfio-users mailing list<br>
          <a moz-do-not-send="true" href="mailto:vfio-users@redhat.com"
            target="_blank">vfio-users@redhat.com</a><br>
          <a moz-do-not-send="true"
            href="https://www.redhat.com/mailman/listinfo/vfio-users"
            rel="noreferrer" target="_blank">https://www.redhat.com/mailman/listinfo/vfio-users</a><br>
        </blockquote>
      </div>
    </blockquote>
    <br>
  </body>
</html>