tag:blogger.com,1999:blog-3155908228127841862.post5936251093955245713..comments2023-06-28T10:04:44.463-06:00Comments on The Perils of Parallel: How Hardware Virtualization Works (Part 2)Greg Pfisterhttp://www.blogger.com/profile/12651996181651540140noreply@blogger.comBlogger2125tag:blogger.com,1999:blog-3155908228127841862.post-72972175303088164802011-01-30T17:36:57.613-07:002011-01-30T17:36:57.613-07:00Hi, Heshsham. Thanks!
Assuming by GPU virtualizat...Hi, Heshsham. Thanks!<br /><br />Assuming by GPU virtualization you mean cutting up a single GPU attached to a server, giving part to each of several VMs running on the server, then... no. <br /><br />Since GPUs are IO devices, this is IO virtualization; see part 4 of this set of posts about that. It can be done and will be, eventually, but in general isn't there yet.<br /><br />Part 4 is at http://goo.gl/ozuVz .<br /><br />Also, I don't think GPUs are likely to do the kind of trap-and-map time slicing discussed here. They're more likely to be partitioned, with separate parts dedicated to the VMs they serve; that's how most IO devices that do this do it now (mainly network adapters).<br /><br />This was briefly touched on in part 1 of this series. It's expanded in a later post about the varieties of virtualization: http://goo.gl/6kt6E .<br /><br />GregGreg Pfisterhttps://www.blogger.com/profile/12651996181651540140noreply@blogger.comtag:blogger.com,1999:blog-3155908228127841862.post-76299748708377140752011-01-30T08:11:00.618-07:002011-01-30T08:11:00.618-07:00Nice explanation!
By the way why is GPU virtualiza...Nice explanation!<br />By the way why is GPU virtualization far more difficult?Heshshamhttps://www.blogger.com/profile/15527342334504631800noreply@blogger.com