Open vans163 opened 8 years ago
From reading more of the mailing list archives this seems to be a prefetch size issue? Prefetch size of my iGPU is 256mb, apparently the Ubuntu live environment host takes 128mb, that leaves 128mb for the VMs to use. Allowing you to run 1 windows or 2 Linux guests.
Does this mean that the amount of vms you can run is limited by the prefetch size?
Kind of yes, currently the physical aperture size imposes a limitation of NR_VMs. The good side is, we are working on removing that limitation.
Thanks, Jike
On 08/22/2016 10:53 PM, vans163 wrote:
From reading more of the mailing list archives this seems to be a prefetch size issue? Prefetch size of my iGPU is 256mb, apparently the Ubuntu live environment host takes 128mb, that leaves 128mb for the VMs to use. Allowing you to run 1 windows or 2 Linux guests.
Does this mean that the amount of vms you can run is limited by the prefetch size?
— You are receiving this because you are subscribed to this thread. Reply to this email directly, view it on GitHub https://github.com/01org/Igvtg-kernel/issues/17#issuecomment-241438700, or mute the thread https://github.com/notifications/unsubscribe-auth/AG1H7HoNfDGYHTK8U9UZ2K-HpmSjPMvHks5qibfPgaJpZM4Jpfm6.
@jikesong Aha, I see there is also an api forwarding mode, GVT-s. Which should be similar to the way VMWare does SVGA 3D? But that would need a custom driver for the guest? So I am guessing that is way far off.
Is there anywhere to be able to contribute to the project. I see the github repo seems to just be a placeholder, not the place where the actual development is going on.
On 08/23/2016 09:47 PM, vans163 wrote:
@jikesong https://github.com/jikesong Aha, I see there is also an api forwarding mode, GVT-s. Which should be similar to the way VMWare does SVGA 3D? But that would need a custom driver for the guest? So I am guessing that is way far off.
Speaking for myself not for Intel, GVT-s is something that I only heard about, and I don't think there an opensource distribution :-)
Is there anywhere to be able to contribute to the project. I see the github repo seems to just be a placeholder, not the place where the actual development is going on.
Yes, discussions and patches are welcome to the maillist:
https://lists.01.org/mailman/listinfo/igvt-g
Thanks, Jike
@jikesong About the development I mean all the commits here seem to come in the form of collected chunks such as https://github.com/01org/Igvtg-kernel/commit/83df1d4181c634180deea1beac17f5d84fecace3. I don't see any commits from the last month even here. I mean is there a repo where things are moving almost daily?
On 08/24/2016 08:30 PM, vans163 wrote:
@jikesong https://github.com/jikesong About the development I mean all the commits here seem to come in the form of collected chunks such as 83df1d4 https://github.com/01org/Igvtg-kernel/commit/83df1d4181c634180deea1beac17f5d84fecace3. I don't see any commits from the last month even here. I mean is there a repo where things are moving almost daily?
— You are receiving this because you were mentioned. Reply to this email directly, view it on GitHub https://github.com/01org/Igvtg-kernel/issues/17#issuecomment-242045678, or mute the thread https://github.com/notifications/unsubscribe-auth/AG1H7LLkbks553r1vWWGtu-eDkEnqaHHks5qjDmAgaJpZM4Jpfm6.
Yes, there had been big commits but it's because some feature that cannot be separated neatly. It will be separate commits in the future :)
Thanks, Jike
Like for example the only two main branches seem to be https://github.com/01org/Igvtg-kernel/tree/2016q2-4.3.0 and perhaps https://github.com/01org/Igvtg-kernel/tree/gvt-upstream-rfc. But I hear work is being done to release 2016Q3 release. For example if I jump into one of those branches now, the entire landscape of what I did could change by the time Q3 release comes.
I am using the Q1 ISO. I see the Q2 release states that "2D/3D/Media workloads can run simultaneously in multiple guests". Maybe I need to just compile Q2 from source to fix this?
Using a windows 7 SP1 32 bit guest, running on a 4030U. If you start the guest without -vgt_low_gm_sz 128 qemu param, you get a blue screen on boot. If you start the guest with -vgt_low_gm_sz 128 then everything works great, OpenGL + DirectX but you cannot start more than 1 guest.
Setting -vgt_low_gm_sz 64 on both guests allows you to start 2 but you get BSOD on boot.
Does anyone know what is going on?
The error I see in stdout when starting the second vm with -vgt_low_gm_sz 128 is
Qemu gets to output "Use flink to get guest framebuffer" before terminating.