qemu-discuss
[Top][All Lists]
Advanced

[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

[Qemu-discuss] Mouse / keyboard input latency in VM


From: Tom Tom
Subject: [Qemu-discuss] Mouse / keyboard input latency in VM
Date: Mon, 31 Dec 2018 01:14:44 +0100

Thread: https://github.com/gnif/LookingGlass/issues/117

Hello, what are possibilities for passing input into VM (which I know):

   - KVM switch (baremetal performance / latency)
   - Pass whole USB port into VM (baremetal performance / latency)
   - Synergy (over network, for example localhost)
   - Spice (as far as I know, LookingGlass uses this as default)
   - Evdev (this is what I found, it should be something like event handler
   which is built into Linux kernel and qemu can work with it and I read that
   it should be low latency)

So KVM switch is the best from performance view, but it is not seamless,
because you have to always press button to redirect input back to host.
Synergy is seamless, but there is probably some kind of latency and
overhead, because it is over network (even localhost). What I heard about
Spice is that it have some latency + overhead, because it will setup
virtual display.

The last one is Evdev what I found recently, here is article about it
https://passthroughpo.st/using-evdev-passthrough-seamless-vm-input/ it
looks like the best option because it is built into Linux kernel and qemu
and as far as I know there is no latency, but disadvantages what I read is
that you bind and unbind input devices from VM by pressing both Ctrl at
same time, I would like to have it on Scroll lock (because it has led
indicator) and another is that it supports (or supported) only basic mouse
input (like left click, right lick, scroll wheel and movement).


So question is what is the best method (lowest latency --> for gaming) for
passing mouse / keyboard input into VM running in localhost, I don't want
to use KVM switch or pass whole USB into VM, so I am interested in Spice,
Synergy or Evdev. Are there any measured numbers of latency?


Thank you very much


reply via email to

[Prev in Thread] Current Thread [Next in Thread]