Comment 26 for bug 1788665

David, your suggestion seemed helpful, at least there is a difference in the pattern of vmentries and vmexits. See the snapshot attached.

Explanation of snapshot_1:
Two windows of kernelshark with trace.dats obtained using the command from above; the left window (trace.dat) is with spec-ctrl feature disabled, the right window with spec-ctrl enabled (default).
CPU9 runs the emulator (emulatorpin), and the spikes seen are kvm_set_irq every 16ms. CPUs 2-7 and 19-15 run the vcpu threads.
Halfway through the trace in the snapshot the test begins (passmark 2d image rendering). The VM without spec-ctrl triggers vmentries/vmexits much more often than the VM with spec-ctrl. I could not spot a difference in the pattern of the vmentries/vmexits themselves (snapshot_2 below), only their frequency seems to differ.

Does anybody have an idea of what is going on?

George