[vbox-dev] ICH9 devices
apallapothu at gmail.com
Thu Jan 7 02:25:11 UTC 2016
It is not the same VM but an identical VM with the same settings. Base
Memory, CPU's, Nesting .... exactly same settings as used on PIIX3.
Reason I am inclined to think in terms of devices is because, while
experimenting with MSI when I enabled logging with DEV_PCI, PDM,
PDM_DEVICE, DEV_APIC following which I saw bunch of calls as below (apart
from my passthrough device)
pdmR3DevHlp_PhysRead: caller='ahci'/0 .....
pdmR3DevHlp_PhysRead: caller='e1000'/0 ......
pdmR3DevHlp_PhysRead: caller='VMMDev'/0 .....
At this point I am not really expecting anything to happen over network, so
thought devices could be causing the issues.
On Wed, Jan 6, 2016 at 4:22 PM, Michal Necasek <michal.necasek at oracle.com>
> There's no architected way to disable the bridges or the LPC device
> because no user needs to do it. You could probably do it by modifying the
> source code.
> You should be able to disable network and USB. Beyond that you will have
> trouble booting anything.
> I can only repeat again that I know of no reason why the ICH9 chipset
> should be significantly slower or faster than the default chipset. If you
> think there is extra code being executed, you should be able to find it
> with a debugger or a profiler. I can't help you because I never heard of
> such behavior.
> Did you compare the exact same VM where only the chipset was different?
> Or are you comparing VMs that aren't really otherwise identical?
> ----- Original Message -----
> From: apallapothu at gmail.com
> To: michal.necasek at oracle.com
> Cc: vbox-dev at virtualbox.org
> Sent: Wednesday, January 6, 2016 12:04:54 AM GMT +01:00 Amsterdam / Berlin
> / Bern / Rome / Stockholm / Vienna
> Subject: Re: [vbox-dev] ICH9 devices
> Right, AHCI isn't behind bridge, I misread.
> I certainly need network and disk.
> Devices I see in my ICH9, VGA-CustomPassThrough, VGA, E1000, VMMDev, ACPI,
> LPC, ich9pcibridge 1, ich9pcibridge 2, AHCI, USB-OHCI
> Can I disable LPC, 2 bridges ? I can't seem to find the commands in user
> manual though.
> I am thinking slowness can be because CPU has to execute more instructions
> coming from "other extra" devices in ICH9 but when I am fairly using same
> setup, drivers as I used on PIIX3 ... not sure why ICH9 would take longer
> which infact we expect it to be faster. Please comment.
> On Fri, Jan 1, 2016 at 1:04 PM, Michal Necasek <michal.necasek at oracle.com>
>> Not sure what you're referring to. Your driver specifically? Or the
>> entire guest OS? In general, there is not known to be any significant
>> performance difference between the chipsets either way. Specific devices
>> might perform faster with ICH9 (e.g. xHCI with MSIs is faster than xHCI
>> without) but the difference isn't huge.
>> If it's your driver specifically then... well, why is it behaving so
>> differently? Only you can find out.
>> Most of the devices aren't "required" as far as VirtualBox is concerned,
>> but there's only so far you can get without disk or network.
>> FYI, in my VM, AHCI definitely isn't behind a bridge. It's at bus/dev/fn
>> - Michal
>> ----- Original Message -----
>> From: apallapothu at gmail.com
>> To: vbox-dev at virtualbox.org
>> Sent: Thursday, December 31, 2015 2:52:49 AM GMT +01:00 Amsterdam /
>> Berlin / Bern / Rome / Stockholm / Vienna
>> Subject: [vbox-dev] ICH9 devices
>> Hello Experts,
>> My "custom" driver execution takes long with ICH9 as compared to
>> PIIX3, ~4X difference. Custom device is a pluggable device registered.
>> Apart from 2 "ich9pcibridge" devices I don't see any thing else "new"
>> in ICH9.
>> With ICH9, AHCI is connected on other side of the bridge (on bus 2),
>> while on PIIX3, it is on the root port. Could this be the reason for
>> slowness ?
>> I used same settings between the two chipsets, have disabled "audio"
>> in both instances since I don't need them. Is there a way to reduce overall
>> traffic ?
>> I presume all current devices in ICH9 are required (?) but is there a
>> way to improve performance ? I am not much bothered about other devices,
>> apart from having functional chipset.
-------------- next part --------------
An HTML attachment was scrubbed...
More information about the vbox-dev