Moblin config kernel initialisation performance
 

One of our goals from UDS was to analyse the kernel initialisation
performance. The moblin kernel was measured to initialise in about .9s
whereas the default Ubuntu kernel was booting in about 1.5s (both on the
10v).

While testing the async_populate_rootfs patches I noted did a little
experimentation to see what else was slowing boot and comparing the
configuration for those components to the moblin kernel. The major costs
are populate_rootfs, pnpisa_init, piix_init and associated port probes.
We know that moblin does not use an initramfs, and as we can see below
isapnp is turned off in their config:

config-generic:# CONFIG_ISAPNP is not set

If I boot my 10v with no initramfs and with isapnp disabled on the
command line I get a kernel init time of approximatly 0.8s. In short
we can account in full for the kernel init performance difference from
disabling both the initramfs and isapnp.

-apw

--
kernel-team mailing list
kernel-team@lists.ubuntu.com
https://lists.ubuntu.com/mailman/listinfo/kernel-team

Moblin config kernel initialisation performance
 

Andy Whitcroft wrote:
> One of our goals from UDS was to analyse the kernel initialisation
> performance. The moblin kernel was measured to initialise in about .9s
> whereas the default Ubuntu kernel was booting in about 1.5s (both on the
> 10v).
>
> While testing the async_populate_rootfs patches I noted did a little
> experimentation to see what else was slowing boot and comparing the
> configuration for those components to the moblin kernel. The major costs
> are populate_rootfs, pnpisa_init, piix_init and associated port probes.
> We know that moblin does not use an initramfs, and as we can see below
> isapnp is turned off in their config:
>
> config-generic:# CONFIG_ISAPNP is not set
>
> If I boot my 10v with no initramfs and with isapnp disabled on the
> command line I get a kernel init time of approximatly 0.8s. In short
> we can account in full for the kernel init performance difference from
> disabling both the initramfs and isapnp.
>
Right so we have:
populating the initramfs taking about .4s and pnpisa taking some where over .2s

We can speed the initramfs portion up by
- reducing initramfs size
- parallelizing populate_rootfs
- possibly switching to LZO decompression which is faster but at a trade off
of some size increase

What kind of measures do we have, and how accurate are they, for the boot loader
loading the initramfs and kernel. So we can evaluate the whether switching to
LZO is a win.

Having not looked at it at all what can we do to improve pnpisa?


--
kernel-team mailing list
kernel-team@lists.ubuntu.com
https://lists.ubuntu.com/mailman/listinfo/kernel-team

Moblin config kernel initialisation performance
 

On Wed, Dec 02, 2009 at 02:40:45PM -0800, John Johansen wrote:

> Right so we have:
> populating the initramfs taking about .4s and pnpisa taking some where over .2s
>
> We can speed the initramfs portion up by
> - reducing initramfs size
> - parallelizing populate_rootfs
> - possibly switching to LZO decompression which is faster but at a trade off
> of some size increase

I am relying on you to help us out making the initramfs smaller.
csurbhi has some promising work to async'ise the populate_rootfs part.

> What kind of measures do we have, and how accurate are they, for the boot loader
> loading the initramfs and kernel. So we can evaluate the whether switching to
> LZO is a win.

None that I am aware of as its before we see time. There was some
rumours we could get grub2 to pass a 'I started at this tsc tick' or
something to allow the kernel to work out the delta.

> Having not looked at it at all what can we do to improve pnpisa?

A lot of the time is in the form of 'send this' 'wait 250us' 'read
that', loops which must be repeated to scan for and find pnp devices.
Its not clear how much this can be improved.

I am hoping we can avoid needing to by the speed ups in initramfs both
in unpack and run time.

-apw

--
kernel-team mailing list
kernel-team@lists.ubuntu.com
https://lists.ubuntu.com/mailman/listinfo/kernel-team

Moblin config kernel initialisation performance
 

On Wed, 2009-12-02 at 23:25 +0000, Andy Whitcroft wrote:

> > Having not looked at it at all what can we do to improve pnpisa?
>
> A lot of the time is in the form of 'send this' 'wait 250us' 'read
> that', loops which must be repeated to scan for and find pnp devices.
> Its not clear how much this can be improved.
>
> I am hoping we can avoid needing to by the speed ups in initramfs both
> in unpack and run time.
>
Can we not make the isapnp_init async too?

Scott
--
Scott James Remnant
scott@ubuntu.com
--
kernel-team mailing list
kernel-team@lists.ubuntu.com
https://lists.ubuntu.com/mailman/listinfo/kernel-team