Wen -

It's a localized issue then.

Just tried a git clone, scons c=24, and the image here:
http://marss86.org/~marss86/index.php/Disk_Images

When GRUB came up, I edited the kernel line and added 'rootwait'.

24 cores booted successfully on the first attempt. Note that it was *very*
slow -- took at least ~5 minutes on a Intel i7 4700-based machine to get
to the login prompt and there are long delays between when I press a key
in the VM window and when it registers on the screen.

I also tried 60 cores and after waiting several minutes, the system
mounted the root filesystem. I'm absolutely confident that it will
continue booting if you're patient enough, but I am not. :-)

Why does the emulation take so long? Unfortunately, qemu only executes one
VCPU at a time, so performance degrades linearly with respect to the
number of contexts that you add to the system.

If it helps, I'm running Debian "jessie".

Tyler

> Now I use rootwait instead of rootdelay=N,  24-core is still not stable
> and
> more than that doesn't boot.
> This is a failure case of 24-core
>
> ​
>
> On Wed, Aug 19, 2015 at 8:41 PM, <[email protected]> wrote:
>
>> rootwait doesn't take an argument, and should be used *in place of*
>> rootdelay=N.
>>
>> Try *just* 'rootwait'. Nothing else.
>>
>> Tyler
>>
>> > I added both rootdelay=2000 and rootwait=2000 in the menu.lst file.
>> > Still booting 24-core is not always successful.
>> > 64-core doesn't boot as before.
>> >
>> > On Wed, Aug 19, 2015 at 9:32 AM, <[email protected]> wrote:
>> >
>> >> Try 'rootwait' instead of 'rootdelay=N'.
>> >>
>> >> The former will wait indefinitely, whereas the latter will timeout
>> and
>> >> is
>> >> probably resulting in the inconsistency that you are observing.
>> >>
>> >> Tyler
>> >>
>> >> > Hi
>> >> >
>> >> > I modified the rootdelay to 2000. And compiled MARSS with 16, 24
>> and
>> >> 64
>> >> > cores respectively.
>> >> >
>> >> > ​
>> >> >
>> >> > The result is:
>> >> > 16-core is quite fine and is working well.
>> >> > 24-core is not stable, not every time the system can boot,
>> sometimes
>> >> it
>> >> > doesn't boot.
>> >> > For 64-core, the boot fails and the screen output is like
>> >> >
>> >> >
>> >> > Best
>> >> > Wen
>> >> > ​
>> >> >
>> >> >
>> >> > On Tue, Aug 18, 2015 at 1:48 AM, Brendan Fitzgerald
>> >> > <[email protected]
>> >> >> wrote:
>> >> >
>> >> >> http://marss86.org/~marss86/index.php/QEMU_Tips
>> >> >>
>> >> >> On Mon, Aug 17, 2015 at 1:17 PM, Wen Zong
>> <[email protected]>
>> >> >> wrote:
>> >> >>
>> >> >>> Hi
>> >> >>>
>> >> >>> Finally I was able to increase the size of the image and copied
>> all
>> >> >>> parsec files to the image and now try to boot the 64-core system.
>> >> >>>
>> >> >>> But the guest Ubuntu 12.04 (precise)  cannot boot with 64-core
>> >> setup.
>> >> >>> And the output is like this
>> >> >>> [image: Inline image 1]
>> >> >>> This seems to be the problem of the guest kernel itself.
>> >> >>>
>> >> >>>
>> >> >>> The ubuntu-natty.qcow2 also doesn't work with 64-core  whose
>> output
>> >> is
>> >> >>> like this
>> >> >>> [image: Inline image 2]
>> >> >>>
>> >> >>> I guess the guest kernel needs to be modified to make it work on
>> >> >>> 64-core
>> >> >>> processor.
>> >> >>>
>> >> >>> Best
>> >> >>> Wen
>> >> >>>
>> >> >>>
>> >> >>>
>> >> >>>
>> >> >>>
>> >> >>> On Sun, Aug 16, 2015 at 10:29 PM, Brendan Fitzgerald <
>> >> >>> [email protected]> wrote:
>> >> >>>
>> >> >>>> http://marss86.org/~marss86/index.php/QEMU_Tips
>> >> >>>>
>> >> >>>> On Sun, Aug 16, 2015 at 10:13 AM, Wen Zong
>> >> <[email protected]>
>> >> >>>> wrote:
>> >> >>>>
>> >> >>>>> Hi
>> >> >>>>>
>> >> >>>>> My desktop connects the network through a proxy which makes the
>> >> >>>>> network
>> >> >>>>> setup tricky for me, so I gave up this morning.
>> >> >>>>>
>> >> >>>>> I'm now trying another approach by creating a image of ubuntu
>> >> 12.04
>> >> >>>>> (precise) and then copy the parsec files and
>> ./create_checkpoint
>> >> etc.
>> >> >>>>> into
>> >> >>>>> my new image.
>> >> >>>>> I first copy the content from parsecROI.img to my host with the
>> >> help
>> >> >>>>> kpartx command.
>> >> >>>>>
>> >> >>>>> Now the problem is that when I copy the parsec folder from my
>> host
>> >> to
>> >> >>>>> the new image, it doesn't have enough space.
>> >> >>>>> I've tried to increase the rootsize option when invoking
>> >> >>>>> ubuntu-vm-builder but the image size is the same as when it is
>> set
>> >> to
>> >> >>>>> 5120.
>> >> >>>>> And after mounting, the image is a 4.1GB volume.
>> >> >>>>> Here's my configuration:
>> >> >>>>>
>> >> >>>>> [DEFAULT]
>> >> >>>>> arch = amd64
>> >> >>>>> suite = precise
>> >> >>>>> rootsize = *10120 # I increased this but the raw image size
>> >> doesn't
>> >> >>>>> change compared with 5120*
>> >> >>>>> addpkg = openssh-server, openssh-client, vim
>> >> >>>>>
>> >> >>>>>
>> >> >>>>> Do you know how to increase the size of the image?
>> >> >>>>>
>> >> >>>>> Many thanks
>> >> >>>>> Wen
>> >> >>>>>
>> >> >>>>>
>> >> >>>>> On Sun, Aug 16, 2015 at 9:43 PM, Brendan Fitzgerald <
>> >> >>>>> [email protected]> wrote:
>> >> >>>>>
>> >> >>>>>> You need to first convert the image to qcow2, since the
>> >> >>>>>> parsecROI.img
>> >> >>>>>> is read-only.
>> >> >>>>>>
>> >> >>>>>> You'll also need to enable networking, there are a bunch of
>> sites
>> >> >>>>>> that
>> >> >>>>>> say how to do this.
>> >> >>>>>>
>> >> >>>>>> On Sun, Aug 16, 2015 at 5:13 AM, Wen Zong
>> >> <[email protected]
>> >> >
>> >> >>>>>> wrote:
>> >> >>>>>>
>> >> >>>>>>> Hi Brendan
>> >> >>>>>>>
>> >> >>>>>>> Thank you for the suggestion.
>> >> >>>>>>>
>> >> >>>>>>> Do you mean I do the upgrade in the guest OS?
>> >> >>>>>>> I tried this in the provided parsecROI.img and cannot upgrade
>> >> >>>>>>> anything.
>> >> >>>>>>> The IP address in parsecROI.img is 127.0.0.1 which is
>> different
>> >> >>>>>>> from
>> >> >>>>>>> the default qemu network setting.
>> >> >>>>>>> I'm not sure if this is this the problem of the network
>> >> >>>>>>> configuration.
>> >> >>>>>>>
>> >> >>>>>>> Best
>> >> >>>>>>> Wen
>> >> >>>>>>>
>> >> >>>>>>>
>> >> >>>>>>>
>> >> >>>>>>> On Sat, Aug 15, 2015 at 11:10 PM, Brendan Fitzgerald <
>> >> >>>>>>> [email protected]> wrote:
>> >> >>>>>>>
>> >> >>>>>>>> You can update the image using apt-get dist-upgrade.
>> >> >>>>>>>>
>> >> >>>>>>>> I know kernel version 3.0 works fine with Marss, I can't
>> >> remember
>> >> >>>>>>>> off the top of my head if 3.2 works, but I know 3.16
>> doesn't,
>> >> so
>> >> >>>>>>>> just be
>> >> >>>>>>>> aware of how recent of an Ubuntu release you go to.
>> >> >>>>>>>>
>> >> >>>>>>>> On Sat, Aug 15, 2015 at 4:35 AM, Wen Zong
>> >> >>>>>>>> <[email protected]>
>> >> >>>>>>>> wrote:
>> >> >>>>>>>>
>> >> >>>>>>>>> Hi all
>> >> >>>>>>>>>
>> >> >>>>>>>>> I'm try to simulate a 64-core system and run Parsec
>> benchmark
>> >> to
>> >> >>>>>>>>> test its performance.
>> >> >>>>>>>>> However the provided parsec image is using ubuntu 9.04,
>> which
>> >> >>>>>>>>> supports at most 16 cores.
>> >> >>>>>>>>> Could someone give some suggestion to increase the maximal
>> >> number
>> >> >>>>>>>>> of supported cores to 64?
>> >> >>>>>>>>>
>> >> >>>>>>>>> One possible solution is that, I can create the disk image
>> >> using
>> >> >>>>>>>>> higher version of ubuntu which is supposed to support more
>> 64
>> >> >>>>>>>>> cores. But I
>> >> >>>>>>>>> don't have the modified Parsec benchmark in which the
>> PTLcalls
>> >> >>>>>>>>> are
>> >> >>>>>>>>> inserted to mark the Region of Interest.
>> >> >>>>>>>>>
>> >> >>>>>>>>> Very appreciated if you can give some advices.
>> >> >>>>>>>>>
>> >> >>>>>>>>> Best regards
>> >> >>>>>>>>> Wen
>> >> >>>>>>>>>
>> >> >>>>>>>>> _______________________________________________
>> >> >>>>>>>>> http://www.marss86.org
>> >> >>>>>>>>> Marss86-Devel mailing list
>> >> >>>>>>>>> [email protected]
>> >> >>>>>>>>> https://www.cs.binghamton.edu/mailman/listinfo/marss86-devel
>> >> >>>>>>>>>
>> >> >>>>>>>>>
>> >> >>>>>>>>
>> >> >>>>>>>
>> >> >>>>>>
>> >> >>>>>
>> >> >>>>
>> >> >>>
>> >> >>
>> >> > _______________________________________________
>> >> > http://www.marss86.org
>> >> > Marss86-Devel mailing list
>> >> > [email protected]
>> >> > https://www.cs.binghamton.edu/mailman/listinfo/marss86-devel
>> >> >
>> >>
>> >>
>> >>
>> >
>>
>>
>>
>



_______________________________________________
http://www.marss86.org
Marss86-Devel mailing list
[email protected]
https://www.cs.binghamton.edu/mailman/listinfo/marss86-devel

Reply via email to