Philip Machanick wrote:
I have a cross-compiler so I suppose I could do this. Presumably this
would remove the need for the hack since addresses would now be
translated by the VM (though it would be obviously a whole lot slower
than a quick and dirty solution to getting multiple processes running,
one per CPU).
I guess it will not be a huge amount of effort to put my own binaries on
a disk image mounted on the simulated machine?
I don't think so... there are instructions on the web site somewhere
(let us know if you can't find them).
What I am trying to do is to look at performance of N processes running
on N CPUs vs. those same N processes running sequentially on a single
more aggressive CPU. Not including the OS is obviously a significant
omission but it would be nice to scope the problem with a simple faster
prototype.
The effect of the simple hack of changing the ASID is the simulator
stops pretty quickly claiming it has hit a halt instruction on a binary
which it otherwise executes to completion. Most likely the problem is
that it is not consistently using the ASID "translation".
Actually the ASID is only used by the timing cache model, so it couldn't
cause this problem. In syscall emulation mode, each process has its own
independent virtually addressed functional memory, so the real problem
is likely that you have one process fetching instructions out of a
different process's functional memory object.
I'll take a look at building an SMP-capable kernel.
On 10/03/2006, at 2:56 AM, Steve Reinhardt wrote:
Hi Philip,
Philip Machanick wrote:
I would like to model simple multitasking on a multiprocessor
configuration. In
the archive, there is some discussion of the need for a different
ASID for each
process so address spaces don't overlap:
http://sourceforge.net/mailarchive/message.php?msg_id=14328385
We are still in the middle of a major memory system restructuring that
should take care of this problem in v2.0, but that doesn't help you
now. I don't know if the original poster (or anyone else) has hacked
this in to the current version.
plus some discussion of why this hack may not work:
http://sourceforge.net/mailarchive/message.php?msg_id=14066609
I just skimmed this message quickly, but I don't see anything in there
that affects what you want to do (though I may have missed it).
Has there been any progress on this?
(I only want to run SE, not FS, since no doubt we also don't have an
SMP-capable
Alpha Linux.)
It's easy to get an SMP-capable Alpha Linux... if you build a
cross-compiler with crosstools that's all you need to cross-compile a
full kernel and libraries on an x86 machine. No actual Alpha machines
are needed.
Steve
--
Philip Machanick, School of ITEE, University of Queensland
-------------------------------------------------------
This SF.Net email is sponsored by xPML, a groundbreaking scripting
language
that extends applications into web and mobile media. Attend the live
webcast
and join the prime developer group breaking into this new coding
territory!
http://sel.as-us.falkag.net/sel?cmd=lnk&kid=110944&bid=241720&dat=121642
_______________________________________________
m5sim-users mailing list
[email protected]
https://lists.sourceforge.net/lists/listinfo/m5sim-users
Philip Machanick School of ITEE, University of Queensland
Brisbane, Qld 4072, Australia http://www.itee.uq.edu.au/~philip/
phone +61(7) 3365 2766 fax +61(7) 3365 4999 mailto:[EMAIL PROTECTED]
-------------------------------------------------------
This SF.Net email is sponsored by xPML, a groundbreaking scripting language
that extends applications into web and mobile media. Attend the live
webcast
and join the prime developer group breaking into this new coding territory!
http://sel.as-us.falkag.net/sel?cmd=lnk&kid=110944&bid=241720&dat=121642
_______________________________________________
m5sim-users mailing list
[email protected]
https://lists.sourceforge.net/lists/listinfo/m5sim-users
-------------------------------------------------------
This SF.Net email is sponsored by xPML, a groundbreaking scripting language
that extends applications into web and mobile media. Attend the live webcast
and join the prime developer group breaking into this new coding territory!
http://sel.as-us.falkag.net/sel?cmd=lnk&kid=110944&bid=241720&dat=121642
_______________________________________________
m5sim-users mailing list
[email protected]
https://lists.sourceforge.net/lists/listinfo/m5sim-users
