On Tue, 24 Jul 2012, Romulo Curty Cerqueira wrote:
Hi Nilay
I am a very new user and after your answer I found this:
(from FAQ at
http://www.m5sim.org/Frequently_Asked_Questions#How_many_CPUs_can_M5_run.3F)
How many CPUs can M5 run?
- There is no inherent limit in M5 (other than simulation speed). In SE
mode there are no obstacles to simulating as many CPUs as you want.
However, in FS mode, the real-world Alpha platform we model (Tsunami) only
supports up to 4 processors. To get around this limit, we defined and
implemented a variant of the Tsunami platform (which we call BigTsunami)
that can take up to 64 processors. Note that BigTsunami does not correspond
to any real system. BigTsunami support is included in the standard M5 Alpha
build, but booting with more than 4 CPUs requires modifications to the PAL
code and kernel as well. Take a look at the
Download<http://www.m5sim.org/Download>page for our Linux patches and
modified PAL code. Note that even with the
BigTsunami changes, simulating 64 processors will be quite slow, and the
Linux scheduler doesn't seem particularly good at scheduling a large number
of processors.
Because of Alpha was the basis architecture for M5 development (now
included in GEM5), I suppose it is also valid for any cpu architecture
simulated at GEM5. In practice I really received error messages when I
tried to simulated more than 4 CPUs in ARM FS mode, as follows:
...
ARM ==> gem5.opt: build/ARM/dev/arm/a9scu.cc:66: virtual Tick
A9SCU::read(PacketPtr): Assertion `sys->numContexts() <= 4' failed.
...
So I suppose my instance of GEM5 simulator is not working well and I should
review its compiled code. Is it correct?
If not, what should I do to simulate more than 4 ARM CPUs?
Somebody more familiar with ARM implementation should respond to this.
--
Nilay
_______________________________________________
gem5-users mailing list
[email protected]
http://m5sim.org/cgi-bin/mailman/listinfo/gem5-users
