If that assert is removed, there is a timing assert fail caused in tport.cc
assert(when > curTick)
I do a hack setting when = curTick + 1 to get around this and it seems ro tun
but I don't know if this makes sense. Can you reproduce this error?
- Sujay
From: Lisa Hsu
Sent: Wednesday, November 18, 2009 1:37 PM
To: M5 users mailing list
Subject: Re: [m5-users] fast-forwarding and switching cpus
I have a vague recollection about this error. I just looked into it and it
came from a commit that cleaned up some stuff regarding full-system. I was
deep in the throes of thesis finishing and did a really stopgap fix - I
reinserted the #if FULL_SYSTEM around that assert that had been removed in that
commit.
This is really not a true solution but it might get you by.
Nate, this comes from your changeset 5803 on Jan. 19, 2009, subject
thread_context. Clues for a fix/solution?
Lisa
On Thu, Nov 12, 2009 at 4:50 PM, Korey Sewell <[email protected]> wrote:
I'm not a fast-forwarding wiz myself (Lisa?), but it loooks like if things
arent working you are getting to have to get your hands dirty in the code and
figure the internals of how fast forwarding is supposed to work, so you can get
fix the problem.
Once you start looking at code and debugging, it might be a good idea to send
emails to m5-dev, so that once you detect the problem, a patch can be generated
and the repo can be updated.
Alternatively, check the mailing list archives as I'm sure there have been
people asking about fast forwarding in the past...
On Tue, Nov 10, 2009 at 6:51 PM, Sujay Phadke <[email protected]> wrote:
Hi Steve,
Here is a run with unmodified m5-stable using se.py
spha...@lowpower:~/research/m5/m5-stable$ ./build/ALPHA_SE/m5.opt
configs/example/se.py -d --caches --l2cache --fast-forward=1
M5 Simulator System
Copyright (c) 2001-2008
The Regents of The University of Michigan
All Rights Reserved
M5 compiled Nov 8 2009 21:09:22
M5 revision 15e5581286b6+ 6285+ default tip
M5 started Nov 10 2009 18:44:15
M5 executing on lowpower
command line: ./build/ALPHA_SE/m5.opt configs/example/se.py -d --caches
--l2cache --fast-forward=1
Global frequency set at 1000000000000 ticks per second
0: system.remote_gdb.listener: listening for remote gdb #0 on port 7000
Switch at instruction count:1
info: Entering event queue @ 0. Starting simulation...
hack: be nice to actually delete the event here
Switched CPUS @ cycle = 1000
Changing memory mode to timing
switching cpus
m5.opt: build/ALPHA_SE/cpu/o3/thread_context_impl.hh:57: void
O3ThreadContext< <template-parameter-1-1>
>::takeOverFrom(ThreadContext*) [with Impl = O3CPUImpl]: Assertion
`getSystemPtr() == old_context->getSystemPtr()' failed.
Program aborted at cycle 1000
Aborted
I get the same error. I also tried giving the "-s" for standard switch
and changing the value of FF. Any idea why this is happening?
Thanks,
Sujay
On Tue, 2009-11-10 at 15:30 -0800, Steve Reinhardt wrote:
> I don't know any details about what works and what doesn't off the top
> of my head, but I can tell you that if something "works" in m5.fast
> but hits an assertion error in m5.opt, that's only because assertions
> are compiled out in m5.fast. So either the assertion is asserting
> something that doesn't necessarily need to be true (which happens) or
> the assertion is valid and m5.fast is merrily proceeding to do
> something bogus that just happens not to crash the simulation.
>
> That particular assertion looks like it should hold; what it is saying
> is that you're substituting a detailed CPU in system A for a simple
> CPU in system B, which would possibly cause problems if the CPU ever
> did something based on the system it's a member of. That could be a
> configuration error.
>
> Steve
>
> On Tue, Nov 10, 2009 at 12:00 PM, Sujay Phadke <[email protected]>
wrote:
> > Does any one know about this? Is FF working with the latest m5.opt in SE
> > mode, and is it possible to FF with multiple CPUS?
> >
> > Thanks,
> > Sujay
> >
> > ----- Original Message -----
> > From: "Sujay Phadke" <[email protected]>
> > To: <[email protected]>
> > Sent: Saturday, November 07, 2009 5:41 PM
> > Subject: [m5-users] fast-forwarding and switching cpus
> >
> >
> >> Hello,
> >> I was going through the previous threads about fastforwarding
> >> support, but couldnt get a clear answer. When we specify
> >> --fast-forward=FF, is if FF ticks or FF instructions? If it is ticks,
> >> then is there a new way for FF instructions?
> >>
> >> As an aside, I am able to use --fast-forward with m5.fast. With m5.opt,
> >> it crashes with the error:
> >>
> >> M5 compiled Nov 6 2009 17:34:43
> >> M5 revision 15e5581286b6 6285 default tip
> >> M5 started Nov 7 2009 17:40:10
> >> M5 executing on lowpower
> >> command line: ./build/ALPHA_SE/m5.opt -d ../output/test/
> >> configs/spec2k6/runspec2k6.py -b bzip2 -d --caches --l2cache
> >> --fast-forward=5000000 --max-inst=100000
> >> Global frequency set at 1000000000000 ticks per second
> >> 0: system.remote_gdb.listener: listening for remote gdb #0 on port 7000
> >> Switch at instruction count:5000000
> >> info: Entering event queue @ 0. Starting simulation...
> >> info: Increasing stack size by one page.
> >> hack: be nice to actually delete the event here
> >> Switched CPUS @ cycle = 5079299000
> >> switching cpus
> >> m5.opt: build/ALPHA_SE/cpu/o3/thread_context_impl.hh:57: void
> >> O3ThreadContext< <template-parameter-1-1>
> >>>::takeOverFrom(ThreadContext*) [with Impl = O3CPUImpl]: Assertion
> >> `getSystemPtr() == old_context->getSystemPtr()' failed.
> >> Program aborted at cycle 5079299000
> >> Aborted
> >>
> >> Is there still the same bug about no valid threadcontext in o3?
> >>
> >> - Sujay
> >>
> >>
> >> _______________________________________________
> >> m5-users mailing list
> >> [email protected]
> >> http://m5sim.org/cgi-bin/mailman/listinfo/m5-users
> >>
> >
> > _______________________________________________
> > m5-users mailing list
> > [email protected]
> > http://m5sim.org/cgi-bin/mailman/listinfo/m5-users
> >
> _______________________________________________
> m5-users mailing list
> [email protected]
> http://m5sim.org/cgi-bin/mailman/listinfo/m5-users
_______________________________________________
m5-users mailing list
[email protected]
http://m5sim.org/cgi-bin/mailman/listinfo/m5-users
--
- Korey
_______________________________________________
m5-users mailing list
[email protected]
http://m5sim.org/cgi-bin/mailman/listinfo/m5-users
--------------------------------------------------------------------------------
_______________________________________________
m5-users mailing list
[email protected]
http://m5sim.org/cgi-bin/mailman/listinfo/m5-users_______________________________________________
m5-users mailing list
[email protected]
http://m5sim.org/cgi-bin/mailman/listinfo/m5-users
