Depends on how your multi-threaded program is written. If it's a
general pthreads program, then you'll need to run it in full-system
mode. This means copying it onto a disk image (see the FAQ) and then
running it from the command prompt. Once you get this working you can
automate it with an rcS file (see
http://www.m5sim.org/wiki/index.php/Running_M5_in_Full-System_Mode).
Steve
Rajdeep Bhowmik wrote:
Hi,
Thank you very much for your reply. I have been able
to run two programs in two different cores with the
help of your suggestion.
My objective is to run different threads of a
multi-threaded program in different cores. For this to
be done, do I have to use the full-system mode or the
system emulation mode ? And how can this be achieved
in either in M5 1.1 or M5 2.0 ?
Regards-
Rajdeep
--- Steve Reinhardt <[EMAIL PROTECTED]> wrote:
Rajdeep Bhowmik wrote:
Hi Steve,
Thanks very much for your reply.
You're welcome. Please reply to the mailing list
and not directly to
me. That way if others have similar problems they
can learn from the
mailing list archives.
Both of my programs are running successfully. Now
I have compiled both
the programs with the '-static' option and they
are running okay when i
am not using the printf() in the code. But when I
am using printf(), it
is giving me an error " System call 427 out of
range". I checked tthe
code and found that the function pointer for
printf() is not
implemented. So do I have to implement printf()
and similarly other
functions like fork(), pthread_create() etc on my
own if I use them in
my code?
System call 427 is fstat64. This is implemented in
m5 2.0 (but not in 1.1).
printf() is implemented in libc, which should be
statically linked in to
your binary. I don't know what you mean by "the
function pointer for
printf() is not implemented".
pthread_create()/fork() is a different issue; if you
want to run general
pthreads programs then you'll need to run them under
full system mode,
as there isn't (and probably never will be) a thread
scheduler built
into m5. (There's a perfectly good one in the Linux
kernel, and since
we can run that, there's not a lot of motivation to
do another one.) If
you just want to run simple SPLASH-like programs,
that's not
unthinkable, but right now the only way to do that
in SE mode is if you
can compile them under Tru64. Linux SE-mode PARMACS
support is on my
to-do list but unfortunately behind a few other
things.
Can you tell me the way to run two different
programs in two different
cores?
Set the workload parameter of each cpu to the
workload you want to run,
e.g.:
cpu[0].workload = Benchmarks.AnagramLongCP()
cpu[1].workload = Benchmarks.GCCLongCP()
Steve
regards-
Rajdeep
*/Steve Reinhardt <[EMAIL PROTECTED]>/* wrote:
Getting the deadlock message at cycle 0
basically means that the CPU
never really started executing instructions.
One common cause for this
is that your programs are not statically
linked; see
http://www.m5sim.org/wiki/index.php/Frequently_Asked_Questions#I_compiled_an_executable_to_run_in_syscall_emulation_mode_but_it_doesn.27t_work.
You should verify that both programs run
successfully by themselves
before trying to run them both together.
BTW, specifying two programs in a list like
you have there is
specifying
two programs for a two-way multithreaded CPU,
not for two different
cores.
Steve
Rajdeep Bhowmik wrote:
> Hi All,
>
> I am trying to run two independent
workloads -
> "HelloWorld" and "MatrixMultiplication" in
two
> different cores. I am using m5 1.1 version.
I have
> written a run.py file similar to the run.py
file given
> in m5-tests/test3/ i.e.
> "BaseCPU.workload =
> [
Benchmarks.AnagramLongCP(),Benchmarks.GCCLongCP() ]"
>
> But the simulator is returning an error
saying
> "Deadlock.. Simulation terminated at cycle
100000".
>
> My programs are single-threaded
applications. Is this
> the reason for the error ?
>
> Can anyone tell me what I should do to make
it run
> successfully.
>
> Regards-
> Rajdeep
>
>
>
>
>
__________________________________________________
> Do You Yahoo!?
> Tired of spam? Yahoo! Mail has the best
spam protection around
> http://mail.yahoo.com
>
_______________________________________________
> 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
------------------------------------------------------------------------
Yahoo! Messenger with Voice. Make PC-to-Phone
Calls
<http://us.rd.yahoo.com/mail_us/taglines/postman1/*http://us.rd.yahoo.com/evt=39663/*http://voice.yahoo.com>
to the US (and 30+ countries) for 2ยข/min or less.
__________________________________________________
Do You Yahoo!?
Tired of spam? Yahoo! Mail has the best spam protection around
http://mail.yahoo.com
_______________________________________________
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
