I'm not very familiar with SimpleScalar, but my understanding is that
the encumbered FullCPU is not just a direct conversion of sim-outorder,
so you probably can't expect similar performance out of the two. M5 has
a much more flexible simulator infrastructure, some of which comes at
the cost of speed; it also has a much more detailed memory system and
various other structures surrounding the CPU. Similarly the encumbered
FullCPU was made to be able to do SMT and full-system simulation (with a
large focus on various IQ's), so there may be a fair bit of extra code
over sim-outorder. So unfortunately there's no quick and easy way to
disable those things that you don't need. There are a few things you
can do for speed over the default DetailedUniConfig, such as scaling
back the branch predictor (possibly not using a hybrid predictor),
scaling back the number of functional units (see FuncUnitConfig.py),
using a normal cache for the dcache (see MemConfig.py, by default it
uses the Indirect-Indexed Cache), and scaling back resources like the
ROB and IQ in general. It's not going to make a 10-11x difference, but
it might help speed things up a bit.
Kevin
Patrick Meredith wrote:
Nope, I'm using m5.fast ;-)
On May 30, 2006, at 6:00 PM, Korey Sewell wrote:
Have you tried building 'm5.opt' instead of 'm5.debug' ?
The optimized M5 binary would run faster than the "debugging" binary
and you would use a command line of:
"scons ALPHA_SE/m5.opt ..."
instead of
"scons ALPHA_SE/m5.debug ..."
(However, the build time could increase a bit when you build a
optimized binary)
On 5/30/06, Patrick Meredith <[EMAIL PROTECTED]> wrote:
I'm trying to run a large number of tests, and I want to use the out
of order core.
I use DetailedUniConfig and I end up with run times around 11-12x that
of simple scalar (the real one, not the simple scalar based m5 core).
It seems DetailedUniConfig is doing a lot of things I really
don't need but I'm not sure how to specify a configuration which is
faster
while using the encumbered simple scalar core (I tried using
FullO3Config
but it appears to use the newer o3 cpu, and also asserts on every
test I've tried it on). The encumbered core itself is
very close to simple scalar so it seems I should be able to get at
least 50% (g++ isn't the best)
the speed of simple scalar if I can pare down the extra details.
Basically,
does anyone have such a config file? It's sort of lazy but I have
a lot
to do in a very short amount of time ;-)
-------------------------------------------------------
All the advantages of Linux Managed Hosting--Without the Cost and
Risk!
Fully trained technicians. The highest number of Red Hat
certifications in
the hosting industry. Fanatical Support. Click to learn more
http://sel.as-us.falkag.net/sel?
cmd=lnk&kid=107521&bid=248729&dat=121642
_______________________________________________
m5sim-users mailing list
[email protected]
https://lists.sourceforge.net/lists/listinfo/m5sim-users
--
--
Korey LaMar Sewell
University of Michigan
Computer Science & Engineering
Ph.D Student
------
``Experience is not what happens to you. It is what you do with what
happens
to you.''
-------------------------------------------------------
All the advantages of Linux Managed Hosting--Without the Cost and Risk!
Fully trained technicians. The highest number of Red Hat
certifications in
the hosting industry. Fanatical Support. Click to learn more
http://sel.as-us.falkag.net/sel?cmd=lnk&kid=107521&bid=248729&dat=121642
_______________________________________________
m5sim-users mailing list
[email protected]
https://lists.sourceforge.net/lists/listinfo/m5sim-users
-------------------------------------------------------
All the advantages of Linux Managed Hosting--Without the Cost and Risk!
Fully trained technicians. The highest number of Red Hat certifications in
the hosting industry. Fanatical Support. Click to learn more
http://sel.as-us.falkag.net/sel?cmd=lnk&kid=107521&bid=248729&dat=121642
_______________________________________________
m5sim-users mailing list
[email protected]
https://lists.sourceforge.net/lists/listinfo/m5sim-users
