Date: Wed, 13 Feb 2008 11:25:29 -0500
From: "Vilas Sridharan" <[EMAIL PROTECTED]>
Subject: Re: [m5-users] M5 Tutorial
To: "M5 users mailing list" <m5-users@m5sim.org>
Hi Ali,
Sorry for the delayed response to this. I took a
look at the ISCA-33 slides and am basing my comments
mostly on those.
Basically, I think what you present should depend on
the audience, and you are in the best position to
know this. If the audience is mostly people who are
currently using M5 (i.e. people on this list), then,
I think you can do several things:
* Condense the intro/overview section to ~5-10
slides (from 18 currently) -- there's a lot of stuff
here that we'd already know or is relatively easy to
figure out.
* Completely omit the "Compiling / Running M5"
section (12 slides) -- or, at least, condense it
down to ~1 slide and have the details as backup for
later reference. If we're already using M5, we know
how to do this.
* Re-organize the "Current M5 Object Models" section
to trace a detailed example of a modification that
would affect all three areas (I/O, Memory, CPU) --
adding some sort of new networking interface /
protocol, perhaps? -- which would also eliminate the
need for the "Extending M5" section (since it's
integral to the previous section).
* Add a "Putting It All Together" section that ties
together the whole thing -- as Philip suggested, you
may not actually need to spend a lot of time on
this.
I would think that should give you plenty of time to
actually work through an example.
If the audience is mostly people who *aren't*
currently using M5, then you may want to spend more
time on overview (as you said, I would describe most
of the ISCA-33 tutorial as overview) -- but again,
you're in the best position to know this.
I hope this helps!
-Vilas
On Feb 11, 2008 9:18 PM, Philip Machanick
<[EMAIL PROTECTED]> wrote:
Time is why I suggested building up the complete
example through
examples for each concept as you cover the rest of
the content. The
"putting it all together" section could be covered
fast in the
tutorial but would be a nice summary to go back
and read.
On Feb 12, 2008 11:12 AM, Alexandra (Sasha)
Fedorova <[EMAIL PROTECTED]> wrote:
> Ali, would it be possible to discuss
heterogeneous multicore
> configurations (perhaps with examples) in the
section where you plan to
> discuss configurations?
>
> -- Sasha
>
>
> On Mon, 11 Feb 2008, Ali Saidi wrote:
>
>> Thanks for the comments Vilias and Philip.
>>
>> I have a few more questions. Looking at the
previous tutorial slides, it
>> doesn't seem like there is much history (just
a couple of slides), am I
>> missing something? By overview I'm assuming
that you mean the majority of the
>> presentation (and not just the first section
titled as such). Without the
>> overview slides about various objects do you
think a section on, "adding
>> feature X to component Y," would make sense? I
think it's a good idea to do
>> something like that, I'm just trying to figure
out what can be cut to make
>> room (since we're limited to about the same
amount of time).
>>
>> Thanks,
>> Ali
>>
>>
>> On Feb 11, 2008, at 12:47 PM, Vilas Sridharan
wrote:
>>
>>> I'm not sure what 'typical' usage of M5 looks
like, but I for one spend a
>>> lot of time understanding the existing code
(both C++ and Python) in order
>>> to modify / extend it -- as opposed to either
using it as is or writing
>>> entirely new code. I think more emphasis in
a tutorial on how to use M5 in
>>> that capacity (and less on history /
overview) would be helpful.
>>>
>>> I personally wasn't familiar with Python, so
I had to spend a lot of time
>>> understanding (relatively simple) Python code
-- for example, the code
>>> provided in se.py or fs.py -- in order to
change it to suit my needs. It
>>> would be great if you could cover the basics
of the Python infrastructure
>>> -- and perhaps use one of those files as part
of the example (since they
>>> are likely the most commonly used Python
files, from a user perspective).
>>>
>>> Echoing Philip's suggestion, perhaps the
'complete example' he suggests
>>> could be a 'How to' for a (simple)
modification to the existing code? You
>>> wouldn't need to actually dwell on the
details of the new feature, but it
>>> would allow you to highlight a variety of
areas in the simulator in the
>>> process.
>>>
>>> -Vilas
>>>
>>> On Feb 11, 2008 12:35 AM, Philip Machanick
<[EMAIL PROTECTED]>
>>> wrote:
>>> I took a quick look at the ISCA 2006 tutorial
and it reads well as an
>>> overview of a manual but a complete example
including a good fraction
>>> (obvious not all) of the features at the end
would be a useful
>>> addition to put it all together.
>>>
>>> A possible approach: use pieces of this
complete example to introduce
>>> the features, then go over the whole thing at
the end.
>>>
>>> On 2/11/08, Ali Saidi <[EMAIL PROTECTED]>
wrote:
>>>> Everyone,
>>>>
>>>> Below is roughly the outline we've used for
previous M5 tutorials.
>>>> We're in the process of updating our slides
for the ASPLOS tutorial in
>>>> a few weeks and would like any input you
have. Irrespective of if you
>>>> plan to attend the tutorial or not we would
like input about what
>>>> should be covered in more detail and what
we're spending too much time
>>>> on.
>>>>
>>>> Thanks,
>>>> Ali
>>>>
>>>>
>>>> Introduction & Overview
>>>> History
>>>> Attributes
>>>> Objects/Events/Modes
>>>> Current work (lots has change)
>>>> Future Work (lots has changed)
>>>> Compiling and Running M5
>>>> Source Tree
>>>> Building
>>>> Output files
>>>> Checkpoints
>>>> Simpoints
>>>> Full system workloads
>>>> Network workloads
>>>> Parts of a workload
>>>> Current M5 Object models
>>>> CPU Models
>>>> Time buffers
>>>> Templates
>>>> Checker
>>>> State
>>>> Memory System
>>>> Requests/Packets
>>>> Access modes
>>>> Interconnects
>>>> Caches
>>>> Coherence
>>>> I/O
>>>> Device overview
>>>> Disks
>>>> NICs
>>>> Extending M5
>>>> M5 Internals
>>>> Configurations
>>>> Serialization
>>>> Events
>>>> ISA description
>>>> Statistics
>>>> Debugging M5
>>>> Tracing
>>>> Debugger
>>>> Remote debugger
>>>> Wrapup
>>>>
_______________________________________________
>>>> m5-users mailing list
>>>> m5-users@m5sim.org
>>>>
http://m5sim.org/cgi-bin/mailman/listinfo/m5-users
>>>>
>>>
>>>
>>> --
>>> Philip Machanick (Visiting Fellow, School of
ITEE, University of
>>> Queensland)
>>> 39 Cunningham Street, Taringa, Qld 4068,
Australia
>>> http://opinion-nation.blogspot.com/
>>> +61-7-3871-0963 (office 3365-1190) skype
philipmach
>>>
_______________________________________________
>>> m5-users mailing list
>>> m5-users@m5sim.org
>>>
http://m5sim.org/cgi-bin/mailman/listinfo/m5-users
>>>
>>>
_______________________________________________
>>> m5-users mailing list
>>> m5-users@m5sim.org
>>>
http://m5sim.org/cgi-bin/mailman/listinfo/m5-users
>>
>>
_______________________________________________
>> m5-users mailing list
>> m5-users@m5sim.org
>>
http://m5sim.org/cgi-bin/mailman/listinfo/m5-users
> _______________________________________________
> m5-users mailing list
> m5-users@m5sim.org
>
http://m5sim.org/cgi-bin/mailman/listinfo/m5-users
>
--
Philip Machanick (Visiting Fellow, School of ITEE,
University of Queensland)
39 Cunningham Street, Taringa, Qld 4068, Australia
http://opinion-nation.blogspot.com/
+61-7-3871-0963 (office 3365-1190) skype
philipmach
_______________________________________________
m5-users mailing list
m5-users@m5sim.org
http://m5sim.org/cgi-bin/mailman/listinfo/m5-users
________________
_______________________________________________
m5-users mailing list
m5-users@m5sim.org
http://m5sim.org/cgi-bin/mailman/listinfo/m5-users
