Subrata and Mike,
Thank you both for the feedback.  I will take notes through out my
effort clean them up and post them along with any patches that I have
had to add.

quote:
>> I think longterm there needs to be a max static (flash) and max
>> dynamic (RAM) arguments that you can set somewhere in ltp

I was a little hasty to make the above statement.  When I sent you
that email, I had spent several days trying to get ltp to work and
would fix one memory test only to have ltp fail again with an OOM
several tests later .  Now that I'm further along and have a better
perspective on what really is involved with getting ltp working on an
embedded system I realize the OOM issues really are isolated to a
handful of tests.

I think that if the few tests that use large memory and don't
currently contain an input argument to limit it are patched and that
somewhere there is good documentation (wiki probably) on how to
configure ltp to work nicely on a small memory (embedded) system
everyone will be happy.

I guess the tests that use considerable amounts of memory could also
just be re-written to detect the system RAM making sure to only use
some percentage of it, if the memory limit makes the test useless then
it should return some message stating that.  I will look at the tests
I have had issues with and see if this is an easy thing to implement.
Regards,
Shane

On Fri, Apr 25, 2008 at 3:41 AM, Mike Frysinger <[EMAIL PROTECTED]> wrote:
> On Friday 25 April 2008, Subrata Modak wrote:
>  > But i am not that much aware of the embedded environment, and, so will
>  > also leave on others who have worked on it (May be Carmelo, Mike,
>  > others) to review all that you have proposed below, and give their
>  > suggestions on this as well. Hoping to work with you for making LTP work
>  > on Embedded platforms as well.
>
>  we run LTP on no-mmu systems with 64megs of ram (which means more like 
> ~50megs
>  actually available) and really dont see OOM issues.  we see stack overflows,
>  but that is a completely different topic.  on mmu systems, i see it not
>  really being an issue at all.  but again, we really only do the kernel subdir
>  (and it looks like we're talking about the math subdir).
>
>  for tests that use a lot of memory, instead of bemoaning the issue, fixup the
>  tests to address your needs and post the patch.
>  -mike
>



-- 
Registered Linux User: #293401

-------------------------------------------------------------------------
This SF.net email is sponsored by the 2008 JavaOne(SM) Conference 
Don't miss this year's exciting event. There's still time to save $100. 
Use priority code J8TL2D2. 
http://ad.doubleclick.net/clk;198757673;13503038;p?http://java.sun.com/javaone
_______________________________________________
Ltp-list mailing list
[email protected]
https://lists.sourceforge.net/lists/listinfo/ltp-list

Reply via email to