[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

CORBA for RTEMS




Rosimildo DaSilva wrote:
>
> The snapshots are provided by OARCorp, they are not
> publicly available. All this new stuff will be available with the
> next release of RTEMS. Joel, and his team,  are working
> extremely hard to get this release out *real soon*.
> I'll let Joel to fill in about availability of the snapshots.

That's about the size of the situation right now.  There were a lot
of features that needed to stabilize.  This last snapshot should
have been a feature freeze.  Now to stablize enough to make a public 
release.
 
> >I have also a second question : the required code space (.text)
> >for CORBA seems to be about 600 kB. However in an article from
> >Niall Murphy (Introduction to CORBA for Embedded Systems,
> >http://204.33.180.24/98/9810fe2.htm ) I read that a full ORB
> >library is typically in the order of 150 kB... Does this 600 kB
> >contain more than just the ORB library or is this typical number
> >not so correct ?
> 
> The 600K is a combination of many packages. I'll try to
> give some hints where this number comes from.
> In the example below I have used the CDTEST as a small test case,
> where I compiled it, with just RTEMS, and added peaces to it. These
> peaces are: IOSTREAM + Networking + ORB.

When some of the RTEMS folks emailed with the TAO group, we learned that
although we may include EVERYTHING in the footprint number, they tend
to focus on the code in just the CORBA implementation.  Think about their
perspective.  Under UNIX, you don't worry about the size of the OS (which
includes the network stack, or the size of the C Library, etc.  You focus
on strictly your part.  That is a large source of the discrepancy.


>    text    data     bss     dec     hex filename
>  121589    4983   11684  138256   21c10 cdtest.exe
> 
>    CODE --> 121K
>    DATA -->   5K
>    BSS  -->  12K
> 
> I'll assume this number as the minimun size of a RTEMS application. It might
> not be
> precise, but it gives you an idea of the footprint of RTEMS itself.

It is pretty far above a functional minimal application (about 60K on the
SPARC) but is a fair representation of a C++ application that actually
prints something. :)

I like to point out that minimum footprint and minimum USEFUL footprint
are totally different.  I can produce somethign very tiny but it is 
very likely that it does not include enough 

> CASE #2 - RTEMS' cdtest example with iostream support:
> 
>    text    data     bss     dec     hex filename
>  175861    5943   12164  193968   2f5b0 cdtest.exe
> 
> CODE --> 176K
> DATA -->   6K
> BSS  -->  13K
> 
> It is important to notice that Iostream alone adds a fooprint of 55K to the
> code space
> and 2K to the data space.
> 
> CASE #3 - RTEMS' cdtest example with iostream + Networking support
> 
>    text    data     bss     dec     hex filename
>  291957   8503   17052  317512   4d848 cdtest.exe
> 
>    CODE --> 292K
>    DATA -->   9K
>    BSS  -->    18K
> 
> As you can see, the network stack adds a footprint of 116K to the code
> space,
> and 8K to the data space. Network stack requires a lot of memory to that are
> allocated dynamically and do not appear on these numbers.
> 
> CASE #4 - omniORB eg3_impl example:   --- Full support ( all dynamic stuff
> from CORBA )
> 
>    text    data     bss     dec     hex filename
>  750301   33855   19044  803200   c4180 eg3_impl.elf32
> 
>    CODE --> 751K
>    DATA -->  34K
>    BSS  -->  20K
> 
> CASE #5 - omniORB eg3_impl example compiled with RtemsMinimumCorba=1.
> 
>    text    data     bss     dec     hex filename
>  611645   24415   18788  654848   9fe00 eg3_impl.elf32
> 
> CODE --> 612K
> DATA -->  25K
> BSS  -->  19K
> 
> For a full blown ORB, with all dynamic stuff, it adds around 459K of code
> space
> plus 27K of data space.
> 
> For a minimum ORB, without the dynamic stuff, it adds around 320K of code
> space
> plus 17K of data space.

I think your presentation is quite good.  Each time you add something, it
has a cost. :)

> Regards, Rosimildo.
> 
> +---------------------------------------------------------------------
> | ConnectTel, Inc. - Austin, Texas.
> | Phone : 512-338-1111 - Fax : 512-918-0449
> | Email : devl at connecttel.com  or mkting at connecttel.com
> | Site  : http://www.connecttel.com
> +---------------------------------------------------------------------

-- 
Joel Sherrill, Ph.D.             Director of Research & Development
joel at OARcorp.com                 On-Line Applications Research
Ask me about RTEMS: a free RTOS  Huntsville AL 35805
Support Available                (256) 722-9985