This is the mail archive of the
ecos-devel@sources.redhat.com
mailing list for the eCos project.
Re: CYGSEM_HAL_VIRTUAL_VECTOR_SUPPORT
- From: Gary Thomas <gary at mlbassoc dot com>
- To: Bart Veer <bartv at ecoscentric dot com>
- Cc: jani at iv dot ro, eCos development <ecos-devel at ecos dot sourceware dot org>
- Date: Fri, 03 Sep 2004 09:15:33 -0600
- Subject: Re: CYGSEM_HAL_VIRTUAL_VECTOR_SUPPORT
- Organization: MLB Associates
- References: <20040903173326.4d811df2.jani@iv.ro> <20040903150918.1BA45EC10C@delenn.bartv.net>
On Fri, 2004-09-03 at 09:09, Bart Veer wrote:
> >>>>> "Jani" == Jani Monoses <jani@iv.ro> writes:
>
> Jani> Hello
> Jani> are there any platforms not implementing virtual vectors?
> Jani> How long is non VV code planned to be supported?
>
> What code in particular are you concerned about? I am not sure if
> there any interesting platforms left where virtual vectors are
> completely unsupported. However there are certainly platforms where
> you may want to disable the virtual vector support.
>
> On some systems the application will boot directly from ROM, and
> with no RedBoot in the system there is no point in enabling virtual
> vectors. Booting directly from the ROM means that you can keep down
> the hardware costs, although it may also mean that some of the
> development has to happen on special boards with extra memory.
>
> An important subset are systems where there is just not enough memory
> for both RedBoot and an application. For example there are chips like
> the MCF5282 with 512K on-chip flash and 64K on-chip RAM, and some
> people want the application to run in just those two areas of memory
> so that the boards don't have to be populated with any external memory
> at all. Fitting just an application into the internal memory can prove
> tricky. If you need a ROM monitor as well it becomes impossible.
>
> Another example is when you are debugging via jtag or BDM. Even if
> there is a ROM monitor programmed into flash as well it won't get a
> chance to execute, so the application must run without virtual
> vectors.
I disagree. I think that the virtual vectors are useful in all
environments, even totally self-contained ones like your ROM example.
By abstracting these operations [console I/O, delay timers, etc],
applications (and even the kernel itself) are simplified.
--
Gary Thomas <gary@mlbassoc.com>
MLB Associates