BSP for a new platform

Hi All,

I am new to WinCE BSP development. Can someone tell me how long it is
going to take to develop a BSP for a customized hardware such as AMD NX
DB -1500 Geode processor. I searched and I found out that Microsoft
provides BSP for AMD GX DB -1500 processors but not for NX series.
Please let me know how long it will take to develop using Platform
builder.

Thanks in Advance,
Ashok

Valter
Thu Sep 28 05:08:09 CDT 2006

"ashok" <ashokrajshekar@gmail.com> wrote in
news:1159437412.244497.79280@e3g2000cwe.googlegroups.com:

> Hi All,
>
> I am new to WinCE BSP development. Can someone tell me how long it
> is going to take to develop a BSP for a customized hardware such
> as AMD NX DB -1500 Geode processor. I searched and I found out
> that Microsoft provides BSP for AMD GX DB -1500 processors but not
> for NX series. Please let me know how long it will take to develop
> using Platform builder.

The development time depends on many things:
- how compatible is the NX series with the GX? If you can keep all
the "old" peripherals driver this will shorten the porting time.
- how good is the BSP provided by the manufacturer? Many
manufacturers release BSPs that are not exactly "production code"
(even if they follows the PQD standards...).
- how many of the features of the new processor you need to support
in your new device? For some of those features you may be able to
start from the other BSP and add only the code needed to support the
new features or the differences with the other version. If the
manufacture provides the BSP in binary form you're out of luck (or
almost out of luck) in modifying existing drivers. For other
features you may need to write a new driver from scratch and this
could be a short (days) or long (monhts) task, depending on the
complexity of the peripheral device you need to support, the quality
of the documentation provided by the device manufacture and the
quality of the samples of the same kind of driver you'll be able to
find inside PB or on the net.

I'll start by cloning the existing BSP (do that as your first step,
don't modify the original one or porting your platforms to other PCs
will became a painful process), building an image that contains all
the features you need and check if it runs and how many of those
features already work the way you want.

--
Valter Minute
(the reply address of this message is invalid)
(l'indirizzo di reply di questo messaggio non è valido)

Dean
Thu Sep 28 08:52:49 CDT 2006

Add your experience level on to that. The learning curve for CE BSP
development is quite high, someone with a good deal of experience could do
this job in far less time than it would take someone with little experience.

--
Dean Ramsier - eMVP
BSQUARE Corporation


"Valter Minute" <v_a_l_t_e_r.m_i_n_u_t_e@g_m_a_i_l.com> wrote in message
news:Xns984C7B73D821AVALTERMINUTE@207.46.248.16...
> "ashok" <ashokrajshekar@gmail.com> wrote in
> news:1159437412.244497.79280@e3g2000cwe.googlegroups.com:
>
>> Hi All,
>>
>> I am new to WinCE BSP development. Can someone tell me how long it
>> is going to take to develop a BSP for a customized hardware such
>> as AMD NX DB -1500 Geode processor. I searched and I found out
>> that Microsoft provides BSP for AMD GX DB -1500 processors but not
>> for NX series. Please let me know how long it will take to develop
>> using Platform builder.
>
> The development time depends on many things:
> - how compatible is the NX series with the GX? If you can keep all
> the "old" peripherals driver this will shorten the porting time.
> - how good is the BSP provided by the manufacturer? Many
> manufacturers release BSPs that are not exactly "production code"
> (even if they follows the PQD standards...).
> - how many of the features of the new processor you need to support
> in your new device? For some of those features you may be able to
> start from the other BSP and add only the code needed to support the
> new features or the differences with the other version. If the
> manufacture provides the BSP in binary form you're out of luck (or
> almost out of luck) in modifying existing drivers. For other
> features you may need to write a new driver from scratch and this
> could be a short (days) or long (monhts) task, depending on the
> complexity of the peripheral device you need to support, the quality
> of the documentation provided by the device manufacture and the
> quality of the samples of the same kind of driver you'll be able to
> find inside PB or on the net.
>
> I'll start by cloning the existing BSP (do that as your first step,
> don't modify the original one or porting your platforms to other PCs
> will became a painful process), building an image that contains all
> the features you need and check if it runs and how many of those
> features already work the way you want.
>
> --
> Valter Minute
> (the reply address of this message is invalid)
> (l'indirizzo di reply di questo messaggio non è valido)

Valter
Thu Sep 28 09:14:41 CDT 2006

"Dean Ramsier" <ramsiernospam@nospam.com> wrote in
news:uimt0Tw4GHA.2144@TK2MSFTNGP04.phx.gbl:

> Add your experience level on to that. The learning curve for CE
> BSP development is quite high, someone with a good deal of
> experience could do this job in far less time than it would take
> someone with little experience.
>

That's true.
And it's also likely that someone with a good experience already has a
good library of code that can be used to build the "missing" parts of
the BSP more quickly.

--
Valter Minute
(the reply address of this message is invalid)
(l'indirizzo di reply di questo messaggio non è valido)

Paul
Thu Sep 28 11:11:42 CDT 2006

Our experience level here was for an older x86 process for which we had to
develop a ROM bootloader to load the nk.bin from a PCMCIA card into RAM,
then customize the driver set and, of course, the internal peripherals in
the processor. It took about six months to get everything working. The
tools from MS are better now, so you might get some benefit from that and a
good BSP, as mentioned, from the silicon vendor would help a *lot*, but, if
you do it yourself, plan for months, not weeks.

Paul T.

"Valter Minute" <v_a_l_t_e_r.m_i_n_u_t_e@g_m_a_i_l.com> wrote in message
news:Xns984CA53FA617AVALTERMINUTE@207.46.248.16...
> "Dean Ramsier" <ramsiernospam@nospam.com> wrote in
> news:uimt0Tw4GHA.2144@TK2MSFTNGP04.phx.gbl:
>
>> Add your experience level on to that. The learning curve for CE
>> BSP development is quite high, someone with a good deal of
>> experience could do this job in far less time than it would take
>> someone with little experience.
>>
>
> That's true.
> And it's also likely that someone with a good experience already has a
> good library of code that can be used to build the "missing" parts of
> the BSP more quickly.
>
> --
> Valter Minute
> (the reply address of this message is invalid)
> (l'indirizzo di reply di questo messaggio non è valido)

Valter
Thu Sep 28 11:29:55 CDT 2006

"Paul G. Tobey [eMVP]" <p space tobey no spam AT no instrument no
spam DOT com> wrote in news:ea2uKjx4GHA.1496@TK2MSFTNGP05.phx.gbl:

> Our experience level here was for an older x86 process for which
> we had to develop a ROM bootloader to load the nk.bin from a
> PCMCIA card into RAM, then customize the driver set and, of
> course, the internal peripherals in the processor. It took about
> six months to get everything working. The tools from MS are
> better now, so you might get some benefit from that and a good
> BSP, as mentioned, from the silicon vendor would help a *lot*,
> but, if you do it yourself, plan for months, not weeks.

I agree with you and, as you are talking about the bootloader, this
is one of the most critical part of the porting process.
Adding a different booting media (ex: SD card, USB mass-storage
etc.) is a complex process and on some kinds of hardware (non-PC
hardware) a jtag debugger is mandatory.
Knowing well the OS is very helpful and will allow you to re-use
some code (ex: the code needed to handle a FAT filesystem) from the
CE codebase of from your (or your company's) codebase.
I also did some very quick portings for boards that were very very
similar to the orginal ones (ex: geode-based boards used for
industrial PCs) and this should be considered the opposite side of
the range, requiring not much more than some registry tweaking and
application integration.
But also this kind of easy jobs may hide some evil problems...
As I tried to point out in my message (and I'm sorry if my poor
english doesn't allow me to be very understandable) is that there
are so many factors that could determine the amount of work needed
for a BSP that answering to the original question is almost
impossible. And also after having evaluated all those factors you
may stumb in some nasty problems or bugs that can require days (or
weeks) to work to solve/fix.
I don't know any programmer that knows in advance how long it will
take to find a bug :)
I also have to notice that sometimes the message that comes from the
marketing is that CE is an "easy" OS and that building a new device
from scratch requires a few weeks of work.
I hope that our discussion gave our friend a good "checklist" of
things to evaluate to be able to estimate the effort required for
the porting and to compare it to the cost of a porting made by some
third party company or professional.

--
Valter Minute
(the reply address of this message is invalid)
(l'indirizzo di reply di questo messaggio non è valido)

Paul
Thu Sep 28 12:09:26 CDT 2006

I agree 100%. I found no problems with your replies at all. I just wanted
to put a timeframe on a real-world project, done be relatively new users, in
the thread.

Paul T.

"Valter Minute" <v_a_l_t_e_r.m_i_n_u_t_e@g_m_a_i_l.com> wrote in message
news:Xns984CBC2D3FFC8VALTERMINUTE@207.46.248.16...
> "Paul G. Tobey [eMVP]" <p space tobey no spam AT no instrument no
> spam DOT com> wrote in news:ea2uKjx4GHA.1496@TK2MSFTNGP05.phx.gbl:
>
>> Our experience level here was for an older x86 process for which
>> we had to develop a ROM bootloader to load the nk.bin from a
>> PCMCIA card into RAM, then customize the driver set and, of
>> course, the internal peripherals in the processor. It took about
>> six months to get everything working. The tools from MS are
>> better now, so you might get some benefit from that and a good
>> BSP, as mentioned, from the silicon vendor would help a *lot*,
>> but, if you do it yourself, plan for months, not weeks.
>
> I agree with you and, as you are talking about the bootloader, this
> is one of the most critical part of the porting process.
> Adding a different booting media (ex: SD card, USB mass-storage
> etc.) is a complex process and on some kinds of hardware (non-PC
> hardware) a jtag debugger is mandatory.
> Knowing well the OS is very helpful and will allow you to re-use
> some code (ex: the code needed to handle a FAT filesystem) from the
> CE codebase of from your (or your company's) codebase.
> I also did some very quick portings for boards that were very very
> similar to the orginal ones (ex: geode-based boards used for
> industrial PCs) and this should be considered the opposite side of
> the range, requiring not much more than some registry tweaking and
> application integration.
> But also this kind of easy jobs may hide some evil problems...
> As I tried to point out in my message (and I'm sorry if my poor
> english doesn't allow me to be very understandable) is that there
> are so many factors that could determine the amount of work needed
> for a BSP that answering to the original question is almost
> impossible. And also after having evaluated all those factors you
> may stumb in some nasty problems or bugs that can require days (or
> weeks) to work to solve/fix.
> I don't know any programmer that knows in advance how long it will
> take to find a bug :)
> I also have to notice that sometimes the message that comes from the
> marketing is that CE is an "easy" OS and that building a new device
> from scratch requires a few weeks of work.
> I hope that our discussion gave our friend a good "checklist" of
> things to evaluate to be able to estimate the effort required for
> the porting and to compare it to the cost of a porting made by some
> third party company or professional.
>
> --
> Valter Minute
> (the reply address of this message is invalid)
> (l'indirizzo di reply di questo messaggio non è valido)