[mc1322x] CodeSourcery , flash loader

[Jon Smirl]

On Fri, Apr 30, 2010 at 9:30 PM, Andrew Pullin <pullin at berkeley.edu> wrote:
> Here's a couple of things that I've discovered in my tinkering:
>
> CodeSourcery released a new version of their ARM toolchain recently, and it
> supposedly fixes a lot of issue for minimal ARM targets, specifically the
> atexit() calls. It's pretty nice to just be able to unpack a TAR file and
> not have to build the whole toolchain!
>
> It builds all of the RAM applications, and they all seem to work, including
> printf (need to comment out declarations of stderr).
> Math libraries seem to work and only add a few KB to code size. I've read
> that float-point ops work by compiling with normal FPU instructions, and
> then will handle problems in the interrupt if it turns out to be an illegal
> instruction; I don't know if that's happening here or not. Once I get my
> J-Link JTAG pod I should be able to step through it, I think.

Did you notice Mar's econtags that have the JTAG pod built into them?
http://redwirellc.com/store/index.php?route=product/product&product_id=56

Just stick the tiny thing into a USB port on your PC and you have JTAG
and a serial console. It's a really nice dev evironment and cheaper
than buying an external  JTAG pod. He was hoping that they would get
picked up for classroom use.


>
> The downside: none of application that reference NVM code seem to compile.
> The compiler segfaults when trying to assemble start-romvars.o . No idea
> why, though.
>
>
> On a separate front: up until now, I was running mc1322x-load inside VMware,
> and it was very slow. As an exercise, I implemented it in Python. Comparing
> them now in linux, it seems the python one is faster, by 2x. Maybe it's a
> fluke? If anyone cares, or loves python more than perl:
> http://bmi.berkeley.edu/~pullin/red-bsl.py
> Although, I haven't put in the function where it drops to a serial terminal
> afterwards.
>
> On the same topic of the boot loader, why not use the Second Stage Loader
> that is provided by Freescale? It's a very small binary, ~3K, and it writes
> to the flash really fast, since it uses palletized transfers.
>
> -Andrew
>
> _______________________________________________
> mc1322x mailing list
> mc1322x at devl.org
> http://devl.org/cgi-bin/mailman/listinfo/mc1322x
>



-- 
Jon Smirl
jonsmirl at gmail.com