[En-Nut-Discussion] Question regarding missing examples

Ulrich Prinz ulrich.prinz at googlemail.com
Fri Oct 28 00:05:06 CEST 2011


Hi Uwe,

you're right. As project pressure rose, I had to do some ugly
quick-hacks and to postpone some other good ones as I had no time to
carefully test them. The merge I am doing is a bit more work than I told
in my last mail.

To be precise I am actually doing this:

1) merge devnut_m3n and trunk.
Test on all available test-kits I have
2) merge my private and my work copy to devnut_m3n
Test on all available test-kits I have
3) merge devnut_m3n to trunk
Test on all available test-kits I have

Test kits are different STM32 kits, AT91SAM7X-EK and EIR.

The biggest part of work will be the upgrade on ARM architecture. There
are some big changes deep down in the drivers. But they are worth the
work as they give great benefit.

Examples are the abstraction of GPIOs where the ports are assigned to
addresses, not numbers. This speeds up GPIO access a lot compared to the
switch/case constructs we have.
Second USARTs, SPI and I2C got a device description struct and their
code is only compiled in once. You have to pay with some 10 bytes of RAM
but you spare hundreds of bytes of FLASH. You even increase speed.

All changes where made with almost no modification to the upper API.
Existing code could be recycled with new nutos with almost no changes.

There is one important change, GPIO_CFG_OUTPUT will init to
open-collector / open-drain. You have to set Push/Pull explicitly as
that may kill something on your board. MULTIDRIVE will be obsolete.

And yes, the whole porting was pretty much work and I had to do that
99.5% on my own. But to be honest the porting of Cortex was not that
problematic. The project behind was the thing... It does not only
consist of some STM32 but involves a i.MX based Linux box, non working
MMC drivers, CAN bus with CANopen protocol, device drivers, device
managers... and only three people working on it and a sub-contractor
that never delivered what he was paid for.

So if there is again a small stop in the merge, you find me writing an
eMMC driver for linux and U-Boot... Must be ready on mid of November.

Best regards
Ulrich



More information about the En-Nut-Discussion mailing list