Google Groups no longer supports new Usenet posts or subscriptions. Historical content remains viewable.
Dismiss
CC3200 observations
4 views
Skip to first unread message
Thomas Zimmermann
Oct 28, 2016, 10:58:00 AM10/28/16
to mozilla-de...@lists.mozilla.org, Dave Hylands, fde...@mozilla.com
Hi
FYI, here are some observations from my work with the CC3200.
* FreeRTOS is only available at version 8.0.1 in the SDK. This version
does some internal dynamic memory allocation. This works for now, but at
some point we should update to Version 9, which comes with
'CreateStatic' APIs for most data structures. These allow to avoid
dynamic allocation completely.
* When calling xTaskCreate(), *never* give an address on the task's
stack as the created task's parameter. Even though the creator task and
its stack is still present, the new task will not be able to access the
address. It took me a while to figure out why my newly created tasks
aborted. Allocating the task parameter's data on heap memory works.
* Depending on the features in use, C++ requires a number of stub
functions to be implemented by the application. That even happens for
the most trivial classes. Before we deal with this, let's just use C for
now.
Best regards
Thomas
FYI, here are some observations from my work with the CC3200.
* FreeRTOS is only available at version 8.0.1 in the SDK. This version
does some internal dynamic memory allocation. This works for now, but at
some point we should update to Version 9, which comes with
'CreateStatic' APIs for most data structures. These allow to avoid
dynamic allocation completely.
* When calling xTaskCreate(), *never* give an address on the task's
stack as the created task's parameter. Even though the creator task and
its stack is still present, the new task will not be able to access the
address. It took me a while to figure out why my newly created tasks
aborted. Allocating the task parameter's data on heap memory works.
* Depending on the features in use, C++ requires a number of stub
functions to be implemented by the application. That even happens for
the most trivial classes. Before we deal with this, let's just use C for
now.
Best regards
Thomas
Dave Hylands
Oct 28, 2016, 4:45:44 PM10/28/16
to Thomas Zimmermann, fde...@mozilla.com, mozilla-de...@lists.mozilla.org
Fabrice and I both think that we can do most of the work in a mixed rust/C environment.
After working on FoxLink using rust, I've become a much stronger proponent of using rust for this type of thing. It is a bit of a learning curve, but the result is much more reliable code.
For the early prototypes we did on FoxLink the only crash I ever experienced was from calling C code from rust with incorrect parameters.
So I think both fabrice and I would really like to do this in rust, with C mixed in where needed, and eventually rewriting the C into rust as well (down the road). Interfacing rust to C is fairly straight-forward.
I think that if you want to do a portion in C (perhaps at the task/thread level) then that should be fine.
Thomas Zimmermann
Oct 31, 2016, 10:01:52 AM10/31/16
to Dave Hylands, fde...@mozilla.com, mozilla-de...@lists.mozilla.org
Hi Dave, Fabrice,
can we have a chat about this after today's stand-up meeting? I'd like
to understand the benefits and implications of using Rust. I'll be in my
Vidyo room.
Best regards
Thomas
can we have a chat about this after today's stand-up meeting? I'd like
to understand the benefits and implications of using Rust. I'll be in my
Vidyo room.
Best regards
Thomas
Fabrice Desré
Oct 31, 2016, 11:17:59 AM10/31/16
to Thomas Zimmermann, Dave Hylands, fde...@mozilla.com, mozilla-de...@lists.mozilla.org
I have to go to the DMV this morning to get my driver license renewed so
I can't tell if I'll be online at this time. But in any case, Dave knows
it all so proceed without me.
--
Fabrice Desré
Connected Devices
Mozilla Corporation
I can't tell if I'll be online at this time. But in any case, Dave knows
it all so proceed without me.
Fabrice Desré
Connected Devices
Mozilla Corporation
0 new messages