Skip to main content
dale
dale
Associate III
July 20, 2015
Question

STM32F746 lacks bit-banding?

  • July 20, 2015
  • 4 replies
  • 1117 views
Posted on July 20, 2015 at 21:53

Does the new STM32F7 series omit bit-banding support?

There is no mention of bit-banding in the STM32F7 reference manual.  Attempting to use bit-banding results in a fault condition.

I'm surprised by this omission.  Or maybe I'm just missing something?

Thanks,

Dale Wheat

#nothing-of-value-was-lost #stm32f7-bit-banding
This topic has been closed for replies.

4 replies

waclawek.jan
waclawek.jan
Super User
July 20, 2015
Posted on July 20, 2015 at 22:12

It's a ''feature'' of the Cortex-M7 core. You are supposed to study the ARM documents.

Migration: http://community.arm.com/docs/DOC-9831

TRM: http://infocenter.arm.com/help/index.jsp?topic=/com.arm.doc.ddi0489b/index.html

JW

Tesla DeLorean
Tesla DeLorean
Guru
July 20, 2015
Posted on July 20, 2015 at 22:24

I'm sure it gets in the way of a rational cache coherency model, and it's pretty dangerous and non-atomic in the peripheral space.

Tips, Buy me a coffee, or three.. PayPal Venmo (See Profile) Up vote any posts that you find helpful, it shows what's working..
jpeacock
jpeacock
Associate
July 20, 2015
Posted on July 20, 2015 at 22:30

Bit Band is an optional feature in the IP core definition.  It is up to the vendor to decide if it's included or not.  I have noticed that not many Cortex M vendors offer bit banding on new parts, maybe it's not worth the space on the chip and the IP royalty.

Too bad if it's being phased out, it's a very nice feature and one I used frequently.  ST (and ARM too) didn't offer much help in how or where it could be used.  One place where it is a real joy to use is with single bit objects in Modbus, no bit masks, no shifting, no word boundaries, just direct, word aligned addressing for the bit.

  Jack Peacock

Tesla DeLorean
Tesla DeLorean
Guru
July 20, 2015
Posted on July 20, 2015 at 22:47

Bit Band is an optional feature in the IP core definition.

Apparently not in the M7, where ARM has basically said you can back-door this in your own memory mapped implementation if you want, but we're not doing it, thanks.

Over the years ARM has been pretty good at eschewing things that add more complexity and transistors than they are worth. With the caching being pulled into core functionality I can see strong drivers for this to be dropped.

Tips, Buy me a coffee, or three.. PayPal Venmo (See Profile) Up vote any posts that you find helpful, it shows what's working..
Terms & ConditionsAccessibility statement