• Locked Locked
  • Replies 3 replies
  • Subscribers 119 subscribers
  • Views 2775 views
  • Users 0 members are here
2025 survey

We are running a survey to help us improve the experience for all of our members. If you see the survey appear, please take the time to tell us about your experience if you can.

Options
Related
How was your experience today?
This discussion has been locked.
You can no longer post new replies to this discussion. If you have a question you can start a new discussion

What happens if ITCM and DTCM overlap on ARM926EJ-S?

Frédéric Hoguin
Note: This was originally posted on 23rd January 2012 at http://forums.arm.com

Hello,

I noticed a bug in ITCM and DTCM initialization in some bootloader code.

The result is that the same value is written in the ITCM region register  and in the DTCM region register, so both are configured with the same  address and size.

It looks like in this case, ITCM gets priority over DTCM since  everything is working correctly when executing code stored in this  region of memory. If DTCM had priority, fetching an instruction from  this region of memory would cause a prefetch abort exception.

Now, that's my understanding of what happens, but I couldn't find anything about it in the ARM926EJ-S documentation.

Can anybody confirm my theory, or explain what happens in detail?

Thanks,

Frédéric.
Parents
  • Note: This was originally posted on 23rd January 2012 at http://forums.arm.com

    Thanks for your answer.

    Actually, there are literal pools in this ITCM code, but I recall reading somewhere that the ITCM can contain data precisely because of PC-relative literal pools (however the DTCM cannot contain instructions). This would mean that for data accesses to PC-relative literal pools, the ITCM is used anyway. I cannot imagine how the ITCM code would work otherwise.

    I hope someone can confirm this or explain what happens in this case.

    Anyway, I agree that it is probably some kind of undefined behavior, and it shouldn't be done even if it seems to be working.
Reply
  • Note: This was originally posted on 23rd January 2012 at http://forums.arm.com

    Thanks for your answer.

    Actually, there are literal pools in this ITCM code, but I recall reading somewhere that the ITCM can contain data precisely because of PC-relative literal pools (however the DTCM cannot contain instructions). This would mean that for data accesses to PC-relative literal pools, the ITCM is used anyway. I cannot imagine how the ITCM code would work otherwise.

    I hope someone can confirm this or explain what happens in this case.

    Anyway, I agree that it is probably some kind of undefined behavior, and it shouldn't be done even if it seems to be working.
Children
No data