Device bus cycles (STM32F427)
- Mark as New
- Bookmark
- Subscribe
- Mute
- Subscribe to RSS Feed
- Permalink
- Email to a Friend
- Report Inappropriate Content
‎2018-11-18 7:23 AM
Is there a document that discusses the buses?
I was wondering how much Chrome-Art might affect Flash-Art and CPU, if a picture is stored in flash.
How many bus cycles Chrome-Art takes for a transfer and how does Flash-Art access the flash.
I'm working on something that the bus timing might be important for.
My current guess is that Chrome-Art takes about 10 cycles per transfer and maybe 2 cycles of them are such that Chrome-Art is not using the bus and Flash-Art can use those cycles.
Solved! Go to Solution.
- Labels:
-
STM32F4 Series
Accepted Solutions
- Mark as New
- Bookmark
- Subscribe
- Mute
- Subscribe to RSS Feed
- Permalink
- Email to a Friend
- Report Inappropriate Content
‎2018-11-18 7:38 AM
Figure it is designed for code execution and the flash line reads takes 35ns
You can use the cycle timer to time and benchmark things accurately. ​
Up vote any posts that you find helpful, it shows what's working..
- Mark as New
- Bookmark
- Subscribe
- Mute
- Subscribe to RSS Feed
- Permalink
- Email to a Friend
- Report Inappropriate Content
‎2018-11-18 7:38 AM
Figure it is designed for code execution and the flash line reads takes 35ns
You can use the cycle timer to time and benchmark things accurately. ​
Up vote any posts that you find helpful, it shows what's working..
- Mark as New
- Bookmark
- Subscribe
- Mute
- Subscribe to RSS Feed
- Permalink
- Email to a Friend
- Report Inappropriate Content
‎2018-11-18 10:17 AM
So no such documents...
Well, thanks.
- Mark as New
- Bookmark
- Subscribe
- Mute
- Subscribe to RSS Feed
- Permalink
- Email to a Friend
- Report Inappropriate Content
‎2018-11-18 3:54 PM
DMA2D_AMTCR is there to limit the impact of DMA2D to other bus masters.
I'd suggest to benchmark.
JW
- Mark as New
- Bookmark
- Subscribe
- Mute
- Subscribe to RSS Feed
- Permalink
- Email to a Friend
- Report Inappropriate Content
‎2018-11-18 3:54 PM
DMA2D_AMTCR is there to limit the impact of DMA2D to other bus masters.
I'd suggest to benchmark.
JW
- Mark as New
- Bookmark
- Subscribe
- Mute
- Subscribe to RSS Feed
- Permalink
- Email to a Friend
- Report Inappropriate Content
‎2018-11-18 5:12 PM
>>So no such documents... Well, thanks.
I don't know, I've seen presentations about the ART, and I have a reasonable understanding of the underlying architectures in an SoC context. The CM4 doesn't have an architecturally defined cache, ST bolted something on the side to provide a fast prefetch path based on widening the flash read line width, 128-bit as I recall.
It really isn't geared to one-time, non-temporal use, or access via DMA, etc. In the worst case it is going to cause bus contention, likely blocking for 6 cycles or more, per access.
Where documentation is lacking, or of questionable value, one has to read between the lines, apply knowledge of other ARM implementations, and formulate tests to prove or disprove the viability of different methods.
Up vote any posts that you find helpful, it shows what's working..
- Mark as New
- Bookmark
- Subscribe
- Mute
- Subscribe to RSS Feed
- Permalink
- Email to a Friend
- Report Inappropriate Content
‎2018-11-19 12:07 PM
Yes, I have searched ARM pages if something is missing from the STM32-documentations. Sometimes I have found something.
Like the description of bus matrix IP.
- Mark as New
- Bookmark
- Subscribe
- Mute
- Subscribe to RSS Feed
- Permalink
- Email to a Friend
- Report Inappropriate Content
‎2018-11-19 12:08 PM
It seems to be the only way...
The chip even has the trace cell, but I don't have any trace-capable POD. :(
