CurieEEPROM refactor #281
CurieEEPROM refactor #281
Conversation
the new version will be based on AVR EEPROM lib
Since FLASH_STTS.ER_DONE and FLASH_STTS.WRITE_DONE are not being set, the delays are hardcoded.
calvinatintel
force-pushed
the
master
branch
from
ce97bd7 to
aaaf3db
Compare
calvinatintel
force-pushed
the
master
branch
from
aaaf3db to
1beffde
Compare
|
This fix passes my local tests :) I had been struggling for many hours with CurieEEPROM 1.0.6 |
| uint32_t data32 = (currentDword & ~(uint32_t)(0xFF << ((3-offset)*8))); | ||
| data32 = data32 | (data << ((3-offset)*8)); | ||
|
|
||
| if (currentDword != 0xFFFFFFFF) { |
There was a problem hiding this comment.
maybe instead of checking the value of CurieRead32 to be 0xFFFFFFFF
check the value of CurieRead8 to be 0xFF instead?
That way we are not clearing and restoring the entire block 4x as much as before.
You can take a look at the implementation of CurieEEPROM::write8
I basically read the dword and insert the byte inside the dword depending on where it needs to be.
There was a problem hiding this comment.
Great! I believe I misinterpreted the datasheet, so I thought that the whole dword needed to be erased before rewriting. Testing with the "reduced" version works too in my test sketches. I'm pushing the modification right now.
@zeptochain , could you give a try to the updated patch with your test sketches? Thanks
There was a problem hiding this comment.
@facchinm will try this over the weekend and report back
There was a problem hiding this comment.
just looked at the commit - not really confident, but will try it - i have a feeling that Martino got it right first time - i'll first write further test sketches to prove out 281 before trying 291 for comparison...
There was a problem hiding this comment.
OK take it all back -- this update passes my test sketches including a complex config job I needed this library to accomplish. Has my thumbs up @facchinm
Previously, if any of the bytes composing a dword was not 0xFF, a clear was executed.
calvinatintel
force-pushed
the
master
branch
2 times, most recently
from
5340b0c to
83c76a6
Compare
|
Cherry-picked into the 1.0.7 branch. Will be closed once 1.0.7 is final. |
calvinatintel
force-pushed
the
master
branch
from
83c76a6 to
77d9086
Compare
|
cherry-picked |
This PR refactors CurieEEPROM library, syncing it with AVR version.
Since EEPROM on Curie chip is based on OTP flash storage, the minimum write size is 4 bytes. The current implementation brings some problems like http://forum.arduino.cc/index.php?topic=421348.0 , and the examples are quite broken.
The refactor also changes the name, from CurieEEPROM to EEPROM, so code for AVR sketches compiles out of the box. The new library resolution engine (from IDE 1.6.8 onward) handles this perfectly but we can stick to older name if we need compatibility with older IDEs.