Interesting result using spin2cpp on code with fullduplexserial in it...

Rayman

Just tried using spin2cpp on this test program for RamPage2 memory module.
RP2_test6 - Archive [Date 2013.03.02 Time 05.45].zip

I had a dream that maybe it would just work, but I know that's asking a lot...

Anyway, I first tried LMM mode and got garbage on the terminal window.
When I reverted to the non-P2 version of propgcc I got more legible text, but still not right.

Went to CMM mode and serial terminal text looked almost right.
Turned off all optimization (-O0 None) and now the text looks right.

Is this kind of the expected behavior when trying to directly port fullduplexserial to propgcc?

jazzed

Does it work with this manually ported package?
http://forums.parallax.com/attachment.php?attachmentid=99241&d=1360890415

Rayman

Ok, I did notice that in this forum. Guess that is the recommended replacement?
I'll keep that in mind.

But, because I want this to work on DE0, I think I need to replace all the serial stuff with printf.
That's probably a better approach anyway, right?

SRLM

You should try adding {++volatile} to the inter-cog variables. After that, I've been able to convert FSRW and FFDS1. For example:

var
    long {++volatile} shared_pointer

Rayman wrote: »

But, because I want this to work on DE0, I think I need to replace all the serial stuff with printf.
That's probably a better approach anyway, right?

On the Propeller 1, I prefer to use a standalone serial driver for most things, although printf is handy for quick testing programs. I can't speak for the Propeller 2.

David Betz

jazzed wrote: »

Does it work with this manually ported package?
http://forums.parallax.com/attachment.php?attachmentid=99241&d=1360890415

Is this a new improved version of FDS for PropGCC? If so, could you check it into the main PropGCC repository so it can be used by all users of PropGCC, not just SimpleIDE users?

Thanks!
David

jazzed

David Betz wrote: »

Is this a new improved version of FDS for PropGCC? If so, could you check it into the main PropGCC repository so it can be used by all users of PropGCC, not just SimpleIDE users?

Thanks!
David

You make many assumptions That program is over 5 years old now.
It was committed once, but got morphed into our current FDX handler with stdio support.

Rayman

nice tip SRLM... That seems to have fixed it... It appears that another way, besides adding "{++volatile}" to the spin is to find them in the .h file and add "volatile" there...

David Betz

jazzed wrote: »

You make many assumptions That program is over 5 years old now.
It was committed once, but got morphed into our current FDX handler with stdio support.

Okay thanks for the history lesson! :-)
I'm going to need to look at this soon to port it to P2 so I guess I should know this code better than I seem to. Sorry!

ersmith

Rayman wrote: »

Went to CMM mode and serial terminal text looked almost right.
Turned off all optimization (-O0 None) and now the text looks right.

Is this kind of the expected behavior when trying to directly port fullduplexserial to propgcc?

FullDuplexSerial works fine with -O0 and -O1, but sometimes has problems with -O2 and -Os (depending on timing). In general you'll have better luck with spin2cpp with -O1 rather than -Os.

Alternatively, as others have suggested here, you can annotate the Spin source code with {++volatile} declarations to let the compiler know that variables will be modified behind its back. Using {++volatile} is better than manually editing the .h file, because it allows you to make changes to the .spin and re-run without having to edit the .h file every time.

Eric