RE: [avr-libc-dev] Sleep.h not appropriate for AT94K devices

[Wouter van Gulik]

> >
> Great! I saw the bug report come across just before you sent this. Thanks!
>
No problem

> >There are several C compilers (for evaluation) on the CD.
> >WinAVR is on it as well (not for eval. of course)
> >This got me wondering why ATMEL puts WinAVR on it when it's not
> compatible
> >(or did the old version work?)
> >
> >
> Since WinAVR is a collection of open source tools, and binutils, gcc,
> and avr-libc say they support the AT94K devices, they probably just
> slapped it on the CD. As I mentioned, you're the first, AFAIK, to give
> the toolset support for the device a good testing.
>
> Do you happen to have an email address of the group at Atmel that puts
> together the CD? I can go bug them to upgrade.

Well the e-mail address that is on almost every file, pdf, CD and eval kit
is:

address@hidden

I suggest you try this one.
Please note that the current cd I have (SystemDesigner 3.0) is dated 2003.
I requested the latest version from my local resaler and I got this version.
So they might be aware of newer version.


> >
> Keep it up!
> Let us know how you progress. And if you run into further problems let
> us know, and fill out bug reports.

Well there is this one thing that really bothers me but it's more a compiler
issue I guess.

If I make a construction like this:

 PORTB = 1<< someUnsignedCharVariable;

Then the compiler promotes the "1" to an int and therefore generates some
extra code:
Example:

eor r25, r25
ldi r24, 0x01
add r24, r24
adc r25, r25

But finally it only uses r24, leaving r25 unused.

out PORTB , r24.

I tried various other options but I just don't seem to get the compiler to
handle the 1 as a char.
Shifting with a constant however works fine. (no promotion to an integer)
Any suggestions?

Thanks in advance,

Wouter