Possibly the WiKi is not quite up to
date.
Please correct me, but it was my impression that
the general opinion was that the 4-byte alignment method will be less
efficient and more difficult to implement.
1.) IMO the fastest way will be to insert jump stubs
to the function entry point in the lower 128k as part of the
prologues
2.) Change the tablejump and casesi instruction patterns so
that they work also on the larger devices
3.) Correct the call patterns so that they directly call
the function body in the higher memory areas and not the jump
stubs
4.) add support within the binutils for a new class
avr6
5.) make the necessary modifications in avr-libc for the
larger devices (include header files and change the loop for the initialized ram
variables such that it uses 3-byte LPM because the initialization data end up at
the very end of the flash.
I am working on 1-3.
Bjoern.
_______________________________________________ AVR-GCC-list mailing list AVR-GCC-list@nongnu.org http://lists.nongnu.org/mailman/listinfo/avr-gcc-list
