Re: [avr-gcc-list] Xmega class confusion?

[Erik Walthinsen]

On 06/03/2010 05:05 PM, Weddington, Eric wrote:
> Also take a look at the comments for avr.c in the patch. There it shows the classification 
> of the devices. Those devices that are in an architecture that says ">  64K 
> RAM" means the possibility of using external RAM.
Right, that's the table I copied into my original post.  There are two 
inconsistencies I'm trying to understand and make sure are correct:

1) the 32a4 is listed as avrxmega3, which is 8-64KB Flash, and >64KB 
RAM.  Problem is, *only* the A1 series chips have an EBI, so the 32a4 is 
not capable of more than the 4KB it comes with.  The 32d4 is correctly 
listed as avrxmega2, <=64KB RAM.

2) all the class boundaries are very clearly delineated as far as which 
boundaries are *in* the window, and which are *outside*.  avrxmega5 is 
listed as > 64KB and <= 128KB (and >64KB RAM), yet the only chip ever 
put in that range is the 64a1.  That's inconsistent, as 64KB is not > 64KB.

!!!!! *brainstorm* I think I might have realized where the confusion is, 
and how it can be rectified in comments and documentation:

The > <= window for flash *includes* the bootloader space.  That means 
the 64a4 is actually 64KB + 4KB, so it falls within the "> 64KB" rule. 
If that's the reason, then the comments and docs should indicate that 
"flash size includes bootloader", and that will clear up any confusion. 
 I can draft a patch to that effect.

> And yes this patch will be in the next toolchain release, of course.
Do you mean the mainline binutils/gcc release, or WinAVR et al?

I'll make a point of finding/filing and following Debian and Ubuntu bugs 
to make sure the next versions of the binutils-avr, gcc-avr, and 
avr-libc packages have the proper device support.