Lpc2000

--- In lpc2000@yahoogroups.com, "dsidlauskas1" <dsidlauskas@...> wrote:
>
> Jaya,
> 
> Since the compiler knows that buf is a byte aligned object, woutldn't
> it be resonable for the compiler to treat
> 
> (int *)&buf[i]
> 
> as 
> 
> (__packed int *)&buf[i]
> 
> In general, if non-naturaly aligned objects are promoted by a cast,
> shouldn't, by default, the cast be of the proper non-aligned type?
> 
> BTW. Really appreciate your discussion of interupts on your web site.
> 
> Dave
> 

I guess I don't see why the compiler should do anything.  You asked it
to cast the ADDRESS of a byte in memory to an int * which is an
address (by definition) and put it in another int * which is also an
address.  No problem...  So far everything is just an address.

Until you actually use the pointer to grab the contents of memory
there is no alignment issue.  And, on the i386 there still wouldn't be
a problem because the i386 doesn't care about alignment, just
endianness.  It is just a thing with the ARM (and others, I suppose)
that alignment is a concern.

As to porting: well, it is likely that nearly every line of C code is
suspect as almost everything is being done with pointers or structures
or is hardware specific.  Getting the program to work again will be
more of a challenge than simply rewriting it.  How big could it be if
it originally ran on an 8 bit processor?  Unless it is a full O/S like
CP/M and all of the utilities...

The best Fortran programs are written after you spill the box of
cards; rewriting the code may be faster and better.

Richard