🚀 go-pugleaf

In article <1992Sep28.191208.10169@ulowell.ulowell.edu> jrichard@cs.ulowell.edu (John 'MacGyver' Richardson) writes:

>	While searching though the config .h files I noticed 
>there is code for 680x0 CPUs.  Would it be hard to convert gcc to the
>6811 from a 68000 (I assume it would be a 'scaling down' of addressing
>modes)?
 
There is no GCC port to the 6811 that I am aware of.  There are some major
architectural differences between the 68xxx and 68yy families which make
such a port non-trivial.  You've noted the reduction in addressing modes,
but not specifically the lack of indexed mode off the stack pointer.  There
are some gymnastics involved in working with parameters passed on the stack
when such a mode is missing.  These gymnasitics negatively impact performance.

A second major difference difference is the register set model.  The 68xxx
family has 8 32-bit data registers and 8 32-bit address registers.  The
68yy family has a 16 bit accumulator (or non-simultaneously two 8 bit
accumulators) and two index registers.  The 68HC16 has two 16 bit accumulators
and three index registers.  The 68xxx family registers nicely support
temporary values from expression evaluation, frame pointers, array element
address calculation, and structure field offsets.  One runs out of registers
to do all this nicely on the 68yy family - especially if all these have to
be done in the same source code statement.

A third difference is instruction set regularity and orthogonality.  Not
all 68yy operations work with all operand types and addressing modes.
Multiplication and division are particularly poor in this respect.

I'd be interested in seeing comments from other people that have studied
code generation from HLLs for the 68yy family....



-- 
Noel W. Anderson                   nanders@plains.nodak.edu (Internet)
Asst Prof, Computer Engineering         "I speak for myself, not NDSU"
EE Dept, North Dakota State University         nanders@plains (BITNET)