🚀 go-pugleaf

> The one I tried is called Version 1.03.  When you try to compile
> something with it, it bombs when it gets to the point of looking
> for the non-existent optimizer.  The illegal instruction sequences
> include using an address register on the instruction after loading it,
> and doing a subroutine return on the instruction after modifying
> the system stack (i.e. popping the status).

How hard is this? Motorola has a product named DSP56KCC. It is a C 
compiler system that generates executable programs for the Motorola 
DSP5600X family of chips. Much of the aforementioned system is based 
gcc. The sources to the gcc based portion have been made available.
You have part of the system, and you expect everything to work? The 
compiler system doesn't generate illegal code sequences as you say;
nop insertion is performed later in the code generation process, and 
by non-FSF code. No, it isn't performed by the assembly language 
optimizer, it is performed by the assembler.

BTW, I really appreciate it when people point out real problems with 
the compiler. When we find out about bugs we fix them. I'm sorry if 
I sound grumpy, but hey, the thing works. If you later discover any 
real bugs, I would really appreciate hearing about them.

	no hard feelings,

	Jeff Enderwick
	jeff@dsp.sps.mot.com

PS: I would think that spending $709 (this is *not* a price quote) for
a compiler kit that includes an assembler, linker, librarian, libraries
(both ANSI and runtime support), a manual and phone support would be 
cheaper than fooling around with the sources. But I guess that's not my
decision to make.