paulkoning at comcast.net
Mon Jul 25 11:34:19 CDT 2016
> On Jul 25, 2016, at 12:19 PM, Chuck Guzis <cclist at sydex.com> wrote:
> On 07/25/2016 02:31 AM, Peter Corlett wrote:
>> Eliminating condition codes just moves the complexity from the ALU to
>> the branch logic (which now needs its own mini-ALU for comparisons),
>> and there's not much in it either way. Where it *does* win is that
>> the useful instructions are all single-output and so one can use the
>> noddy code generators found in undergraduate-level compiler
>> construction textbooks such as the Dragon Book.
> Sorry, I'm not following this bit. I am speaking about a three-address
> ISA here, BTW.
> Simply adding a flag to each register reflecting its zero/nonzero
> content should do the job. The high-order (sign) bit is the only other
> bit necessary. Branch instructions need only inquire if either, neither
> or both flags are set.
> I suppose that one could also have several flags registers, being set
> selectively as dictated by a field in the instruction.
The point is that instructions that set condition codes have two outputs: their output register, and the condition code register.
That creates both software and hardware complexity. Software complexity in the compiler, which has to track the assignments to both these outputs and do instruction rearranging accordingly. Hardware complexity, because the instruction issue logic has to issue the instructions for the correct outcome in both these outputs.
More information about the cctalk