An issue for the SC: horrible documentation quality of GCC

[David Edelsohn]

>>>>> Michael Matz writes:

Michael> But this needs to be determined at a later point, instead of cluttering
Michael> all the optimization passes with machine dependend checks.

	Not every processor has an x86 cost model.  Generating fewer
instructions usually is an accurate, generic metric.  However, choosing
one sequence of instructions over another is very machine-dependent.

David