C switch performance

[Kurt Akeley]

Regarding the performance of switch statements on Silicon Graphics
R3000-based processors, there really is a performance loss when the
cases are sorted.  An unnecessary jump instruction is included in
this case.  Apparantly this bug had already been eliminated in the
SGI-internal compiler being prepared for a subsequent release.

Thanks for the many notes suggesting other explanations, and
special thanks to Dave Ciemiewicz, who had the real explanation.

-- kurt