the "new" instruction provided by i686 is "cmov", which, btw, isn't actually faster anymore compared to explict test and jump with p4's and athlons.
What about on P6 core machines? It seems Intel is heading down the path of multi-P6-core machines.
For Transmeta processors, the dynamic translator turns a jump+mov combination into a select micro-op (the native version of cmov), so it doesn't matter there; and AFAIK VIA don't have them.
-hpa
