My impression so far is that the -QaxW and -QxW flags break binary compatibility with MS's linker, and require the use of intel's xilink. Is that correct?

If so, is it possible to get theP4-specificSSE loopvectorization without breaking binary compatibility?