What ever happened to Tandem and NonStop OS 2012

Morten Reistad

I don't think they are mindsets, I think they are skillsets. The naming is utter tripe, she shouldn't use it IMO. It also makes absolutely zero sense when you apply it to something like OCCAM & the Transputer. If they were mutually exclusive mindsets (BAH's usage implies that) then OCCAM & the Transputer would never have existed at theory, silicon, compiler or OS level. I suspect that BAH hasn't had much exposure to MPP, and that maybe why I end up having cross words with her on this topic.

The folks who write highly parallel applications *HAVE* to think in both frames of mind and just about every permutation in between. That is why parallelism is often considered "hard", but in practice I don't believe it *has* to be hard. People seem to make a hash out of it because they are too scared or unwilling to straddle the mindsets. In practice MPP programming sharpens your reasoning about dynamic systems very quickly. If it doesn't you quit or get fired very quickly. ;)

"OS thinkers" and compiler writers don't get to think in terms of well defined states these days either. Branch prediction is a given on most platforms these days, yet the majority of OS writers don't even get to know what the prediction rules are. Therefore the OS writers don't have a well-defined "physical" state to work with because they have an incomplete model of the hardware.

I can see what you are driving at MRR, but I think you're going to have to think it through a little more. :)

Cheers, Rupert