The Itanic Saga: The History of VLIW and Itanium

Article note: Because I know too much about this, I like comparing tellings. This includes some international context that most tellings don't, which is fun, but skips some of the domestic context like Apollo, Multiflow, and Convex/Convey. Marks but intentionally does not discuss some of the other VLIW fad things like Transmeta. It makes an amusing clear-but-understated presentation of the degree to which Itanium was a bulldozer in the industry. It makes the back-ties to i860 and PA-RISC, and somehow does Elbrus but not iAPX423 (which was an unmitigated disaster) or Burroughs Large Systems (if you want an example that wasn't a disaster) when talking about HLL support in processors. It doesn't touch the (hilarious) thread that the Alpha team leaving for AMD after HP killed it in favor of Itanium is _very directly_ the genesis of the initial AMD64 designs, which were the stake through the heart in the market. And it disappointingly doesn't really talk about _why_ the performance was consistently underwhelming in a semi-technical way. The great lesson of the era is that "You can't statically schedule dynamic behavior." Some of the HN comments get there in nice detail.