I wonder how would people feel about itanium failure had it been developed by anyone other than Intel; I mean, would there be that lingering feeling it could have been something had only the right amount of resources put into it etc etc?
Itanium was actually developed by HP starting in 1989 and then Intel joined in 1994.
Not only that but in the early 2000s HP effectively gave Intel all their hardware design team to help with the effort.
Maybe? The architecture is weird -- it required user programs to be aware of and utilize superscalar pipelining. Superscalar pipelining was a great bet, the assumption that programmers and application writers could explicitly take advantage of it was not. It's a really COOL architecture, but I'm not sure it could ever have been a commercially successful one.
For comparison, how do you feel about Transmeta?
Probably not, it felt a lot like an experiment that was picked to win and then still didn't. Almost like a reverse i960, which ended up being really good at a few niche applications that certain problem spaces kept it in at least some demand long after it was effectively abandoned.
no. in order VLIW dependent on handwaving compiler magic never works.
I still think there is a medium between the two that could work - if you look at uops per port measurements there's still a lot of room left for improvements i.e. the mantra is "The compiler will never know more than the chip" but sometimes the CPU is pretty dumb too