Comment by vladms

> Given the above predicate: Isn’t everything described in the article as it should be?

I think the real trick question is "as it should be for whom?".

Reading the comments I think people underestimate the complex interaction between:

- engineers that design hardware (they don't care much about the compiler, except when it has to fix their mistakes)

- engineers that do the compiler (they have to struggle with all quirks of the new architecture and all of the complaints of the users)

- users of the new system (hardware + compiler) that just want to take their 100k lines of code (libraries) and just use it on the new system with better performance (as that's what the hardware people promissed!)

- users working on one architecture all their lives

For the compiler people, yes, probably most what is described is as it should be. For the users (that care about performance and not making porting efforts), probably no.

Now, even when I was doing compiler work we had a hard time explaining our users why we couldn't do some things they wanted (while also improving performance and not changing code that was writting), so explaining that on the internet seems to me a lost battle.

I am sure there are things that can be improved, and standards evolve. But the problem is very complex given the sheer amount of code written and the strange architectures out there.