On Mon, Jun 8, 2020 at 10:14 AM Andy Kosela <akosela(a)andykosela.com> wrote:
The real question is can Go be really faster than
That's really a matter of economics more than architecture. Obviously how
to measure it and a ton of things like that will have to be agreed upon.
But in the end, it has to have a wide enough use that it started to be a
financial driver for 'enough' of the business. If people or important
customers start to use it enough, firms like my own will start to say, it's
in our best interest to help create a better code generator. As I said, we
pay more people to work on LLVM than anyone else at this point and when it
finally can handle Fortran as well or better, we have told the world, we
will move to that compiler. We got to that point when it was clear, that
it was the compiler technology of future and 'enough' customer demand it
(BTW: we still have a ton of working to gcc also, but the effort there is
much less than when I started at Intel).
BTW: Apple ships clang - but guess what product compiler Apple's product
teams use for things like Aperture and GarageBand. Intel spends a ton of
money making sure icc(1) is the best for them [And again, when clang's
C/C++ can match icc - our execs will want to go there in a heartbeat I
A couple of years ago at an HPC workshop, I was talking to the folks.
They all screamed for a parallel language, but was all over the map.
Chapel turns to have a small backing (Cray wrote it for a DoD contract),
but even the spooks don't use it for production (that all Fortran an C).
After a enough digging, we just rolled the dice with DPC++ because the Venn
diagram of 'enough' of our HPC customers showed it would be helpful. We
shall see (that took some convincing in management to make that investment)
and right now, its a parallel effort to C++.
As I have said before Cole's law: 'inexpensive economics beats
If Go (or Rust - whatever) starts driving silicon sales for enough folks,
you see IBM, Intel et al start getting really excited.