3 ms·

> To write parallel programs for such hardware, researchers have converged on using implicit parallelism, where the programmer expresses all opportunities for p

by klabb3 5y ago

> To write parallel programs for such hardware, researchers have converged on using implicit parallelism, where the programmer expresses all opportunities for parallelism, and the compiler and run-time system then work together to manage parallel execution, relieving the programmer from an otherwise tedious and difficult task. This idea would be hugely controversial in all places I've worked at, and is often the opposite direction of where many languages, runtimes etc are moving. Maybe it's just innocent wording, but "research converging" on something that few in the industry believe in makes me think academia may be navel-gazing beyond healthy levels, even if the research itself is very interesting.

choeger 5y ago
> Maybe it's just innocent wording, but "research converging" on something that few in the industry believe in makes me think academia may be navel-gazing beyond healthy levels Did the industry believe in the value of linear or affine types? Sound static typing? Pure functional programming? It's the typical arrogance of the practionier to rule out computer science research as irrelevant just because the application doesn't come in a nicely advertised package yet. Of course, the scientific community can get lost on wrong paths or mix up fashion with established knowledge. I don't know enough about this particular topic to form an opinion here. But in general, when there has been a scientific consensus regarding programming languages it turned out to be relevant for the field.
- klabb3 5y ago
  > I don't know enough about this particular topic to form an opinion here. To be fair, I don't either, depending on what's enough. I base this on what I've read, heard and experienced regarding GCs, query planning, and a lot of runtime "magic", that often end up being over generalized black boxes in many practical cases. In turn, this led to a Renaissance of simpler systems that are predictable, easy to reason about, and tweakable for the specific domain. Just my observations. > But in general, when there has been a scientific consensus regarding programming languages it turned out to be relevant for the field. I have a different view of academia than most people, so not expecting agreement here, but scientific consensus in a field like programming language theory should be rare, in fact I'd expect much more vigorous debate. This field is so young, incredibly complex and has common surface area with industry use cases and our mental models.

by klabb3 5y ago

choeger 5y ago

> Maybe it's just innocent wording, but "research converging" on something that few in the industry believe in makes me think academia may be navel-gazing beyond healthy levels Did the industry believe in the value of linear or affine types? Sound static typing? Pure functional programming? It's the typical arrogance of the practionier to rule out computer science research as irrelevant just because the application doesn't come in a nicely advertised package yet. Of course, the scientific community can get lost on wrong paths or mix up fashion with established knowledge. I don't know enough about this particular topic to form an opinion here. But in general, when there has been a scientific consensus regarding programming languages it turned out to be relevant for the field.

klabb3 5y ago
> I don't know enough about this particular topic to form an opinion here. To be fair, I don't either, depending on what's enough. I base this on what I've read, heard and experienced regarding GCs, query planning, and a lot of runtime "magic", that often end up being over generalized black boxes in many practical cases. In turn, this led to a Renaissance of simpler systems that are predictable, easy to reason about, and tweakable for the specific domain. Just my observations. > But in general, when there has been a scientific consensus regarding programming languages it turned out to be relevant for the field. I have a different view of academia than most people, so not expecting agreement here, but scientific consensus in a field like programming language theory should be rare, in fact I'd expect much more vigorous debate. This field is so young, incredibly complex and has common surface area with industry use cases and our mental models.