How does the future of ParallelAccelerator look like?

It seems I cannot add ParallelAccelerator to Julia 1.0 because it requires version 0.6. That being said, the last commit to the package is from October 2017. Are there any plans for:

  1. making the package compatible with Julia 1.0,
  2. continue the development (i.e. adding features / improving performance),

or is the project dead?

It’s possible they would accept pull requests. However, I think it’s fair to say that Julia itself now has all the features that were “farmed out” to gcc for ParallelAccelerator, so you could just use Julia itself to accomplish the same task.

4 Likes

It would be nice though if someone made a @performance macro kind of like ParallelAccelerator, but with a Julia base. For example, take naive array expressions and insert some cache arrays to reduce allocations, multithread the loops, etc. ParallelAccelerator’s macro did some cool stuff, and now all of the features that it built are in Julia, but some of the macro translations itself were still interesting.

(I guess in v1.0 since array expressions like A+B require broadcast there are even less cases where the macro is truly useful than before though…)

2 Likes

I found the @runStencil macro very useful. I was using it for advancing a heat equation in time. Having the macro readily available saved me the hassle of implementing stencil computations in parallel myself. Should not be too difficult, but still, why reinvent the wheel.