Skip to content
New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Refactor to use StructsOfArrays.jl #19

Closed
mweastwood opened this issue Jul 30, 2015 · 2 comments
Closed

Refactor to use StructsOfArrays.jl #19

mweastwood opened this issue Jul 30, 2015 · 2 comments

Comments

@mweastwood
Copy link
Owner

I should be able to eek out some more performance by using StructsOfArrays.jl to SIMD vectorize the inner calibration loop.

@mweastwood
Copy link
Owner Author

This turns out to not be useful for polcal:

  • A StructOfArrays(JonesMatrix, dims...) creates 4 Array{Complex64}s. What we really want is 8 Array{Float32}s.
  • I experimented with a SplayedJonesMatrix type that stores the real and imaginary components of the elements separately (to fix the previous problem). However this makes the code pretty ugly.
  • Furthermore, I couldn't get the inner calibration loop to vectorize. Possibly because expanding the complex matrix multiplies in terms of real operations spooks the compiler with too much complexity. Not 100% sure though (I did make sure as much as possible was getting inlined..)

This may still be useful for the unpolarized calibration.

@mweastwood
Copy link
Owner Author

Closing because I no longer think this is the right thing to do to get SIMD vectorization. My current thinking is that maybe SIMD.jl might be helpful for this, but it will require some work that I don't think is terribly important right now.

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Labels
None yet
Projects
None yet
Development

No branches or pull requests

1 participant