7.3 Numba: A Python compiler using LLVM (invited)

Wednesday, 9 January 2013: 9:30 AM
Room 12B (Austin Convention Center)
Mark Florisson, Continuum Analytics, Austin, TX; and S. K. Lam

Numba is a NumPy aware optimizing compiler for Python, which uses LLVM to speed up numerical Python code operating on NumPy arrays or scalars, without needing to code in low-level languages like C or Fortran, or type annotations like Cython to get good performance. Functions can be decorated to use just-in-time specialization for the input argument types, or they can be specified statically. Numba furthermore allows the user to write custom NumPy ufuncs in plain Python. Numba can also use minivect to support efficient arrayexpressions that exceed even the performance of Fortran, and more efficient application of user-defined ufuncs.
- Indicates paper has been withdrawn from meeting
- Indicates an Award Winner