Numba 0.28.1 released

[Stanley Seibert]

Hi Numba Users,

It's time for another release, and this is the first release supported by the Moore Foundation grant that was announced in July.  This release moves us forward to llvmlite 0.13, which depends on LLVM 3.8.  (If you are wondering what happened to Numba 0.28.0, we needed to tag a new release to fix a setuptools dependency.)

There's a bunch of new features and fixes in this release, but here are a few:

• The Numba RNG is now fork and thread-safe!  The RNG state is thread-local and automatically initialized as needed.
  
• Added documentation of floating point quirks and pitfalls you may run into with statically typed code: http://numba.pydata.org/numba-doc/latest/reference/fpsemantics.html
• Fixed performance regression for parallel target (very noticeable on Windows)
• Reduced function call overhead when using default arguments
• Support for numpy.concatenate() and stack()
• Support for numpy.argsort()
  
• Major reduction in overhead when calling numba.cuda functions that access the CUDA driver
  
• Support for structured dtypes that have "titles" for their fields.  (Required for the vector types in PyOpenCL.)
• Support for more numpy.linalg functions: matrix_rank, cond, norm, det, slogdet
• Many bug fixes for issues reported by the Numba community
  

More details here: http://numba.pydata.org/numba-doc/latest/release-notes.html#version-0-28-0

You can download the Numba source release from PyPI:

https://pypi.python.org/pypi/numba/0.28.1

or installed with conda:

conda update numba

As always, thanks for your questions, bug reports, and feature requests!

[Zhang Richard]

The next version(0.29) support CUDA 8.0 or not ?

[Stanley Seibert]

We hope so.  CUDA 8 has not been released yet (just the release candidate), but when it is, we plan to move to it.  Look for CUDA 8 in the October Numba release.

On Fri, Sep 2, 2016 at 3:50 AM, Zhang Richard <foxpr...@gmail.com> wrote:

The next version(0.29) support CUDA 8.0 or not ?

--
You received this message because you are subscribed to the Google Groups "Numba Public Discussion - Public" group.
To unsubscribe from this group and stop receiving emails from it, send an email to numba-users+unsubscribe@continuum.io.
To post to this group, send email to numba...@continuum.io.
To view this discussion on the web visit https://groups.google.com/a/continuum.io/d/msgid/numba-users/a6e57c95-625c-49bf-913f-6d19968916a5%40continuum.io.
For more options, visit https://groups.google.com/a/continuum.io/d/optout.