Add support for subsampling, PERMANOVA and multi-unifrac

What's Changed

• Add support for multi unifrac to file by @sfiligoi in #151 and #152
• Add PERMANOVA and multi-object h5 reads by @sfiligoi in #150 and #153

Full Changelog: 1.2...1.3

Change default precision and deprecate the threads argument

What's Changed

• Deprecate threads and add n_substeps by @sfiligoi in #146
• Switch from condensed to matrix API calls by @sfiligoi in #147
• Change default precision to fp32 and add explicit fp64 functions by @sfiligoi in #148

Full Changelog: 1.1.1...1.2

Bug fix for compatibility with biom-format 2.1.12

What's Changed

• Expand build matrix by @wasade in #144

Full Changelog: 1.1.0...1.1.1

Provide in-memory support against unifrac-binaries

What's Changed

• Support for passing biom and the tree in memory by @wasade in #142

Full Changelog: 1.0.0...1.1.0

Release associated with manuscript describing OpenACC improvements

This is a historical release documenting the code used for benchmarking the OpenACC release. We advise using the latest version of this software. Details for install, including a functional example, can be found here.

Python-only repo, depends on unifrac-binaries

Since the python binding and the backend implementation evolve independently, the sucpp code was moved into an independent repo (unifrac-binaries).
This repo now only contains the python bindings.

Bug fix for massive PCoA and simplified build

As the title says, thanks @sfiligoi!!

Automatic PCoA compute

This release improves documentation and automatic PCoA computation

Cache locality, and other performance improvements

This release brings substantial performance improvements to UniFrac for CPUs and GPUs thanks to @sfiligoi's incredible rework of the underlying algorithms and implementations, as well as a few minor bug fixes.

GPU optimization, performance improvements, binary output

With this release, we expose the ability to apply UniFrac to GPUs with upwards of 500x performance gains. More detail on the optimization can be found in an arXiv post, made possible thanks to the incredible efforts from @sfiligoi! While the primary focus was the adaption to GPUs, many of the improvements extend to CPUs as well.

Thank you as well to @gwarmstrong and @ChrisKeefe for contributions and testing over the past year since the last release!