v1.0.0

Version 1.0.0 is a breaking release in a minor way: the chkbnds option has been removed, reflecting its upstream removal (inputs are no longer bounded). Support has also been improved for recent versions of JAX.

The finufft dependency is at 2.4.

What's Changed

• Updating the installation documentation by @dfm in #85
• Simplifying lowering and abstract evaluation rules by @dfm in #89
• Update the upstream version of finufft by @dfm in #107
• Fix 1D GPU warning, and a few build improvements by @lgarrison in #126
• ci: bump nvidia container image by @lgarrison in #128
• Use uv on CI by @dfm in #136
• Update for recent versions of JAX. by @dfm in #137
• pyproject.toml: set default pytest path by @lgarrison in #141
• finufft: update to 2.4 by @lgarrison in #143
• readme: add notes on cuda versions by @lgarrison in #144

Full Changelog: v0.1.0...v1.0.0

v0.1.0

What's Changed

• Adding more sensible shape handling by @dfm in #15
• update ShapedArray import by @mlefkir in #18
• Update build-backend to scikit-build-core by @lgarrison in #19
• Update vendored finufft and add GPU support by @lgarrison in #20
• Updating pre-commit rules by @dfm in #23
• Replacing deprecated XLA translation rules with MLIR lowering rules by @dfm in #21
• Making backwards compatible with older versions of JAX by @dfm in #24
• Updating CMake infrastructure to use upstream CMakeLists by @dfm in #22
• Using jax test_utils for dtype-aware test tolerances by @dfm in #30
• Add Jenkins CI by @lgarrison in #31
• readme: update jenkins badge by @lgarrison in #33
• Remove backend_config arg in CPU lowering by @lgarrison in #38
• Update README.md by @lgarrison in #41
• Updating CUDA build to be opt in by @dfm in #43
• Fix indexing of nupts for batched nufft (n_tot > 1) by @lgarrison in #47
• Fixing #54 by @dfm in #56
• ci: use cuda from pip by @lgarrison in #63
• vendor: bump finufft to 2.2.0 by @lgarrison in #61
• installation: update readme and tweak dependencies by @lgarrison in #70
• Adding advanced interface by @dfm in #68
• Update README to include warning about input ranges and info about setting tuning parameters by @dfm in #72
• Fixing clone URL in README by @dfm in #80
• Add check for OpenMP compile state to options by @dfm in #82
• Updating publishing workflow by @dfm in #84
• Replacing pybind11 with nanobind by @dfm in #83
• Fixing nanobind compilation issue on older mac OS by @dfm in #86

New Contributors

• @mlefkir made their first contribution in #18
• @lgarrison made their first contribution in #19

Full Changelog: v0.0.3...v0.1.0

v0.1.0rc2

Fixing nanobind compilation issue on older mac OS (#86)

* Fixing nanobind compilation issue

* Don't debug

v0.1.0rc1

Replacing pybind11 with nanobind (#83)

* Replacing pybind11 with nanobind

* removing extra unused namespace

* build: nanobind doesn't enable LTO by default, no need to globally disable

* Including @lgarrison's suggestion

---------

Co-authored-by: Lehman Garrison <lgarrison@flatironinstitute.org>

jax-finufft v0.0.3

What's Changed

• Fix segfault when batching multiple transforms by @dfm in #11
• Generalize the behavior of vmap by @dfm in #12

Full Changelog: v0.0.2...v0.0.3

jax-finufft v0.0.2

• Faster differentiation using stacked transforms
• Better error checking for vmap

Full Changelog: v0.0.1...v0.0.2

jax-finufft v0.0.1

Initial release