GT4Py v1.0.7
Summary of changes since v1.0.6
Cartesian
- Introduce switches for the default
intandfloatprecision. - Introduce
erfanderfcfunctions. - Make CUDA compilation thread-safe.
Development
- Add MacOS to daily CI.
Next
See commit history.
All changes
- feat[next][dace]: Updated GPU Transformations by @philip-paul-mueller in #2197
- ci: add MacOS to daily ci by @havogt in #2153
- feat[cartesian]: Expose
erf()anderfc()functions by @romanc in #2199 - feat[cartesian]: literal precision (float and int) by @romanc in #2187
- fix[next][dace]: Wrong memlet subset in lowering of deref by @edopao in #2200
- feat[next]: Indexing field in local dimension by @havogt in #2190
- fix[cartesian]: fix setuptools for parallel CUDA compilation by @SF-N in #2204
- build[dace][next]: Updated DaCe Dependency by @philip-paul-mueller in #2206
- fix[cartesian]: respect literal precision in dace backends by @romanc in #2208
- fix[dace][next]: "Fixed" Asynchronous SDFG by @philip-paul-mueller in #2201
- bug[next]: change passes to get rid of neighbors in list_get operations by @havogt in #2212
- Releasing v1.0.7. by @philip-paul-mueller in #2207
Full Changelog: v1.0.6...v1.0.7