• add mmrprj.back_prj(div_sino=None) CUDA convenience
• add & expose projector sync=True CUDA optimisation
• add bprj(output=None) convenience (follow-up to #36, part of #33)
• remove unneeded aw2ali
• add mmraux.remgaps CuVec support
• clean .cuvec usages (AMYPAD/CuVec#17)
• use nimpa.dcm2nii
• drop PATHTOOLS & REGPATH (#40)
• support py3.10
• misc fixes
• misc tests updates

• cmake: allow linking shared libs for external builds
• cmake: expose NIPET_CU_THREADS=1024
• cmake: fix multi-arch CUDA support
• setup: move hardware attenuation maps logic to runtime
• optimisation: use CuVec forward & back projectors (#36)
• update logging usage
• update resources
• update GIF parcellation file recognition
• fix memory leaks (#39)
• fix device ID detection with NInst
• fix deprecation warnings
• fix RTX & reduced rings bugs
• fix & update get_norm()
• fix span-1 sinogram numbers in scatter and pipe recon
• fix generating aligned mu-maps when loading existing ones
• fix np.bool deprecation
• fix editable installation
• docs: update examples/demo
• format: misc tidy & linting
• major build & test framework updates (#32)
  • automate code formatting & basic bug detection
  • isolate test env
  • separate check & deploy
  • add test runner dependencies
  • update pre-commit hooks
  • cmake: static CUDA lib linking
  • use clang-format

Features

• support Python3.6+ (#15 <- #14)
• add customisable PSF resolution modelling kernel (#31, #25)
  • support all of: scanner measured, manual Gaussian, and custom separable PSF kernels
• prevent floating point division errors (#25, #23)
• improve classify_input
  • add optional directory recursion (#24)
  • basic pathlib support
• fix & upgrade logging (#9, #5)
  • provide & use new LOG_FORMAT
• fix & update dependencies
• fix & update example notebook (#13, #9)
• fix nonexistent paths (#22)
• add progressbar (#4)

Installation/build framework

• support Python3.6+ (#15 <- #14)
  • drop Python 2 compatibility
  • drop py<=3.5 which are unsupported by ninst<-miutil[cuda]<-pynvml and no longer receive security fixes anyway
• add tests (#29)
  • including helpers for downloading data
  • add cloud testing
  • add coverage reporting
• fix warnings (#28, #27)
  • update CMake policies & fix Python deprecation warnings
• improve CUDA code (#28, #27)
  • better error handling
  • de-deuplication
  • better linking
  • fix compiler warnings
  • remove unneeded variables
  • put > stream[%d] messages on one line
  • only create poisson distribution if needed (#19)
• fix conda install
• allow non-interactive installation (#17, #9)
  • allow setting PATHTOOLS & HMUDIR from env vars
• new build framework (#30, #27)
  • cmake/cuda build overhaul
  • move metadata setup.py -> setup.cfg
  • allows pip install nipet on a completely clean system (no numpy, no miutil[cuda], no -e, no --no-binary :all:)
  • replace dinf module => miutil.cuinfo (#26 <- #18)
  • rely on numpy, miutil[cuda], ninst, scikit-build, cmake>=3.18, ninja
    • all auto-installed at setup-time!
    • get rid of cudasetup.py
    • get rid of install_tools.py
    • get rid of resources.py
  • use new CMAKE_CUDA_ARCHITECTURES policy
  • use new project LANGUAGES specification
  • support python setup.py clean
• basic code tidy (particularly import ordering)
• update .gitignore, README.rst badges (#9, #4)

Final v1 (Python2) release