Changelog¶
v2.0.5 (2026-03-25)¶
Fixed¶
Fix bug in TOF listmode forward and backprojector such that only image planes are taken that where the ray intersects the image volume. Now the behavior of sinogram and LM TOF projectors should be very close for all LORs and the sum over TOF should be very close to non-TOF.
v2.0.4 (2026-03-13)¶
Fixed¶
remove cuda arch auto detection magic (conda-forge sets CUDAARCHS env variable)
improve install docs
v2.0.3 (2026-03-05)¶
Fixed¶
bugfix to support all cuda 12.x versions (cudaMemLocation)
improved auto version detection from git in build process
v2.0.2 (2026-03-04)¶
Fixed¶
better default install directories of python bindings compatible with conda-forge
v2.0.1 (2026-03-04)¶
Fixed¶
add VERSION and SOVERSION number to build of shared library libparallelproj
v2.0.0 (2026-03-03)¶
Added¶
direct support for host arrays, cuda managed arrays and cuda device arrays in the C API
support for even number of TOF bins in TOF projectors
nanobind Python API using python abi3 (python 3.12+ required) for better compatibility and easier installation
Breaking Changes¶
listmode TOF forward and backward projectors now require the extra argument
num_tof_binsto specify the number of TOF bins, which can be even or odd.renaming:
joseph3d_fwd_tof_sinotojoseph3d_tof_sino_fwdrenaming:
joseph3d_back_tof_sinotojoseph3d_tof_sino_backrenaming:
joseph3d_fwd_tof_lmtojoseph3d_tof_lm_fwdrenaming:
joseph3d_back_tof_lmtojoseph3d_tof_lm_backIn the Python API to all TOF projectors, the order of the TOF related arguments has changed.
num_sigmasis now last and has a default value of3.0