Parla.py v0.2.0

Release for SC22 Artifact submission.

What's Changed since 0.0.0

• Adding simple kokkos example by @dialecticDolt in #1
• Working virtual execution environments by @insertinterestingnamehere in #4
• Start running threads after the whole threadpool got initialized by @bozhiyou in #5
• Merge auto data move by @zhangsiyu1103 in #6
• Clean up changes for remove clone/clone_here from inner/blocked_cholesky by @bozhiyou in #7
• Add initialize_thread to component instances. by @arthurp in #20
• Blocked cyclic based multi-GPU blocked cholesky + clone_here bug fix by @nicelhc13 in #21
• Added number of iterations for tests on the multi-GPU Cholesky by @nicelhc13 in #27
• Fix data movement calls for partitioners by @insertinterestingnamehere in #33
• Create a new stream for each task. by @arthurp in #32
• Automatic data movement based on partitioners w/ moving semantics by @bozhiyou in #22
• Automate data movement in QR benchmark by @bozhiyou in #35
• Fix our asarray workaround. by @insertinterestingnamehere in #41
• Initial working version of persistent data reservation. by @insertinterestingnamehere in #40
• Fix Persistent Memory Reservation by @insertinterestingnamehere in #48
• Change CPU architecture to have one device. by @arthurp in #45
• Fix parla cpu/multiload/numpy dependency cycle. by @insertinterestingnamehere in #58
• Add tutorial section for data partitioning/movement by @bozhiyou in #86
• Tutorial 2 update by @ag1548 in #92
• Tutorial 0 update by @ag1548 in #91
• Tutorial 1 update by @ag1548 in #90
• Tutorial by @sestephens73 in #94
• feat: single device PArray with automatic datamovement by @yinengy in #96
• fix: hide cupy in system without GPU by @yinengy in #97
• fix: update setup.py for submodule and also requirements.txt by @yinengy in #101
• feat: support coherence protocol on PArray for multi devices by @yinengy in #100
• fix: add locks to prevent race in updating coherence protocol by @yinengy in #102
• Merge dev-scheduler by @nicelhc13 in #103
• fix: remove the use of gpu module by @yinengy in #104
• Dev fix by @nicelhc13 in #105
• Refactoring task classes by @nicelhc13 in #107
• Runahead scheduling by @nicelhc13 in #108
• Mapper memory tracking and policy updates by @sestephens73 in #110
• Fix data race among multiple readers by @nicelhc13 in #111
• Fix parray tracking, removing over-deallocation bug by @sestephens73 in #113
• Attempt to fix intialization crash by making initialization seqential by @dialecticDolt in #114
• Mapper dependency load balance by @sestephens73 in #115
• Handle task colocations through two queues by @nicelhc13 in #117
• Memory tracking fixes by @sestephens73 in #120
• Fix counter + free worker thread bugs by @nicelhc13 in #121
• New Parray: support fine-grained data movement by @yinengy in #119
• Syncing dev with main by @sestephens73 in #122
• Miscellaneous refactoring and robustness enhancement by @bozhiyou in #112
• Miscellaneous refactoring, continued by @sestephens73 in #124
• Callback dev wlr deadlock by @dialecticDolt in #129
• Add Parla-corresponding Dask Cholesky on GPU and CPU by @nicelhc13 in #127
• Add cholesky implementations including the Dask version, and fix a data movement task related bug. by @nicelhc13 in #130

New Contributors

• @insertinterestingnamehere made their first contribution in #4
• @bozhiyou made their first contribution in #5
• @zhangsiyu1103 made their first contribution in #6
• @arthurp made their first contribution in #20
• @ag1548 made their first contribution in #92
• @yinengy made their first contribution in #96

Full Changelog: https://github.com/ut-parla/Parla.py/commits/v0.2.0