| Name | Modified | Size | Downloads / Week |
|---|---|---|---|
| Parent folder | |||
| README.md | 2026-07-24 | 776 Bytes | |
| v0.21.0 source code.tar.gz | 2026-07-24 | 856.7 kB | |
| v0.21.0 source code.zip | 2026-07-24 | 970.4 kB | |
| Totals: 3 Items | 1.8 MB | 0 | |
Dagger v0.21.0
Improvements
- Add MPI-based distributed-memory acceleration, including SPMD execution support in DataDeps.
- Add deterministic MPI processor grids and rank-uniform task placement, extensible to other SPMD backends.
- Add support for MPI execution with CUDA, ROCm, oneAPI, Metal, and OpenCL backends. Extensible to other backends.
- Add Acceleration selection for multiplexing between communication and execution backends.
- Add MPI testing infrastructure, CI coverage, documentation, and benchmark utilities.
Breaking changes
None. Yay!!!!!!
Merged pull requests:
- MPI: Final merge sprint of the whole MPI infrastructure (#720) (@fda-tome)