Many new contributions and features, the more important of them are described in my dissertation.
The baseline is: You can run really large scenarios with OpenMP + MPI while using most of the memory for your simulation data. The combination can be tweaked in various ways (basis transform operators, reduction algorithms, chunk sizes).
What's Changed
- Feature: Broadcast parameters by @freifrauvonbleifrei in #77
- Fix fill memset by @freifrauvonbleifrei in #79
- revert d86233b for combine_selalib_diagnostics by @freifrauvonbleifrei in #84
- Feature example worker by @freifrauvonbleifrei in #85
- Restructure DistributedSparseGridUniform by @freifrauvonbleifrei in #86
- update READMEs by @freifrauvonbleifrei in #78
- Feature: Never copy big tensors (unnecessarily) by @freifrauvonbleifrei in #83
- Feature restructure worker by @freifrauvonbleifrei in #87
- Feature: subspace reduce by @freifrauvonbleifrei in #82
- Feature: OpenMP support by @freifrauvonbleifrei in #88
- Have fewer MPI comms by @freifrauvonbleifrei in #90
- No volatile, but compile w/ precise float operations by @freifrauvonbleifrei in #91
- Chunked combine: no need to allocate all the sparse grid data at once by @freifrauvonbleifrei in #92
- CombiMinMaxScheme: add static load balancing by @freifrauvonbleifrei in #93
- Feature chunked readreduce by @freifrauvonbleifrei in #94
- Clear DSG after chunked reduce; fix out-of-bounds-writing error from extra DSG by @freifrauvonbleifrei in #95
- Uftp by @freifrauvonbleifrei in #98
- Fix allreduce error by @freifrauvonbleifrei in #97
- selalib: workers only by @freifrauvonbleifrei in #96
- Feature jenkins pipeline by @freifrauvonbleifrei in #100
- partitioned sparse grid I/O by @freifrauvonbleifrei in #99
- jenkins: use official spack repo (which now has discotec) by @freifrauvonbleifrei in #108
- Subspace vote for more groups by @freifrauvonbleifrei in #106
- README.md : add license badge by @freifrauvonbleifrei in #115
- Feature selalib nonlinear by @freifrauvonbleifrei in #113
- More expressive functions/function names in ProcessGroupWorker by @freifrauvonbleifrei in #107
- The statistics files are only written by the root MPI process, stolen from #114 by @freifrauvonbleifrei in #116
- Cmake optimizations by @freifrauvonbleifrei in #117
- three systems for file-based widely-distributed combination by @freifrauvonbleifrei in #118
- Compression w/ LZ4 and MPI-IO by @freifrauvonbleifrei in #119
- tidy DFG code structure and fix most warnings by @freifrauvonbleifrei in #120
- Update fault tolerance by @freifrauvonbleifrei in #123
- MPI-FT.h : update includes by @freifrauvonbleifrei in #124
- Fix missing boost include by @freifrauvonbleifrei in #125
- Jenkinsfile: update spack syntax by @freifrauvonbleifrei in #126
- Extend README by @freifrauvonbleifrei in #122
Full Changelog: 0.1.1...1.0.0
Contributors
freifrauvonbleifrei