[Experimental] Parallelize Graphs on multiple numa nodes
Open
research
- Dominant language
- C++
- Stars
- 54
- Forks
- 16
- PR merge metrics
- No merged PRs in 30d
Description
- [x] Added transform pass for splitting graphs
- [x] Add split op in tensor dialect
- [x] alloc test buffers on asigned numa node via python scripts
- [x] update deep tiled matmul loop generation to adapt to numa control
- [ ] bind deep tiled matmul to the expected numa node
- [ ] bufferize tensor.concat / tensor.split
Contributor guide
Assessment
This issue has not been assessed yet.