Here we are discussing cM extensions and possible collaborative projects/internships
Contents
Adding past research techniques, tools and data to cM
- cM microkernel/codelet/benchmark and dataset creator
- cM reproducible execution
- cM reproducible experimental pipeline
- Program decomposition into codelets
- Add more tools for auto-tuning (source-to-source polyhedral tools, HMPP, etc)
- Add/generate more program/architecture features using Alchemist
Validating and extending past research techniques
- Implement various focused search techniques
- Improve optimization predictions
- Find representative benchmarks and datasets
- Characterization of program and architecture behavior by breaking code semantics
- Add universal modeling of program behavior (performance, power, CPI, etc)
- Run-time adaptation (scheduling) through static multi-versioning and dynamic reaction to changes in program/system behavior and dataset parameters
- Add agent-based whole system optimization and adaptation)
Improving cM
- Improve data visualization using JavaScript to avoid page scrolling or showing all unnecessary info
- Support python3
- Create Linux package (Debian, etc)
- Add more function and uni tests
- Improve cM bootstraping and kernel (remove data search duplications)
- Improve/speed up searching in repository
- Improve visualization of tables with experimental results (add universal multi-column sorting, extend export function)
- Improve handling of dependencies between data and modules (speed it up) including in cM web front-end (add AJAX or Java Script)
- Improve P2P data/module exchange
- Improve data browsing and online pruning by classes
- Speed up visualization of graphs and tables for aggregated experiments
- Perform online data pruning (using AJAX) for aggregated experiments
- Improve registration process and password reminders
Setting up new publication model
- Prepare new ADAPT workshop/conference/journal with validation of experiments (for HiPEAC or CGO, etc) - collaborators are very welcome!