Community
Participate
Working Groups
For large model runs, we're running into memory and performance issues when writing to adata. We need to investigate: 1. Whether to use another EObjImpl such as MinimalEObject. 2. Using binary resource factory (a good idea anyway). 3. Changes in actual structure. 4. Whether something in XText / APar is contributing to the issue. We'll need to run profiling on a session to figure that out.
We changed to Binary Resource Factory. Vastly improved performance and memory footprint. Downside is that artifacts are no longer human readable and failures are less recoverable, but worth trade off in this case. NOTE: This will break compatibility with existing adata files. Let me know if this is a problem for anyone. Eventually we probably should make Binary and Text/XML preference options with ability to open either, but that's a good bit of engineering.
