Releases: LLNL/CallFlow
v.1.3.0
This release comprises of the following improvements.
- Decouple auxiliary computation architecture to a request-response architecture for individual pieces of data.
- Introduce a new view called the timeline view, that summarizes the performance of multiple executions across different modules.
- Improve module-callsite mapping strategies using regex.
- Comparison views for all the components.
- Improve gradient view computations.
v.1.2.0
This version improves the performance of the tool and improves processing speeds.
v1.1.2
CallFlow 1.1.2 brings improvements and new features, including
- Enable handling of .cali files.
- Automate the config file generation process
- Add dist/* to the repo
- Single server ecosystem
- Consistent processing into .callflow directory.
- Restores the projection view
- Creates an Argparser class replacing the ConfigFileGenerator.
v1.1.1
The release introduces the following features.
- Consistent license headers (see #15)
- Introduces Jupyter notebook integration (see #17)
- Adds documentation (see docs)
- Adds CI/CD for the project. (see #16, #19)
- Improve argument parsing (see #21, #22)
- Adds
manager.py
to control how notebooks launch the server and app. - Improves code quality/linting issues/minor interaction issues (see #18, #23, #24, #25, #26)
- Improves setup.py to fix installation (see PR #28)
- Improves splitting interaction UI (see PR #27)
- Combines HTML and CSS for Vue components (see #29)
- Fixes the npm production scripts (see PR #29)
v1.1.0
The release introduces the following features.
-
Introduces encodings for visualizing performance variation in multiple datasets.
- Ensemble Sankey view
- CCT views to explore the entire CCT.
- Comparing two datasets pair-wise
- Histograms, Scatterplots, and Boxplots for data analysis
- Module hierarchy view for exploring the module-level supergraph.
-
Exposes 2 visualization interfaces, one for single call graph, and an ensemble of call graphs along with a landing page for displaying basic information.
-
Introduces ensemble processing of multiple HPCToolkit, Caliper datasets.
-
CallFlow interface that exposes operations (e.g., creating, filtering, grouping, and union Hatchet's GraphFrames), data structures for handling multiple call graphs, and modules that interface the interactions performed from the UI).
-
Improved web UI for visualizing the performance data collected in the form of CallGraphs.
-
Migrates from the previous D3 javascript UI framework to a Vue based framework, where each view in the visualization is an independent component.
v1.1.0-beta
This release pertains to the code for the VIS-2020 submission.
v1.1.0-alpha
CallFlow v1.1.0 Alpha release.
This release translates the previous codebase from the node.js-D3 framework to the Python-Vue framework.