BruceForstall
commented
4 years ago Including the issues in https://github.com/dotnet/coreclr, this is the current state of CodeGen issues:
| Row Labels | Count of Category |
|---|---|
| cq | 462 |
| correctness | 109 |
| implementation | 104 |
| testing | 35 |
| throughput | 35 |
| design | 17 |
| proposal | 16 |
| documentation | 8 |
| eng-sys | 8 |
| reach | 5 |
| security | 1 |
| Grand Total | 800 |
| Row Labels | Count of Theme |
|---|---|
| basic-cq | 96 |
| optimization | 66 |
| vector-codegen | 48 |
| register-allocator | 48 |
| intrinsics | 45 |
| jit-coding-style | 40 |
| inlining | 39 |
| ir | 38 |
| testing | 34 |
| structs | 28 |
| tail-call | 26 |
| loop-opt | 23 |
| benchmarks | 21 |
| emitter | 15 |
| bounds-checks | 13 |
| gc-stress | 10 |
| helpers | 9 |
| runtime | 9 |
| calling-convention | 9 |
| importer | 9 |
| assertion-prop | 9 |
| debug-dumps | 9 |
| super-pmi | 8 |
| value-numbering | 8 |
| throughput | 8 |
| barriers | 8 |
| gc-info | 8 |
| range-check | 8 |
| floating-point | 7 |
| cse | 6 |
| flowgraph | 6 |
| block-opts | 6 |
| msil | 5 |
| devirtualization | 5 |
| alignment | 4 |
| big-bets | 4 |
| pinning | 4 |
| ref-counts | 4 |
| stack-allocation | 3 |
| minopts | 3 |
| copy-prop | 3 |
| profile-feedback | 3 |
| prolog-epilog | 3 |
| large-methods | 3 |
| lower | 3 |
| md-arrays | 2 |
| morph | 2 |
| volatile | 2 |
| eh | 2 |
| type-intrinsics | 2 |
| liveness | 2 |
| interpreter | 2 |
| div-mod-rem | 2 |
| zero-init | 2 |
| debug-info | 2 |
| pinvoke | 2 |
| invalid-il | 2 |
| varargs | 2 |
| delegates | 1 |
| memory-usage | 1 |
| jit-ee-interface | 1 |
| block-layout | 1 |
| addressing-modes | 1 |
| elt-hooks | 1 |
| ready-to-run | 1 |
| generics | 1 |
| tiering | 1 |
| ngen | 1 |
| Grand Total | 800 |
| Row Labels | Count of SkillLevel |
|---|---|
| expert | 368 |
| intermediate | 340 |
| beginner | 92 |
| Grand Total | 800 |
| Row Labels | Count of Cost |
|---|---|
| medium | 415 |
| small | 230 |
| large | 112 |
| extra-large | 43 |
| Grand Total | 800 |
There is certainly room for improvement in classification, e.g.,
- fixing the classification of individual issues
- refining categories and themes
- adding some notion of priority
CarolEidt
JulieLeeMSFT
The CodeGen team attempts to classify GitHub issues with label
area-CodeGen-coreclrwith additional classifications beyond the GitHub labels. Each issue is given a "category", "theme", "skill level", "cost", and optionally "impact". These are added as a block of text at the bottom of the main issue description, e.g.:We have a tool that allows us to extract these to Excel, modify and analyze them there, and update the GitHub issues that have been modified. We update periodically, but not necessarily regularly, to classify new issues.
The goal is to be able to get a better idea of what kinds of issues we have outstanding, and think about ways to address them.
We chose to use this mechanism instead of creating new GitHub labels, as we didn't want to bloat the GitHub label set in a way that would overwhelm and negatively affect other teams in the repo. You can still query for issues using specific classification tags.
As of now, the following are the used classifications:
Details of the breakdown of these classifications follows in subsequent comments to this issue.
@dotnet/jit-contrib