Problem
The dotnet-performance-patterns skill (PR #6) covers customer-actionable API usage patterns. But deeper JIT-aware optimization — understanding what the JIT will actually do with specific code shapes — requires runtime-internal knowledge that generic AI lacks.
What expert .NET knowledge would this encode?
- When the JIT will actually elide bounds checks vs. when it will not
- Which patterns trigger tier-1 vs. tier-2 compilation
- How
[MethodImpl(AggressiveInlining)] interacts with PGO
- When
stackalloc helps vs. hurts (stack depth, async contexts)
- The actual allocation profiles of LINQ operators and when
CollectionsMarshal tricks are safe
Relationship to existing skills
Skill concept
Deepen the existing dotnet-jit-optimization skill or create a companion that goes beyond "use the right API" into "write code shapes the JIT can optimize" — with version-specific JIT behavior knowledge.
Why customers would value this
For customers with true hot-path code (game engines, high-frequency trading, cloud infrastructure), the difference between "the JIT inlined this" and "the JIT did not" is the difference between shipping and not shipping. This knowledge currently requires reading JIT source code or asking the runtime team directly.
Problem
The
dotnet-performance-patternsskill (PR #6) covers customer-actionable API usage patterns. But deeper JIT-aware optimization — understanding what the JIT will actually do with specific code shapes — requires runtime-internal knowledge that generic AI lacks.What expert .NET knowledge would this encode?
[MethodImpl(AggressiveInlining)]interacts with PGOstackallochelps vs. hurts (stack depth, async contexts)CollectionsMarshaltricks are safeRelationship to existing skills
dotnet-performance-patterns(PR Add .NET performance analysis skill and optimizing agent #6): Customer-actionable API patterns — complements, does not overlapdotnet-jit-optimization: Already exists — evaluate whether it needs deepening with this contentSkill concept
Deepen the existing
dotnet-jit-optimizationskill or create a companion that goes beyond "use the right API" into "write code shapes the JIT can optimize" — with version-specific JIT behavior knowledge.Why customers would value this
For customers with true hot-path code (game engines, high-frequency trading, cloud infrastructure), the difference between "the JIT inlined this" and "the JIT did not" is the difference between shipping and not shipping. This knowledge currently requires reading JIT source code or asking the runtime team directly.