[efficiency-improver] perf: avoid redundant GetParameters() allocations in test-invocation hot path

[github-actions]

Summary

MethodInfo.GetParameters() is a CLR-guaranteed allocation: it always returns a freshly allocated ParameterInfo[] for safety (even for repeated calls on the same method). TestMethodInfo already caches this result in the ParameterTypes property (added in PR #9514), but GetInvokeResultAsync never sees that cache — it receives only the raw MethodInfo and therefore re-calls GetParameters() on every single test invocation.

For generic test methods the situation is worse: GetParameters() was called twice per invocation — once at the top of GetInvokeResultAsync and once inside ConstructGenericMethod.

Changes

Branch: efficiency/pass-cached-params-to-invoke (1 commit: 0bc0bb6)

src/Adapter/MSTestAdapter.PlatformServices/Extensions/MethodInfoExtensions.cs   | +37 / -12
src/Adapter/MSTestAdapter.PlatformServices/Execution/TestMethodInfo.Execution.cs|  +2 /  -2

MethodInfoExtensions.cs

• Added hot-path overload GetInvokeResultAsync(MethodInfo, object?, object?[]?, ParameterInfo[]) that accepts pre-cached parameters.
• Extracted private GetInvokeResultAsyncCore(...) so both overloads share one implementation (DRY, no behaviour change).
• ConstructGenericMethod now accepts ParameterInfo[] cachedParameters instead of calling methodInfo.GetParameters() internally — eliminates the double-allocation for generic tests.

TestMethodInfo.Execution.cs

• Both hot-path call sites (lines 147 and 164) updated to pass ParameterTypes (the already-cached property) to the new overload.

Energy Efficiency Evidence

Proxy metric: heap allocation / GC pressure (maps to energy via: fewer short-lived allocations → fewer GC collections → fewer CPU stalls due to GC → less energy drawn from DRAM and CPU during GC sweeps).

Scenario	Before	After
Plain test (non-generic, non-data-driven)	1 ParameterInfo[] alloc per call	0 allocations (served from cache)
Data-driven test, N rows, non-generic	N ParameterInfo[] allocs per run	0 allocations
Data-driven test, N rows, generic method	2N ParameterInfo[] allocs per run	0 allocations

For a typical parameterized suite with 1000 data rows, this removes ~1000+ unnecessary short-lived heap allocations per test run.

GSF Principle: Hardware Efficiency — reducing GC frequency makes better use of the CPU and DRAM that are physically powering the test host.

Relationship to existing work

• Complements PR perf: reduce allocations in data-driven test execution hot path #9617 (perf: reduce allocations in data-driven test execution hot path), which caches ReflectionTestMethodInfo and removes the TCS bridge. This PR targets a separate call site that perf: reduce allocations in data-driven test execution hot path #9617 does not cover (MethodInfoExtensions.GetInvokeResultAsync).
• Builds on PR [efficiency-improver] perf: cache MethodInfo.GetParameters() to avoid per-row array allocations in data-driven tests #9514 (merged 2026-06-30) which introduced TestMethodInfo.ParameterTypes; this PR threads that cache all the way to the invocation call.

Validation

The SDK required to build locally is .NET 11.0.100-preview (Arcade-provisioned via ./build.sh). CI will run the full test suite on PR open.

To review: main...efficiency/pass-cached-params-to-invoke

🤖 Automated content by GitHub Copilot. Posted via a maintainer's GitHub token, so it appears under their account — the account owner did not write or approve this content personally. Generated by the Efficiency Improver workflow. · 400.5 AIC · ⌖ 11.3 AIC · ⊞ 13.6K · [◷]( · ◷)

Add this agentic workflows to your repo

To install this agentic workflow, run

gh aw add githubnext/agentics/workflows/efficiency-improver.md@main