[perf-improver] perf: add GetLongestRunningTask to avoid list allocation in SimpleTerminal

[Evangelink]

🤖 This is an automated contribution from Perf Improver.

Goal and Rationale

SimpleTerminalBase.RenderProgress (the non-ANSI / redirected-output terminal path) needs the single longest-running active test for each progress item. The current call:

TestDetailState? activeTest = p.TestNodeResultsState?.GetRunningTasks(1).FirstOrDefault();

calls GetRunningTasks(1) which:

1. Allocates a new List<TestDetailState> with capacity _testNodeProgressStates.Count
2. Iterates the entire ConcurrentDictionary to populate the list
3. Sorts the list descending by elapsed time (O(n log n))
4. Returns the list — then FirstOrDefault() takes only the first element

The list is immediately eligible for GC. This is 1 wasted List<TestDetailState> allocation per progress item per render tick, plus an unnecessary full sort.

Approach

Add GetLongestRunningTask() to TestNodeResultsState: a single O(n) linear scan that tracks the maximum-elapsed entry directly, returning the TestDetailState? with no heap allocation.

Update SimpleTerminalBase.RenderProgress to call it instead.

The existing GetRunningTasks(int maxCount) is unchanged — it's still needed by AnsiTerminalTestProgressFrame.GenerateLinesToRender which requests multiple items.

Performance Evidence

Metric	Before	After
List<TestDetailState> allocs per progress item per render tick (SimpleTerminal)	1	0
Algorithm for finding the max	O(n log n) sort + take first	O(n) linear scan

For a 5-minute run with N=4 assemblies at ~5 fps:

• ~6,000 List<TestDetailState> allocations eliminated
• Sort overhead replaced by a single pass

Methodology: code inspection + allocation analysis. The SimpleTerminal path is used when output is redirected (CI pipelines, log capture) or when ANSI is not available.

Trade-offs

• None. GetLongestRunningTask() is a strict simplification: the "sort then take first" pattern is overkill for a single-element request.
• The new method's semantics match the existing call exactly: when the dictionary is empty it returns null (same as FirstOrDefault() on an empty list).

Test Status

• Microsoft.Testing.Platform.UnitTests (net8.0): 1103 passed, 0 failed, 3 skipped ✅
• Build (all TFMs): 0 warnings, 0 errors ✅

Reproducibility

./build.sh -build -c Debug
artifacts/bin/Microsoft.Testing.Platform.UnitTests/Debug/net8.0/Microsoft.Testing.Platform.UnitTests

Generated by Perf Improver

Generated by Perf Improver · sonnet46 4M · ◷

Add this agentic workflows to your repo

To install this agentic workflow, run

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

[Copilot]

Pull request overview

This PR targets Microsoft.Testing.Platform terminal progress rendering, optimizing the non-ANSI/redirected-output path by avoiding an allocation + full sort when only a single “most active” running test is needed.

Changes:

• Added TestNodeResultsState.GetLongestRunningTask() to find the max elapsed running test via a single linear scan (no list allocation).
• Updated SimpleTerminal.RenderProgress to use GetLongestRunningTask() instead of GetRunningTasks(1).FirstOrDefault().

Show a summary per file

File	Description
src/Platform/Microsoft.Testing.Platform/OutputDevice/Terminal/TestNodeResultsState.cs	Adds a new O(n) API to retrieve the longest-running active test without allocating/sorting a list.
src/Platform/Microsoft.Testing.Platform/OutputDevice/Terminal/SimpleTerminalBase.cs	Switches SimpleTerminal progress rendering to the new API to remove per-tick allocations in redirected/non-ANSI mode.

Copilot's findings

• Files reviewed: 2/2 changed files
• Comments generated: 1