[martincostello]

Pull Request

The issue or feature being addressed

Improve the performance of the hot path when executing resilience pipelines.

Details on the issue fix or feature implementation

Profiling a pipeline of five strategies showed that a significant part of the time Polly itself spends per execution is generic virtual method (GVM) dispatch between the components of the pipeline chain (each hop costs roughly 7 ns and there were up to three per strategy), plus the circuit breaker taking its lock twice per execution.

• Avoid GVM dispatch between delegating components in the pipeline chain. DelegatingComponent exposes a non-virtual Execute() method that the composite root and the previous link in the chain call directly when the next component is a DelegatingComponent.
• Execute non-generic strategies directly rather than through their BridgeComponent, which is a trivial forwarder. DelegatingComponent unwraps the strategy at construction time.
• Skip the circuit breaker lock on pre-execution when the circuit is closed. A closed circuit cannot transition to another state in OnActionPreExecuteAsync(), so a volatile read is sufficient and a concurrent transition after the read is indistinguishable from one that happens right after the lock would have been released.

There are no public API changes and no changes to allocations.

Benchmarks

Benchmarks were run against main (3ab3ec0) in a separate worktree and against this branch, pinned to the performance cores (--affinity 4095) with two interleaved rounds per side. The figures below are the mean of the two rounds, and Ratio is branch/main.

Benchmark	Parameters	main (ns)	branch (ns)	Ratio	Allocated
PipelineBenchmark.ExecutePipeline_V8	Components=1	37.2	34.7	0.93	-
PipelineBenchmark.ExecutePipeline_V8	Components=2	50.8	46.8	0.92	-
PipelineBenchmark.ExecutePipeline_V8	Components=5	112.6	90.9	0.81	-
PipelineBenchmark.ExecutePipeline_V8	Components=10	236.4	200.3	0.85	-
MultipleStrategiesBenchmark.ExecuteStrategyPipeline_Generic_V8		619.6	569.7	0.92	40 B
MultipleStrategiesBenchmark.ExecuteStrategyPipeline_NonGeneric_V8		742.4	649.6	0.88	40 B
CircuitBreakerBenchmark.ExecuteCircuitBreaker_V8		130.3	115.8	0.89	-
TimeoutBenchmark.ExecuteTimeout_V8		126.2	121.0	0.96	-
RetryBenchmark.ExecuteRetry_V8		105.8	99.1	0.94	-

The retry strategy is generic and is not affected by these changes, so RetryBenchmark acts as a control for run-to-run noise on this machine (its two rounds were 110.0/101.5 ns on main and 99.1/99.2 ns on this branch). The 40 B allocated by MultipleStrategiesBenchmark is the ConcurrencyLease from System.Threading.RateLimiting and is unchanged.

BenchmarkDotNet v0.15.8, Windows 11 (10.0.26300.9711)
13th Gen Intel Core i7-13700H 2.90GHz, 1 CPU, 20 logical and 14 physical cores
.NET SDK 10.0.402
  [Host] : .NET 10.0.13 (10.0.13, 10.0.1326.50504), X64 RyuJIT x86-64-v3
Job=MediumRun  Toolchain=InProcessEmitToolchain  IterationCount=15  LaunchCount=2  WarmupCount=10

An attempt to skip the retry strategy's GetTimestamp() calls when neither telemetry nor OnRetry can observe the duration was measured at ~1.7 ns and discarded as noise.

Confirm the following

• I started this PR by branching from the head of the default branch
• I have targeted the PR to merge into the default branch
• I have included unit tests for the issue/feature
• I have successfully run a local build
• I have the experience and knowledge necessary to understand, review, and validate all content in this PR.
• This PR has content that I did not fully write myself.
  • I used AI to assist in writing or reviewing the content of this PR.

Note

This PR and its description were generated with Claude Code.

[codecov]

Codecov Report

✅ All modified and coverable lines are covered by tests.
✅ Project coverage is 98.03%. Comparing base (3ab3ec0) to head (4dd7b9b).
✅ All tests successful. No failed tests found.

Additional details and impacted files

@@           Coverage Diff           @@
##             main    #3287   +/-   ##
=======================================
  Coverage   98.02%   98.03%           
=======================================
  Files         321      321           
  Lines        7103     7131   +28     
  Branches     1089     1099   +10     
=======================================
+ Hits         6963     6991   +28     
  Misses         98       98           
  Partials       42       42           

Flag	Coverage Δ
linux	98.02% <100.00%> (+0.05%)	⬆️
macos	98.02% <100.00%> (+<0.01%)	⬆️
windows	98.03% <100.00%> (+<0.01%)	⬆️

Flags with carried forward coverage won't be shown. Click here to find out more.

☔ View full report in Codecov by Harness.
📢 Have feedback on the report? Share it here.