Repository navigation
[Perf] Regressions in System.Buffers.Tests.RentReturnArrayPoolTests<Object> #56555
Description
Run Information
| Architecture | arm64 |
|---|---|
| OS | Windows 10.0.19041 |
| Baseline | ec3f9fa03f72ceadfa0610196a12bab3b37b63f9 |
| Compare | 775ecaefe7aa3be85858e612fef782f0e1991f51 |
| Diff | Diff |
Regressions in System.Buffers.Tests.RentReturnArrayPoolTests<Object>
| Benchmark | Baseline | Test | Test/Base | Test Quality | Edge Detector | Baseline IR | Compare IR | IR Ratio | Baseline ETL | Compare ETL |
|---|---|---|---|---|---|---|---|---|---|---|
| MultipleSerial - Duration of single invocation | 814.29 ns | 881.75 ns | 1.08 | 0.07 | False |
Historical Data in Reporting System
_1.png)
Repro
git clone https://github.com/dotnet/performance.git
py .\performance\scripts\benchmarks_ci.py -f netcoreapp5.0 --filter 'System.Buffers.Tests.RentReturnArrayPoolTests<Object>*'Details
Payloads
Histogram
System.Buffers.Tests.RentReturnArrayPoolTests<Object>.MultipleSerial(RentalSize: 4096, ManipulateArray: False, Async: False, UseSharedPool: True)
Docs
Profiling workflow for dotnet/runtime repository
Benchmarking workflow for dotnet/runtime repository
Activity
[-][Perf] Changes at 7/20/2021 1:51:45 PM[/-][+][Perf] Regressions in System.Buffers.Tests.RentReturnArrayPoolTests<Object>[/+]DrewScoggins commentedon Jul 29, 2021
Seems related to #55959
stephentoub commentedon Jul 29, 2021
@DrewScoggins, how do I see other generic instantations of this same test for the same period? The test runs for both object and byte:
https://github.com/dotnet/performance/blob/d7dac8a7ca12a28d099192f8a901cf8e30361384/src/benchmarks/micro/libraries/System.Buffers/ArrayPoolTests.cs#L14-L15
DrewScoggins commentedon Jul 29, 2021
So the link is here, https://pvscmdupload.blob.core.windows.net/reports/allTestHistory/refs/heads/main_arm64_Windows%2010.0.19041/System.Buffers.Tests.RentReturnArrayPoolTests(Byte).MultipleSerial(RentalSize%3a%204096%2c%20ManipulateArray%3a%20False%2c%20Async%3a%20False%2c%20UseSharedPool%3a%20True).html
I got this by just clicking on the link in the table above and replacing the Object inside parens with Byte.
tannergooding commentedon Aug 9, 2021
@DrewScoggins, @stephentoub, was this addressed by #56316 ?
stephentoub commentedon Aug 9, 2021
Apparently partially but not fully:
#56316 (comment)
tannergooding commentedon Aug 9, 2021
A 2.56% regression (20ns) isn't bad. Do we think there are additional gains that can still be won back?
If so, I'll put this in future; otherwise, I think we should probably close the issue and accept the regression (given the wins we otherwise got).
stephentoub commentedon Aug 12, 2021
I think future (or closed) here is fine. It's a tiny hit in one use case on one architecture, and apparently only for
objectbut notbyte/char.