-
Notifications
You must be signed in to change notification settings - Fork 4.7k
New issue
Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.
By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.
Already on GitHub? Sign in to your account
System.Numerics.Tests.Perf_Matrix4x4.MultiplyByScalarOperatorBenchmark has regressed #65191
Comments
Tagging subscribers to this area: @dotnet/area-system-numerics Issue Details
Repro: git clone https://github.com/dotnet/performance.git
py .\performance\scripts\benchmarks_ci.py -f net6.0 net7.0 --filter System.Numerics.Tests.Perf_Matrix4x4.MultiplyByScalarOperatorBenchmark It seems that @performanceautofiler has missed the regression (cc @DrewScoggins ): https://pvscmdupload.blob.core.windows.net/reports/allTestHistory%2frefs%2fheads%2fmain_x64_Windows%2010.0.18362%2fSystem.Numerics.Tests.Perf_Matrix4x4.MultiplyByScalarOperatorBenchmark.html System.Numerics.Tests.Perf_Matrix4x4.MultiplyByScalarOperatorBenchmark
|
Same goes for
|
And |
There are more, I won't list them all here. |
I'll have to look at the codegen for these. They haven't been changed in a long time and so I expect this is something that's changed in CSE or similar. Notably, they could/should likely be simplified to use the xplat helper intrinsics now, rather than individual AdvSimd/Sse code paths |
Some benchmarks discussed here appeared in the 6.0 vs 7.0-rc2 perf report. Are they still under investigation? The historical data only goes back until october 2021 so we are unable to determine if this is noise nor to see what might've caused it.
System.Numerics.Tests.Perf_Matrix4x4.MultiplyByScalarOperatorBenchmark
System.Numerics.Tests.Perf_Matrix4x4.NegateBenchmark
System.Numerics.Tests.Perf_Matrix4x4.NegationOperatorBenchmark
System.Numerics.Tests.Perf_Matrix4x4.SubtractOperatorBenchmark
System.Numerics.Tests.Perf_Matrix4x4.Transpose
|
The .NET 7 GA report is also showing potential regression in the following. I ran into the same problem where the historical data does not go back far enough to help triage this. @DrewScoggins is this something we can change? System.Numerics.Tests.Perf_Matrix4x4.AddOperatorBenchmark
|
We should figure out something that we can do. Naively, we could just extend the window of that history page from 12 to 18 months. However, that would not be feasible with the way those pages are currently generated, they would take too long to retrieve all of the data. For these individual tests that we need more data for we can go ahead and generate longer trends, and just do them on a case by case basis, just get me a list of the tests that we need this for. However, I think the right way to fix this going forward is to update the generation method for the data on these pages so that we are not redoing work over and over (Currently, we regenerate the data every day, even though all we need to do is drop the earliest and grab the latest) |
This was resolved with my rewrite of Matrix4x4 so be able to utilize promotion, field enregistration, and general SIMD/vectorization |
System.Numerics.Tests.Perf_Matrix4x4.MultiplyByScalarOperatorBenchmark
has regressed, it might be related to #59415Repro:
It seems that @performanceautofiler has missed the regression (cc @DrewScoggins ): https://pvscmdupload.blob.core.windows.net/reports/allTestHistory%2frefs%2fheads%2fmain_x64_Windows%2010.0.18362%2fSystem.Numerics.Tests.Perf_Matrix4x4.MultiplyByScalarOperatorBenchmark.html
System.Numerics.Tests.Perf_Matrix4x4.MultiplyByScalarOperatorBenchmark
cc @tannergooding
The text was updated successfully, but these errors were encountered: