-
Notifications
You must be signed in to change notification settings - Fork 4.8k
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
Investigate Throughtput regression in Microsoft.CodeAnalysis #8706
Comments
With this change we create a better more accurate crossgen native image (when IBC data is present) I tried
|
The (small) increase is expected as this change fixes an issue where we were failing to compile some methods that the IBC data specified should be compiled into the native image. |
Additional methods that I see getting compiled are:
etc... |
@briansull is there any way to verify that the set of methods that are now prejitted is the set we expect? Assuming it is, there is no bug here -- the Roslyn assemblies we use in these tests have embedded IBC data, so we're seeing the impact of what is presumably a desirable IBC only fix. |
Yes, There is no bug here, we now generate extra methods with IBC data. With R2R It is not easy to verify that the exact set of methods we expect to get prejjited are in fact prejitted. Before R2R we were able to check that the JIT wasn't needed and and that every method needed was prejitted. We can't do that for R2R native images. There are quite a few random reasons where we simply omit a method from a R2R native image and fall back to using the Jit. |
Benchview link: https://benchview/compare?jobid=48095&resulttypeid=956&archid=4&testid=63143&configid=2706&machinepoolid=1292&comparejobids=[48074]&
b5c3e56 appears to have caused a 4% regression in throughput of crossgening Microsoft.CodeAnalysis. We see this regression in both x64 and x86 Windows.
The text was updated successfully, but these errors were encountered: