Skip to content
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

Enable RuntimeType.GenericCache to hold multiple types of cache entries #102034

Merged
merged 17 commits into from
May 31, 2024

Conversation

jkoritzinsky
Copy link
Member

Enable GenericCache to hold multiple types of cache entries in preparation for some boxing optimizations that can use the cache. Try to encapsulate the design so the users of the GenericCache don't need to think about multiple cache entries.

Copy link
Contributor

Tagging subscribers to this area: @dotnet/area-system-runtime
See info in area-owners.md if you want to be subscribed.

@jkoritzinsky

This comment was marked as outdated.

This comment was marked as outdated.

This comment was marked as outdated.

@jkoritzinsky

This comment was marked as outdated.

This comment was marked as outdated.

This comment was marked as outdated.

@jkoritzinsky
Copy link
Member Author

/benchmark micro aspnet-perf-win runtime --variable filter=System.Tests.Perf_Array

@jkoritzinsky
Copy link
Member Author

/benchmark micro aspnet-perf-win runtime --variable filter=System.Tests.Perf_Enum

@jkoritzinsky
Copy link
Member Author

/benchmark micro aspnet-perf-win runtime --variable filter=System.Reflection.Activator

Copy link

pr-benchmarks bot commented May 14, 2024

Benchmark started for micro on aspnet-perf-win with runtime and arguments --variable filter=*System.Reflection.Activator*. Logs: link

Copy link

pr-benchmarks bot commented May 14, 2024

micro - aspnet-perf-win

| benchmark                                    | mean (micro.base) | mean (micro.pr) | ratio | allocated (micro.base) | allocated (micro.pr) | ratio |
| -------------------------------------------- | ----------------- | --------------- | ----- | ---------------------- | -------------------- | ----- |
| Activator<EmptyClass>.CreateInstanceGeneric  |          7.302 ns |       70.649 ns |  9.67 |                   24 B |                 56 B |  2.33 |
| Activator<EmptyClass>.CreateInstanceNames    |            1.5 μs |          1.5 μs |  1.00 |                  920 B |                920 B |  1.00 |
| Activator<EmptyClass>.CreateInstanceType     |          6.434 ns |        8.138 ns |  1.26 |                   24 B |                 24 B |  1.00 |
| Activator<EmptyStruct>.CreateInstanceGeneric |         0.0001 ns |       0.0000 ns |  0.20 |                    0 B |                  0 B |       |
| Activator<EmptyStruct>.CreateInstanceNames   |            1.3 μs |          1.3 μs |  1.04 |                  744 B |                744 B |  1.00 |
| Activator<EmptyStruct>.CreateInstanceType    |          5.630 ns |        6.893 ns |  1.22 |                   24 B |                 24 B |  1.00 |

Copy link

pr-benchmarks bot commented May 14, 2024

Benchmark started for micro on aspnet-perf-win with runtime and arguments --variable filter=*System.Tests.Perf_Enum*. Logs: link

Return a ref into the boxed composite entry to make sure we're actually updating the composite entry.
@jkotas
Copy link
Member

jkotas commented May 22, 2024

I have run this micro-benchmark on Win x64 to see the perf of composite cache:

using System;
using System.Diagnostics;
using System.Runtime.CompilerServices;

var sw = new Stopwatch();
for (;;)
{
   sw.Restart();
   Work();
   Internal.Console.WriteLine(sw.ElapsedMilliseconds.ToString());
}

static void Work()
{
    for (int i = 0; i < 100000000; i++)
    {
        GC.KeepAlive(Activator.CreateInstance(typeof(object)));
        GC.KeepAlive(RuntimeHelpers.GetUninitializedObject(typeof(object)));
    }
}

It runs about 1.3x-1.4x slower compared to current main. These seems to be a lot of perf lost in the generics. Can we do better?

@jkoritzinsky
Copy link
Member Author

I'll see if I can make it faster.

I've been focusing on matching performance with our existing (single-entry-focused) microbenchmarks. I've been able to get within single-digit percentage (even with a possible improvement in one run) for Activator.CreateInstance. I'll spend some time on the composite case.

@jkotas
Copy link
Member

jkotas commented May 22, 2024

I have tried to replace the InlineArray with explicit fields jkotas@1a96a78 . It fixed the perf for my micro-benchmark. (The change is not complete - it is missing handling of EnumInfo.)

Copy link
Member

@jkotas jkotas left a comment

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

LGTM. Thanks!

@jkoritzinsky jkoritzinsky requested a review from jkotas May 30, 2024 22:32
@jkoritzinsky
Copy link
Member Author

@jkotas can you take one last review pass on this? Looks like Build Analysis is green so it’s ready to merge once it’s approved.

Copy link
Member

@jkotas jkotas left a comment

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

LGTM. Thank you!

@jkoritzinsky jkoritzinsky merged commit 9fca0c3 into dotnet:main May 31, 2024
146 checks passed
@jkoritzinsky jkoritzinsky deleted the generic-cache-composition branch May 31, 2024 05:23
@github-actions github-actions bot locked and limited conversation to collaborators Jun 30, 2024
Sign up for free to subscribe to this conversation on GitHub. Already have an account? Sign in.
Labels
area-System.Runtime runtime-coreclr specific to the CoreCLR runtime
Projects
None yet
Development

Successfully merging this pull request may close these issues.

2 participants