Tagging subscribers to this area: @JulieLeeMSFT, @jakobbotsch
See info in area-owners.md if you want to be subscribed.

IEnumerable<int> _enum1 = new int[10];
IEnumerable<int> _enum2 = new int[10];

[Benchmark]
public bool SequenceEqual() => _enum1.SequenceEqual(_enum2);

[Benchmark]
public int ElementAt() => _enum1.ElementAt(5);

[Benchmark]
public int Count() => _enum1.Count();

// Promote IsList to Tier1 with PGO
for (int i = 0; i < 200; i++)
{
    IsList(new int[10]); // source is always int[]
    Thread.Sleep(10);
}

[MethodImpl(MethodImplOptions.NoInlining)]
bool IsList<T>(IEnumerable<T> source) => source is IList<T>;

; Assembly listing for method IsList
G_M12490_IG01:
       sub      rsp, 40
G_M12490_IG02:
+      mov      rax, rcx
+      test     rax, rax                 ; if (source == null)
+      je       SHORT G_M12490_IG05
+G_M12490_IG03:
+      mov      rdx, 0x7FF941DE9750
+      cmp      qword ptr [rax], rdx
+      je       SHORT G_M12490_IG05      ; if (source is int[])
+G_M12490_IG04:
       mov      rdx, rcx
       mov      rcx, 0x7FF942439A60
       call     CORINFO_HELP_ISINSTANCEOFINTERFACE ; slow fallback
+G_M12490_IG05:
       test     rax, rax
       setne    al
       movzx    rax, al
-G_M12490_IG03:
+G_M12490_IG06:
       add      rsp, 40
       ret      
; Total bytes of code 59

/azp run runtime-coreclr pgo, runtime-coreclr libraries-pgo, runtime-coreclr pgostress

Thanks. If this is isinst/castclass only, I assume it doesn't cover a pattern like this?

Thanks. If this is isinst/castclass only, I assume it doesn't cover a pattern like this?

runtime/src/libraries/System.Linq/src/System/Linq/Enumerable.cs

Lines 39 to 59 in 21b4a85

	// Use `GetType() == typeof(...)` rather than `is` to avoid cast helpers. This is measurably cheaper
	// but does mean we could end up missing some rare cases where we could get a span but don't (e.g. a uint[]
	// masquerading as an int[]). That's an acceptable tradeoff. The Unsafe usage is only after we've
	// validated the exact type; this could be changed to a cast in the future if the JIT starts to recognize it.
	// We only pay the comparison/branching costs here for super common types we expect to be used frequently
	// with LINQ methods.
	bool result = true;
	if (source.GetType() == typeof(TSource[]))
	{
	span = Unsafe.As<TSource[]>(source);
	}
	else if (source.GetType() == typeof(List<TSource>))
	{
	span = CollectionsMarshal.AsSpan(Unsafe.As<List<TSource>>(source));
	}
	else
	{
	span = default;
	result = false;
	}

(or is that covered separately?)

This code looks like it's already devirtualized by hands so there is nothing JIT can do additionally here. Although, you can convert it back to is TSource[] arr if you want (but it might regress NAOT that typically doesn't have any PGO data and Mono).

One benefit of is TSource[] arr would be that uint[]/int[] case 🙂

you can convert it back to is TSource[] arr if you want
One benefit of is TSource[] arr would be that uint[]/int[] case 🙂

It was written this way to explicitly avoid the variance costs

PGO pipelines fail with #96612 (unrelated issue)

As a follow up you might look at whether cloning can profitably key off of one of these casts. Right now type test cloning requires OMF_HAS_GUARDEDDEVIRT.

As a follow up you might look at whether cloning can profitably key off of one of these casts. Right now type test cloning requires OMF_HAS_GUARDEDDEVIRT.

Ah good point, will file an issue for it!

Improvements:

• [Perf] Linux/arm64: 67 Improvements on 1/9/2024 2:49:36 AM perf-autofiling-issues#27131
• [Perf] Windows/arm64: 54 Improvements on 1/9/2024 2:49:36 AM perf-autofiling-issues#27117
• [Perf] Windows/arm64: 6 Improvements on 1/8/2024 10:15:44 PM perf-autofiling-issues#27116
• [Perf] Linux/x64: 60 Improvements on 1/9/2024 2:49:36 AM perf-autofiling-issues#27274
• [Perf] Linux/arm64: 66 Improvements on 1/9/2024 2:49:36 AM perf-autofiling-issues#28336
• [Perf] Linux/arm64: 66 Improvements on 1/9/2024 2:49:36 AM perf-autofiling-issues#28781