string.IndexOf(string) poor performance in other cultures

[bennil]

Description

After migrating an old .NET Framework 4.7.2 windows app to .NET 8.0 I was suprised by it's poor performance.
I traced it down to an 11x times slower string.IndexOf(string) call on a large string (500 kilobyte).

But this happens only if the current culture is set other than English/Invariant.

Benchmark running in German culture context

BenchmarkDotNet v0.13.11, Windows 11 (10.0.22621.2861/22H2/2022Update/SunValley2)
Intel Core i7-8650U CPU 1.90GHz (Kaby Lake R), 1 CPU, 8 logical and 4 physical cores
  [Host]               : .NET Framework 4.8.1 (4.8.9181.0), X64 RyuJIT VectorSize=256
  .NET 5.0             : .NET 5.0.7 (5.0.721.25508), X64 RyuJIT AVX2
  .NET 8.0             : .NET 8.0.0 (8.0.23.53103), X64 RyuJIT AVX2
  .NET Framework 4.7.2 : .NET Framework 4.8.1 (4.8.9181.0), X64 RyuJIT VectorSize=256

Method	Job	Runtime	Mean	Error	StdDev	Ratio	RatioSD
IndexOfOfNotContainingStringInLargeString	.NET 5.0	.NET 5.0	20.210 ms	0.3075 ms	0.2877 ms	12.79	0.16
IndexOfOfNotContainingStringInLargeString	.NET 8.0	.NET 8.0	17.440 ms	0.2713 ms	0.2538 ms	11.04	0.15
IndexOfOfNotContainingStringInLargeString	.NET Framework 4.7.2	.NET Framework 4.7.2	1.578 ms	0.0098 ms	0.0087 ms	1.00	0.00

Benchmark running in US culture context (perf as expected)

Method	Job	Runtime	Mean	Error	StdDev	Ratio	RatioSD
IndexOfOfNotContainingStringInLargeString	.NET 5.0	.NET 5.0	1,133.2 μs	7.46 μs	6.98 μs	0.71	0.02
IndexOfOfNotContainingStringInLargeString	.NET 8.0	.NET 8.0	789.0 μs	3.95 μs	3.30 μs	0.50	0.01
IndexOfOfNotContainingStringInLargeString	.NET Framework 4.7.2	.NET Framework 4.7.2	1,587.1 μs	31.59 μs	31.03 μs	1.00	0.00

Reproduce

BenchmarkDotNet StringIndexOfFrameworkPerfTest repo

You can easily switch the culture context in the BenchmarkDotNet setup method.

[ghost]

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

Issue Details

---

Description

After migrating an old .NET Framework 4.7.2 windows app to .NET 8.0 I was suprised by it's poor performance.
I traced it down to an 11x times slower string.IndexOf(string) call on a large string (500 kilobyte).

But this happens only if the current culture is set other than English/Invariant.

Benchmark running in German culture context

BenchmarkDotNet v0.13.11, Windows 11 (10.0.22621.2861/22H2/2022Update/SunValley2)
Intel Core i7-8650U CPU 1.90GHz (Kaby Lake R), 1 CPU, 8 logical and 4 physical cores
  [Host]               : .NET Framework 4.8.1 (4.8.9181.0), X64 RyuJIT VectorSize=256
  .NET 5.0             : .NET 5.0.7 (5.0.721.25508), X64 RyuJIT AVX2
  .NET 8.0             : .NET 8.0.0 (8.0.23.53103), X64 RyuJIT AVX2
  .NET Framework 4.7.2 : .NET Framework 4.8.1 (4.8.9181.0), X64 RyuJIT VectorSize=256

Method	Job	Runtime	Mean	Error	StdDev	Ratio	RatioSD
IndexOfOfNotContainingStringInLargeString	.NET 5.0	.NET 5.0	20.210 ms	0.3075 ms	0.2877 ms	12.79	0.16
IndexOfOfNotContainingStringInLargeString	.NET 8.0	.NET 8.0	17.440 ms	0.2713 ms	0.2538 ms	11.04	0.15
IndexOfOfNotContainingStringInLargeString	.NET Framework 4.7.2	.NET Framework 4.7.2	1.578 ms	0.0098 ms	0.0087 ms	1.00	0.00

Benchmark running in US culture context (perf as expected)

Method	Job	Runtime	Mean	Error	StdDev	Ratio	RatioSD
IndexOfOfNotContainingStringInLargeString	.NET 5.0	.NET 5.0	1,133.2 μs	7.46 μs	6.98 μs	0.71	0.02
IndexOfOfNotContainingStringInLargeString	.NET 8.0	.NET 8.0	789.0 μs	3.95 μs	3.30 μs	0.50	0.01
IndexOfOfNotContainingStringInLargeString	.NET Framework 4.7.2	.NET Framework 4.7.2	1,587.1 μs	31.59 μs	31.03 μs	1.00	0.00

Reproduce

BenchmarkDotNet StringIndexOfFrameworkPerfTest repo

You can easily switch the culture context in the BenchmarkDotNet setup method.

Author:	bennil
Assignees:	-
Labels:	area-System.Globalization, tenet-performance, untriaged
Milestone:	-

[tarekgh]

This is tracked by #40942