Vectorized HttpCharacters (and used IndexOfAnyValues in other places found)

[gfoidl]

The title sounds like a complicated change, but thanks to dotnet/runtime#68328 it got trivial and is a way better alternative to doing it manually as in #44041.

Unfortunately the .NET daily builds are way behind at the moment, so these new APIs aren't available to run benchmarks in an easy way.

Edit: searched for other places where applicable as well. Some of the found were analyzers (NS2.0) or built for older targets where this API isn't available.

[ghost]

Thanks for your PR, @gfoidl. Someone from the team will get assigned to your PR shortly and we'll get it reviewed.

[gfoidl]

/cc: @MihaZupan -- it gets really nice that way 😄.

[MihaZupan]

Beautiful

[MihaZupan]

Unfortunately the .NET daily builds are way behind at the moment, so these new APIs aren't available to run benchmarks in an easy way.

Grabbing some numbers from dotnet/runtime#78863 as an example:

Chars

Method	Length	Mean	Error	StdDev
IndexOfAnyAsciiChar	100000	3.936 us	0.0525 us	0.0491 us
IndexOf	100000	4.925 us	0.0186 us	0.0174 us
IndexOfAny2Values	100000	5.754 us	0.0196 us	0.0174 us

Bytes

Method	Length	Mean	Error	StdDev
IndexOfAnyAsciiByte	100000	3.189 us	0.0171 us	0.0151 us
IndexOf	100000	1.971 us	0.0186 us	0.0174 us
IndexOfAny2Values	100000	2.233 us	0.0093 us	0.0078 us

[ghost]

@gfoidl, this change will be considered for inclusion in the blog post for the release it'll ship in. Nice work!

Please ensure that the original comment in this thread contains a clear explanation of what the change does, why it's important (what problem does it solve?), and, if relevant, include things like code samples and/or performance numbers.

This content may not be exactly what goes into the blog post, but it will help the team putting together the announcement.

Thanks!

[adityamandaleeka]

Nice! Another example of goodness coming together across multiple layers of the stack.

[MihaZupan]

cc: @stephentoub @danmoseley for this before and after

[gfoidl]

... and how it would look if vectorized w/o runtime (not showing the generator part).

[stephentoub]

thanks to dotnet/runtime#68328 it got trivial and is a way better alternative to doing it manually as in #44041.

Very glad to see it's panning out as planned. Thanks.

[JamesNK]

I believe verifying perf improvement is blocking this, which is itself blocked by build issues. I added blocked label.

[amcasey]

Very nice!

[adityamandaleeka]

/benchmark plaintext,json kestrel

[pr-benchmarks]

Crank Pull Request Bot

/benchmark <benchmark[,...]> <profile[,...]> <component,[...]> <arguments>

Benchmarks:

• plaintext: TechEmpower Plaintext Scenario - ASP.NET Platform implementation
• json: TechEmpower JSON Scenario - ASP.NET Platform implementation
• fortunes: TechEmpower Fortunes Scenario - ASP.NET Platform implementation
• yarp: YARP - http-http with 10 bytes
• mvcjsoninput2k: Sends 2Kb Json Body to an MVC controller

Profiles:

• aspnet-perf-lin: INTEL/Linux 12 Cores
• aspnet-perf-win: INTEL/Windows 12 Cores
• aspnet-citrine-lin: INTEL/Linux 28 Cores
• aspnet-citrine-win: INTEL/Windows 28 Cores
• aspnet-citrine-arm: ARM/Linux 32 Cores
• aspnet-citrine-amd: AMD/Linux 48 Cores

Components:

• kestrel
• mvc

Arguments: any additional arguments to pass through to crank, e.g. --variable name=value

[sebastienros]

/benchmark plaintext,json aspnet-citrine-lin kestrel

[pr-benchmarks]

Benchmark started for plaintext, json on aspnet-citrine-lin with kestrel. Logs: link

[pr-benchmarks]

plaintext - aspnet-citrine-lin

application	plaintext.base	plaintext.pr
CPU Usage (%)	99	99	0.00%
Cores usage (%)	2,784	2,773	-0.40%
Working Set (MB)	118	118	0.00%
Private Memory (MB)	642	643	+0.16%
Build Time (ms)	3,152	3,282	+4.12%
Start Time (ms)	217	235	+8.29%
Published Size (KB)	96,234	96,234	0.00%
.NET Core SDK Version	8.0.100-alpha.1.22607.4	8.0.100-alpha.1.22607.4

load	plaintext.base	plaintext.pr
CPU Usage (%)	91	92	+1.10%
Cores usage (%)	2,559	2,564	+0.20%
Working Set (MB)	119	119	0.00%
Private Memory (MB)	370	370	0.00%
Start Time (ms)	0	0
First Request (ms)	95	96	+1.05%
Requests/sec	10,972,281	10,830,568	-1.29%
Requests	165,670,123	163,541,936	-1.28%
Mean latency (ms)	1.23	1.17	-4.88%
Max latency (ms)	75.49	64.51	-14.54%
Bad responses	0	0
Socket errors	0	0
Read throughput (MB/s)	1,320.96	1,300.48	-1.55%
Latency 50th (ms)	0.76	0.77	+1.45%
Latency 75th (ms)	1.14	1.15	+0.88%
Latency 90th (ms)	1.76	1.75	-0.57%
Latency 99th (ms)	13.76	12.43	-9.67%

json - aspnet-citrine-lin

application	json.base	json.pr
CPU Usage (%)	99	99	0.00%
Cores usage (%)	2,777	2,782	+0.18%
Working Set (MB)	85	88	+3.53%
Private Memory (MB)	610	612	+0.33%
Build Time (ms)	3,146	3,167	+0.67%
Start Time (ms)	217	228	+5.07%
Published Size (KB)	96,234	96,234	0.00%
.NET Core SDK Version	8.0.100-alpha.1.22607.4	8.0.100-alpha.1.22607.4

load	json.base	json.pr
CPU Usage (%)	79	80	+1.27%
Cores usage (%)	2,221	2,231	+0.45%
Working Set (MB)	119	119	0.00%
Private Memory (MB)	363	363	0.00%
Start Time (ms)	0	0
First Request (ms)	99	99	0.00%
Requests/sec	1,207,837	1,213,658	+0.48%
Requests	18,238,050	18,325,094	+0.48%
Mean latency (ms)	0.78	0.74	-4.80%
Max latency (ms)	44.28	39.21	-11.45%
Bad responses	0	0
Socket errors	0	0
Read throughput (MB/s)	168.17	168.99	+0.49%
Latency 50th (ms)	0.38	0.38	-0.27%
Latency 75th (ms)	0.45	0.44	-0.67%
Latency 90th (ms)	0.60	0.58	-3.67%
Latency 99th (ms)	10.96	10.40	-5.11%

[adityamandaleeka]

Numbers look good. The RPS deltas appear within the usual margin of noise for these benchmarks.

[adityamandaleeka]

Thanks @gfoidl!