Vectorize BMP string

[vcsjones]

As a follow up to #131109, this introduces vectorization for the BMPEncoding.

BMPEncoding is big-endian UCS-2 encoding. It is a subset of UTF-16-BE, with the exception that it is truly fixed width and only supports plane 0 without the surrogate pairs.

Performance is largely improved across the board. Small inputs, like country codes, regress ~0.5ns simply because the method body itself has slightly more register pressure. Other inputs like common names, etc. see notable performance improvements.

This also incorporates some feedback about making the SIMD more JIT-friendly into the existing IA5 and VisibleString encoding.

Performance Benchmarks

Source: https://gist.github.com/vcsjones/4b91a33bcae337333cb26d46fd087bf9

BenchmarkDotNet v0.15.8, macOS Tahoe 26.5.2 (25F84) [Darwin 25.5.0]
Apple M1 Ultra, 1 CPU, 20 logical and 20 physical cores
  [Host] : .NET 11.0.0 (11.0.0-dev, 42.42.42.42424), Arm64 RyuJIT armv8.0-a
  Main   : .NET 11.0.0 (11.0.0-dev, 42.42.42.42424), Arm64 RyuJIT armv8.0-a
  Vector : .NET 11.0.0 (11.0.0-dev, 42.42.42.42424), Arm64 RyuJIT armv8.0-a

Runtime=.NET 10.0  IterationCount=15  WarmupCount=5  

Method	Job	Toolchain	Alignment	Input	Mean	Error	StdDev	Ratio	RatioSD	Gen0	Gen1	Allocated	Alloc Ratio
Decode	Main	Main	Aligned	CountryCode	14.27 ns	0.037 ns	0.031 ns	1.00	0.00	0.0051	-	32 B	1.00
Decode	Vector	Vector	Aligned	CountryCode	14.97 ns	0.031 ns	0.029 ns	1.05	0.00	0.0051	-	32 B	1.00
TryDecode	Main	Main	Aligned	CountryCode	11.34 ns	0.012 ns	0.011 ns	1.00	0.00	-	-	-	NA
TryDecode	Vector	Vector	Aligned	CountryCode	11.74 ns	0.007 ns	0.006 ns	1.04	0.00	-	-	-	NA
Write	Main	Main	Aligned	CountryCode	16.13 ns	0.043 ns	0.040 ns	1.00	0.00	0.0127	-	80 B	1.00
Write	Vector	Vector	Aligned	CountryCode	16.79 ns	0.044 ns	0.039 ns	1.04	0.00	0.0127	-	80 B	1.00
Decode	Main	Main	Aligned	CommonName	56.85 ns	0.073 ns	0.068 ns	1.00	0.00	0.0127	-	80 B	1.00
Decode	Vector	Vector	Aligned	CommonName	26.71 ns	0.040 ns	0.035 ns	0.47	0.00	0.0127	-	80 B	1.00
TryDecode	Main	Main	Aligned	CommonName	52.77 ns	0.033 ns	0.029 ns	1.00	0.00	-	-	-	NA
TryDecode	Vector	Vector	Aligned	CommonName	21.71 ns	0.038 ns	0.034 ns	0.41	0.00	-	-	-	NA
Write	Main	Main	Aligned	CommonName	43.31 ns	0.067 ns	0.060 ns	1.00	0.00	0.0216	-	136 B	1.00
Write	Vector	Vector	Aligned	CommonName	26.72 ns	0.394 ns	0.349 ns	0.62	0.01	0.0217	-	136 B	1.00
Decode	Main	Main	Aligned	UnicodeName	28.28 ns	0.078 ns	0.069 ns	1.00	0.00	0.0076	-	48 B	1.00
Decode	Vector	Vector	Aligned	UnicodeName	19.37 ns	0.037 ns	0.034 ns	0.68	0.00	0.0076	-	48 B	1.00
TryDecode	Main	Main	Aligned	UnicodeName	24.90 ns	0.039 ns	0.032 ns	1.00	0.00	-	-	-	NA
TryDecode	Vector	Vector	Aligned	UnicodeName	15.75 ns	0.024 ns	0.020 ns	0.63	0.00	-	-	-	NA
Write	Main	Main	Aligned	UnicodeName	25.70 ns	0.397 ns	0.371 ns	1.00	0.02	0.0166	-	104 B	1.00
Write	Vector	Vector	Aligned	UnicodeName	20.71 ns	0.040 ns	0.037 ns	0.81	0.01	0.0166	-	104 B	1.00
Decode	Main	Main	Aligned	Length8	23.64 ns	0.012 ns	0.011 ns	1.00	0.00	0.0063	-	40 B	1.00
Decode	Vector	Vector	Aligned	Length8	14.63 ns	0.044 ns	0.041 ns	0.62	0.00	0.0063	-	40 B	1.00
TryDecode	Main	Main	Aligned	Length8	20.37 ns	0.020 ns	0.019 ns	1.00	0.00	-	-	-	NA
TryDecode	Vector	Vector	Aligned	Length8	11.74 ns	0.020 ns	0.016 ns	0.58	0.00	-	-	-	NA
Write	Main	Main	Aligned	Length8	22.18 ns	0.062 ns	0.058 ns	1.00	0.00	0.0153	-	96 B	1.00
Write	Vector	Vector	Aligned	Length8	17.24 ns	0.064 ns	0.057 ns	0.78	0.00	0.0153	-	96 B	1.00
Decode	Main	Main	Aligned	Length16	36.07 ns	0.010 ns	0.008 ns	1.00	0.00	0.0089	-	56 B	1.00
Decode	Vector	Vector	Aligned	Length16	16.58 ns	0.027 ns	0.025 ns	0.46	0.00	0.0089	-	56 B	1.00
TryDecode	Main	Main	Aligned	Length16	32.38 ns	0.046 ns	0.043 ns	1.00	0.00	-	-	-	NA
TryDecode	Vector	Vector	Aligned	Length16	13.21 ns	0.030 ns	0.025 ns	0.41	0.00	-	-	-	NA
Write	Main	Main	Aligned	Length16	30.14 ns	0.054 ns	0.051 ns	1.00	0.00	0.0178	-	112 B	1.00
Write	Vector	Vector	Aligned	Length16	19.41 ns	0.356 ns	0.315 ns	0.64	0.01	0.0179	-	112 B	1.00
Decode	Main	Main	Aligned	Length32	61.89 ns	0.310 ns	0.290 ns	1.00	0.01	0.0139	-	88 B	1.00
Decode	Vector	Vector	Aligned	Length32	21.06 ns	0.647 ns	0.605 ns	0.34	0.01	0.0140	-	88 B	1.00
TryDecode	Main	Main	Aligned	Length32	58.35 ns	0.375 ns	0.332 ns	1.00	0.01	-	-	-	NA
TryDecode	Vector	Vector	Aligned	Length32	16.27 ns	0.343 ns	0.268 ns	0.28	0.00	-	-	-	NA
Write	Main	Main	Aligned	Length32	47.59 ns	0.814 ns	0.680 ns	1.00	0.02	0.0229	-	144 B	1.00
Write	Vector	Vector	Aligned	Length32	23.47 ns	0.717 ns	0.671 ns	0.49	0.02	0.0229	-	144 B	1.00
Decode	Main	Main	Aligned	Length64	123.09 ns	1.822 ns	1.704 ns	1.00	0.02	0.0241	-	152 B	1.00
Decode	Vector	Vector	Aligned	Length64	30.67 ns	0.297 ns	0.278 ns	0.25	0.00	0.0242	-	152 B	1.00
TryDecode	Main	Main	Aligned	Length64	117.53 ns	0.437 ns	0.409 ns	1.00	0.00	-	-	-	NA
TryDecode	Vector	Vector	Aligned	Length64	23.48 ns	0.103 ns	0.091 ns	0.20	0.00	-	-	-	NA
Write	Main	Main	Aligned	Length64	83.22 ns	0.652 ns	0.578 ns	1.00	0.01	0.0331	-	208 B	1.00
Write	Vector	Vector	Aligned	Length64	31.31 ns	0.323 ns	0.286 ns	0.38	0.00	0.0331	-	208 B	1.00
Decode	Main	Main	Aligned	Length256	460.07 ns	0.758 ns	0.709 ns	1.00	0.00	0.0854	-	536 B	1.00
Decode	Vector	Vector	Aligned	Length256	73.86 ns	0.125 ns	0.104 ns	0.16	0.00	0.0854	-	536 B	1.00
TryDecode	Main	Main	Aligned	Length256	447.23 ns	0.333 ns	0.311 ns	1.00	0.00	-	-	-	NA
TryDecode	Vector	Vector	Aligned	Length256	54.62 ns	0.059 ns	0.055 ns	0.12	0.00	-	-	-	NA
Write	Main	Main	Aligned	Length256	297.00 ns	0.250 ns	0.234 ns	1.00	0.00	0.0939	-	592 B	1.00
Write	Vector	Vector	Aligned	Length256	71.91 ns	0.080 ns	0.071 ns	0.24	0.00	0.0943	0.0001	592 B	1.00
Decode	Main	Main	Aligned	Length1024	1,769.30 ns	1.602 ns	1.420 ns	1.00	0.00	0.3300	-	2072 B	1.00
Decode	Vector	Vector	Aligned	Length1024	270.31 ns	0.690 ns	0.611 ns	0.15	0.00	0.3300	-	2072 B	1.00
TryDecode	Main	Main	Aligned	Length1024	1,693.69 ns	1.338 ns	1.252 ns	1.00	0.00	-	-	-	NA
TryDecode	Vector	Vector	Aligned	Length1024	195.24 ns	0.092 ns	0.086 ns	0.12	0.00	-	-	-	NA
Write	Main	Main	Aligned	Length1024	1,119.74 ns	1.917 ns	1.700 ns	1.00	0.00	0.3376	0.0019	2128 B	1.00
Write	Vector	Vector	Aligned	Length1024	250.16 ns	1.218 ns	1.080 ns	0.22	0.00	0.3390	0.0024	2128 B	1.00
Decode	Main	Main	Unaligned	CountryCode	14.42 ns	0.127 ns	0.119 ns	1.00	0.01	0.0051	-	32 B	1.00
Decode	Vector	Vector	Unaligned	CountryCode	15.08 ns	0.040 ns	0.033 ns	1.05	0.01	0.0051	-	32 B	1.00
TryDecode	Main	Main	Unaligned	CountryCode	11.31 ns	0.021 ns	0.018 ns	1.00	0.00	-	-	-	NA
TryDecode	Vector	Vector	Unaligned	CountryCode	11.63 ns	0.021 ns	0.020 ns	1.03	0.00	-	-	-	NA
Write	Main	Main	Unaligned	CountryCode	16.89 ns	0.182 ns	0.170 ns	1.00	0.01	0.0127	-	80 B	1.00
Write	Vector	Vector	Unaligned	CountryCode	17.38 ns	0.043 ns	0.038 ns	1.03	0.01	0.0127	-	80 B	1.00
Decode	Main	Main	Unaligned	CommonName	56.72 ns	0.067 ns	0.056 ns	1.00	0.00	0.0127	-	80 B	1.00
Decode	Vector	Vector	Unaligned	CommonName	26.71 ns	0.057 ns	0.053 ns	0.47	0.00	0.0127	-	80 B	1.00
TryDecode	Main	Main	Unaligned	CommonName	52.68 ns	0.068 ns	0.064 ns	1.00	0.00	-	-	-	NA
TryDecode	Vector	Vector	Unaligned	CommonName	21.68 ns	0.018 ns	0.017 ns	0.41	0.00	-	-	-	NA
Write	Main	Main	Unaligned	CommonName	43.62 ns	0.035 ns	0.033 ns	1.00	0.00	0.0216	-	136 B	1.00
Write	Vector	Vector	Unaligned	CommonName	26.82 ns	0.068 ns	0.064 ns	0.61	0.00	0.0217	-	136 B	1.00
Decode	Main	Main	Unaligned	UnicodeName	28.22 ns	0.061 ns	0.057 ns	1.00	0.00	0.0076	-	48 B	1.00
Decode	Vector	Vector	Unaligned	UnicodeName	19.32 ns	0.033 ns	0.031 ns	0.68	0.00	0.0076	-	48 B	1.00
TryDecode	Main	Main	Unaligned	UnicodeName	24.74 ns	0.022 ns	0.021 ns	1.00	0.00	-	-	-	NA
TryDecode	Vector	Vector	Unaligned	UnicodeName	15.72 ns	0.023 ns	0.022 ns	0.64	0.00	-	-	-	NA
Write	Main	Main	Unaligned	UnicodeName	25.81 ns	0.032 ns	0.030 ns	1.00	0.00	0.0166	-	104 B	1.00
Write	Vector	Vector	Unaligned	UnicodeName	21.11 ns	0.033 ns	0.031 ns	0.82	0.00	0.0166	-	104 B	1.00
Decode	Main	Main	Unaligned	Length8	23.47 ns	0.032 ns	0.030 ns	1.00	0.00	0.0063	-	40 B	1.00
Decode	Vector	Vector	Unaligned	Length8	14.49 ns	0.036 ns	0.034 ns	0.62	0.00	0.0063	-	40 B	1.00
TryDecode	Main	Main	Unaligned	Length8	20.27 ns	0.025 ns	0.022 ns	1.00	0.00	-	-	-	NA
TryDecode	Vector	Vector	Unaligned	Length8	11.39 ns	0.004 ns	0.003 ns	0.56	0.00	-	-	-	NA
Write	Main	Main	Unaligned	Length8	22.72 ns	0.080 ns	0.062 ns	1.00	0.00	0.0153	-	96 B	1.00
Write	Vector	Vector	Unaligned	Length8	18.17 ns	0.627 ns	0.586 ns	0.80	0.03	0.0153	-	96 B	1.00
Decode	Main	Main	Unaligned	Length16	36.14 ns	0.537 ns	0.476 ns	1.00	0.02	0.0089	-	56 B	1.00
Decode	Vector	Vector	Unaligned	Length16	16.50 ns	0.147 ns	0.114 ns	0.46	0.01	0.0089	-	56 B	1.00
TryDecode	Main	Main	Unaligned	Length16	32.57 ns	0.478 ns	0.447 ns	1.00	0.02	-	-	-	NA
TryDecode	Vector	Vector	Unaligned	Length16	12.89 ns	0.022 ns	0.020 ns	0.40	0.01	-	-	-	NA
Write	Main	Main	Unaligned	Length16	31.11 ns	0.646 ns	0.604 ns	1.00	0.03	0.0178	-	112 B	1.00
Write	Vector	Vector	Unaligned	Length16	19.64 ns	0.068 ns	0.064 ns	0.63	0.01	0.0179	-	112 B	1.00
Decode	Main	Main	Unaligned	Length32	61.77 ns	0.095 ns	0.084 ns	1.00	0.00	0.0139	-	88 B	1.00
Decode	Vector	Vector	Unaligned	Length32	20.65 ns	0.032 ns	0.028 ns	0.33	0.00	0.0140	-	88 B	1.00
TryDecode	Main	Main	Unaligned	Length32	57.64 ns	0.049 ns	0.046 ns	1.00	0.00	-	-	-	NA
TryDecode	Vector	Vector	Unaligned	Length32	15.87 ns	0.013 ns	0.012 ns	0.28	0.00	-	-	-	NA
Write	Main	Main	Unaligned	Length32	47.41 ns	0.314 ns	0.294 ns	1.00	0.01	0.0229	-	144 B	1.00
Write	Vector	Vector	Unaligned	Length32	23.44 ns	0.072 ns	0.068 ns	0.49	0.00	0.0229	-	144 B	1.00
Decode	Main	Main	Unaligned	Length64	123.07 ns	0.224 ns	0.199 ns	1.00	0.00	0.0241	-	152 B	1.00
Decode	Vector	Vector	Unaligned	Length64	29.79 ns	0.068 ns	0.063 ns	0.24	0.00	0.0242	-	152 B	1.00
TryDecode	Main	Main	Unaligned	Length64	117.86 ns	1.290 ns	1.077 ns	1.00	0.01	-	-	-	NA
TryDecode	Vector	Vector	Unaligned	Length64	23.68 ns	0.010 ns	0.008 ns	0.20	0.00	-	-	-	NA
Write	Main	Main	Unaligned	Length64	85.83 ns	0.225 ns	0.211 ns	1.00	0.00	0.0331	-	208 B	1.00
Write	Vector	Vector	Unaligned	Length64	31.51 ns	0.062 ns	0.058 ns	0.37	0.00	0.0331	-	208 B	1.00
Decode	Main	Main	Unaligned	Length256	462.01 ns	0.529 ns	0.495 ns	1.00	0.00	0.0854	-	536 B	1.00
Decode	Vector	Vector	Unaligned	Length256	73.40 ns	0.129 ns	0.114 ns	0.16	0.00	0.0854	-	536 B	1.00
TryDecode	Main	Main	Unaligned	Length256	436.59 ns	0.388 ns	0.344 ns	1.00	0.00	-	-	-	NA
TryDecode	Vector	Vector	Unaligned	Length256	53.95 ns	0.054 ns	0.051 ns	0.12	0.00	-	-	-	NA
Write	Main	Main	Unaligned	Length256	299.77 ns	0.426 ns	0.398 ns	1.00	0.00	0.0939	-	592 B	1.00
Write	Vector	Vector	Unaligned	Length256	73.73 ns	0.133 ns	0.124 ns	0.25	0.00	0.0943	0.0001	592 B	1.00
Decode	Main	Main	Unaligned	Length1024	1,770.61 ns	1.237 ns	1.033 ns	1.00	0.00	0.3300	-	2072 B	1.00
Decode	Vector	Vector	Unaligned	Length1024	269.57 ns	0.657 ns	0.583 ns	0.15	0.00	0.3300	-	2072 B	1.00
TryDecode	Main	Main	Unaligned	Length1024	1,692.96 ns	1.859 ns	1.739 ns	1.00	0.00	-	-	-	NA
TryDecode	Vector	Vector	Unaligned	Length1024	194.38 ns	0.181 ns	0.169 ns	0.11	0.00	-	-	-	NA
Write	Main	Main	Unaligned	Length1024	1,127.43 ns	1.654 ns	1.548 ns	1.00	0.00	0.3376	0.0019	2128 B	1.00
Write	Vector	Vector	Unaligned	Length1024	251.62 ns	0.766 ns	0.717 ns	0.22	0.00	0.3390	0.0024	2128 B	1.00

[azure-pipelines]

Azure Pipelines:
Successfully started running 3 pipeline(s).
13 pipeline(s) were filtered out due to trigger conditions.
There may be pipelines that require an authorized user to comment /azp run to run.

[dotnet-policy-service]

Tagging subscribers to this area: @dotnet/area-system-formats-asn1, @bartonjs, @vcsjones
See info in area-owners.md if you want to be subscribed.

[Copilot]

Pull request overview

This PR introduces a vectorized implementation of ASN.1 BMPString (big-endian UCS-2 without surrogates) encode/decode logic in BMPEncoding, aiming to improve throughput on larger inputs while preserving scalar behavior for validation and exact error indexing.

Changes:

• Add Vector<ushort>-based fast paths for BMPString encode (GetBytes) and decode (GetChars), including big-endian handling and surrogate detection.
• Split BMPEncoding into partial implementations for net vs downlevel builds, keeping max-count logic in the shared file.
• Add reader/writer tests covering vector boundary lengths, bounds safety (BoundedMemory), and surrogate detection index correctness.

Show a summary per file

File	Description
src/libraries/System.Formats.Asn1/src/System/Formats/Asn1/AsnCharacterStringEncodings.net.cs	Adds vectorized BMPEncoding encode/decode with endian swap + surrogate scanning.
src/libraries/System.Formats.Asn1/src/System/Formats/Asn1/AsnCharacterStringEncodings.downlevel.cs	Restores scalar BMPEncoding implementation for downlevel targets.
src/libraries/System.Formats.Asn1/src/System/Formats/Asn1/AsnCharacterStringEncodings.cs	Makes BMPEncoding partial and retains max byte/char count logic.
src/libraries/System.Formats.Asn1/tests/Writer/WriteBMPString.cs	Adds vector-boundary + surrogate-index writer regression tests and bounds checks.
src/libraries/System.Formats.Asn1/tests/Reader/ReadBMPString.cs	Adds vector-boundary + surrogate-index reader regression tests and bounds checks.

Copilot's findings

• Files reviewed: 5/5 changed files
• Comments generated: 1

[EgorBo]

@vcsjones not sure if it matters, but currently for loop shapes that load more than one element (either SIMD or BinaryOperations.*) it's more jit friendly to use code like

while (span.Length >= <const>)
{
    var vec = Vector.Create(span); // or any other operation that loads from/saves to span slice.
    span = span.Slice(const);
}

the ones you used will lead to redundant checks (hopefully will be optimized in .NET 12.0).

Regardless, thanks for using safe code for that 🙂

[vcsjones]

not sure if it matters, but currently for loop shapes that load more than one element (either SIMD or BinaryOperations.*) it's more jit friendly to use code like

Hm yeah that gave a nice 20~25% performance bump.

[Copilot]

Copilot's findings

• Files reviewed: 5/5 changed files
• Comments generated: 2