[jnunn-aws]

Description

Replace the byte-at-a-time Ionic.Zlib CRC32 loop with Crc32Ieee: PCLMULQDQ
folding on x86/x64 and ARM64 CRC32 instructions on .NET 8+, slicing-by-16
tables on netstandard2.0/netcoreapp3.1/net472. Covers Crc32Managed (flexible
checksums), CrcCalculatorStream (x-amz-crc32, event streams). Output is
bit-identical to the previous implementation.

PCLMULQDQ folding on x86/x64 and ARM64 CRC32 instructions on .NET 8+; slicing-by-16 tables on netstandard2.0/netcoreapp3.1/net472. The vectorized code is adapted from dotnet/runtime's System.IO.Hashing (MIT, under ThirdParty/RuntimeBackports).

Motivation and Context

The byte-at-a-time CRC32 dominated checksum cost on large uploads and x-amz-crc32 validation.

	x86	ARM64
ops/CPU-sec, this change alone (71-case geomean)	+2.3%	+3.1%

The gain is concentrated in the large-payload cases.

Measurement: Numbers come from the serde ops/CPU-sec runner (sdk/test/Performance/SerdeBenchmarks, --cpu-time, stub transport), geomean of 71 cases, 3 interleaved runs. Instances: m7i.metal-24xl (x86) and m7g.metal (ARM64). Settings: net8.0, 20,000-iteration warmup, measured on development 8956121, which is the same code as this branch for these files. Each change is measured alone against that base. They will be re-measured at the harness's current default (100,000 warmup, net10.0, stub and https) before the PRs leave draft.

Testing

• New Crc32IeeeTests (20): known vectors; slicing-by-16, PCLMULQDQ and ARM64 paths against a reference implementation over all lengths and offsets; incremental chunking vs one-shot; Crc32Managed, the checksum factory and CrcCalculatorStream against the reference.
• Real-AWS check: S3's stored CRC32 matched an independently computed CRC32 for every upload that stores one.
• Core unit tests pass on net8.0 and netcoreapp3.1.

Breaking Changes Assessment

None. Output is bit-identical.

Dry-runs

Combined dry-run of this series (jonatnun/perf-all-v3 @ 2a8cec6: #4557–#4563 together).

• DotNet Dry-run ID: 438fa1b0-130d-4670-b74e-2df6c79bb7a1
  • Completed successfully
• PowerShell Dry-run ID: c0049dbb-e11a-4a90-9fcd-087f2dc26e93
  • Completed successfully

Checklist

• My code follows the code style of this project
• I have added tests to cover my changes
• All new and existing tests passed

License

• I confirm that this pull request can be released under the Apache 2 license

[jnunn-aws]

Closing: 4.1 covers this with System.IO.Hashing (#4492), plus slicing-by-16 for targets before .NET 10 (#4533).