Ensure that applicable intrinsics have their ranges tracked by vn/assertionprop/rngchk

[tannergooding]

Found some cases that were getting pessimized in #128677 because we didn't consistently handle different intrinsics across the various places that can produce range information.

[dotnet-policy-service]

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

[Copilot]

Pull request overview

This PR updates CoreCLR JIT range reasoning so that more intrinsic operations contribute consistent non-negative/range information across value numbering, assertion propagation, and range check elimination.

Changes:

• Adds a ValueNumStore helper to extract SIMD size/base type from HW-intrinsic VN function apps.
• Extends IsVNNeverNegative and RangeCheck’s VN-based range extraction to recognize additional HW-intrinsic VNFuncs and bit-counting VNFuncs.
• Extends RangeCheck and AssertionProp tree-based range modeling for GT_INTRINSIC and GT_HWINTRINSIC bit-counting / extract-msb / element extraction patterns.

Reviewed changes

Copilot reviewed 4 out of 4 changed files in this pull request and generated 3 comments.

File	Description
src/coreclr/jit/valuenum.h	Declares new HW-intrinsic VN helper for SIMD size/base type extraction.
src/coreclr/jit/valuenum.cpp	Uses new helper in IsVNNeverNegative and refactors HW-intrinsic VN decoding.
src/coreclr/jit/rangecheck.cpp	Adds new intrinsic/HW-intrinsic range modeling in VN-based and tree-based range computation.
src/coreclr/jit/assertionprop.cpp	Extends IntegralRange modeling for long constants, primitive intrinsics, and HW intrinsics.

[Copilot]

Pull request overview

Copilot reviewed 4 out of 4 changed files in this pull request and generated 4 comments.

[tannergooding]

CC. @dotnet/jit-contrib, @EgorBo for review. Another PR around the various range/limit support we have.

Diffs are generally positive, with about half of the -2k bytes of improvements being in tests and the other half being in other collections.

The general improvements come from us being able to drop otherwise unnecessary casts and a few from redundant comparisons, some of which were found in #128677 and inspired this PR.

The ExtractMostSignificantBits helper is the new thing and is particularly prominent since we can know its never negative for any V64/V128 and almost any V256 (barring V256<byte> which has 32 elements).

For the rest, we had some handling already, and so this merges the Arm64/x64 code paths together to reduce duplication in those spots and is otherwise ensuring consistent handling across the locations that produce range/limit information, as we had a split of the handling before which could cause things to be missed.

There is no way for the JIT to determine this range information without us being explicit about it

[Copilot]

Pull request overview

Copilot reviewed 4 out of 4 changed files in this pull request and generated 2 comments.

[Copilot]

Pull request overview

Copilot reviewed 4 out of 4 changed files in this pull request and generated 1 comment.