Skip to content

Scalar vector operations can use NEON instructions, which have incorrect subnormal behavior #129853

New issue
New issue
Open
Open
Scalar vector operations can use NEON instructions, which have incorrect subnormal behavior#129853
Labels
A-LLVMArea: Code generation parts specific to LLVM. Both correctness bugs and optimization-related issues.A-floating-pointArea: Floating point numbers and arithmeticC-bugCategory: This is a bug.T-compilerRelevant to the compiler team, which will review and decide on the PR/issue.
@RalfJung

Description

@RalfJung
RalfJung
opened on Sep 1, 2024

This is the Rust version of llvm/llvm-project#106909: scalar f32/f64 operations can produce the wrong result for subnormals on ARM if NEON instructions are used.

This just got fixed in upstream LLVM, but we should also have a test covering this.

Cc @nikic

Metadata

Metadata

Assignees

No one assigned

    Labels

    A-LLVMArea: Code generation parts specific to LLVM. Both correctness bugs and optimization-related issues.A-floating-pointArea: Floating point numbers and arithmeticC-bugCategory: This is a bug.T-compilerRelevant to the compiler team, which will review and decide on the PR/issue.

    Type

    No type

    Projects

    No projects

    Milestone

    No milestone

    Relationships

    None yet

    Development

    No branches or pull requests

    Issue actions