[sjfeng1999]

Motivation

Technical Details

Test Plan

Test Result

Submission Checklist

• Look over the contributing guidelines at https://github.com/ROCm/ROCm/blob/develop/CONTRIBUTING.md#pull-requests.

[Copilot]

Pull request overview

Adds user-facing rounding-mode control for narrowing float-to-float casts in the FlyDSL expression layer, wiring the requested rounding mode through to MLIR arith.truncf and documenting/testing the new behavior.

Changes:

• Add rounding_mode= keyword parameter to Numeric.to(...) and Vector.to(...), with validation and rejection of non-float casts / compile-time constants.
• Extend fp_to_fp(...) to (a) route same-width sub-32-bit float casts through f32 and (b) attach the MLIR roundingmode attribute on narrowing arith.truncf.
• Update language spec docs and add conformance tests for rounding-mode behavior.

Reviewed changes

Copilot reviewed 7 out of 7 changed files in this pull request and generated 3 comments.

Show a summary per file

File	Description
tests/language/test_arithmetic_types.py	Adds IR-level tests for same-width float casts via f32 and for rounding_mode= propagation/rejection.
python/flydsl/expr/utils/arith.py	Plumbs rounding_mode into float-to-float lowering and adds f32 detour for same-width sub-32-bit casts.
python/flydsl/expr/typing.py	Extends Vector.to(...) API to accept/validate rounding_mode= for float-to-float casts.
python/flydsl/expr/numeric.py	Extends Numeric.to(...) API to accept/validate rounding_mode= and enforce runtime-only narrowing semantics.
python/flydsl/expr/arith.py	Re-exports RoundingMode in the public arith API surface.
docs/language/arithmetic_types.md	Documents the new rounding_mode= keyword and rounding-mode semantics.
docs/index.rst	Renames the toctree caption for the language docs section.

---

💡 Add Copilot custom instructions for smarter, more guided reviews. Learn how to get started.