[Optimization]: Reduce branching when possible in casting.hpp

[zacharyvincze]

Details

• Removes branching where possible to the casting helper functions seen in casting.hpp. Aims to reduce divergence on GPU kernel implementations.
• Includes fixes to some float -> integer saturation casts, especially for 32/64-bit integer cases that are not represented exactly as 32-bit floats.

[Copilot]

Pull request overview

This PR updates the core casting helpers to reduce branching (especially for GPU code paths) and adjusts saturation behavior for some float→integer conversions, alongside adding a small test and extending supported type traits.

Changes:

• Refactors ScalarSaturateCast / ScalarRangeCast logic in casting.hpp to use more branchless/min-max based clamping and special-case small integer widths.
• Extends type traits support to include long/ulong vectorized types.
• Adds a new C++ test covering basic SaturateCast behavior and a few limit/vector cases.
• Adjusts the GPU block dimensions for the Composite operator kernel launch.

Reviewed changes

Copilot reviewed 4 out of 4 changed files in this pull request and generated 5 comments.

File	Description
include/core/detail/casting.hpp	Refactors saturate/range cast implementations to reduce branching and adjust clamping/rounding logic.
include/core/detail/type_traits.hpp	Adds long / ulong to the type-traits macro set.
tests/roccv/cpp/src/tests/core/detail/test_saturate_cast.cpp	Introduces a basic unit test for SaturateCast, including a couple of vectorized casts.
src/op_composite.cpp	Changes GPU kernel launch block dimensions for the composite operator.

---

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

[Copilot]

The device-side float->integer paths use float-specific intrinsics (rintf, fminf/fmaxf, __saturatef, __float2int_rn) even though U is any floating-point type. If U is double, this will downcast to float and can change rounding/saturation behavior. Consider either constraining these branches to U == float (static_assert / if constexpr) or adding double-correct implementations (rint, fmin/fmax, __double2int_rn, etc.).

[codecov-commenter]

Codecov Report

❌ Patch coverage is 54.54545% with 15 lines in your changes missing coverage. Please review.

Files with missing lines	Patch %	Lines
include/core/detail/casting.hpp	54.55%	12 Missing and 3 partials ⚠️

Additional details and impacted files

@@             Coverage Diff             @@
##           develop     #117      +/-   ##
===========================================
+ Coverage    73.51%   73.59%   +0.08%     
===========================================
  Files           77       77              
  Lines         2956     2969      +13     
  Branches       640      635       -5     
===========================================
+ Hits          2173     2185      +12     
+ Misses         338      337       -1     
- Partials       445      447       +2     

Files with missing lines	Coverage Δ
include/core/detail/type_traits.hpp	87.50% <ø> (ø)
include/core/detail/casting.hpp	78.26% <54.55%> (+2.31%)	⬆️

🚀 New features to boost your workflow:

• ❄️ Test Analytics: Detect flaky tests, report on failures, and find test suite problems.

[simonCatBot]

Review: [Optimization] Reduce branching in casting.hpp

Kernel optimization focusing on GPU divergence:

Changes:

• Branch reduction in casting helper functions
• Fixes for float->integer saturation casts (32/64-bit cases)
• 4 files changed, +129/-33 lines

Assessment: Needs Review - Performance optimization.

Reducing branching in GPU kernels is always good for warp efficiency. The fixes for 32/64-bit integer saturation casts sound important - precision issues in type conversion can be subtle bugs.

Would benefit from:

1. Performance benchmarks showing divergence reduction
2. Verification that precision is maintained for edge cases
3. Review of the saturation logic changes

Solid optimization PR.