Summary
The README currently uses wording that implies la-stack supports f64 only “today,” which suggests possible future expansion to other floating-point types such as f32 or f16.
That no longer reflects the crate’s intended scope.
Problem
la-stack is focused on small, fixed-size linear algebra, especially for numerical and computational geometry workloads where f64 and exact arithmetic are the relevant choices. Wording that implies future reduced-precision support can mislead users about the crate’s goals and roadmap.
In practice, likely users of f32 or f16 are working in throughput-oriented or large-matrix settings, which are outside this crate’s target domain.
Proposed change
Revise README language so that it presents f64 as the intended floating-point type rather than a temporary implementation detail.
For example, replace wording like:
today, floating-point operations use f64
with wording closer to:
Floating-point operations in la-stack use f64, consistent with the crate’s focus on small, robustness-sensitive numerical and computational geometry workloads.
Acceptance criteria
- README no longer implies planned expansion to f32 or f16
- README clearly states that f64 is the intended floating-point type
- Messaging is consistent with the crate’s stated scope and anti-goals
Summary
The README currently uses wording that implies la-stack supports f64 only “today,” which suggests possible future expansion to other floating-point types such as f32 or f16.
That no longer reflects the crate’s intended scope.
Problem
la-stack is focused on small, fixed-size linear algebra, especially for numerical and computational geometry workloads where f64 and exact arithmetic are the relevant choices. Wording that implies future reduced-precision support can mislead users about the crate’s goals and roadmap.
In practice, likely users of f32 or f16 are working in throughput-oriented or large-matrix settings, which are outside this crate’s target domain.
Proposed change
Revise README language so that it presents f64 as the intended floating-point type rather than a temporary implementation detail.
For example, replace wording like:
today, floating-point operations use f64
with wording closer to:
Floating-point operations in la-stack use f64, consistent with the crate’s focus on small, robustness-sensitive numerical and computational geometry workloads.
Acceptance criteria