Skip to content

feat: Add global fast_math flag to CudaBuilder #331

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Open
koreaygj wants to merge 3 commits into
base: main
Choose a base branch
from
Open

feat: Add global fast_math flag to CudaBuilder #331

Changes from all commits
Commits
Show all changes
3 commits
Select commit Hold shift + click to select a range
888373b
feat: Add global fast_math flag to CudaBuilder
koreaygj Dec 1, 2025
a8b5dae
Merge branch 'Rust-GPU:main' into feat/add-fast-math
koreaygj Mar 11, 2026
02f3a46
refactor: fast_math flag implements
koreaygj Mar 11, 2026
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
15 changes: 15 additions & 0 deletions crates/cuda_builder/src/lib.rs
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -266,6 +266,21 @@ impl CudaBuilder {
self
}

/// Enable fast math approximations globally (equivalent to NVCC's `--use_fast_math`).
/// Sets `ftz=true`, `fast_sqrt=true`, `fast_div=true`, and `fma_contraction=true`.
/// Individual flags can still be overridden afterward.
///
/// Note: this sacrifices IEEE 754 compliance for performance. Single-precision
/// division and square root will have up to 2 ULP error, and denormal values
/// will be flushed to zero.
pub fn fast_math(mut self) -> Self {
self.ftz = true;
self.fast_sqrt = true;
self.fast_div = true;
self.fma_contraction = true;
self
}

/// Use a fast approximation for single-precision floating point square root.
pub fn fast_sqrt(mut self, fast_sqrt: bool) -> Self {
self.fast_sqrt = fast_sqrt;
Expand Down