Skip to content

Fix ldexp to prevent math range errors #6216

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

Merged
youknowone merged 5 commits into from
Oct 26, 2025
Merged

Fix ldexp to prevent math range errors #6216

Changes from 1 commit
Commits
Show all changes
5 commits
Select commit Hold shift + click to select a range
4c88621
Fix ldexp to prevent math range errors
YashSuthar983 Oct 25, 2025
728b7b4
Fix ldexp to handle denormalized inputs correctly
YashSuthar983 Oct 25, 2025
6aaf3d2
tests(math): remove expectedFailure from testLdexp_denormal
YashSuthar983 Oct 25, 2025
e029a01
Update Lib/test/test_math.py
YashSuthar983 Oct 26, 2025
8acd6c5
fixed formate in math.rs
YashSuthar983 Oct 26, 2025
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
Prev Previous commit
Next Next commit
Fix ldexp to handle denormalized inputs correctly 
Detect denormalized input values below f64::MIN_POSITIVE
Scale subnormal inputs by 2^54 to bring them into normalized range

Signed-off-by: Yash Suthar <[email protected]>
  • Loading branch information
@YashSuthar983
YashSuthar983 committed Oct 25, 2025
commit 728b7b420ac3fc430b7f66a8f2c815f9bc903c95
10 changes: 9 additions & 1 deletion stdlib/src/math.rs
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -570,7 +570,15 @@ mod math {
// Direct multiplication would overflow for large i values, especially when computing
// the largest finite float (i=1024, x<1.0). By directly modifying the exponent bits,
// we avoid intermediate overflow to infinity.
let (mant, exp0) = float_ops::decompose_float(value);

// Scale subnormals to normal range first, then adjust exponent.
let (mant, exp0) = if value.abs() < f64::MIN_POSITIVE {
let scaled = value * (1u64 << 54) as f64; // multiply by 2^54
let (mant_scaled, exp_scaled) = float_ops::decompose_float(scaled);
(mant_scaled, exp_scaled - 54) // adjust exponent back
} else {
float_ops::decompose_float(value)
};

let i_big = i.as_bigint();
let overflow_bound = BigInt::from(1024_i32 - exp0); // i > 1024 - exp0 => overflow
Expand Down