Fix a bunch more float16 and fma_emulated edge cases #63525
oscardssmith
merged
12 commits into
JuliaLang:master
from
oscardssmith:os/float-edge-cases
math: Fix `fma_emulated` rounding near underflow and overflow
3460fe72
runtime: Round the product separately in 16-bit `muladd_float`
b58335cd
codegen: Avoid double rounding in BFloat16 `fma_float` and int conver…
2ad5c812
float: Throw `InexactError` when truncating `-Inf16` to wide integers
f41d8a43
Printf: Fix `%a` with a precision of 4 or more for Float16
b70ce77f
Printf: Put zero padding after the sign of `-0.0`
381028c6
math: Fix `NaN^n` for large `n` and the sign of `(-0.0)^n`
493e70bb
gmp: Don't overflow in `Float64(::BigInt, RoundToZero)`
af22f396
vtjnash
commented
on 2026-10-02
vtjnash
commented
on 2026-10-02
math: Return subnormal `sech` and `csch` once `cosh` and `sinh` overflow
ef2721d1
math: Compute Float16 degree-based trig functions in Float32
9186d7d7
ranges: Scale rationals exactly when constructing float ranges
07b4c010
mpfr: Keep `precision` and `rounding` when parsing a BigFloat with tr…
aa8ce9ea
vtjnash
approved these changes
on 2026-10-06
topolarity
approved these changes
on 2026-10-06
Reviewers
Assignees
No one assigned
Labels
maths
bugfix
don't squash
Milestone
No milestone
Login to write a write a comment.
Login via GitHub