Tolerances
|
Страница в процессе перевода. |
The absolute tolerance used for floating point comparisons is hard-coded in the project to 1e-10 for Float64 and to 1f-5 for Float32. You can use ScopedValues.jl to customize these tolerance values in specific computations:
using Meshes
using ScopedValues
with(Meshes.ATOL64 => 1e-9, Meshes.ATOL32 => 1f-4) do
# do your computations with custom tolerances
end