Fix exp10 fallback and exp2 and exp10 for Float16 #22716
Conversation
test/math.jl
Outdated
| @@ -661,3 +661,17 @@ end | |||
| @test sincos(big(1)) == (sin(big(1)), cos(big(1))) | |||
| @test sincos(big(1.0)) == (sin(big(1.0)), cos(big(1.0))) | |||
| end | |||
|
|
|||
| @testset "Float16 fallbacks" begin | |||
There was a problem hiding this comment.
Perhaps integrate with
Line 252 in e5cc4d7
base/math.jl
Outdated
| @@ -263,14 +263,27 @@ cbrt(x::AbstractFloat) = x < 0 ? -(-x)^(1//3) : x^(1//3) | |||
| """ | |||
| exp2(x) | |||
|
|
|||
| Compute ``2^x``. | |||
| Compute the base `2` exponential of `x`, in other words ``2^x``. | |||
|
|
|||
| ```jldoctest | |||
base/math.jl
Outdated
|
|
||
| Compute the base `10` exponential of `x`, in other words ``10^x``. | ||
|
|
||
| ```jldoctest |
test/math.jl
Outdated
| end | ||
|
|
||
| # test AbstractFloat fallback | ||
| struct Float32432{T<:AbstractFloat} <: AbstractFloat |
There was a problem hiding this comment.
what is this number referring to?
There was a problem hiding this comment.
Nothing, just a random string of numbers so it doesn't clash with another name
There was a problem hiding this comment.
use the issue number or a word, less confusing that way
There was a problem hiding this comment.
I will change to use the PR number since I didn't open an issue for this
|
Appveyor looks missing, restarting |
|
Anything else to do here? |
base/math.jl
Outdated
| @@ -263,14 +263,29 @@ cbrt(x::AbstractFloat) = x < 0 ? -(-x)^(1//3) : x^(1//3) | |||
| """ | |||
| exp2(x) | |||
|
|
|||
| Compute ``2^x``. | |||
| Compute the base `2` exponential of `x`, in other words ``2^x``. | |||
There was a problem hiding this comment.
2 should probably not be code highlighted? Likewise 10 in the docstring below.
There was a problem hiding this comment.
I don't see a harm in highlighting the number, maybe others have a different thought on this? Will update accordingly.
There was a problem hiding this comment.
don't think it needs it for just the number on its own
* Fix exp10 fallback and exp2 and exp10 for Float16 * Update docstring and test * Change float struct name * Update math.jl
* Fix exp10 fallback and exp2 and exp10 for Float16 * Update docstring and test * Change float struct name * Update math.jl (cherry picked from commit 3ff6eac)
exp2was not added to the Float16 definitions and hence calls the very slow fallback to2^x.exp10does not have any fallback definitions and thus things likelead to a StackOverFlow or with other
AbstractFloat