|
|
|
|
|
|
by markkitti
1239 days ago
|
|
|
This becomes an important part of optimizing Julia code. There is some tooling for this. Below, identity is a type stable function because we know that an Int64 argument results in a Int64 output. The macro code_warntype reveals the type analysis: ```
julia> @code_warntype identity(5)
MethodInstance for identity(::Int64)
from identity(x) @ Base operators.jl:513
Arguments
#self#::Core.Const(identity)
x::Int64
Body::Int64
1 ─ nothing
└── return x
``` This is type unstable and results in dynamic dispatch because we are not sure if the argument to identity will be an Int64 or a Float64. ```
julia> f(x) = identity(x ≥ 0 ? x : x + 0.0)
f (generic function with 1 method) julia> @code_warntype f(5) MethodInstance for f(::Int64)
from f(x) @ Main REPL[4]:1
Arguments
#self#::Core.Const(f)
x::Int64
Locals
@_3::Union{Float64, Int64}
Body::Union{Float64, Int64}
1 ─ %1 = (x ≥ 0)::Bool
└── goto #3 if not %1
2 ─ (@_3 = x)
└── goto #4
3 ─ (@_3 = x + 0.0)
4 ┄ %6 = @_3::Union{Float64, Int64}
│ %7 = Main.identity(%6)::Union{Float64, Int64}
└── return %7
``` |
|
|