How to convert Symbolics.jl expressions to functions while keeping speed?

General Usage
, ,
1

Hello!

I’m using Symbolics.jl to perform some symbolic computations and then convert the resulting expressions to functions for further use. However, when I use build_function to generate functions, the speed is only a quarter of that when I directly define the function. How can I optimize this?

Thank you all in advance!

Following is an example:

using Symbolics
using BenchmarkTools

@variables x y
f(x,y) = x^2 + sin(x+y)
D = Differential(x)
expr = expand_derivatives(D(f(x,y)))  # output: 2x + cos(x + y)

# generate the function from the expression
f_expr = build_function(expr, x, y, expression = Val{false})
# define the function directly
f_defn(x, y) = 2x + cos(x + y)

# benchmarks
@btime for x in rand(100), y in rand(100)
	f_expr(x, y)
end
# output: 239.600 μs (30101 allocations: 557.12 KiB)

@btime for x in rand(100), y in rand(100)
	f_defn(x, y)
end
# output: 60.300 μs (101 allocations: 88.38 KiB)
2 
const fexpr = build_function(expr, x, y, expression = Val{false})

f_defn is const, while fexpr is not.

4 Likes
3

Thank you!