I am new to Julia. While choosing a programming language for our next project wrote a short program in Julia: using LinearAlgebra using Statistics using Random import Base: +, - struct Point x::Float64 y::Float64 z::Float64 end +(a::Point, b::Point) = Point(a.x + b.x, a.y + b.y, a.z…

[image] DNF: Do you really want this thing to print all the time? It’s going to spend its time printing mostly. It seems printing to a terminal in Julia is pretty slow compared to Python, so that will be noticeable for lots of output to a terminal: paulm@cmstorm 09:25:/data/examples/julia$ c…

I would not use my own definition of Point. Instead I would just use StaticArrays. You can roll your own, like you have, but then you should add some more functionality. You should absolutely avoid the super-expensive conversion from points to matrix, and I would not generate the whole set of batch…

Here’s a multithreaded version that scales pretty well (3.5x faster with 4 threads in my laptop): function main_threaded() MIN = -2000 MAX = 2000 BATCHES_COUNT = 1000000 POINTS_PER_BATCH = 8 generatePoint() = rand(SVector{3, Float64}) .* (MAX - MIN) .+ MIN generateBatch()…

[image] paulmelis: It seems printing to a terminal in Julia is pretty slow compared to Python, so that will be noticeable for lots of output to a terminal: I have a hard time believing that it’s useful to print 1 million calculated Planes objects to screen. Printing to file, ok, but I don’t t…

At least one should avoid printing inside the hot loop.

If the objective is to just speed up the posted script, then I think the conversation is going just fine. But if an objective is to compare versions of the script in different languages, then I think the Python and Haskell code should also be posted. I myself won’t be able to make language compariso…

A note: In my posted example, roughly 80%-90% of the time is now spent inside svd. If you want more speed-ups, you should consider whether you can do better than the standard svd.

using LinearAlgebra, Statistics, Random, StaticArrays struct Plane point::SVector{3,Float64} normal::SVector{3,Float64} end function mean_center(points::SMatrix) centroid = mean(points,dims=2) points_centered = points .- centroid centroid, points_centered end function best_fit…

@Yury_Kotlyarov if you are working with computational geometry, consider Meshes.jl . It should be a performant api for geometrical objects and meshes. Your Point type is available there for example, as well as many other types.

Very slow execution time in comparison even to Python

General Usage Performance

paulmelis October 29, 2020, 9:10am 14

See my post in Why is printing to a terminal slow? - #22 by paulmelis just yet, e.g. piping through less is also very slow

Topic		Replies	Views
Why Julia and linear algebra module are incredibly slow, compare to C++ and Eigen Performance question , package , performance , linearalgebra	14	5462	March 1, 2019
SVD 2x slower than in Matlab and how to get best performance on Windows10 Performance	20	3643	October 9, 2024
Benchmark MATLAB & Julia for Matrix Operations Performance	148	20821	October 15, 2019
Numpy 10x faster than Julia ?! What am I doing wrong ?! [solved - julia faster now] Performance question	37	11321	October 15, 2019
Why is this simple function twice as slow as its Python version Performance question	97	4957	April 12, 2021

Very slow execution time in comparison even to Python

Related topics