Back to Questions
3.5k
views1
comment
WebSharper collections performance
websharperf#

I've written a version of Conway's Game of Life that runs in the browser, for experimenting with WebSharper. I also wrote a pure Javascript version for comparison purposes, basing the JS code closely on the WebSharper code (though I did e.g., have to write my own JS implementation of Set). I'm noticing much poorer performance from the WebSharper code than the pure JS code; I'm not sure what to attribute that to or if there are general performance improvement tips for WebSharper. Since one of my interests in WebSharper is to use it for mobile development, the speed and efficiency of execution in the browser is an issue.

The WebSharper code is up on AppHarbor here:
[link:conwaylife.apphb.com]

And the pure JS version is here:
[link:phobot.net]

For testing purposes, I placed an R-pentomino in the center of each of the four sides of the canvas, near the outer edge. (The game representation uses an infinite grid, so patterns continue evolving properly even if they extend off the sides of the viewable region.) The game clock ticks every half-second, but after the WebSharper version gets up to a few hundred cells, the game slows down perceptibly and my laptop fan starts coming on. The pure Javascript version doesn't miss a beat. I'd be happy to create more precise benchmarks and/or run a JS profiler on the code if that's useful.

Another thing I'm noticing is that occasionally patterns in the game that are supposed to be stable will unpredictably disintegrate. I'd love to report this as an issue, but I can't reliably recreate the problem; it happens at random, mainly when there is a large number of cells on the screen.

I'm using WebSharper.2.4.85.235.

Source code is publicly available on github at [link:github.com] ; the version used for comparison between WebSharper and JS is 0.9.0.

.

Hi Paul,

It appears your JavaScript Set uses something similar to a HashSet while WebSharper code uses a purely functional Set. The later simply can't compete on a use case where HashSet would suffice. Can't compete even in F#, let alone in JavaScript, Set is too general. Even if you used fast sets, your code has some other performance anti-patterns, for example this allocates a list (slow):

1
2
3
4

       [   for i in -1..1 do
           for j in -1..1 do
               yield a+i, b+j
       ] |> Set.ofList

You can always match pure JS performance in WebSharper by dropping to pure JavaScript in performance-critical code. Our collections are primarily slower because they are more general.

That being said, it certainly might be the case that some of our collections / standard library code can be improved. I will be glad to review specific code for performance if you could narrow it down a bit.

Same goes for the crash problem, it would be nice to know more details.

Anton

By on 10/30/2012 8:03 PM ()

    Topic tags

    IntelliFactory Offices Copyright (c) 2011-2012 IntelliFactory. All rights reserved.
    Home | Products | Consulting | Trainings | Blogs | Jobs | Contact Us | Terms of Use | Privacy Policy | Cookie Policy     		Built with WebSharper