Press flag to sort, and set "visualize?" to 0 for timing. Or click "Benchmark" to test various list lengths. I invented a new sorting algorithm which is up to 25% faster than @GarboMuffin's Quicksort at lower array sizes (e.g. 20,000 items) but slower at larger sizes. This method is much slower than Quicksort on Turbowarp, though. The algorithm uses only 83 blocks, making it very compact for use in 3D projects. What it does is: 1. For each item of the unsorted set, it places that item into a pre-processing list ("Approximation") using a linear interpolation function between the largest and smallest items in the Approximation list. 2. Once all items have been pre-processed, a weird quicksort-adjacent method attempts to efficiently correct all of the errors in the Approximation list, and it places the corrected items into the Sorted list. From my testing, this is 100% reliable with both integers and floats. In other words, this method uses a very fast O(n) pre-processing pass to make the post-processing pass (a normal sorting algorithm) way faster. I don't have an exact idea of how to calculate the worst-case complexity for this algorithm, but it's really fast for most realistic use cases. For 200,000 items, this method is about 30% slower than @GarboMuffin's Quicksort. I don't think this is a very big problem though, as LerpSort is meant more for Z-sorting in 3D, where you usually won't have many more than 5,000 vertices (and even that's pushing it). @GarboMuffin's Quicksort for comparison: https://scratch.mit.edu/projects/310372816/ I think there's a good amount of potential in this method to go even faster. If you want to try and improve this, by all means, go ahead :D Edit: I've tried improving the method but I can't seem to do it without just recreating Bucket Sort.
