AcceleratedArrays.jl
AcceleratedArrays.jl copied to clipboard
andyferris
Arrays with acceleration indices
https://github.com/andyferris/AcceleratedArrays.jl/blob/a77ffdee3bd1c15bcf12d731d09a579d9ef1a0b2/src/predicates.jl#L40-L42 Should be `isgreater(a, b) = isless(b, a)`. It's correct in the README though! https://github.com/andyferris/AcceleratedArrays.jl/blob/a77ffdee3bd1c15bcf12d731d09a579d9ef1a0b2/README.md?plain=1#L120 In practice I think the tests don't catch this because they're only looking at the...
```jl using AcceleratedArrays using BenchmarkTools n = 1_000_000 xs = rand(1:big(n)^4, n) a = accelerate(xs, UniqueHashIndex) b = accelerate(xs, UniqueSortIndex) ab = accelerate(a, UniqueSortIndex) ba = accelerate(b, UniqueHashIndex) @btime 1...
I saw there is a `spatialindex` branch with one commit and was curious what the background/status of that branch is? I asked about it on Slack a while back but...
For indexing distributed arrays, there are two invariants to think of: 1. A DArray is a chain of subarrays 2. The subarrays are on other processes Maybe the indexes are...
It's super common to count or search for or filter data using predicates other than `isequal` or `isless`, and we should really support things like `==` (e.g. `count(==(0), vector)`) The...
Performance was much worse than expected for the UniqueHashIndex that I was using, so I started digging into it. I could be doing something wrong, but from the below it...
```julia julia> a = accelerate([1, 2, 3], UniqueSortIndex) 3-element Array{Int64,1} + UniqueSortIndex: 1 2 3 julia> copy(a) 3-element Array{Int64,1}: 1 2 3 ``` Unless I'm missing some subtlety where this...
This is useful to me, and could be useful to others. I often have UniqueHashIndexes that never get indexed into into (either the indexing is more often happening on another...
Even if this package's interval isn't the same as that in IntervalSets.jl, could it at least be an appropriate subtype? Ref https://github.com/mcabbott/AxisKeys.jl/issues/13, where that package could use this information to...
Hi @andyferris - this project looks great, and is something I am considering extending / building on. I already have some hacky implementations of something similar, but not as nicely...