Kristoffer Carlsson

This package takes a massive amount of time to precompile (~100s) and most of that is due to the precompile workload. Removing the precompile workload makes it take 15s (and...

High-level stable API

4

I'm in the progress of upgrading FemtoCleaner (the GitHub bot that fixes deprecations) from CSTParser 0.2.1 to the current version of CSTParser. However, progress is very slow because since CSTParser...

See https://github.com/JuliaLang/julia/issues/53591 for details.

```julia # 1.11 julia> :UUID in propertynames(Base) true # 1.10 julia> :UUID in propertynames(Base) false ``` Causes this test to fail: https://github.com/JuliaData/StructTypes.jl/blob/47c4691d9b5f72455566658a50d950bfb1522a75/test/runtests.jl#L263-L281

add a missing dot in front of an `include` and fix include path

5

I guess this was missed.

Type for `broadcast(-, 1:2:∞, 0.3)` changed in Julia 1.8

2

```jl # 1.8 julia> broadcast(-, 1:2:∞, 0.3) |> typeof StepRangeLen{Infinities.Infinity, Float64, Int64, Integer} # 1.7 julia> broadcast(-, 1:2:∞, 0.3) |> typeof InfiniteArrays.InfStepRange{Float64, Float64} ``` This breaks some stuff later: ```jl...

Don't force loop unrolling in reductions

11

Instead of unrolling the whole loop in the reduction, we can just keep the loop and let LLVM do its job with loop unrolling (since the number of loop iterations...

The package should not hard error just because a precompile statement gets stale.

3x3 matrix multiply could potentially be faster

28

This issue is mostly to document that there is a potential speedup to be gain in e.g. 3x3 matrix multiply (and possible other operations where the matrix size is "odd")....