Chris Elrod

I think this may work: ```julia run(`make -j$(Sys.CPU_THREADS)`) code = raw""" using BenchmarkTools, ThreadsX, Random seed = Ref(0) while true @btime ThreadsX.sort($(rand(MersenneTwister(@show(seed[] += 1)), 0:0.01:1, 1_000_000))) end """ const EXIT_CODE...

Could this be a problem? ```julia signal (11): Segmentation fault in expression starting at REPL[3]:1 - at ./int.jl:52 [inlined] unsafe_length at ./range.jl:517 [inlined] unsafe_indices at ./abstractarray.jl:99 [inlined] _indices_sub at ./subarray.jl:409...

> Which Julia revision you used to get the segmentation fault? That was 1.5.0-DEV.300. https://github.com/JuliaLang/julia/commit/65b8e7ed72541d1dd5bc1f4dcabc5245e1dba718 I have no idea why it would segfault either. In my ignorance, a broken stack...

Julia can SIMD just fine without `@avx` by relying on LLVM. The reason for macros like `@simd` is the same as adding compiler flags like `--ffast-math`, which are often required...

Seems it did already exist: https://github.com/JuliaArrays/ArrayInterface.jl/issues/10#issuecomment-1133809007 Trial/error isn't nearly as nice of a way of finding things out as just being able to look at the information.

Seems ArrayInterface 6 restricts it to 0.1.0: https://github.com/JuliaSIMD/VectorizationBase.jl/runs/6541055845?check_suite_focus=true

This continues to be annoying. In order to see if this change is actually okay, how do I easily look up what changed in ArrayInterfaceOffsetArrays 0.1.2? https://github.com/JuliaSIMD/LoopVectorization.jl/pull/412#discussion_r893860162 It'd be trivial...

What's the best way to test the parses depending on CUDA?

Yes. Or at least, I can merge. What needs to happen?

I'm not sure what exactly you have in mind. When it comes to allocating memory, I've almost always used Julia APIs instead of going the `llvm` route (and the `llvm`...