Chris Elrod

That looks good, although JuliaHub should probably be added as an author also. I left Eli Lilly and joined JuliaHub in August, 2020.

```julia julia> function foo_debug(y, x, b) return LoopVectorization.@turbo_debug for i in axes(y, 1), c in axes(y, 2), n in axes(y, 3) y[i, c, n] = x[i, c, n] + b[c]...

BTW, have you seen VectorizedRNG.jl? On the M1: ```julia julia> vrandn_stridearray(size...) = StrideArray(randn(local_rng(), Float32, size...), static.(size)) vrandn_stridearray (generic function with 1 method) julia> @benchmark randn_stridearray(64, 32, 256) BenchmarkTools.Trial: 2344 samples...

> But yeah, I am trying to train a convolutional neural network for chess on the CPU and memory seems to be quite the bottleneck. I'll have to check with...

Do you need to distribute the image? What happens if you do `cpu_target="native"`? Haswell supports BZHI, but Sandybridge does not (and obviously generic doesn't either). Does the project get precompiled...

@Tobriand mind upgrading to HostCPUFeatures 0.1.11 once it is released (should be in about 15 minutes), and letting me know if it fixes the problem? https://github.com/JuliaRegistries/General/pull/72190

https://julialang.slack.com/archives/C6A044SQH/p1668053535015429

``` ERROR: ArgumentError: type does not have a definite number of fields Stacktrace: [1] fieldcount(t::Any) @ Base ./reflection.jl:895 [2] _append_fields!(t::Expr, body::Expr, sym::Symbol, ::Type{Core.Box}) ``` You have a `Core.Box` in your...

As one aside ```julia function integrate_avx(f::Function, xmin::SVector{2,T}, xmax::SVector{2,T}, x::AbstractVector{T}, w::AbstractVector{T}, h::T) where {T

It works for me. I don't get the warning, and I see `@turbo` is getting used.