Add some CUDA-specific tests for EnzymeTestUtils

[kshyatt]

I'm really the dog-with-chemistry-set meme here but tried to actually test some of this and see if it works, it seems there is some stuff that isn't quite functioning?

[github-actions]

Your PR requires formatting changes to meet the project's style guidelines.
Please consider running Runic (git runic main) to apply these changes.

Click here to view the suggested changes.

diff --git a/lib/EnzymeTestUtils/ext/EnzymeTestUtilsGPUArraysCoreExt.jl b/lib/EnzymeTestUtils/ext/EnzymeTestUtilsGPUArraysCoreExt.jl
index ab3231ef..e50566f1 100644
--- a/lib/EnzymeTestUtils/ext/EnzymeTestUtilsGPUArraysCoreExt.jl
+++ b/lib/EnzymeTestUtils/ext/EnzymeTestUtilsGPUArraysCoreExt.jl
@@ -57,7 +57,7 @@ function EnzymeTestUtils.to_vec(x::AbstractGPUArray{<:Complex{<:EnzymeTestUtils.
         is_const && return x
         x_new = Core.Typeof(x)(undef, sz)
         x_vec_new_real = view(x_vec_new, 1:length(x))
-        x_vec_new_imag = view(x_vec_new, (length(x) + 1):(2*length(x)))
+        x_vec_new_imag = view(x_vec_new, (length(x) + 1):(2 * length(x)))
         x_vec_complex = reshape(complex.(x_vec_new_real, x_vec_new_imag), sz)
         x_new .= x_vec_complex
         seen_xs[x] = x_new

[github-actions]

Benchmark Results

	main	894c622...	main / 894c622...
basics/make_zero/namedtuple	0.0554 ± 0.0041 μs	0.0547 ± 0.0032 μs	1.01 ± 0.096
basics/make_zero/struct	0.251 ± 0.0062 μs	0.249 ± 0.0064 μs	1.01 ± 0.036
basics/overhead	4.89 ± 0.001 ns	4.89 ± 0.001 ns	1 ± 0.00029
basics/remake_zero!/namedtuple	0.229 ± 0.0084 μs	0.235 ± 0.0092 μs	0.976 ± 0.052
basics/remake_zero!/struct	0.23 ± 0.01 μs	0.234 ± 0.01 μs	0.983 ± 0.062
fold_broadcast/multidim_sum_bcast/1D	10.4 ± 0.3 μs	10.5 ± 2.2 μs	0.985 ± 0.21
fold_broadcast/multidim_sum_bcast/2D	12.7 ± 0.39 μs	12.8 ± 0.37 μs	0.995 ± 0.042
time_to_load	1.09 ± 0.0068 s	1.09 ± 0.002 s	0.999 ± 0.0065

Benchmark Plots

A plot of the benchmark results has been uploaded as an artifact at https://github.com/EnzymeAD/Enzyme.jl/actions/runs/27195096884/artifacts/7503162974.

[wsmoses]

alternatively you could maybe use jlarrays here to test on cpu?

[kshyatt]

why not both 😈

[wsmoses]

Fair nuff

[kshyatt]

lol ok the complex one needs help if anyone can spare a moment, it's cool this extension was just totally not working for months haha