Valentin Churavy

`CombinedAdjointThunk` should also return ret value

4

``` julia> autodiff(f, Active(2.0)) (4.0,) ``` Should maybe be: ``` julia> (4.0, (4.0,)) (4.0, (4.0,)) ```

Right now they are available at https://docs.juliahub.com/Enzyme/G1p5n, but we should actually host them from `https://enzyme.mit.edu/julia/stable`

Both type-parameters in thunks `Width` and `RetPrimal` should just be raw values instead of `Val{Width}` and `Val(retprimal)` respectively.

``` ~/.julia/packages/GPUCompiler/XyxTy/src/utils.jl:35 ERROR: LoadError: Conversion of boxed type Vector{Float64} is not allowed Stacktrace: [1] error(s::String) @ Base ./error.jl:33 [2] convert(::Type{LLVM.LLVMType}, typ::Type; ctx::LLVM.Context, allow_boxed::Bool) @ LLVM.Interop ~/.julia/packages/LLVM/szqwr/src/interop/base.jl:92 [3] create_abi_wrapper(enzymefn::LLVM.Function, F::Type, argtypes::Vector{DataType},...

I always forget which one to use to not get the "typtree to deep warning"

`floatrange` causes "unkown" binary operator

2

``` using Enzyme f(x) = first(1.0:0.1:x) autodiff(f, Active(20.0)) ``` cc: @boriskaus

Use test infrastructure with multiple threads

2

For the reverse CI it would be good to use a multi-threaded testsuite (x-ref https://github.com/EnzymeAD/Enzyme/pull/531) I think https://github.com/RelationalAI-oss/XUnit.jl is the only alternative to `Test` that supports this.

``` julia> f(cond, x) = cond ? x : 0 f (generic function with 1 method) julia> using Enzyme julia> autodiff(f, Active, true, Active(1.0)) ERROR: LLVM error: broken function Stacktrace:...