Generated kernel argument order (#82)

* Swap order of arguments from (in, out) to idiomatic (out, in)

* Update to next breaking version

Author: AntonReinhard

Project.toml

@@ -1,6 +1,6 @@
 name = "ComputableDAGs"
 uuid = "62933717-1c9d-4a3f-b06f-7ab7f17ca32d"
-version = "0.4.0"
+version = "0.5.0"
 authors = ["Anton Reinhard <[email protected]>", "Uwe Hernandez Acosta"]
 
 [deps]

ext/KernelAbstractionsExt.jl

@@ -8,7 +8,7 @@ using Random
 include("kernel_wrapper.jl")
 
 function ComputableDAGs.init_kernel(mod::Module)
-    mod.eval(Meta.parse("@kernel inbounds = true function _ka_broadcast!(@Const(in::AbstractVector), out::AbstractVector, val::Val)
+    mod.eval(Meta.parse("@kernel inbounds = true function _ka_broadcast!(out::AbstractVector, @Const(in::AbstractVector), val::Val)
         id = @index(Global)
         @inline out[id] = _compute_expr(in[id], val)
     end"))

src/devices/interface.jl

@@ -50,13 +50,13 @@ function measure_device! end
     kernel(dag::DAG, instance, context_module::Module)
 
 !!! warn
-    Before calling this function, make sure to have called [`init_kernel`](@ref) in your session!
+    Before calling this function, make sure to have called [`ComputableDAGs.init_kernel`](@ref) in your session!
 
-For a [`DAG`](@ref), and a problem instance, return a KernelAbstractions kernel of signature `kernel(input::AbstractVector, output::AbstractVector; ndranges::Int64)`, which will return the result of the DAG computation of the input on the given output vector on each index (like a broadcast). This function is only available as an extension when KernelAbstractions is loaded.
+For a [`DAG`](@ref), and a problem instance, return a KernelAbstractions kernel of signature `kernel(output::AbstractVector, input::AbstractVector; ndranges::Int64)`, which will return the result of the DAG computation of the input on the given output vector on each index (like a broadcast). This function is only available as an extension when KernelAbstractions is loaded.
 
 A simple example call for a kernel generated from this might look like the following:
 ```Julia
-cuda_kernel(get_backend(inputs), 256)(inputs, outputs; ndrange=length(inputs))
+cuda_kernel(get_backend(inputs), 256)(outputs, inputs; ndrange=length(inputs))
 ```
 
 The internal index used is `@index(Global)` as provided by KernelAbstractions. For more details, please refer to the documentation of KernelAbstractions.jl.

test/gpu/fibonacci.jl

@@ -15,14 +15,14 @@ function barrier(fib::Fibonacci, backend::Backend, type::Type, size::Int)
     dag = graph(fib)
     k = kernel(dag, fib, @__MODULE__)
 
-    rand_inputs = [(rand(type), rand(type)) for i in 1:size]
+    rand_inputs = [(rand(type), rand(type)) for _ in 1:size]
 
     in = allocate(backend, Tuple{type, type}, size)
     out = allocate(backend, type, size)
 
     copyto!(in, rand_inputs)
 
-    k(backend, 32)(in, out; ndrange = length(in))
+    k(backend, 32)(out, in; ndrange = length(in))
 
     gt_out = fib_gt.(fib.n, getindex.(rand_inputs, 1), getindex.(rand_inputs, 2))
     return @test count(isapprox.(Vector(out), gt_out)) == size