push

Author: William Moses

ext/ReactantCUDAExt.jl

@@ -410,6 +410,7 @@ function (func::LLVMFunc{F,tt})(args...; convert=Val(false), blocks::CuDim=1, th
 	arg = transpose_val(arg)
 	@show arg
 	push!(restys, MLIR.IR.type(arg))
+	push!(mlir_args, arg)
 	push!(aliases,
 	      MLIR.IR.Attribute(MLIR.API.stablehloOutputOperandAliasGet(
 		MLIR.IR.context(),

test/cuda.jl

@@ -18,6 +18,7 @@ end
 @testset "Square Kernel" begin
     oA = collect(1:1:64)
     A = Reactant.to_rarray(oA)
+    @show @code_hlo optimize=false square!(A)
     @show @code_hlo square!(A)
     func = @compile square!(A)
     @test all(Array(A) .≈ (oA .* oA))