reduce doesn't query src backend if backend specified

src/reduce/mapreduce_1d.jl

@@ -125,14 +125,14 @@ function mapreduce_1d(
     blocks = (len + num_per_block - 1) ÷ num_per_block
 
     if !isnothing(temp)
-        @argcheck get_backend(temp) === get_backend(src)
+        @argcheck get_backend(temp) === backend
         @argcheck eltype(temp) === typeof(init)
         @argcheck length(temp) >= blocks * 2
         dst = temp
     else
         # Figure out type for destination
         dst_type = typeof(init)
-        dst = similar(src, dst_type, blocks * 2)
+        dst = KernelAbstractions.allocate(backend, dst_type, blocks * 2)
     end
 
     # Later the kernel will be compiled for views anyways, so use same types

test/Project.toml

@@ -2,6 +2,7 @@
 Aqua = "4c88cf16-eb10-579e-8560-4a9242c79595"
 InteractiveUtils = "b77e0a4c-d291-57a0-90e8-8db25a27a240"
 KernelAbstractions = "63c18a36-062a-441e-b654-da1e3ab1ce7c"
+MappedArrays = "dbb5928d-eab1-5f90-85c2-b9b0edb7c900"
 Pkg = "44cfe95a-1eb2-52ea-b672-e2afdf69b78f"
 Random = "9a3f8284-a2c9-5f02-9a11-845980a1fd5c"
 Test = "8dfed614-e22c-5e08-85e1-65c5234f0b40"

test/runtests.jl

@@ -2,6 +2,7 @@ import AcceleratedKernels as AK
 using KernelAbstractions
 using Test
 using Random
+using MappedArrays
 import Pkg
 
 
@@ -756,6 +757,12 @@ end
         max_tasks=16,
         min_elems=1000,
     )
+
+    # Range input
+    @test AK.sum(1234:100_000, backend) == sum(1234:100000)
+
+    # MappedArrays input
+    @test AK.sum(mappedarray(x->x*x, 1234:100_000), backend) == sum(x->x*x, 1234:100000)
 end
 
 
@@ -778,6 +785,7 @@ end
         end
     end
 
+
     # Fuzzy correctness testing
     for _ in 1:100
         for dims in 1:3
@@ -964,6 +972,12 @@ end
     v = array_from_host([Point(rand(Float32), rand(Float32)) for _ in 1:10_042])
     temp = similar(v, Tuple{Float32, Float32})
     f(v, temp)
+
+    # Range input
+    @test AK.mapreduce(x->x*x, +, 1234:100_000, backend, init=Int64(10)) == sum(x->x*x, 1234:100000, init=Int64(10))
+
+    # MappedArrays input
+    @test AK.mapreduce(x->2*x, +, mappedarray(x->x*x, 1234:100_000), backend, init=Int64(10)) == sum(x->2*x*x, 1234:100000, init=Int64(10))
 end