Update Project.toml

Project.toml

@@ -1,6 +1,6 @@
 name = "Strided"
 uuid = "5e0ebb24-38b0-5f93-81fe-25c709ecae67"
-version = "2.3.2"
+version = "2.3.3"
 authors = ["Lukas Devos <lukas.devos@ugent.be>", "Maarten Van Damme <maartenvd1994@gmail.com>", "Jutho Haegeman <jutho.haegeman@ugent.be>"]
 
 [deps]
@@ -28,7 +28,7 @@ JLArrays = "0.3.1"
 GPUArrays = "11.4.1"
 LinearAlgebra = "1.6"
 Random = "1.6"
-StridedViews = "0.4.5"
+StridedViews = "0.4.6"
 Test = "1.6"
 TupleTools = "1.6"
 julia = "1.6"

ext/StridedGPUArraysExt.jl

@@ -12,4 +12,11 @@ function Base.Broadcast.BroadcastStyle(gpu_sv::StridedView{T, N, TA}) where {T,
     return typeof(raw_style)(Val(N)) # sets the dimensionality correctly
 end
 
+function Base.copy!(dst::AbstractArray{TD, ND}, src::StridedView{TS, NS, TAS, FS}) where {TD <: Number, ND, TS <: Number, NS, TAS <: AbstractGPUArray{TS}, FS <: ALL_FS}
+    bc_style = Base.Broadcast.BroadcastStyle(TAS)
+    bc = Base.Broadcast.Broadcasted(bc_style, identity, (src,), axes(dst))
+    GPUArrays._copyto!(dst, bc)
+    return dst
+end
+
 end

ext/StridedJLArraysExt.jl

@@ -13,11 +13,4 @@ function Base.copy!(dst::StridedView{TD, ND, TAD, FD}, src::StridedView{TS, NS,
     return dst
 end
 
-function Base.copy!(dst::AbstractArray{TD, ND}, src::StridedView{TS, NS, TAS, FS}) where {TD <: Number, ND, TS <: Number, NS, TAS <: JLArray{TS}, FS <: ALL_FS}
-    bc_style = Base.Broadcast.BroadcastStyle(TAS)
-    bc = Base.Broadcast.Broadcasted(bc_style, identity, (src,), axes(dst))
-    GPUArrays._copyto!(dst, bc)
-    return dst
-end
-
 end

test/amd.jl

@@ -7,12 +7,15 @@ for T in (Float32, Float64, Complex{Float32}, Complex{Float64})
                 A1 = ROCMatrix(randn(T, (m1, m2)))
             end
             A2 = similar(A1)
+            zA1 = ROCMatrix(f1(zeros(T, (m1, m2))))
+            zA2 = ROCMatrix(f2(zeros(T, (m1, m2))))
             A1c = copy(A1)
             A2c = copy(A2)
             B1 = f1(StridedView(A1c))
             B2 = f2(StridedView(A2c))
             axes(f1(A1)) == axes(f2(A2)) || continue
             @test collect(ROCMatrix(copy!(f2(A2), f1(A1)))) == AMDGPU.Adapt.adapt(Vector{T}, copy!(B2, B1))
+            @test copy!(zA1, f1(A1)) == copy!(zA2, B1)
         end
     end
 end

test/cuda.jl

@@ -3,12 +3,15 @@ for T in (Float32, Float64, Complex{Float32}, Complex{Float64})
         for m1 in (0, 16, 32), m2 in (0, 16, 32)
             A1 = CUDA.randn(T, (m1, m2))
             A2 = similar(A1)
+            zA1 = CuMatrix(f1(zeros(T, (m1, m2))))
+            zA2 = CuMatrix(f2(zeros(T, (m1, m2))))
             A1c = copy(A1)
             A2c = copy(A2)
             B1 = f1(StridedView(A1c))
             B2 = f2(StridedView(A2c))
             axes(f1(A1)) == axes(f2(A2)) || continue
             @test collect(CuMatrix(copy!(f2(A2), f1(A1)))) == CUDA.Adapt.adapt(Vector{T}, copy!(B2, B1))
+            @test copy!(zA1, f1(A1)) == copy!(zA2, B1)
         end
     end
 end

test/jlarrays.jl

@@ -3,12 +3,15 @@ for T in (Float32, Float64, Complex{Float32}, Complex{Float64})
         for m1 in (0, 16, 32), m2 in (0, 16, 32)
             A1 = JLArray(randn(T, (m1, m2)))
             A2 = similar(A1)
+            zA1 = JLArray(f1(zeros(T, (m1, m2))))
+            zA2 = JLArray(f2(zeros(T, (m1, m2))))
             A1c = copy(A1)
             A2c = copy(A2)
             B1 = f1(StridedView(A1c))
             B2 = f2(StridedView(A2c))
             axes(f1(A1)) == axes(f2(A2)) || continue
             @test collect(Matrix(copy!(f2(A2), f1(A1)))) == JLArrays.Adapt.adapt(Vector{T}, copy!(B2, B1))
+            @test copy!(zA1, f1(A1)) == copy!(zA2, B1)
         end
     end
 end