exporting mul!

src/ParallelMergeCSR.jl

@@ -9,6 +9,8 @@ using SparseArrays: AbstractSparseMatrixCSC,
                     getcolptr
 
 
+export mul!
+
 include("parallel_csr_mv.jl")
 include("parallel_csc_mv.jl")
 

src/parallel_csc_mv.jl

@@ -25,7 +25,7 @@ function atomic_add!(a::AbstractVector{T},b::T) where T <: Complex
     end
 end
 
-function SparseArrays.mul!(C::StridedVecOrMat, A::AbstractSparseMatrixCSC, B::DenseInputVecOrMat, α::Number, β::Number)
+function mul!(C::StridedVecOrMat, A::AbstractSparseMatrixCSC, B::DenseInputVecOrMat, α::Number, β::Number)
     size(A, 2) == size(B, 1) || throw(DimensionMismatch())
     size(A, 1) == size(C, 1) || throw(DimensionMismatch())
     size(B, 2) == size(C, 2) || throw(DimensionMismatch())

src/parallel_csr_mv.jl

@@ -121,7 +121,7 @@ end
 # C = transpose(A)B + Cβ
 # C = xABα + Cβ
 for (T, t) in ((Adjoint, adjoint), (Transpose, transpose))
-    @eval function SparseArrays.mul!(C::StridedVecOrMat, xA::$T{<:Any,<:AbstractSparseMatrixCSC}, B::DenseInputVecOrMat, α::Number, β::Number)
+    @eval function mul!(C::StridedVecOrMat, xA::$T{<:Any,<:AbstractSparseMatrixCSC}, B::DenseInputVecOrMat, α::Number, β::Number)
         # obtains the original matrix underneath the "lazy wrapper"
         A = xA.parent
         size(A, 2) == size(C, 1) || throw(DimensionMismatch())

test/merge_csr_mv!.jl

@@ -1,6 +1,6 @@
 using Test
 using ParallelMergeCSR: merge_csr_mv!, Range
-using SparseArrays
+using SparseArrays: sprand, sparse
 
 
 @testset "Range" begin
@@ -32,7 +32,7 @@ end
 
         merge_csr_mv!(0.3, A, x, y, transpose)
 
-        @test Matrix(A) * x * 0.3 == y
+        @test A * x * 0.3 == y
     end
 
     @testset "Singleton (Complex)" begin
@@ -50,7 +50,7 @@ end
     @testset "Single row (Real)" begin
 
         # 10 x 1 converted to 1 x 10
-        A = 10.0 * SparseArrays.sprand(10, 1, 0.3)
+        A = 10.0 * sprand(10, 1, 0.3)
 
         # x needs to be 10 x 1 
         x = rand(size(A, 1))
@@ -59,21 +59,21 @@ end
 
         merge_csr_mv!(1.1, A, x, y, transpose)
 
-        @test (transpose(Matrix(A)) * x) * 1.1 ≈ y
+        @test (transpose(A) * x) * 1.1 ≈ y
     end
 
     @testset "Single row (Complex)" begin
 
         # 10 x 1 is treated as 1 x 10 inside merge_csr_mv!
-        A = SparseArrays.sprand(Complex{Float64}, 10, 1, 0.3)
+        A = sprand(Complex{Float64}, 10, 1, 0.3)
 
         x = rand(eltype(A), size(A, 1))
 
         y = zeros(eltype(A), 1)
 
         merge_csr_mv!(1.1, A, x, y, transpose)
 
-        @test (transpose(Matrix(A)) * x) * 1.1 ≈ y
+        @test (transpose(A) * x) * 1.1 ≈ y
     end
 
 
@@ -92,7 +92,7 @@ end
 
 # test fails, y seems to have lots of zero-entries
 @testset "Square (Real)" begin
-    A = SparseArrays.sprand(10,10,0.3)
+    A = sprand(10,10,0.3)
 
     # 10 x 1
     x = rand(10)
@@ -102,20 +102,20 @@ end
 
     merge_csr_mv!(1.1, A, x, y, adjoint)
 
-    @test (adjoint(Matrix(A)) * x) * 1.1 ≈ y
+    @test (adjoint(A) * x) * 1.1 ≈ y
 
 end
 
 @testset "Square (Complex)" begin
-    A = SparseArrays.sprand(Complex{Float64}, 10, 10, 0.3)
+    A = sprand(Complex{Float64}, 10, 10, 0.3)
 
     x = 10 * rand(Complex{Float64}, 10)
 
     y = zeros(eltype(A), size(x))
 
     merge_csr_mv!(1.1, A, x, y, adjoint)
 
-    @test (adjoint(Matrix(A)) * x) * 1.1 ≈ y
+    @test (adjoint(A) * x) * 1.1 ≈ y
 end
 
 @testset "4x6 (Real)" begin
@@ -138,7 +138,7 @@ end
     merge_csr_mv!(2.0, A, x, y, adjoint)
 
 
-    @test Matrix(adjoint(m)) * x * 2.0 == y
+    @test adjoint(m) * x * 2.0 == y
 end
 
 @testset "4 x 6 (Complex)" begin
@@ -160,13 +160,13 @@ end
     # multiply
     merge_csr_mv!(2.0, A, x, y, adjoint)
 
-    @test Matrix(adjoint(m)) * x * 2.0 == y
+    @test adjoint(m) * x * 2.0 == y
 
 end
 
 @testset "100x100 (Real)" begin
     # create matrix
-    A = SparseArrays.sprand(100, 100, 0.3)
+    A = sprand(100, 100, 0.3)
 
     # create vector
     x = rand(100)
@@ -176,12 +176,12 @@ end
 
     merge_csr_mv!(3.0, A, x, y, transpose)
 
-    @test transpose(Matrix(A)) * x * 3 ≈ y
+    @test transpose(A) * x * 3 ≈ y
 end
 
 @testset "100x100 (Complex)" begin
     # create matrix
-    A = SparseArrays.sprand(Complex{Float64}, 100, 100, 0.3)
+    A = sprand(Complex{Float64}, 100, 100, 0.3)
 
     # create vector
     x = rand(Complex{Float64}, 100)
@@ -191,7 +191,7 @@ end
 
     merge_csr_mv!(3.0, A, x, y, transpose)
 
-    @test transpose(Matrix(A)) * x * 3 ≈ y
+    @test transpose(A) * x * 3 ≈ y
 end
 
 #=
@@ -215,7 +215,7 @@ end
         merge_csr_mv!(α, A, col, Y_view, transpose)
     end
 
-    @test transpose(Matrix(A)) * X * 9.2 ≈ Y 
+    @test transpose(A) * X * 9.2 ≈ Y 
 
 end
 
@@ -236,6 +236,6 @@ end
         merge_csr_mv!(α, A, col, Y_view, adjoint)
     end
 
-    @test adjoint(Matrix(A)) * X * 9.2 ≈ Y 
+    @test adjoint(A) * X * 9.2 ≈ Y 
 
 end

test/mul!.jl

@@ -14,9 +14,7 @@ using SparseArrays
     C_copy = deepcopy(C)
     β = 3.9
 
-    SparseArrays.mul!(C, A, B, α, β)
-
-    @test C ≈ Matrix(A) * B * α + C_copy * β
+    @test ParallelMergeCSR.mul!(C, A, B, α, β) ≈ SparseArrays.mul!(C_copy, A, B, α, β)
 
 end
 
@@ -31,9 +29,7 @@ end
     C_copy = deepcopy(C)
     β = 1.2
 
-    SparseArrays.mul!(C, A, B, α, β)
-
-    @test C ≈ Matrix(A) * B * α + C_copy * β
+    @test ParallelMergeCSR.mul!(C, A, B, α, β) ≈ SparseArrays.mul!(C_copy, A, B, α, β)
 
 end
 # trigger merge_csr_mv!
@@ -49,10 +45,8 @@ end
     C_copy = deepcopy(C)
     β = 1.0
 
-    SparseArrays.mul!(C, A, B, α, β)
-
     # right hand side is correct, left hand side is problematic
-    @test C ≈ Matrix(A) * B * α + C_copy * β
+    @test ParallelMergeCSR.mul!(C, A, B, α, β) ≈ SparseArrays.mul!(C_copy, A, B, α, β)
 
 end
 
@@ -67,9 +61,8 @@ end
     C_copy = deepcopy(C)
     β = 6.11 + 9.2im
 
-    SparseArrays.mul!(C, A, B, α, β)
 
-    @test C ≈ Matrix(A) * B * α + C_copy * β
+    @test ParallelMergeCSR.mul!(C, A, B, α, β) ≈ SparseArrays.mul!(C_copy, A, B, α, β)
 end
 
 @testset "Transpose Rectangular Real" begin
@@ -83,9 +76,7 @@ end
     C_copy = deepcopy(C)
     β = 1.0
 
-    SparseArrays.mul!(C, A, B, α, β)
-
-    @test C ≈ Matrix(A) * B * α + C_copy * β
+    @test ParallelMergeCSR.mul!(C, A, B, α, β) ≈ SparseArrays.mul!(C_copy, A, B, α, β)
 
 end
 
@@ -100,9 +91,7 @@ end
     C_copy = deepcopy(C)
     β = 1.0
 
-    SparseArrays.mul!(C, A, B, α, β)
-
-    @test C ≈ Matrix(A) * B * α + C_copy * β
+    @test ParallelMergeCSR.mul!(C, A, B, α, β) ≈ SparseArrays.mul!(C_copy, A, B, α, β)
 
 end
 
@@ -117,8 +106,7 @@ end
     C_copy = deepcopy(C)
     β = 0.3
 
-    SparseArrays.mul!(C, A, B, α, β)
-    @test C ≈ Matrix(A) * B * α + C_copy * β
+    @test ParallelMergeCSR.mul!(C, A, B, α, β) ≈ SparseArrays.mul!(C_copy, A, B, α, β)
 
 end
 
@@ -133,8 +121,7 @@ end
     C_copy = deepcopy(C)
     β = 0.3
 
-    SparseArrays.mul!(C, A, B, α, β)
-    @test C ≈ Matrix(A) * B * α + C_copy * β
+    @test ParallelMergeCSR.mul!(C, A, B, α, β) ≈ SparseArrays.mul!(C_copy, A, B, α, β)
 
 end
 
@@ -148,9 +135,7 @@ end
     C_copy = deepcopy(C)
     β = 5.2
 
-    SparseArrays.mul!(C, A, B, α, β)
-
-    @test C ≈ Matrix(A) * B * α + C_copy * β
+    @test ParallelMergeCSR.mul!(C, A, B, α, β) ≈ SparseArrays.mul!(C_copy, A, B, α, β)
 end
 
 @testset "Matrix x Vector (Real)" begin
@@ -163,9 +148,7 @@ end
     C_copy = deepcopy(C)
     β = 5.2
 
-    SparseArrays.mul!(C, A, B, α, β)
-
-    @test C ≈ Matrix(A) * B * α + C_copy * β
+    @test ParallelMergeCSR.mul!(C, A, B, α, β) ≈ SparseArrays.mul!(C_copy, A, B, α, β)
 
 end
 
@@ -179,8 +162,6 @@ end
     C_copy = deepcopy(C)
     β = 2.1+0.1im
 
-    SparseArrays.mul!(C, A, B, α, β)
-
-    @test C ≈ Matrix(A) * B * α + C_copy * β
+    @test ParallelMergeCSR.mul!(C, A, B, α, β) ≈ SparseArrays.mul!(C_copy, A, B, α, β)
 
 end