Handle CuSparse CSC case

Author: ChrisRackauckas

src/differentiation/compute_jacobian_ad.jl

@@ -294,6 +294,8 @@ function forwarddiff_color_jacobian!(J::AbstractMatrix{<:Number},
     sparsity = jac_cache.sparsity
     chunksize = jac_cache.chunksize
     color_i = 1
+    adaptedcolorvec = adapt(__parameterless_type(typeof(dx)),colorvec)
+
     maxcolor = maximum(colorvec)
 
     if J isa AbstractSparseMatrix
@@ -357,7 +359,15 @@ function forwarddiff_color_jacobian!(J::AbstractMatrix{<:Number},
                     += means requires a zero'd out start
                     =#
                     if J isa AbstractSparseMatrix
-                        @. void_setindex!(Ref(nonzeros(J)),getindex(Ref(nonzeros(J)),rows_index) + (getindex(Ref(colorvec),cols_index) == color_i) * getindex(Ref(vecdx),rows_index),rows_index)
+                        if J isa SparseMatrixCSC
+                            @. void_setindex!(Ref(nonzeros(J)),getindex(Ref(nonzeros(J)),rows_index) + (getindex(Ref(adaptedcolorvec),cols_index) == color_i) * getindex(Ref(vecdx),rows_index),rows_index)
+                        else
+                            nzval = @view nonzeros(J)[rows_index]
+                            cv = @view adaptedcolorvec[cols_index]
+                            vdx = @view dx[rows_index]
+                            tmp = cv .== color_i
+                            nzval .+= tmp .* vdx
+                        end
                     else
                         @. void_setindex!(Ref(J),getindex(Ref(J),rows_index, cols_index) + (getindex(Ref(colorvec),cols_index) == color_i) * getindex(Ref(vecdx),rows_index),rows_index, cols_index)
                     end

test/test_gpu_ad.jl

@@ -1,18 +1,25 @@
 using SparseDiffTools, CUDA, Test, LinearAlgebra
 using ArrayInterface: allowed_getindex, allowed_setindex!
+using SparseArrays
+
 function f(dx,x)
     dx[2:end-1] = x[1:end-2] - 2x[2:end-1] + x[3:end]
     allowed_setindex!(dx,-2allowed_getindex(x,1) + allowed_getindex(x,2),1)
     allowed_setindex!(dx,-2allowed_getindex(x,30) + allowed_getindex(x,29),30)
     nothing
 end
-
+x = rand(4)
 _J1 = similar(rand(30,30))
 _denseJ1 = cu(collect(_J1))
 x = cu(rand(30))
 CUDA.allowscalar(false)
-forwarddiff_color_jacobian!(_denseJ1, f, x)
-@test_broken forwarddiff_color_jacobian!(_denseJ1, f, x, sparsity = _J1) isa Nothing
-@test_broken forwarddiff_color_jacobian!(_denseJ1, f, x, colorvec = repeat(1:3,10), sparsity = _J1) isa Nothing
+_J2 = sparse(forwarddiff_color_jacobian!(_denseJ1, f, x))
+out = copy(_J2)
+forwarddiff_color_jacobian!(out, f, x, colorvec = repeat(1:3,10), sparsity = _J2)
+
+@test_broken forwarddiff_color_jacobian!(_denseJ1, f, x, sparsity = cu(_J1)) isa Nothing
+@test_broken forwarddiff_color_jacobian!(_denseJ1, f, x, colorvec = repeat(1:3,10), sparsity = cu(_J1)) isa Nothing
 _Jt = similar(Tridiagonal(_J1))
 @test_broken forwarddiff_color_jacobian!(_denseJ1, f, x, colorvec = repeat(1:3,10), sparsity = _Jt) isa Nothing
+_Jt2 = similar(Tridiagonal(cu(_J1)))
+@test_broken forwarddiff_color_jacobian!(_denseJ1, f, x, colorvec = repeat(1:3,10), sparsity = _Jt2) isa Nothing