Merge pull request #468 from SciML/krylov

Avoid constructing any Krylov cache with default dense

Author: ChrisRackauckas

Project.toml

@@ -1,7 +1,7 @@
 name = "LinearSolve"
 uuid = "7ed4a6bd-45f5-4d41-b270-4a48e9bafcae"
 authors = ["SciML"]
-version = "2.23.0"
+version = "2.23.1"
 
 [deps]
 ArrayInterface = "4fba245c-0d91-5ea0-9b3e-6abc04ee57a9"

src/default.jl

@@ -121,7 +121,7 @@ function defaultalg(A::Nothing, b::GPUArraysCore.AnyGPUArray, assump::OperatorAs
 end
 
 # Ambiguity handling
-function defaultalg(A::GPUArraysCore.AnyGPUArray, b::GPUArraysCore.AbstractGPUArray,
+function defaultalg(A::GPUArraysCore.AnyGPUArray, b::GPUArraysCore.AnyGPUArray,
     assump::OperatorAssumptions{Bool})
     if assump.condition === OperatorCondition.IllConditioned || !assump.issq
         DefaultLinearSolver(DefaultAlgorithmChoice.QRFactorization)
@@ -152,7 +152,7 @@ function defaultalg(A, b, assump::OperatorAssumptions{Bool})
         # Special case on Arrays: avoid BLAS for RecursiveFactorization.jl when
         # it makes sense according to the benchmarks, which is dependent on
         # whether MKL or OpenBLAS is being used
-        if (A === nothing && !(b isa GPUArraysCore.AbstractGPUArray)) || A isa Matrix
+        if (A === nothing && !(b isa GPUArraysCore.AnyGPUArray)) || A isa Matrix
             if (A === nothing ||
                 eltype(A) <: BLASELTYPES) &&
                ArrayInterface.can_setindex(b) &&
@@ -296,7 +296,7 @@ cache.cacheval = NamedTuple(LUFactorization = cache of LUFactorization, ...)
     abstol, reltol,
     verbose::Bool, assump::OperatorAssumptions)
     caches = map(first.(EnumX.symbol_map(DefaultAlgorithmChoice.T))) do alg
-        if alg === :KrylovJL_GMRES
+        if alg === :KrylovJL_GMRES || alg === :KrylovJL_CRAIGMR || alg === :KrylovJL_LSMR
             quote
                 if A isa Matrix || A isa SparseMatrixCSC
                     nothing

test/default_algs.jl

@@ -77,8 +77,8 @@ b = rand(m)
 x = rand(n)
 f = (du, u, p, t) -> mul!(du, A, u)
 fadj = (du, u, p, t) -> mul!(du, A', u)
-fo = FunctionOperator(f, x, b; op_adjoint = fadj)
-prob = LinearProblem(fo, b)
+funcop = FunctionOperator(f, x, b; op_adjoint = fadj)
+prob = LinearProblem(funcop, b)
 sol1 = solve(prob)
 sol2 = solve(prob, LinearSolve.KrylovJL_GMRES())
 @test sol1.u == sol2.u
@@ -89,8 +89,8 @@ b = rand(m)
 x = rand(n)
 f = (du, u, p, t) -> mul!(du, A, u)
 fadj = (du, u, p, t) -> mul!(du, A', u)
-fo = FunctionOperator(f, x, b; op_adjoint = fadj)
-prob = LinearProblem(fo, b)
+funcop = FunctionOperator(f, x, b; op_adjoint = fadj)
+prob = LinearProblem(funcop, b)
 sol1 = solve(prob)
 sol2 = solve(prob, LinearSolve.KrylovJL_LSMR())
 @test sol1.u == sol2.u
@@ -101,8 +101,8 @@ b = rand(m)
 x = rand(n)
 f = (du, u, p, t) -> mul!(du, A, u)
 fadj = (du, u, p, t) -> mul!(du, A', u)
-fo = FunctionOperator(f, x, b; op_adjoint = fadj)
-prob = LinearProblem(fo, b)
+funcop = FunctionOperator(f, x, b; op_adjoint = fadj)
+prob = LinearProblem(funcop, b)
 sol1 = solve(prob)
 sol2 = solve(prob, LinearSolve.KrylovJL_CRAIGMR())
 @test sol1.u == sol2.u