Add missing tests for given initial guess

Author: haampie

test/bicgstabl.jl

@@ -18,11 +18,18 @@ n = 20
         x1, his1 = bicgstabl(A, b, l, max_mv_products = 100, log = true, tol = tol)
         @test isa(his1, ConvergenceHistory)
         @test norm(A * x1 - b) / norm(b) ≤ tol
+        
+        # With an initial guess
+        x_guess = rand(T, n)
+        x2, his2 = bicgstabl!(x_guess, A, b, l, max_mv_products = 100, log = true, tol = tol)
+        @test isa(his2, ConvergenceHistory)
+        @test x2 == x_guess
+        @test norm(A * x2 - b) / norm(b) ≤ tol
 
         # Do an exact LU decomp of a nearby matrix
         F = lufact(A + rand(T, n, n))
-        x2, his2 = bicgstabl(A, b, Pl = F, l, max_mv_products = 100, log = true, tol = tol)
-        @test norm(A * x2 - b) / norm(b) ≤ tol
+        x3, his3 = bicgstabl(A, b, Pl = F, l, max_mv_products = 100, log = true, tol = tol)
+        @test norm(A * x3 - b) / norm(b) ≤ tol
     end
 end
 end

test/chebyshev.jl

@@ -22,20 +22,37 @@ srand(1234321)
 @testset "Matrix{$T}" for T in (Float32, Float64, Complex64, Complex128)
     A = randSPD(T, n)
     b = rand(T, n)
-    λ_min, λ_max = approx_eigenvalue_bounds(A)
     tol = √(eps(real(T)))
 
-    x, history = chebyshev(A, b, λ_min, λ_max, tol=tol, maxiter=10n, log=true)
-    @test isa(history, ConvergenceHistory)
-    @test history.isconverged
-    @test norm(A * x - b) / norm(b) ≤ tol
+    # Without a preconditioner
+    begin
+        λ_min, λ_max = approx_eigenvalue_bounds(A)
+        x0 = rand(n)
+        x, history = chebyshev(A, b, λ_min, λ_max, tol=tol, maxiter=10n, log=true)
+        @test isa(history, ConvergenceHistory)
+        @test history.isconverged
+        @test norm(A * x - b) / norm(b) ≤ tol
+    end
 
-    # Preconditioned solve
-    B = randSPD(T, n)
-    B_fact = cholfact!(B)
-    λ_min, λ_max = approx_eigenvalue_bounds(B_fact \ A)
-    x, history = chebyshev(A, b, λ_min, λ_max, Pl = B_fact, tol=tol, maxiter=10n, log=true)
-    @test history.isconverged
-    @test norm(A * x - b) / norm(b) ≤ tol
+    # With an initial guess
+    begin
+        λ_min, λ_max = approx_eigenvalue_bounds(A)
+        x0 = rand(T, n)
+        x, history = chebyshev!(x0, A, b, λ_min, λ_max, tol=tol, maxiter=10n, log=true)
+        @test isa(history, ConvergenceHistory)
+        @test history.isconverged
+        @test x == x0
+        @test norm(A * x - b) / norm(b) ≤ tol
+    end
+
+    # With a preconditioner
+    begin
+        B = randSPD(T, n)
+        B_fact = cholfact!(B)
+        λ_min, λ_max = approx_eigenvalue_bounds(B_fact \ A)
+        x, history = chebyshev(A, b, λ_min, λ_max, Pl = B_fact, tol=tol, maxiter=10n, log=true)
+        @test history.isconverged
+        @test norm(A * x - b) / norm(b) ≤ tol
+    end
 end
 end

test/minres.jl

@@ -27,12 +27,16 @@ n = 15
 @testset "Hermitian Matrix{$T}" for T in (Float32, Float64, Complex64, Complex128)
     A, x, b = hermitian_problem(T, n)
     tol = sqrt(eps(real(T)))
+    x0 = rand(T, n)
 
-    x_approx, hist = minres(A, b, maxiter = 10n, tol = tol, log = true)
+    x1, hist1 = minres(A, b, maxiter = 10n, tol = tol, log = true)
+    x2, hist2 = minres!(x0, A, b, maxiter = 10n, tol = tol, log = true)
 
-    @test isa(hist, ConvergenceHistory)
-    @test norm(b - A * x_approx) / norm(b) ≤ tol
-    @test hist.isconverged
+    @test isa(hist1, ConvergenceHistory)
+    @test norm(b - A * x1) / norm(b) ≤ tol
+    @test hist1.isconverged
+    @test norm(b - A * x2) / norm(b) ≤ tol
+    @test x2 == x0
 end
 
 @testset "Skew-Hermitian Matrix{$T}" for T in (Float32, Float64, Complex64, Complex128)