Test JSOBot

benchmark/Project.toml

@@ -1,6 +1,11 @@
 [deps]
+ArgParse = "c7e460c6-2fb9-53a9-8c5b-16f535851c63"
+BenchmarkTools = "6e4b80f9-dd63-53aa-95a3-0cdb28fa8baf"
+DataFrames = "a93c6f00-e57d-5684-b7b6-d8193f3e46c0"
+Git = "d7ba0133-e1db-5d97-8f8c-041e4b3a1eb2"
 GitHub = "bc5e4493-9b4d-5f90-b8aa-2b2bcaad7a26"
 IterativeSolvers = "42fd0dbc-a981-5370-80f2-aaf504508153"
+JLD2 = "033835bb-8acc-5ee8-8aae-3f567f8a3819"
 JSON = "682c06a0-de6a-54ab-a142-c8b1cf79cde6"
 Krylov = "ba0b0d4f-ebba-5204-a429-3ac8c609bfb7"
 KrylovKit = "0b1a1467-8014-51b9-945f-bf0ae24f4b77"
@@ -9,6 +14,8 @@ LinearAlgebra = "37e2e46d-f89d-539d-b4ee-838fcccc9c8e"
 LinearOperators = "5c8ed15e-5a4c-59e4-a42b-c7e8811fb125"
 MatrixMarket = "4d4711f2-db25-561a-b6b3-d35e7d4047d3"
 PkgBenchmark = "32113eaa-f34f-5b0d-bd6c-c81e245fc73d"
+Plots = "91a5bcdd-55d7-5caf-9e0b-520d859cae80"
 Printf = "de0858da-6303-5e67-8744-51eddeeeb8d7"
+SolverBenchmark = "581a75fa-a23a-52d0-a590-d6201de2218a"
 SparseArrays = "2f01184e-e22b-5df5-ae63-d93ebab69eaf"
 SuiteSparseMatrixCollection = "ac199af8-68bc-55b8-82c4-7abd6f96ed98"

benchmark/benchmarks.jl

@@ -1,47 +1,32 @@
 using BenchmarkTools
-using MatrixMarket
-
-using LinearAlgebra
-using SparseArrays
-
-using Krylov
-using LinearOperators
-using SuiteSparseMatrixCollection
+using LinearAlgebra, SparseArrays, Krylov
+using MatrixMarket, SuiteSparseMatrixCollection
 
 include("../test/get_div_grad.jl")
 
-# ufl_posdef = filter(p -> p.structure == "symmetric" && p.posDef == "yes" && p.type == "real" && p.rows ≤ 2_000, ssmc)
-ufl_posdef = filter(p -> p.structure == "symmetric" && p.posDef == "yes" && p.type == "real" && p.rows ≤ 100, ssmc)
-
-# fetch_ssmc(ufl_posdef, format="MM")
+ssmc = ssmc_db()
+ufl_posdef = ssmc[(ssmc.numerical_symmetry .== 1) .& (ssmc.positive_definite .== true) .& (ssmc.real .== true) .& (ssmc.binary .== false) .& (ssmc.nrows .≤ 500), :]
+paths = fetch_ssmc(ufl_posdef, format="MM")
 
 const SUITE = BenchmarkGroup()
 
-SUITE["CG"] = BenchmarkGroup(["CG", "SPD"])
-
+SUITE["DivGrad"] = BenchmarkGroup()
 for N in [32, 64, 128]
-  SUITE["CG"]["DivGrad N=$N"] = BenchmarkGroup()
   A = get_div_grad(N, N, N)
-  n = size(A, 1)
+  n, m = size(A)
   b = ones(n)
-  op = PreallocatedLinearOperator(A)
-  M = opEye()
-  rtol = 1.0e-6
-  SUITE["CG"]["DivGrad N=$N"]["Krylov"] = @benchmarkable cg($op, $b, M=$M, atol=0.0, rtol=$rtol, itmax=$n)
+  rtol = 1.0e-8
+  SUITE["DivGrad"]["DivGrad N=$N"] = @benchmarkable cg($A, $b, atol=0.0, rtol=$rtol, itmax=$n)
 end
 
-SUITE["CG"]["UFL"] = BenchmarkGroup()
-for matrix in ufl_posdef
-  name = matrix.name
-  A = MatrixMarket.mmread(joinpath(matrix_path(matrix, format="MM"), "$(name).mtx"))
-  if eltype(A) <: Integer
-    A = convert(SparseMatrixCSC{Float64,Int}, A)
+SUITE["UFL"] = BenchmarkGroup()
+for path in paths
+  name = split(path, '/')[end]
+  A = MatrixMarket.mmread(path * "/$name.mtx")
+  n, m = size(A)
+  if eltype(A) == Float64
+    b = ones(n)
+    rtol = 1.0e-8
+    SUITE["UFL"][name] = @benchmarkable cg($A, $b, atol=0.0, rtol=$rtol, itmax=$n)
   end
-  n = size(A, 1)
-  b = ones(eltype(A), n)
-  op = PreallocatedLinearOperator(A)
-  M = opEye()
-  rtol = 1.0e-6
-  SUITE["CG"]["UFL"][name] = BenchmarkGroup()
-  SUITE["CG"]["UFL"][name]["Krylov"] = @benchmarkable cg($op, $b, M=$M, atol=0.0, rtol=$rtol, itmax=$n)
 end