Fix CI: Exclude Enzyme tests from Julia pre

.github/workflows/Tests.yml

@@ -24,6 +24,10 @@ jobs:
           - "pre"
         group:
           - Core
+          - Enzyme
+        exclude:
+          - version: "pre"
+            group: Enzyme
     uses: "SciML/.github/.github/workflows/tests.yml@v1"
     with:
       julia-version: "${{ matrix.version }}"

ext/FastPowerEnzymeExt.jl

@@ -5,9 +5,7 @@ import FastPower: fastpower
 using Enzyme
 using Enzyme.EnzymeRules: FwdConfig
 
-Enzyme.EnzymeRules.@easy_rule(
-    FastPower.fastpower(x, y),
-    ( y * fastpower(x, y - 1), Ω * log(x) )
-)
+Enzyme.EnzymeRules.@easy_rule(FastPower.fastpower(x, y),
+    (y * fastpower(x, y - 1), Ω * log(x)))
 
 end

test/runtests.jl

@@ -1,58 +1,67 @@
 using FastPower
 using FastPower: fastlog2, fastpower
-using Enzyme, EnzymeTestUtils
-using ForwardDiff, ReverseDiff, Tracker, Mooncake
 using Test
 
-@testset "Fast log2" begin
-    for x in 0.001:0.001:1.2 # (0, 1+something] is the domain which a controller uses
-        @test log2(x)≈fastlog2(Float32(x)) atol=1e-3
+const GROUP = get(ENV, "GROUP", "All")
+
+if GROUP == "All" || GROUP == "Core"
+    @testset "Fast log2" begin
+        for x in 0.001:0.001:1.2 # (0, 1+something] is the domain which a controller uses
+            @test log2(x)≈fastlog2(Float32(x)) atol=1e-3
+        end
     end
-end
 
-@testset "Fast pow" begin
-    @test fastpower(1, 1) isa Float64
-    @test fastpower(1.0, 1.0) isa Float64
-    errors = [abs(^(x, y) - fastpower(x, y)) for x in 0.001:0.001:1, y in 0.08:0.001:0.5]
-    @test maximum(errors) < 1e-4
+    @testset "Fast pow" begin
+        @test fastpower(1, 1) isa Float64
+        @test fastpower(1.0, 1.0) isa Float64
+        errors = [abs(^(x, y) - fastpower(x, y)) for x in 0.001:0.001:1, y in 0.08:0.001:0.5]
+        @test maximum(errors) < 1e-4
 
-    errors = [abs(^(x, y) - fastpower(x, y)) for x in 0.001:0.001:1, y in 0.08:0.001:1000.0]
-    @test maximum(errors) < 1e-3
+        errors = [abs(^(x, y) - fastpower(x, y)) for x in 0.001:0.001:1, y in 0.08:0.001:1000.0]
+        @test maximum(errors) < 1e-3
 
-    errors = [abs(^(x, y) - fastpower(x, y)) for x in 0.001:0.001:100, y in 0.08:0.001:1.0]
-    @test maximum(errors) < 1e-2
-end
+        errors = [abs(^(x, y) - fastpower(x, y)) for x in 0.001:0.001:100, y in 0.08:0.001:1.0]
+        @test maximum(errors) < 1e-2
+    end
+
+    using ForwardDiff, ReverseDiff, Tracker, Mooncake
 
-@testset "Fast pow - Enzyme forward rule" begin
-    @testset for RT in (Duplicated, DuplicatedNoNeed),
-        Tx in (Const, Duplicated),
-        Ty in (Const, Duplicated)
-        x = 1.0
-        y = 0.5
-        test_forward(fastpower, RT, (x, Tx), (y, Ty), atol = 1e-4, rtol = 1e-3)
+    function mooncake_derivative(f, x)
+        Mooncake.value_and_gradient!!(Mooncake.build_rrule(f, x), f, x)[2][2]
+    end
+    @testset "Fast pow - Other AD Engines" begin
+        x = 1.5123233245141
+        y = 0.22352354326
+        @test ForwardDiff.derivative(x -> fastpower(x, x + y), x) ≈
+              ForwardDiff.derivative(x -> ^(x, x + y), x)
+        @test Tracker.gradient(x -> fastpower(x, x + y), x)[1] ≈
+              Tracker.gradient(x -> ^(x, x + y), x)[1]
+        @test ReverseDiff.gradient(x -> fastpower(x[1], x[1] + y), [x])[1] ≈
+              ReverseDiff.gradient(x -> ^(x[1], x[1] + y), [x])[1]
+        @test mooncake_derivative(x -> fastpower(x, x + y), x) ≈
+              mooncake_derivative(x -> ^(x, x + y), x)
     end
 end
 
-@testset "Fast pow - Enzyme reverse rule" begin
-    @testset for RT in (Active,), Tx in (Active, Const), Ty in (Active, Const)
-        x = 1.0
-        y = 0.5
-        test_reverse(fastpower, RT, (x, Tx), (y, Ty), atol = 1e-4, rtol = 1e-3)
+if GROUP == "All" || GROUP == "Enzyme"
+    using Enzyme, EnzymeTestUtils
+
+    @testset "Fast pow - Enzyme forward rule" begin
+        @testset for RT in (Duplicated, DuplicatedNoNeed),
+            Tx in (Const, Duplicated),
+            Ty in (Const, Duplicated)
+
+            x = 1.0
+            y = 0.5
+            test_forward(fastpower, RT, (x, Tx), (y, Ty), atol = 1e-4, rtol = 1e-3)
+        end
     end
-end
 
-function mooncake_derivative(f, x)
-    Mooncake.value_and_gradient!!(Mooncake.build_rrule(f, x), f, x)[2][2]
-end
-@testset "Fast pow - Other AD Engines" begin
-    x = 1.5123233245141
-    y = 0.22352354326
-    @test ForwardDiff.derivative(x -> fastpower(x, x + y), x) ≈
-          ForwardDiff.derivative(x -> ^(x, x + y), x)
-    @test Tracker.gradient(x -> fastpower(x, x + y), x)[1] ≈
-          Tracker.gradient(x -> ^(x, x + y), x)[1]
-    @test ReverseDiff.gradient(x -> fastpower(x[1], x[1] + y), [x])[1] ≈
-          ReverseDiff.gradient(x -> ^(x[1], x[1] + y), [x])[1]
-    @test mooncake_derivative(x -> fastpower(x, x + y), x) ≈
-          mooncake_derivative(x -> ^(x, x + y), x)
+    @testset "Fast pow - Enzyme reverse rule" begin
+        @testset for RT in (Active,), Tx in (Active, Const), Ty in (Active, Const)
+            x = 1.0
+            y = 0.5
+            test_reverse(fastpower, RT, (x, Tx), (y, Ty), atol = 1e-4, rtol = 1e-3)
+        end
+    end
 end