Test master

[ChrisRackauckas]

No description provided.

[ChrisRackauckas]

@wsmoses caught this in a downstream test today.

Enzyme Derivative Rules: Error During Test at /home/runner/.julia/packages/SafeTestsets/raUNr/src/SafeTestsets.jl:30
  Got exception outside of a @test
  LoadError: Constant memory is stored (or returned) to a differentiable variable.
  As a result, Enzyme cannot provably ensure correctness and throws this error.
  This might be due to the use of a constant variable as temporary storage for active memory (https://enzyme.mit.edu/julia/stable/faq/#Activity-of-temporary-storage).
  If Enzyme should be able to prove this use non-differentable, open an issue!
  To work around this issue, either:
   a) rewrite this variable to not be conditionally active (fastest, but requires a code change), or
   b) set Enzyme.API.runtimeActivity!(true) immediately after loading Enzyme (which maintains correctness, but may slightly reduce performance).
  Mismatched activity for:   %.fca.0.insert = insertvalue { {} addrspace(10)*, {} addrspace(10)* } poison, {} addrspace(10)* %1, 0, !dbg !19 const val: {} addrspace(10)* %1
   value=Unknown object of type Matrix{Float64}
   llvalue={} addrspace(10)* %1
  
  Stacktrace:
   [1] LinearProblem
     @ ~/.julia/packages/SciMLBase/FfEKU/src/problems/linear_problems.jl:0
  
  Stacktrace:
    [1] LinearProblem
      @ ~/.julia/packages/SciMLBase/FfEKU/src/problems/linear_problems.jl:0 [inlined]

Is there a change to Enzyme I can look at to help me orient around what changed?

[wsmoses]

we realized we we'rent throwing errors for llvm insert instructions that were runtime active [aka storing, via an insert instruction, a contant pointer, into a differentiable data structure]

[ChrisRackauckas]

So it should've errored before and it just didn't? 😅

[ChrisRackauckas]

I guess like the others, LinearProblem might need to be mutable?

[ChrisRackauckas]

Bump, the question is:

I guess like the others, LinearProblem might need to be mutable?

[wsmoses]

@ChrisRackauckas can you mark

LinearSolve.jl/test/enzyme.jl

Line 178 in 33911f6

A = rand(n, n);

as const A

The issue yells that

LinearSolve.jl/test/enzyme.jl

Line 187 in 33911f6

prob = LinearProblem(A, b)

is type unstable, and uses a constant variable (A) being stored into a differentiable LinearProblem. Marking it as const/type stable may optimize out the linear problem, solving the issue

[ChrisRackauckas]

Fixing the type stability and marking things constant didn't seem to effect it.