"Enzyme: The original primal code hits this error condition, thus differentiating it does not make sense" very unhelpful

[vchuravy]

I have never felt so gaslit by an exception: What error, why? how?

ERROR: LoadError: Enzyme execution failed.
Enzyme: The original primal code hits this error condition, thus differentiating it does not make sense

Stacktrace:
  [1] construct
    @ ~/.julia/packages/KernelAbstractions/X5fk1/src/KernelAbstractions.jl:779 [inlined]
  [2] _esvolumeterm_dir!
    @ ~/.julia/packages/KernelAbstractions/X5fk1/src/macros.jl:61 [inlined]
  [3] esvolumeterm_dir!
    @ ~/.julia/packages/KernelAbstractions/X5fk1/src/macros.jl:68 [inlined]
  [4] #launch_volumeterm#8
    @ ~/src/Raven/src/balancelaws/dgsem.jl:208
  [5] launch_volumeterm
    @ ~/src/Raven/src/balancelaws/dgsem.jl:163 [inlined]
  [6] #_#4
    @ ~/src/Raven/src/balancelaws/dgsem.jl:97
  [7] DGSEM
    @ ~/src/Raven/src/balancelaws/dgsem.jl:84 [inlined]
  [8] rhs
    @ ~/src/Raven/ext/RavenSciMLBaseExt.jl:6 [inlined]
  [9] ODEFunction
    @ ~/.julia/packages/SciMLBase/RPW6n/src/scimlfunctions.jl:2597 [inlined]
 [10] F!
    @ ~/src/Ariadne/libs/Theseus/src/rosenbrock/rosenbrock.jl:35 [inlined]
 [11] F!
    @ ~/src/Ariadne/libs/Theseus/src/rosenbrock/rosenbrock.jl:0 [inlined]
 [12] fwddiffejulia_F__27283_inner_19wrap
    @ ~/src/Ariadne/libs/Theseus/src/rosenbrock/rosenbrock.jl:0
 [13] macro expansion
    @ ~/src/Enzyme/src/compiler.jl:5887 [inlined]
 [14] enzyme_call
    @ ~/src/Enzyme/src/compiler.jl:5421 [inlined]
 [15] ForwardModeThunk
    @ ~/src/Enzyme/src/compiler.jl:5323 [inlined]
 [16] autodiff
    @ ~/src/Enzyme/src/Enzyme.jl:673 [inlined]

https://github.com/EnzymeAD/Enzyme.jl/blob/3e48bb3d07cb2087c8decc46786c690d147baaf5/src/compiler.jl#L5157

[wsmoses]

somehow you have made it past what was believed to be an ijl_throw or similar

[wsmoses]

probably the thing to do here is to upgrade the exception [like I did recently for others] to retain the string of the llvm function, and MI for being able to code_typed [in our interp] of the inside

[wsmoses]

or sorry, rephrasing. you hit an unreachable, that did not have an ijl_throw before it.

how you hit the unreachable....is a question

[vchuravy]

In a different program (related though)

I see:

julia +1.11 --project=. semdg_advection_2d.jl
ERROR: LoadError: 
No forward mode derivative found for jl_get_builtin_fptr
 at context:   %133 = call {} addrspace(10)* ({} addrspace(10)*, {} addrspace(10)**, i32)* @jl_get_builtin_fptr({} addrspace(10)* noundef addrspacecast ({}* inttoptr (i64 139742159320480 to {}*) to {} addrspace(10)*)) #27, !dbg !206

Stacktrace:
 [1] construct
   @ ~/.julia/packages/KernelAbstractions/X5fk1/src/KernelAbstractions.jl:779
 [2] _rhs_volume_kernel!
   @ ~/.julia/packages/KernelAbstractions/X5fk1/src/macros.jl:61
 [3] rhs_volume_kernel!
   @ ~/.julia/packages/KernelAbstractions/X5fk1/src/macros.jl:68
 [4] rhs!
   @ ~/src/Raven/examples/advection/semdg_advection_2d.jl:236
 [5] rhs!
   @ ~/src/Raven/examples/advection/semdg_advection_2d.jl:0


Stacktrace:
  [1] construct
    @ ~/.julia/packages/KernelAbstractions/X5fk1/src/KernelAbstractions.jl:779 [inlined]
  [2] _rhs_volume_kernel!
    @ ~/.julia/packages/KernelAbstractions/X5fk1/src/macros.jl:61 [inlined]
  [3] rhs_volume_kernel!
    @ ~/.julia/packages/KernelAbstractions/X5fk1/src/macros.jl:68 [inlined]
  [4] rhs!
    @ ~/src/Raven/examples/advection/semdg_advection_2d.jl:236 [inlined]
  [5] rhs!
    @ ~/src/Raven/examples/advection/semdg_advection_2d.jl:0 [inlined]
  [6] fwddiffejulia_rhs__14377_inner_1wrap
    @ ~/src/Raven/examples/advection/semdg_advection_2d.jl:0
  [7] macro expansion
    @ ~/.julia/packages/Enzyme/LCg4E/src/compiler.jl:5887 [inlined]
  [8] enzyme_call
    @ ~/.julia/packages/Enzyme/LCg4E/src/compiler.jl:5421 [inlined]
  [9] ForwardModeThunk

Which is very similar if my menory doesn't betray me to Joe issue on Oceanangians. This works on 1.10 so something fishy is going on with KA and 1.11