CompatHelper: bump compat for CUDA in [weakdeps] to 6, (keep existing compat)

[github-actions]

This pull request changes the compat entry for the CUDA package from 5.9.1 to 5.9.1, 6.
This keeps the compat entries for earlier versions.

Note: I have not tested your package with this new compat entry.
It is your responsibility to make sure that your package tests pass before you merge this pull request.

Closes #2927.

[codecov]

Codecov Report

❌ Patch coverage is 0% with 11 lines in your changes missing coverage. Please review.
✅ Project coverage is 96.87%. Comparing base (29765b9) to head (79c2ea2).
⚠️ Report is 1 commits behind head on main.

Files with missing lines	Patch %	Lines
ext/TrixiCUDACoreExt.jl	0.00%	11 Missing ⚠️

Additional details and impacted files

@@            Coverage Diff             @@
##             main    #2925      +/-   ##
==========================================
+ Coverage   96.84%   96.87%   +0.03%     
==========================================
  Files         642      642              
  Lines       49824    49808      -16     
==========================================
  Hits        48250    48250              
+ Misses       1574     1558      -16     

Flag	Coverage Δ
unittests	96.87% <0.00%> (+0.03%)	⬆️

Flags with carried forward coverage won't be shown. Click here to find out more.

☔ View full report in Codecov by Harness.
📢 Have feedback on the report? Share it here.

🚀 New features to boost your workflow:

• ❄️ Test Analytics: Detect flaky tests, report on failures, and find test suite problems.

[ranocha]

Looking at the release notes of CUDA.jl, there should be no breaking changes (just internal reorganization of the repository).

[JoshuaLampert]

We need to wait for JuliaDiff/SparseMatrixColorings.jl#313.

[vchuravy]

Eventually we can drop the CUDA dependency and only depend on for our extension CUDACore

[JoshuaLampert]

CI uses CUDA.jl v6.1. So this should be ready to merge.

[ranocha]

CUDA tests fail:

elixir_euler_source_terms.jl Float32 / CUDA: Error During Test at /root/.cache/julia-buildkite-plugin/depots/019658de-efdd-4c0c-a098-7a91a0dafb4d/packages/TrixiTest/6CVoC/src/macros.jl:196
  Got exception outside of a @test
  LoadError: UndefVarError: `local_toolkit` not defined
  Stacktrace:
    [1] trixi_backend_info!(setup::Vector{Pair{String, Any}}, ::CUDACore.CUDAKernels.CUDABackend)
      @ TrixiCUDAExt /var/lib/buildkite-agent/builds/gpuci-7/julialang/trixi-dot-jl/ext/TrixiCUDAExt.jl:38
    [2] initialize_summary_callback(cb::SciMLBase.DiscreteCallback{typeof(Trixi.summary_callback), typeof(Trixi.summary_callback), Trixi.var"#initialize#1676"{Bool}, typeof(Trixi.finalize_summary_callback), Nothing, Tuple{}}, u::CUDACore.CuArray{Float32, 1, CUDACore.DeviceMemory}, t::Float64, integrator::OrdinaryDiffEqCore.ODEIntegrator{OrdinaryDiffEqLowStorageRK.CarpenterKennedy2N54{typeof(OrdinaryDiffEqCore.trivial_limiter!), typeof(OrdinaryDiffEqCore.trivial_limiter!), FastBroadcast.Serial}, true, CUDACore.CuArray{Float32, 1, CUDACore.DeviceMemory}, Nothing, Float64, Trixi.SemidiscretizationHyperbolic{Trixi.P4estMesh{3, 3, Float64, Static.False, P4est.PointerWrappers.PointerWrapper{P4est.LibP4est.p8est}, P4est.PointerWrappers.PointerWrapper{P4est.LibP4est.p8est_ghost_t}, 5, 4}, Trixi.CompressibleEulerEquations3D{Float32}, typeof(Trixi.initial_condition_convergence_test), Trixi.BoundaryConditionPeriodic, typeof(Trixi.source_terms_convergence_test), Trixi.DGSEM{Trixi.LobattoLegendreBasis{Float32, 4, StaticArraysCore.SVector{4, Float32}, CUDACore.CuArray{Float32, 2, CUDACore.DeviceMemory}, CUDACore.CuArray{Float32, 2, CUDACore.DeviceMemory}}, Trixi.LobattoLegendreMortarL2{Float32, 4, CUDACore.CuArray{Float32, 2, CUDACore.DeviceMemory}, CUDACore.CuArray{Float32, 2, CUDACore.DeviceMemory}}, Trixi.SurfaceIntegralWeakForm{Trixi.FluxLaxFriedrichs{typeof(Trixi.max_abs_speed_naive)}}, Trixi.VolumeIntegralWeakForm}, @NamedTuple{elements::Trixi.P4estElementContainer{3, Float32, Float32, 4, 5, 6, CUDACore.CuArray{Float32, 4, CUDACore.DeviceMemory}, CUDACore.CuArray{Float32, 5, CUDACore.DeviceMemory}, CUDACore.CuArray{Float32, 6, CUDACore.DeviceMemory}, CUDACore.CuArray{Float32, 1, CUDACore.DeviceMemory}, CUDACore.CuArray{Float32, 5, CUDACore.DeviceMemory}, CUDACore.CuArray{Float32, 1, CUDACore.DeviceMemory}}, interfaces::Trixi.P4estInterfaceContainer{3, Float64, Float32, 4, 5, CUDACore.CuArray{Float32, 5, CUDACore.DeviceMemory}, Nothing, CUDACore.CuArray{Int64, 2, CUDACore.DeviceMemory}, CUDACore.CuArray{Tuple{Symbol, Symbol, Symbol}, 2, CUDACore.DeviceMemory}, CUDACore.CuArray{Float32, 1, CUDACore.DeviceMemory}, Nothing, CUDACore.CuArray{Int64, 1, CUDACore.DeviceMemory}, CUDACore.CuArray{Tuple{Symbol, Symbol, Symbol}, 1, CUDACore.DeviceMemory}}, boundaries::Trixi.P4estBoundaryContainer{3, Float32, 4, CUDACore.CuArray{Float32, 4, CUDACore.DeviceMemory}, CUDACore.CuArray{Int64, 1, CUDACore.DeviceMemory}, CUDACore.CuArray{Tuple{Symbol, Symbol, Symbol}, 1, CUDACore.DeviceMemory}, CUDACore.CuArray{Float32, 1, CUDACore.DeviceMemory}}, mortars::Trixi.P4estMortarContainer{3, Float32, 4, 6, CUDACore.CuArray{Float32, 6, CUDACore.DeviceMemory}, CUDACore.CuArray{Int64, 2, CUDACore.DeviceMemory}, CUDACore.CuArray{Tuple{Symbol, Symbol, Symbol}, 2, CUDACore.DeviceMemory}, CUDACore.CuArray{Float32, 1, CUDACore.DeviceMemory}, CUDACore.CuArray{Int64, 1, CUDACore.DeviceMemory}, CUDACore.CuArray{Tuple{Symbol, Symbol, Symbol}, 1, CUDACore.DeviceMemory}}, fstar_primary_threaded::Trixi.VecOfArrays{CUDACore.CuArray{Float32, 4, CUDACore.DeviceMemory}}, fstar_secondary_threaded::Trixi.VecOfArrays{CUDACore.CuArray{Float32, 4, CUDACore.DeviceMemory}}, fstar_tmp_threaded::Trixi.VecOfArrays{CUDACore.CuArray{Float32, 3, CUDACore.DeviceMemory}}, u_threaded::Trixi.VecOfArrays{CUDACore.CuArray{Float32, 3, CUDACore.DeviceMemory}}}}, Float64, Float64, Vector{CUDACore.CuArray{Float32, 1, CUDACore.DeviceMemory}}, SciMLBase.ODESolution{Float32, 2, Vector{CUDACore.CuArray{Float32, 1, CUDACore.DeviceMemory}}, Nothing, Nothing, Vector{Float64}, Vector{Vector{CUDACore.CuArray{Float32, 1, CUDACore.DeviceMemory}}}, Nothing, SciMLBase.ODEProblem{CUDACore.CuArray{Float32, 1, CUDACore.DeviceMemory}, Tuple{Float64, Float64}, true, Trixi.SemidiscretizationHyperbolic{Trixi.P4estMesh{3, 3, Float64, Static.False, P4est.PointerWrappers.PointerWrapper{P4est.LibP4est.p8est}, P4est.PointerWrappers.PointerWrapper{P4est.LibP4est.p8est_ghost_t}, 5, 4}, Trixi.CompressibleEulerEquations3D{Float32}, typeof(Trixi.initial_condition_convergence_test), Trixi.BoundaryConditionPeriodic, typeof(Trixi.source_terms_convergence_test), Trixi.DGSEM{Trixi.LobattoLegendreBasis{Float32, 4, StaticArraysCore.SVector{4, Float32}, CUDACore.CuArray{Float32, 2, CUDACore.DeviceMemory}, CUDACore.CuArray{Float32, 2, CUDACore.DeviceMemory}}, Trixi.LobattoLegendreMortarL2{Float32, 4, CUDACore.CuArray{Float32, 2, CUDACore.DeviceMemory}, CUDACore.CuArray{Float32, 2, CUDACore.DeviceMemory}}, Trixi.SurfaceIntegralWeakForm{Trixi.FluxLaxFriedrichs{typeof(Trixi.max_abs_speed_naive)}}, Trixi.VolumeIntegralWeakForm}, @NamedTuple{elements::Trixi.P4estElementContainer{3, Float32, Float32, 4, 5, 6, CUDACore.CuArray{Float32, 4, CUDACore.DeviceMemory}, CUDACore.CuArray{Float32, 5, CUDACore.DeviceMemory}, CUDACore.CuArray{Float32, 6, CUDACore.DeviceMemory}, CUDACore.CuArray{Float32, 1, CUDACore.DeviceMemory}, CUDACore.CuArray{Float32, 5, CUDACore.DeviceMemory}, CUDACore.CuArray{Float32, 1, CUDACore.DeviceMemory}}, interfaces::Trixi.P4estInterfaceContainer{3, Float64, Float32, 4, 5, CUDACore.CuArray{Float32, 5, CUDACore.DeviceMemory}, Nothing, CUDACore.CuArray{Int64, 2, CUDACore.DeviceMemory}, CUDACore.CuArray{Tuple{Symbol, Symbol, Symbol}, 2, CUDACore.DeviceMemory}, CUDACore.CuArray{Float32, 1, CUDACore.DeviceMemory}, Nothing, CUDACore.CuArray{Int64, 1, CUDACore.DeviceMemory}, CUDACore.CuArray{Tuple{Symbol, Symbol, Symbol}, 1, CUDACore.DeviceMemory}}, boundaries::Trixi.P4estBoundaryContainer{3, Float32, 4, CUDACore.CuArray{Float32, 4, CUDACore.DeviceMemory}, CUDACore.CuArray{Int64, 1, CUDACore.DeviceMemory}, CUDACore.CuArray{Tuple{Symbol, Symbol, Symbol}, 1, CUDACore.DeviceMemory}, CUDACore.CuArray{Float32, 1, CUDACore.DeviceMemory}}, mortars::Trixi.P4estMortarContainer{3, Float32, 4, 6, CUDACore.CuArray{Float32, 6, CUDACore.DeviceMemory}, CUDACore.CuArray{Int64, 2, CUDACore.DeviceMemory}, CUDACore.CuArray{Tuple{Symbol, Symbol, Symbol}, 2, CUDACore.DeviceMemory}, CUDACore.CuArray{Float32, 1, CUDACore.DeviceMemory}, CUDACore.CuArray{Int64, 1, CUDACore.DeviceMemory}, CUDACore.CuArray{Tuple{Symbol, Symbol, Symbol}, 1, CUDACore.DeviceMemory}}, fstar_primary_threaded::Trixi.VecOfArrays{CUDACore.CuArray{Float32, 4, CUDACore.DeviceMemory}}, fstar_secondary_threaded::Trixi.VecOfArrays{CUDACore.CuArray{Float32, 4, CUDACore.DeviceMemory}}, fstar_tmp_threaded::Trixi.VecOfArrays{CUDACore.CuArray{Float32, 3, CUDACore.DeviceMemory}}, u_threaded::Trixi.VecOfArrays{CUDACore.CuArray{Float32, 3, CUDACore.DeviceMemory}}}}, SciMLBase.ODEFunction{true, SciMLBase.FullSpecialize, typeof(Trixi.rhs!), LinearAlgebra.UniformScaling{Bool}, Nothing, Nothing, Nothing, Nothing, Nothing, Nothing, Nothing, Nothing, Nothing, Nothing, Nothing, Nothing, typeof(SciMLBase.DEFAULT_OBSERVED), Nothing, Nothing, Nothing, Nothing}, @Kwargs{}, SciMLBase.StandardODEProblem}, OrdinaryDiffEqLowStorageRK.CarpenterKennedy2N54{typeof(OrdinaryDiffEqCore.trivial_limiter!), typeof(OrdinaryDiffEqCore.trivial_limiter!), FastBroadcast.Serial}, OrdinaryDiffEqCore.InterpolationData{SciMLBase.ODEFunction{true, SciMLBase.FullSpecialize, typeof(Trixi.rhs!), LinearAlgebra.UniformScaling{Bool}, Nothing, Nothing, Nothing, Nothing, Nothing, Nothing, Nothing, Nothing, Nothing, Nothing, Nothing, Nothing, typeof(SciMLBase.DEFAULT_OBSERVED), Nothing, Nothing, Nothing, Nothing}, Vector{CUDACore.CuArray{Float32, 1, CUDACore.DeviceMemory}}, Vector{Float64}, Vector{Vector{CUDACore.CuArray{Float32, 1, CUDACore.DeviceMemory}}}, Nothing, OrdinaryDiffEqLowStorageRK.LowStorageRK2NCache{CUDACore.CuArray{Float32, 1, CUDACore.DeviceMemory}, CUDACore.CuArray{Float32, 1, CUDACore.DeviceMemory}, OrdinaryDiffEqLowStorageRK.LowStorageRK2NConstantCache{4, Float32, Float64}, typeof(OrdinaryDiffEqCore.trivial_limiter!), typeof(OrdinaryDiffEqCore.trivial_limiter!), FastBroadcast.Serial}, Nothing}, SciMLBase.DEStats, Nothing, Nothing, Nothing, Nothing}, SciMLBase.ODEFunction{true, SciMLBase.FullSpecialize, typeof(Trixi.rhs!), LinearAlgebra.UniformScaling{Bool}, Nothing, Nothing, Nothing, Nothing, Nothing, Nothing, Nothing, Nothing, Nothing, Nothing, Nothing, Nothing, typeof(SciMLBase.DEFAULT_OBSERVED), Nothing, Nothing, Nothing, Nothing}, OrdinaryDiffEqLowStorageRK.LowStorageRK2NCache{CUDACore.CuArray{Float32, 1, CUDACore.DeviceMemory}, CUDACore.CuArray{Float32, 1, CUDACore.DeviceMemory}, OrdinaryDiffEqLowStorageRK.LowStorageRK2NConstantCache{4, Float32, Float64}, typeof(OrdinaryDiffEqCore.trivial_limiter!), typeof(OrdinaryDiffEqCore.trivial_limiter!), FastBroadcast.Serial}, OrdinaryDiffEqCore.DEOptions{Float32, Float32, Float64, Float64, typeof(DiffEqBase.ODE_DEFAULT_NORM), typeof(LinearAlgebra.opnorm), Nothing, SciMLBase.CallbackSet{Tuple{}, Tuple{SciMLBase.DiscreteCallback{typeof(Trixi.summary_callback), typeof(Trixi.summary_callback), Trixi.var"#initialize#1676"{Bool}, typeof(Trixi.finalize_summary_callback), Nothing, Tuple{}}, SciMLBase.DiscreteCallback{Trixi.var"#1682#1689"{Int64}, Trixi.AnalysisCallback{Trixi.LobattoLegendreAnalyzer{Float64, 7, StaticArraysCore.SVector{7, Float64}, Matrix{Float64}}, Tuple{typeof(Trixi.entropy_timederivative)}, StaticArraysCore.SVector{5, Float64}, @NamedTuple{u_local::StrideArraysCore.StaticStrideArray{Float64, 4, (1, 2, 3, 4), Tuple{Static.StaticInt{5}, Static.StaticInt{7}, Static.StaticInt{7}, Static.StaticInt{7}}, NTuple{4, Nothing}, NTuple{4, Static.StaticInt{1}}, 1715}, u_tmp1::StrideArraysCore.StaticStrideArray{Float64, 4, (1, 2, 3, 4), Tuple{Static.StaticInt{5}, Static.StaticInt{7}, Static.StaticInt{4}, Static.StaticInt{4}}, NTuple{4, Nothing}, NTuple{4, Static.StaticInt{1}}, 560}, u_tmp2::StrideArraysCore.StaticStrideArray{Float64, 4, (1, 2, 3, 4), Tuple{Static.StaticInt{5}, Static.StaticInt{7}, Static.StaticInt{7}, Static.StaticInt{4}}, NTuple{4, Nothing}, NTuple{4, Static.StaticInt{1}}, 980}, x_local::StrideArraysCore.StaticStrideArray{Float64, 4, (1, 2, 3, 4), Tuple{Static.StaticInt{3}, Static.StaticInt{7}, Static.StaticInt{7}, Static.StaticInt{7}}, NTuple{4, Nothing}, NTuple{4, Static.StaticInt{1}}, 1029}, x_tmp1::StrideArraysCore.StaticStrideArray{Float64, 4, (1, 2, 3, 4), Tuple{Static.StaticInt{3}, Static.StaticInt{7}, Static.StaticInt{4}, Static.StaticInt{4}}, NTuple{4, Nothing}, NTuple{4, Static.StaticInt{1}}, 336}, x_tmp2::StrideArraysCore.StaticStrideArray{Float64, 4, (1, 2, 3, 4), Tuple{Static.StaticInt{3}, Static.StaticInt{7}, Static.StaticInt{7}, Static.StaticInt{4}}, NTuple{4, Nothing}, NTuple{4, Static.StaticInt{1}}, 588}, jacobian_local::StrideArraysCore.StaticStrideArray{Float64, 3, (1, 2, 3), Tuple{Static.StaticInt{7}, Static.StaticInt{7}, Static.StaticInt{7}}, Tuple{Nothing, Nothing, Nothing}, Tuple{Static.StaticInt{1}, Static.StaticInt{1}, Static.StaticInt{1}}, 343}, jacobian_tmp1::StrideArraysCore.StaticStrideArray{Float64, 3, (1, 2, 3), Tuple{Static.StaticInt{7}, Static.StaticInt{4}, Static.StaticInt{4}}, Tuple{Nothing, Nothing, Nothing}, Tuple{Static.StaticInt{1}, Static.StaticInt{1}, Static.StaticInt{1}}, 112}, jacobian_tmp2::StrideArraysCore.StaticStrideArray{Float64, 3, (1, 2, 3), Tuple{Static.StaticInt{7}, Static.StaticInt{7}, Static.StaticInt{4}}, Tuple{Nothing, Nothing, Nothing}, Tuple{Static.StaticInt{1}, Static.StaticInt{1}, Static.StaticInt{1}}, 196}}}, typeof(Trixi.initialize!), typeof(SciMLBase.FINALIZE_DEFAULT), Nothing, Tuple{}}, SciMLBase.DiscreteCallback{Trixi.AliveCallback, Trixi.AliveCallback, typeof(Trixi.initialize!), typeof(SciMLBase.FINALIZE_DEFAULT), Nothing, Tuple{}}, SciMLBase.DiscreteCallback{Trixi.SaveSolutionCallback{Int64, typeof(Trixi.cons2prim)}, Trixi.SaveSolutionCallback{Int64, typeof(Trixi.cons2prim)}, typeof(Trixi.initialize_save_cb!), typeof(SciMLBase.FINALIZE_DEFAULT), Nothing, Tuple{}}, SciMLBase.DiscreteCallback{Trixi.StepsizeCallback{Returns{Float64}, Returns{Float64}}, Trixi.StepsizeCallback{Returns{Float64}, Returns{Float64}}, typeof(Trixi.initialize!), typeof(SciMLBase.FINALIZE_DEFAULT), Nothing, Tuple{}}}}, typeof(DiffEqBase.ODE_DEFAULT_ISOUTOFDOMAIN), typeof(DiffEqBase.ODE_DEFAULT_PROG_MESSAGE), typeof(DiffEqBase.ODE_DEFAULT_UNSTABLE_CHECK), DataStructures.BinaryHeap{Float64, DataStructures.FasterForward}, DataStructures.BinaryHeap{Float64, DataStructures.FasterForward}, Nothing, Nothing, Int64, Tuple{}, Tuple{}, Tuple{}, DiffEqBase.DEVerbosity{true, SciMLLogging.Minimal, SciMLLogging.Minimal}, Nothing}, Nothing, Float32, Nothing, DiffEqBase.DefaultInit, Nothing, OrdinaryDiffEqCore.PIControllerCache{Float64, Float32}, Random.TaskLocalRNG, Nothing, Nothing, Nothing, Nothing, Nothing, Nothing}; reset_threads::Bool)
      @ Trixi /var/lib/buildkite-agent/builds/gpuci-7/julialang/trixi-dot-jl/src/callbacks_step/summary.jl:226
    [3] initialize_summary_callback
      @ /var/lib/buildkite-agent/builds/gpuci-7/julialang/trixi-dot-jl/src/callbacks_step/summary.jl:153 [inlined]
    [4] initialize
      @ /var/lib/buildkite-agent/builds/gpuci-7/julialang/trixi-dot-jl/src/callbacks_step/summary.jl:20 [inlined]

see https://buildkite.com/julialang/trixi-dot-jl/builds/6963/list?jid=019e4e9d-4933-4270-afc9-a3bd786e0653&tab=output#019e4e9d-4933-4270-afc9-a3bd786e0653/L2758

CC @vchuravy

[JoshuaLampert]

Looks like that is because CUDA.jl was split into subpackages in CUDA.jl v6. Now local_toolkit lives in CUDACore.jl. So it would be good if the extension only depends on CUDACore.jl.

[JoshuaLampert]

Is everything we need in the extension already defined CUDACore.jl, i.e., is it trivial to only depend on CUDACore.jl by essentially replacing CUDA by CUDACore, @vchuravy?

[JoshuaLampert]

Because the NVML stuff lives in CUDATools.jl, the extension has to be for both CUDACore.jl and CUDATools.jl.

[JoshuaLampert]

From my point of view, this is ready and could be merged. Would you prefer to keep compatibility with CUDA.jl v5 (at least for some more time) @vchuravy? It would probably be messy to try to make Trixi.jl compatible with both CUDA.jl v5 and v6 at the same time since I think then the extension would need to be for CUDA (for the v5 case), CUDACore, and CUDATools (for the v6 case) with some @static ifs for the imports.

[ranocha]

Thanks for updating this to CUDA.jl v6, @JoshuaLampert! I would like to get the opinion of @vchuravy before merging this.

[JoshuaLampert]

Do we loose anything by removing the NVML stuff, @vchuravy?

[ranocha]

Thanks!

@vchuravy and I discussed this offline: No, we don't lose anything in terms of capabilities. It just makes some query functions used in the summary callback a bit more tedious, but this does not matter (also not in terms of performance since it is only used once in a simulation).