Make test/Project.toml a superset of examples/Project.toml

[AntonOresten]

Cuts the number of dependencies in the examples/ environment in half by using CUDA directly, allowing the dependencies of the test/ environment to be a superset with the addition of only NVTX and FFTW. This eliminates the need to reset the active project to the examples/ environment.

I encountered this problem in #239 while wanting to quickly run tests in a fresh cloud environment:

]add https://github.com/JuliaGPU/cuTile.jl#<branch>
]test cuTile

Otherwise I'd have to clone, activate by path, and instantiate.

This should only make running an example / benchmark slightly slower, since cuSOLVER, cuSPARSE, and CUDATools are now indirect dependencies of examples/.

[maleadt]

I encountered this problem in #239 while wanting to quickly run tests in a fresh cloud environment:

]add https://github.com/JuliaGPU/cuTile.jl#<branch>
]test cuTile

Otherwise I'd have to clone, activate by path, and instantiate.

Can you elaborate? I don't understand what the issue is. CI does exactly this, and it works

[AntonOresten]

The issue is specifically when the cuTile workspace hasn't been instantiated. If we only add cuTile to an external environment, FFTW and others (which previously only existed in cuTile/examples) are never added.

[AntonOresten]

i.e. with ]test cuTile I get:

Test Summary:              | Pass  Error  Total   Time
  Overall                  | 2348      1   2349  53.0s
    codegen/cse            |    7             7   4.9s
    codegen/rng_intrinsics |    5             5   7.0s
    codegen/reflection     |    9             9   6.2s
    device/types           |    4             4   9.0s
    device/math            |    8             8   9.5s
    extensions/DLFP8Types  |    5             5   6.6s
    device/gather_scatter  |   16            16   7.0s
    codegen/fpmode         |    7             7   2.7s
    codegen/assume         |   11            11   7.6s
    device/slice           |    7             7   3.4s
    extensions/Microfloats |    8             8   1.2s
    device/views           |    5             5   2.7s
    codegen/views          |    4             4   1.7s
    device/hints           |   11            11  12.1s
    host/broadcast         |   17            17  12.7s
    device/print           |    5             5   1.7s
    device/integration     |    1             1   1.4s
    device/kernel_state    |    3             3   0.6s
    device/control_flow    |   13            13  14.5s
    host/cache             |   30            30   1.2s
    codegen/kernel_state   |    1             1   1.7s
    analysis/dataflow      |   11            11   0.6s
    types                  |   99            99   0.6s
    examples/vadd          |                  0  10.8s
    device/atomics         |   26            26  17.8s
    examples/fft           |           1      1   0.3s
    examples/transpose     |                  0   2.1s
    codegen/slice          |    8             8   4.3s
    examples/batchmatmul   |                  0  10.4s
    codegen/bounds         |   19            19   1.6s
    codegen/no_wrap        |    8             8   1.3s
    examples/softmax       |                  0  11.2s
    examples/matmul        |                  0   5.6s
    examples/layernorm     |                  0  16.9s
    codegen/integration    |   87            87  26.1s
    device/broadcast       |  119           119  26.2s
    examples/fmha          |                  0  18.2s
    device/core            |   39            39  30.1s
    examples/moe           |                  0  20.3s
    device/random          |  480           480  33.5s
    host/mapreduce         |   29            29  31.2s
    device/tile            |  139           139  31.9s
    codegen/operations     |  180           180  40.2s
    device/reductions      |  927           927  43.9s
    FAILURE

Error in testset examples/fft:
Error During Test at /home/anton/.julia/packages/cuTile/DnjqE/test/setup.jl:46
  Got exception outside of a @test
  LoadError: ArgumentError: Package FFTW [7a1cc6ca-52ef-59f5-83cd-3a7055c09341] is required but does not seem to be installed:
   - Run `Pkg.instantiate()` to install all recorded dependencies.

  Stacktrace:
    [1] __require_prelocked(pkg::Base.PkgId, env::String)
      @ Base ./loading.jl:2617
    [2] _require_prelocked(uuidkey::Base.PkgId, env::String)
      @ Base ./loading.jl:2495
    [3] macro expansion
      @ ./loading.jl:2423 [inlined]
    [4] macro expansion
      @ ./lock.jl:376 [inlined]
    [5] __require(into::Module, mod::Symbol)
      @ Base ./loading.jl:2388
    [6] require(into::Module, mod::Symbol)
      @ Base ./loading.jl:2364
    [7] include(mapexpr::Function, mod::Module, _path::String)
      @ Base ./Base.jl:307
    [8] IncludeInto
      @ ./Base.jl:308 [inlined]
    [9] (::Main.var"##examples/fft#494".var"#5#6")()
      @ Main.var"##examples/fft#494" ~/.julia/packages/cuTile/DnjqE/test/runtests.jl:40
   [10] (::Base.RedirectStdStream)(thunk::Main.var"##examples/fft#494".var"#5#6", stream::Base.DevNull)
      @ Base ./stream.jl:1464
   [11] (::Main.var"##examples/fft#494".var"#3#4")()
      @ Main.var"##examples/fft#494" ~/.julia/packages/cuTile/DnjqE/test/runtests.jl:39
   [12] cd(f::Main.var"##examples/fft#494".var"#3#4", dir::String)
      @ Base.Filesystem ./file.jl:112
   [13] macro expansion
      @ ~/.julia/packages/cuTile/DnjqE/test/runtests.jl:35 [inlined]
   [14] macro expansion
      @ ~/.julia/packages/cuTile/DnjqE/test/setup.jl:47 [inlined]
   [15] macro expansion
      @ ~/.julia/juliaup/julia-1.12.6+0.aarch64.linux.gnu/share/julia/stdlib/v1.12/Test/src/Test.jl:1777 [inlined]
   [16] macro expansion
      @ ~/.julia/packages/cuTile/DnjqE/test/setup.jl:47 [inlined]
   [17] macro expansion
      @ ~/.julia/juliaup/julia-1.12.6+0.aarch64.linux.gnu/share/julia/stdlib/v1.12/Test/src/Test.jl:1777 [inlined]
   [18] macro expansion
      @ ~/.julia/packages/cuTile/DnjqE/test/setup.jl:46 [inlined]
   [19] macro expansion
      @ ~/.julia/packages/CUDACore/sIGRL/src/utilities.jl:35 [inlined]
   [20] macro expansion
      @ ~/.julia/packages/CUDACore/sIGRL/src/memory.jl:896 [inlined]
   [21] top-level scope
      @ ~/.julia/packages/cuTile/DnjqE/test/setup.jl:45
   [22] eval(m::Module, e::Any)
      @ Core ./boot.jl:489
   [23] execute(::Type{cuTileTestRecord}, mod::Module, f::Expr, name::String, start_time::Float64, custom_args::@NamedTuple{})
      @ Main ~/.julia/packages/cuTile/DnjqE/test/setup.jl:42
   [24] (::ParallelTestRunner.var"#inner#runtest##0"{Type{cuTileTestRecord}, Expr, String, Expr, Float64, @NamedTuple{}})()
      @ ParallelTestRunner ~/.julia/packages/ParallelTestRunner/anvy6/src/ParallelTestRunner.jl:429
   [25] runtest(RecordType::Type{cuTileTestRecord}, f::Expr, name::String, init_code::Expr, start_time::Float64, custom_args::@NamedTuple{})
      @ ParallelTestRunner ~/.julia/packages/ParallelTestRunner/anvy6/src/ParallelTestRunner.jl:440
   [26] (::var"#handle##0#handle##1"{Sockets.TCPSocket, UInt64, Bool, @Kwargs{}, Tuple{typeof(ParallelTestRunner.runtest), DataType, Expr, String, Expr, Float64, @NamedTuple{}}, typeof(invokelatest)})()
      @ Main ~/.julia/packages/Malt/OteGQ/src/worker.jl:120
  in expression starting at /home/anton/.julia/packages/cuTile/DnjqE/examples/fft.jl:16
ERROR: LoadError: Test run finished with errors
in expression starting at /home/anton/.julia/packages/cuTile/DnjqE/test/runtests.jl:92
ERROR: Package cuTile errored during testing