CuDeviceTexture getindex breaks when executed on the CPU

[zsoerenm]

Describe the bug

I get a stackoverflow when I use a custom struct in combination with Adapt and CuTexture when it is printed to the REPL

I tested CUDA v3.5 and v4.0.1

To reproduce

The Minimal Working Example (MWE) for this bug:

using CUDA, Adapt
struct MyStruct{T}
    code::T
end
mystruct = MyStruct(
    CuTexture(
        CuTextureArray(CuArray(randn(Float32, 4000, 30))),
        address_mode = CUDA.ADDRESS_MODE_WRAP,
        interpolation = CUDA.NearestNeighbour()
    )
)
Adapt.@adapt_structure MyStruct
A = cudaconvert(mystruct); # works fine with semicolon
show(A);

Error:

MyStruct{CuDeviceTexture{Float32, 2, CUDA.ArrayMemory, false, CUDA.NearestNeighbour}}(Float32[Internal error: encountered unexpected error during compilation of LLVMException:
StackOverflowError()

signal (11): Segmentation fault
in expression starting at none:0

signal (6): Aborted
in expression starting at none:0
Aborted (core dumped)

This throws a stackoverflow as well:

julia> A.code[1]
ERROR: StackOverflowError:
Stacktrace:
 [1] getindex(t::CuDeviceTexture{Float32, 2, CUDA.ArrayMemory, false, CUDA.NearestNeighbour}, idx::Float32) (repeats 79984 times)
   @ CUDA ~/.julia/packages/CUDA/ZdCxS/src/device/texture.jl:87

Manifest

[[deps.Adapt]]
deps = ["LinearAlgebra"]
git-tree-sha1 = "0310e08cb19f5da31d08341c6120c047598f5b9c"
uuid = "79e6a3ab-5dfb-504d-930d-738a2a938a0e"
version = "3.5.0"
[[deps.CUDA]]
deps = ["AbstractFFTs", "Adapt", "BFloat16s", "CEnum", "CUDA_Driver_jll", "CUDA_Runtime_Discovery", "CUDA_Runtime_jll", "CompilerSupportLibraries_jll", "ExprTools", "GPUArrays", "GPUCompiler", "LLVM", "LazyArtifacts", "Libdl", "LinearAlgebra", "Logging", "Preferences", "Printf", "Random", "Random123", "RandomNumbers", "Reexport", "Requires", "SparseArrays", "SpecialFunctions"]
git-tree-sha1 = "edff14c60784c8f7191a62a23b15a421185bc8a8"
uuid = "052768ef-5323-5732-b1bb-66c8b64840ba"
version = "4.0.1"

[[deps.CUDA_Driver_jll]]
deps = ["Artifacts", "JLLWrappers", "LazyArtifacts", "Libdl", "Pkg"]
git-tree-sha1 = "75d7896d1ec079ef10d3aee8f3668c11354c03a1"
uuid = "4ee394cb-3365-5eb0-8335-949819d2adfc"
version = "0.2.0+0"

[[deps.CUDA_Runtime_Discovery]]
deps = ["Libdl"]
git-tree-sha1 = "58dd8ec29f54f08c04b052d2c2fa6760b4f4b3a4"
uuid = "1af6417a-86b4-443c-805f-a4643ffb695f"
version = "0.1.1"

[[deps.CUDA_Runtime_jll]]
deps = ["Artifacts", "CUDA_Driver_jll", "JLLWrappers", "LazyArtifacts", "Libdl", "Pkg", "TOML"]
git-tree-sha1 = "d3e6ccd30f84936c1a3a53d622d85d7d3f9b9486"
uuid = "76a88914-d11a-5bdc-97e0-2f5a05c973a2"
version = "0.2.3+2"

[[deps.GPUArrays]]
deps = ["Adapt", "GPUArraysCore", "LLVM", "LinearAlgebra", "Printf", "Random", "Reexport", "Serialization", "Statistics"]
git-tree-sha1 = "4dfaff044eb2ce11a897fecd85538310e60b91e6"
uuid = "0c68f7d7-f131-5f86-a1c3-88cf8149b2d7"
version = "8.6.2"

[[deps.GPUArraysCore]]
deps = ["Adapt"]
git-tree-sha1 = "57f7cde02d7a53c9d1d28443b9f11ac5fbe7ebc9"
uuid = "46192b85-c4d5-4398-a991-12ede77f4527"
version = "0.1.3"

[[deps.GPUCompiler]]
deps = ["ExprTools", "InteractiveUtils", "LLVM", "Libdl", "Logging", "TimerOutputs", "UUIDs"]
git-tree-sha1 = "48832a7cacbe56e591a7bef690c78b9d00bcc692"
uuid = "61eb1bfa-7361-4325-ad38-22787b887f55"
version = "0.17.1"

[[deps.LLVM]]
deps = ["CEnum", "LLVMExtra_jll", "Libdl", "Printf", "Unicode"]
git-tree-sha1 = "b8ae281340f0d3e973aae7b96fb7502b0119b376"
uuid = "929cbde3-209d-540e-8aea-75f648917ca0"
version = "4.15.0"

[[deps.LLVMExtra_jll]]
deps = ["Artifacts", "JLLWrappers", "LazyArtifacts", "Libdl", "Pkg", "TOML"]
git-tree-sha1 = "771bfe376249626d3ca12bcd58ba243d3f961576"
uuid = "dad2f222-ce93-54a1-a47d-0025e8a3acab"
version = "0.0.16+0"

Version info

Details on Julia:

julia> versioninfo()
Julia Version 1.8.3
Commit 0434deb161e (2022-11-14 20:14 UTC)
Platform Info:
  OS: Linux (x86_64-linux-gnu)
  CPU: 4 × Intel(R) Core(TM) i5-2500K CPU @ 3.30GHz
  WORD_SIZE: 64
  LIBM: libopenlibm
  LLVM: libLLVM-13.0.1 (ORCJIT, sandybridge)
  Threads: 1 on 4 virtual cores

Details on CUDA:

julia> CUDA.versioninfo()
CUDA runtime 11.8, artifact installation
CUDA driver 12.0
NVIDIA driver 525.85.5

Libraries: 
- CUBLAS: 11.11.3
- CURAND: 10.3.0
- CUFFT: 10.9.0
- CUSOLVER: 11.4.1
- CUSPARSE: 11.7.5
- CUPTI: 18.0.0
- NVML: 12.0.0+525.85.5

Toolchain:
- Julia: 1.8.3
- LLVM: 13.0.1
- PTX ISA support: 3.2, 4.0, 4.1, 4.2, 4.3, 5.0, 6.0, 6.1, 6.3, 6.4, 6.5, 7.0, 7.1, 7.2
- Device capability support: sm_35, sm_37, sm_50, sm_52, sm_53, sm_60, sm_61, sm_62, sm_70, sm_72, sm_75, sm_80, sm_86

1 device:
  0: NVIDIA GeForce GTX 950 (sm_52, 1.327 GiB / 2.000 GiB available)

[maleadt]

MyStruct{CuDeviceTexture{...}}

Device structures are not intended to be used on the CPU. Their functionality is strictly meant for use on the GPU, and calling functions (like getindex here) may just crash your session. We could make each of those definitions a @device_function, but that's a bit of a bother (and breaks tools like Revise.jl).

Why are you attempting to use this object on the CPU? You generally shouldn't ever need to call cudaconvert.

[maleadt]

Also doesn't need Adapt:

julia> using CUDA

julia> tex =     CuTexture(
               CuTextureArray(CuArray(randn(Float32, 4000, 30))),
               address_mode = CUDA.ADDRESS_MODE_WRAP,
               interpolation = CUDA.NearestNeighbour()
           )
c4000×30 1-channel CuTexture(::CuTextureArray) with eltype Float32

julia> cudaconvert(tex)
4000×30 CuDeviceTexture{Float32, 2, CUDA.ArrayMemory, false, CUDA.NearestNeighbour}:
Internal error: encountered unexpected error during compilation of LLVMException:
StackOverflowError()

signal (11): Segmentation fault
in expression starting at none:0

signal (6): Aborted
in expression starting at none:0

[zsoerenm]

okay my bad. I wanted to print it for debugging purposes, but I guess it's not going to work this way. Thanks for clarifying.