Cannot precompile GPU code with SnoopPrecompile

[marius311]

An MWE as requested by #1865 (comment)

module MWE

using CUDA, SnoopPrecompile

@precompile_all_calls begin
    sqrt.(CUDA.ones(2))
end

end

Then from a session:

julia> using MWE

julia> sqrt.(ones(Float32,2)) # note *CPU* array here
JIT session error: Symbols not found: [ __nv_sqrtf ]
JIT session error: Symbols not found: [ __nv_sqrtf ]
JIT session error: Symbols not found: [ __nv_sqrtf ]
JIT session error: Symbols not found: [ __nv_sqrtf ]
JIT session error: Symbols not found: [ __nv_sqrtf ]
JIT session error: Symbols not found: [ __nv_sqrtf ]
JIT session error: Symbols not found: [ __nv_sqrtf ]
2-element Vector{Float32}:
 1.0
 1.0

Verison info:

Julia 1.9-rc2
CUDA v4.1.4 https://github.com/JuliaGPU/CUDA.jl.git#1dec082
GPUCompiler v0.19.1
LLVM v5.0.0

julia> CUDA.versioninfo()
CUDA runtime 12.1, artifact installation
CUDA driver 12.1
NVIDIA driver 525.105.17, originally for CUDA 12.0

Libraries: 
- CUBLAS: 12.1.0
- CURAND: 10.3.2
- CUFFT: 11.0.2
- CUSOLVER: 11.4.4
- CUSPARSE: 12.0.2
- CUPTI: 18.0.0
- NVML: 12.0.0+525.105.17

Toolchain:
- Julia: 1.9.0-rc2
- LLVM: 14.0.6
- 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, 7.3, 7.4, 7.5
- Device capability support: 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 A100-PCIE-40GB (sm_80, 39.390 GiB / 40.000 GiB available)

[maleadt]

Yeah, that's timholy/SnoopCompile.jl#338.

[vchuravy]

Should be fixed by JuliaLang/julia#49391