Conditional element-wise assignment still won't compile

[zhangcx93]

a = CuArray([1,2,3,4,5])
a[a .> 3] .= 10

this won't compile, while the cpu version does work:

a = [1,2,3,4,5]
a[ a .> 3] .= 10

Error message:

ERROR: GPU compilation of kernel broadcast_kernel(CUDA.CuKernelContext, SubArray{Int64, 1, CuDeviceVector{Int64, 1}, Tuple{Vector{Int64}}, false}, Base.Broadcast.Broadcasted{Nothing, Tuple{Base.OneTo{Int64}}, typeof(identity), Tuple{Int64}}, Int64) failed
KernelError: passing and using non-bitstype argument

Argument 3 to your kernel function is of type SubArray{Int64, 1, CuDeviceVector{Int64, 1}, Tuple{Vector{Int64}}, false}, which is not isbits:
.indices is of type Tuple{Vector{Int64}} which is not isbits.
.1 is of type Vector{Int64} which is not isbits.

Stacktrace:
[1] check_invocation(job::GPUCompiler.CompilerJob, entry::LLVM.Function)
@ GPUCompiler ~.julia\packages\GPUCompiler\XwWPj\src\validation.jl:68
[2] macro expansion
@ ~.julia\packages\GPUCompiler\XwWPj\src\driver.jl:287 [inlined]
[3] macro expansion
@ ~.julia\packages\TimerOutputs\4QAIk\src\TimerOutput.jl:206 [inlined]
[4] macro expansion
@ ~.julia\packages\GPUCompiler\XwWPj\src\driver.jl:286 [inlined]
[5] emit_asm(job::GPUCompiler.CompilerJob, ir::LLVM.Module, kernel::LLVM.Function; strip::Bool, validate::Bool, format::LLVM.API.LLVMCodeGenFileType)
@ GPUCompiler ~.julia\packages\GPUCompiler\XwWPj\src\utils.jl:62
[6] cufunction_compile(job::GPUCompiler.CompilerJob)
@ CUDA ~.julia\packages\CUDA\qEV3Y\src\compiler\execution.jl:306
[7] check_cache
@ ~.julia\packages\GPUCompiler\XwWPj\src\cache.jl:44 [inlined]
[8] cached_compilation
@ ~.julia\packages\GPUArrays\WV76E\src\host\broadcast.jl:60 [inlined]
[9] cached_compilation(cache::Dict{UInt64, Any}, job::GPUCompiler.CompilerJob{GPUCompiler.PTXCompilerTarget, CUDA.CUDACompilerParams, GPUCompiler.FunctionSpec{GPUArrays.var"#broadcast_kernel#12", Tuple{CUDA.CuKernelContext, SubArray{Int64, 1, CuDeviceVector{Int64, 1}, Tuple{Vector{Int64}}, false}, Base.Broadcast.Broadcasted{Nothing, Tuple{Base.OneTo{Int64}}, typeof(identity), Tuple{Int64}}, Int64}}}, compiler::typeof(CUDA.cufunction_compile), linker::typeof(CUDA.cufunction_link))
@ GPUCompiler ~.julia\packages\GPUCompiler\XwWPj\src\cache.jl:0
[10] cufunction(f::GPUArrays.var"#broadcast_kernel#12", tt::Type{Tuple{CUDA.CuKernelContext, SubArray{Int64, 1, CuDeviceVector{Int64, 1}, Tuple{Vector{Int64}}, false}, Base.Broadcast.Broadcasted{Nothing, Tuple{Base.OneTo{Int64}}, typeof(identity), Tuple{Int64}}, Int64}}; name::Nothing, kwargs::Base.Iterators.Pairs{Union{}, Union{}, Tuple{}, NamedTuple{(), Tuple{}}})
@ CUDA ~.julia\packages\CUDA\qEV3Y\src\compiler\execution.jl:294
[11] cufunction
@ ~.julia\packages\CUDA\qEV3Y\src\compiler\execution.jl:288 [inlined]
[12] macro expansion
@ ~.julia\packages\CUDA\qEV3Y\src\compiler\execution.jl:102 [inlined]
[13] #launch_heuristic#280
@ ~.julia\packages\CUDA\qEV3Y\src\gpuarrays.jl:17 [inlined]
[14] launch_heuristic
@ ~.julia\packages\CUDA\qEV3Y\src\gpuarrays.jl:17 [inlined]
[15] copyto!
@ ~.julia\packages\GPUArrays\WV76E\src\host\broadcast.jl:66 [inlined]
[16] copyto!
@ ~.julia\packages\GPUArrays\WV76E\src\host\broadcast.jl:76 [inlined]
[17] materialize!
@ .\broadcast.jl:894 [inlined]
[18] materialize!(dest::SubArray{Int64, 1, CuArray{Int64, 1}, Tuple{Vector{Int64}}, false}, bc::Base.Broadcast.Broadcasted{Base.Broadcast.DefaultArrayStyle{0}, Nothing, typeof(identity), Tuple{Int64}})
@ Base.Broadcast .\broadcast.jl:891
[19] top-level scope
@ REPL[5]:1

Version info:

CUDA toolkit 11.1.1, artifact installation
CUDA driver 11.2.0
NVIDIA driver 461.92.0

Libraries:

• CUBLAS: 11.3.0
• CURAND: 10.2.2
• CUFFT: 10.3.0
• CUSOLVER: 11.0.1
• CUSPARSE: 11.3.0
• CUPTI: 14.0.0
• NVML: 11.0.0+461.92
• CUDNN: 8.10.0 (for CUDA 11.2.0)
• CUTENSOR: 1.2.2 (for CUDA 11.1.0)

Toolchain:

• Julia: 1.6.0
• LLVM: 11.0.1
• 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
• Device 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

1 device:
0: GeForce RTX 2080 Ti (sm_75, 8.513 GiB / 11.000 GiB available)

I just fresh installed the julia 1.6 stable, and installed the CUDA.jl 2.6.2, on Windows 10 if that matters.

I read about this issue is fixed and merged into the 2.6.2 with PR.

[maleadt]

merged into the 2.6.2 with PR.

eeb7027 is only on master. Please try the development branch first.

[zhangcx93]

Please try the development branch first.

It works on master, thanks.

btw according to changelog of 2.6.2, #724 is merged. since it's only on master, maybe it shoud not show up in changelog of that version release?

[maleadt]

I agree; that's a tagbot issue: JuliaRegistries/TagBot#181