oneAPI error when running basic example on Intel internal GPU

Specific Domains GPU
1

Hi there, I just installed oneAPI just as an experiment.

I’m afraid I was not able to run all the examples and searching for a solution was not solving the issue. So I kindly ask for a hint to get this running, or just to give some insights or point to another direction.

When running some code, there is indicated this:

WARNING: both oneL0 and LLVM export “arguments”; uses of it in module oneAPI must be qualified

Also the ] test oneAPI is not running through.

More info and further down the final error:

julia> versioninfo()
Julia Version 1.6.3
Commit ae8452a9e0 (2021-09-23 17:34 UTC)
Platform Info:
  OS: Linux (x86_64-pc-linux-gnu)
  CPU: 11th Gen Intel(R) Core(TM) i7-1185G7 @ 3.00GHz
  WORD_SIZE: 64
  LIBM: libopenlibm
  LLVM: libLLVM-11.0.1 (ORCJIT, tigerlake)

julia> using oneAPI

julia> using oneAPI.oneL0

julia> drv = first(dev
device         device_alloc    devices
device!        device_barrier  devnull
julia> drv = first(drivers());

julia> dev = first(devices(drv))
ZeDevice(GPU, vendor 0x8086, device 0x9a49): Intel(R) Iris(R) Xe Graphics [0x9a49]

julia> using oneAPI, oneAPI.oneL0

julia> drv = first(drivers());

julia> ctx = ZeContext(drv);

julia> dev = first(devices(drv))
ZeDevice(GPU, vendor 0x8086, device 0x9a49): Intel(R) Iris(R) Xe Graphics [0x9a49]

julia> compute_properties(dev)
(maxTotalGroupSize = 512, maxGroupSizeX = 512, maxGroupSizeY = 512, maxGroupSizeZ = 512, maxGroupCountX = 4294967295, maxGroupCountY = 4294967295, maxGroupCountZ = 4294967295, maxSharedLocalMemory = 65536, subGroupSizes = (8, 16, 32))

julia> queue = ZeCommandQueue(ctx, dev);

julia> execute!(queue) do list
         append_barrier!(list)
       end

julia> function kernel()
         barrier()
         return
       end
kernel (generic function with 1 method)

julia> @oneapi items=1 kernel()
oneAPI.HostKernel{kernel, Tuple{}}(ZeKernel(ZeModule(Ptr{oneAPI.oneL0._ze_module_handle_t} @0x000000000601aeb0, ZeContext(Ptr{oneAPI.oneL0._ze_context_handle_t} @0x0000000002563f40, ZeDriver(Ptr{oneAPI.oneL0._ze_driver_handle_t} @0x00000000023f9620)), ZeDevice(Ptr{oneAPI.oneL0._ze_device_handle_t} @0x0000000002ee6c20, ZeDriver(Ptr{oneAPI.oneL0._ze_driver_handle_t} @0x00000000023f9620))), Ptr{oneAPI.oneL0._ze_kernel_handle_t} @0x00000000038145c0))

julia> @device_code_llvm @oneapi items=1 kernel()
; CompilerJob of kernel kernel() for GPUCompiler.SPIRVCompilerTarget
define spir_kernel void @_Z17julia_kernel_3621() local_unnamed_addr {
entry:
;  @ REPL[16]:2 within `kernel'
; ┌ @ /home/bd/.julia/packages/oneAPI/PolvD/src/device/opencl/synchronization.jl:9 within `barrier' @ /home/bd/.julia/packages/oneAPI/PolvD/src/device/opencl/synchronization.jl:9
; │┌ @ /home/bd/.julia/packages/oneAPI/PolvD/src/device/utils.jl:77 within `macro expansion' @ /home/bd/.julia/packages/LLVM/wnejv/src/interop/pointer.jl:219
; ││┌ @ /home/bd/.julia/packages/LLVM/wnejv/src/interop/pointer.jl:120 within `_typed_llvmcall'
; │││┌ @ /home/bd/.julia/packages/LLVM/wnejv/src/interop/pointer.jl:120 within `macro expansion' @ /home/bd/.julia/packages/LLVM/wnejv/src/interop/base.jl:45
      call void @_Z7barrierj(i32 0)
      ret void
; └└└└
}

julia> @device_code_spirv @oneapi items=1 kernel()
// CompilerJob of kernel kernel() for GPUCompiler.SPIRVCompilerTarget

; SPIR-V
; Version: 1.0
; Generator: Khronos LLVM/SPIR-V Translator; 14
; Bound: 9
; Schema: 0
               OpCapability Addresses
               OpCapability Kernel
          %1 = OpExtInstImport "OpenCL.std"
               OpMemoryModel Physical64 OpenCL
               OpEntryPoint Kernel %4 "_Z17julia_kernel_3659"
               OpSource OpenCL_C 200000
               OpName %entry "entry"
       %uint = OpTypeInt 32 0
     %uint_2 = OpConstant %uint 2
     %uint_0 = OpConstant %uint 0
       %void = OpTypeVoid
          %3 = OpTypeFunction %void
          %4 = OpFunction %void None %3
      %entry = OpLabel
               OpControlBarrier %uint_2 %uint_2 %uint_0
               OpReturn
               OpFunctionEnd

Here the error:

julia> a = oneArray(rand(Float32, 2,2))
2×2 oneArray{Float32, 2}:
 0.149671  0.32529
 0.948333  0.172246

julia> a .+ 1
ERROR: UndefVarError: arguments not defined
Stacktrace:
  [1] onecall(::ZeKernel, ::Type, ::oneDeviceMatrix{Float32, 1}, ::Vararg{Any, N} where N; groups::Int64, items::Int64, queue::ZeCommandQueue)
    @ oneAPI ~/.julia/packages/oneAPI/PolvD/src/compiler/execution.jl:169
  [2] macro expansion
    @ ~/.julia/packages/oneAPI/PolvD/src/compiler/execution.jl:121 [inlined]
  [3] call(::oneAPI.HostKernel{GPUArrays.var"#broadcast_kernel#17"(), Tuple{oneAPI.oneKernelContext, oneDeviceMatrix{Float32, 1}, Base.Broadcast.Broadcasted{Nothing, Tuple{Base.OneTo{Int64}, Base.OneTo{Int64}}, typeof(+), Tuple{Base.Broadcast.Extruded{oneDeviceMatrix{Float32, 1}, Tuple{Bool, Bool}, Tuple{Int64, Int64}}, Int64}}, Int64}}, ::oneAPI.oneKernelContext, ::oneDeviceMatrix{Float32, 1}, ::Base.Broadcast.Broadcasted{Nothing, Tuple{Base.OneTo{Int64}, Base.OneTo{Int64}}, typeof(+), Tuple{Base.Broadcast.Extruded{oneDeviceMatrix{Float32, 1}, Tuple{Bool, Bool}, Tuple{Int64, Int64}}, Int64}}, ::Int64; call_kwargs::Base.Iterators.Pairs{Symbol, Int64, Tuple{Symbol, Symbol}, NamedTuple{(:items, :groups), Tuple{Int64, Int64}}})
    @ oneAPI ~/.julia/packages/oneAPI/PolvD/src/compiler/execution.jl:93
  [4] (::oneAPI.HostKernel{GPUArrays.var"#broadcast_kernel#17"(), Tuple{oneAPI.oneKernelContext, oneDeviceMatrix{Float32, 1}, Base.Broadcast.Broadcasted{Nothing, Tuple{Base.OneTo{Int64}, Base.OneTo{Int64}}, typeof(+), Tuple{Base.Broadcast.Extruded{oneDeviceMatrix{Float32, 1}, Tuple{Bool, Bool}, Tuple{Int64, Int64}}, Int64}}, Int64}})(::oneAPI.oneKernelContext, ::Vararg{Any, N} where N; kwargs::Base.Iterators.Pairs{Symbol, Int64, Tuple{Symbol, Symbol}, NamedTuple{(:items, :groups), Tuple{Int64, Int64}}})
    @ oneAPI ~/.julia/packages/oneAPI/PolvD/src/compiler/execution.jl:179
  [5] macro expansion
    @ ~/.julia/packages/oneAPI/PolvD/src/compiler/execution.jl:53 [inlined]
  [6] gpu_call(::oneAPI.oneArrayBackend, f::Function, args::Tuple{oneArray{Float32, 2}, Base.Broadcast.Broadcasted{Nothing, Tuple{Base.OneTo{Int64}, Base.OneTo{Int64}}, typeof(+), Tuple{Base.Broadcast.Extruded{oneArray{Float32, 2}, Tuple{Bool, Bool}, Tuple{Int64, Int64}}, Int64}}, Int64}, threads::Int64, blocks::Int64; name::Nothing)
    @ oneAPI ~/.julia/packages/oneAPI/PolvD/src/gpuarrays.jl:30
  [7] #gpu_call#1
    @ ~/.julia/packages/GPUArrays/3sW6s/src/device/execution.jl:76 [inlined]
  [8] copyto!
    @ ~/.julia/packages/GPUArrays/3sW6s/src/host/broadcast.jl:69 [inlined]
  [9] copyto!
    @ ./broadcast.jl:936 [inlined]
 [10] copy
    @ ~/.julia/packages/GPUArrays/3sW6s/src/host/broadcast.jl:47 [inlined]
 [11] materialize(bc::Base.Broadcast.Broadcasted{oneAPI.oneArrayStyle{2}, Nothing, typeof(+), Tuple{oneArray{Float32, 2}, Int64}})
    @ Base.Broadcast ./broadcast.jl:883
 [12] top-level scope
    @ REPL[25]:1

Thank you so much for any information on this matter, kind regards.

2

I have tested with the latest master:
add https://github.com/JuliaGPU/oneAPI.jl#master

And my result ( gen8 cpu; julia 1.6.1; ubuntu 20.04 )

julia> using oneAPI

julia> a = oneArray(rand(Float32, 2,2))
2×2 oneArray{Float32, 2}:
 0.233732  0.297938
 0.924481  0.541995

julia> a .+ 1
2×2 oneArray{Float32, 2}:
 1.23373  1.29794
 1.92448  1.54199

julia>
3

Oh yeah!! Can confirm. Thank you very much, I will further investigate.

Kind regards!

1 Like