Support for kernels with keyword arguments

[gmgunter]

Hi,

Thanks for this nice package. I'm just getting started but ran into an issue when trying to GPU-ize an existing function that takes keyword arguments. Is this possible? I'm getting a syntax error when passing such a function as an argument to the @cuda macro (MWE below).

This simple kernel works fine when I pass a positional argument:

julia> using CUDA: @cuda, @cushow

julia> function kernel1(foo::Integer)
           @cushow foo
           return
       end
kernel1 (generic function with 1 method)

julia> @cuda kernel1(1)
foo = 1
CUDA.HostKernel for kernel1(Int64)

But I get a syntax error when I pass a keyword argument instead:

julia> function kernel2(; foo::Integer)
           @cushow foo
           return
       end
kernel2 (generic function with 1 method)

julia> @cuda kernel2(foo=1)
ERROR: syntax: invalid syntax foo = 1
Stacktrace:
 [1] top-level scope
   @ REPL[5]:1

Version info

julia> versioninfo()
Julia Version 1.11.4
Commit 8561cc3d68d (2025-03-10 11:36 UTC)
Build Info:
  Official https://julialang.org/ release
Platform Info:
  OS: Linux (x86_64-linux-gnu)
  CPU: 256 × AMD EPYC 7713 64-Core Processor
  WORD_SIZE: 64
  LLVM: libLLVM-16.0.6 (ORCJIT, znver3)
Threads: 1 default, 0 interactive, 1 GC (on 256 virtual cores)

julia> CUDA.versioninfo()
CUDA runtime 12.4, local installation
CUDA driver 12.5
NVIDIA driver 555.42.6

CUDA libraries:
- CUBLAS: 12.4.2
- CURAND: 10.3.5
- CUFFT: 11.2.0
- CUSOLVER: 11.6.0
- CUSPARSE: 12.3.0
- CUPTI: 2024.1.0 (API 22.0.0)
- NVML: 12.0.0+555.42.6

Julia packages:
- CUDA: 5.7.0
- CUDA_Driver_jll: 0.12.0+0
- CUDA_Runtime_jll: 0.16.0+0
- CUDA_Runtime_Discovery: 0.3.5

Toolchain:
- Julia: 1.11.4
- LLVM: 16.0.6

Preferences:
- CUDA_Runtime_jll.local: true

1 device:
  0: NVIDIA A800 40GB Active (sm_80, 38.224 GiB / 40.000 GiB available)

[maleadt]

No, this is currently not supported.

[gmgunter]

Alright, thanks. Feel free to close this.