Conversion error in drawing the Julia set about cudart.jl

[moon6pence]

I didn't test the code yet, but you have to copy Array to CudaArray to use in GPU.

Use CudaArray like this before calling the kernel:

        d_q = CudaArray(q)
        CUDArt.launch(juliafn, length(d_q), 1, (d_q, out, uint16(maxiter)))

[mindbound]

That fails with:

_ERROR: Invalid value
in checkdrv at /home/???/.julia/v0.3/CUDArt/src/module.jl:6
in launch at /home/???/.julia/v0.3/CUDArt/src/execute.jl:16
in anonymous at /home/???/julia.jl:16
in CuModule at /home/???/.julia/v0.3/CUDArt/src/module.jl:39
in anonymous at /home/???/julia.jl:12
in devices at /home/???/.julia/v0.3/CUDArt/src/device.jl:57
in devices at /home/???/.julia/v0.3/CUDArt/src/device.jl:49
in julia at /home/???/julia.jl:9
in include at /usr/bin/../lib/julia/sys.so
in include_from_node1 at loading.jl:128
in process_options at /usr/bin/../lib/julia/sys.so
in start at /usr/bin/../lib/julia/sys.so
while loading /home/???/julia.jl, in expression starting on line 24

[timholy]

@moon6pence is right that you needed to move the memory to the GPU first. His suggestion seemed to have overcome the first problem, but now there seems to be something else going on. As a first thing to check, does Pkg.test("CUDArt") pass for you?

[mindbound]

Yes, the tests do pass.

[moon6pence]

In launching cuda kernel,

CUDArt.launch(juliafn, length(d_q), 1, (d_q, out, uint16(maxiter)))

You have to specify proper block size rather than 1. ex) 256, 512, 1024, ...
It may be work with block size 1 but definitely run slow.

CUDArt.launch(juliafn, div(length(d_q), 512), 512, (d_q, out, uint16(maxiter)))

However, I couldn't get the julia image.
Keep checking julia.cu generates proper julia set.