Inactive kernel gets executed

[Richardk2n]

This is more of a general OpenCL question rather than a pyopencl question, but you seem to know a lot, so I hope asking here is ok.

Why does this not work:

import pyopencl as cl

plattform = cl.get_platforms()[0]
device = plattform.get_devices()[0]

ctx = cl.Context([device])
queue = cl.CommandQueue(ctx)

src = r"""
kernel void tet() {
    const long m = get_global_id(0);
	printf("ID %ld\n", m);
}
"""

src2 = r"""
kernel void other() {
    const long n = get_global_id(0);
    printf("Do not print %ld\n", n);
}
"""

p = cl.Program(ctx, src).compile()
p2 = cl.Program(ctx, src2).compile()

program = cl.link_program(ctx, [p, p2])

program.tet(queue, (2,), None)

I would expect this to output:

ID 0
ID 1

But what I get is:

ID 0
Do not print 4294967297
(null)

This seems to be specifically related to the printf statements. Commenting out the one in tet removes all output, commenting the one in other results in output as expected. Is what I am doing forbidden in general or should this work?

[inducer]

I can't think of a reason why other would be called, or why the printf in it would execute. In fact, executing your reproducer with pocl 6.0 gives what you list as the expected output.

[Richardk2n]

Thank you for your input. So I installed pocl and ran this code on my CPU, and it worked as expected. So does that mean, that the issue is in the opencl lib from amd/nvidia, and I have to ask them?

[inducer]

That's my read of the situation, yes. Note that pocl can also take to Nvidia GPUs through its CUDA support, though I am uncertain whether printf is supported there.