why does the code below work, but the alternative commented lines fail with
"invalid kernel argument"? i guess i am not understanding the Array class,
but it seems like it is intended to be an easier-to-use replacement for
explicit buffers?
ctx = cl.create_some_context()
queue = cl.CommandQueue(ctx)
a = n.array([0], dtype=n.int32)
# this works
a_dev = cl.Buffer(ctx, cl.mem_flags.WRITE_ONLY, 4)
# this alternative (plus below) fails
#a_dev = cla.to_device(queue, a)
prg = cl.Program(ctx, """
__kernel void test1(__global int* a) {
a[0] = 1;
}
""").build()
event = prg.test1(queue, (1,), None, a_dev)
event.wait()
# this works
cl.enqueue_copy(queue, a, a_dev)
# this alternative (plus above) fails
#a = a_dev.get()
print(a)
thanks,
andrew
_______________________________________________
PyOpenCL mailing list
[email protected]
http://lists.tiker.net/listinfo/pyopencl
