[BUG] LinkableCode not working with CtypesLinker

[gmarkall]

Describe the bug
Use of LinkableCode subclasses (e.g. PTXSource) is only working with pynvjitlink.

Steps/Code to reproduce bug

Executing the following:

from numba import cuda
import numpy as np


ptx = cuda.PTXSource("""
.version 6.5
.target sm_52
.address_size 64

.visible .func  (.param .b32 status) double_an_int32(
        .param .b64 retval,
        .param .b32 arg
)
{
        .reg .b32       %r<4>;
        .reg .b64       %retvalptr;

        ld.param.u64    %retvalptr, [retval];
        ld.param.u32    %r1, [arg];
        shl.b32         %r2, %r1, 1;
        st.u32          [%retvalptr], %r2;
        mov.u32         %r3, 0;
        st.param.b32    [status], %r3;
        ret;
}""".encode())

double_an_int32 = cuda.declare_device("double_an_int32",
                                      "int32(int32)")


@cuda.jit(link=[ptx])
def k(x):
    x[0] = double_an_int32(x[0])


x = np.ones(1, dtype=np.int32)
print(x[0])
k[1, 1](x)
print(x[0])

results in

$ python repro.py 
1
/home/gmarkall/numbadev/numba-cuda/numba_cuda/numba/cuda/dispatcher.py:579: NumbaPerformanceWarning: Grid size 1 will likely result in GPU under-utilization due to low occupancy.
  warn(NumbaPerformanceWarning(msg))
Traceback (most recent call last):
  File "/home/gmarkall/numbadev/issues/9470/repro.py", line 38, in <module>
    k[1, 1](x)
  File "/home/gmarkall/numbadev/numba-cuda/numba_cuda/numba/cuda/dispatcher.py", line 582, in __call__
    return self.dispatcher.call(args, self.griddim, self.blockdim,
           ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
  File "/home/gmarkall/numbadev/numba-cuda/numba_cuda/numba/cuda/dispatcher.py", line 724, in call
    kernel = _dispatcher.Dispatcher._cuda_call(self, *args)
             ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
  File "/home/gmarkall/numbadev/numba-cuda/numba_cuda/numba/cuda/dispatcher.py", line 732, in _compile_for_args
    return self.compile(tuple(argtypes))
           ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
  File "/home/gmarkall/numbadev/numba-cuda/numba_cuda/numba/cuda/dispatcher.py", line 979, in compile
    kernel.bind()
  File "/home/gmarkall/numbadev/numba-cuda/numba_cuda/numba/cuda/dispatcher.py", line 238, in bind
    self._codelibrary.get_cufunc()
  File "/home/gmarkall/numbadev/numba-cuda/numba_cuda/numba/cuda/codegen.py", line 248, in get_cufunc
    cubin = self.get_cubin(cc=device.compute_capability)
            ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
  File "/home/gmarkall/numbadev/numba-cuda/numba_cuda/numba/cuda/codegen.py", line 227, in get_cubin
    self._link_all(linker, cc, ignore_nonlto=False)
  File "/home/gmarkall/numbadev/numba-cuda/numba_cuda/numba/cuda/codegen.py", line 192, in _link_all
    linker.add_file_guess_ext(path, ignore_nonlto)
  File "/home/gmarkall/numbadev/numba-cuda/numba_cuda/numba/cuda/cudadrv/driver.py", line 2758, in add_file_guess_ext
    self.add_data(
    ^^^^^^^^^^^^^
AttributeError: 'CtypesLinker' object has no attribute 'add_data'. Did you mean: 'add_ptx'?

Expected behavior

The code should execute to completion, producing:

$ python repro.py 
1
/home/gmarkall/numbadev/numba-cuda/numba_cuda/numba/cuda/dispatcher.py:579: NumbaPerformanceWarning: Grid size 1 will likely result in GPU under-utilization due to low occupancy.
  warn(NumbaPerformanceWarning(msg))
/home/gmarkall/numbadev/numba-cuda/numba_cuda/numba/cuda/cudadrv/devicearray.py:890: NumbaPerformanceWarning: Host array used in CUDA kernel will incur copy overhead to/from device.
  warn(NumbaPerformanceWarning(msg))
2

which it only does when NUMBA_CUDA_ENABLE_PYNVJITLINK=1.

Environment details (please complete the following information):

numba-cuda 0.2.0, Linux.

Additional context

It's not clear if all the Linkable Code classes should be usable without pynvjitlink, but the following should be done when resolving this issue:

• Audit the Linkable Code classes to determine which can be made to work with the Ctypes linker
• Add tests of Linkable Code subclasses that run with both the Ctypes and pynvjitlink linkers.

[kkraus14]

Closing as not planned since we are actively removing CtypesLinker