[QST] How do I import the `cutlass::gemm::kernel::GemmUniversal`

Following up on #1291, I am finally making the time to properly understand Cutlass. I can build and run the examples in the repo, but I am having a lot of difficulty understanding the type errors I am getting when I try to import the GemmUniversal classes. For example...

#include <iostream>

#include "cutlass/cutlass.h"
#include "cute/tensor.hpp"
#include "cute/atom/mma_atom.hpp"

#include "cutlass/numeric_types.h"

#include "cutlass/gemm/device/gemm_universal_adapter.h"
#include "default_gemm_configuration.hpp"

using namespace cute;

/////////////////////////////////////////////////////////////////////////////////////////////////

using Config = cutlass::gemm::device::DefaultGemmConfigurationToCutlass3Types<
	cutlass::arch::OpClassSimt, cutlass::arch::Sm50,
	double, cutlass::layout::ColumnMajor,
	double, cutlass::layout::ColumnMajor,
	double, cutlass::layout::ColumnMajor,
	double>;

using GemmKernel = cutlass::gemm::kernel::GemmUniversal<
	Shape<int, int, int, int>,
	Config::CollectiveMainloop,
	Config::CollectiveEpilogue
>;

using Gemm = cutlass::gemm::device::GemmUniversalAdapter<GemmKernel>;

I've isolated this piece of code from one of the tests and am trying to bring it into another project without success. Even when I seemingly have all the necessary includes, I see something like this...

Probably I shouldn't be trying to import test harness code, but the trouble is that there are no clean examples how to use the universal kernel.

https://github.com/NVIDIA/cutlass/blob/main/media/docs/quickstart.md#launching-a-gemm-kernel-using-cutlass-30-or-newer

The quickstart one is clean but doesn't typecheck for me.

There are some examples like the 48_hopper_warp_specialized_gemm which do have use examples of it, but when I get rid of the device arch checking macros, I just get a ton of type errors. Those kinds of examples aren't usable.

I am going to try recompiling the Cutlass library for 9.0 devices just to see if something changes.

But otherwise, could I get some help to start me off with Cutlass? Thanks.

Start with media/docs/quickstart.md and compare its universal-kernel example with the 48_hopper_warp_specialized_gemm examples and the included test code. Reproduce the reported type errors in the shown GemmUniversal setup, then clarify the required includes and usage in a standalone example that typechecks.