[SYCL][CUDA][MATRIX] Remove using namespace experimental from headers

[JackAKirk]

Removed using namespace experimental - replaced with fully qualified names.

cc @vladimirlaz @bader

This PR fixes #5213. I chose to use fully qualified names instead of moving local detail to sycl::ext::oneapi::experimental::matrix::detail, although this second option seems the most sensible to me: however I didn't move detail at the moment for consistency with the intel matrix extension namespace use (which also coincides with the standard practice of wider dpc++ as far as I can tell).

Please be aware that using namespace experimental; is also used on line 201 of matrix-aot-amx.hpp, which could lead to similar problems in the future.

Signed-off-by: jack.kirk jack.kirk@codeplay.com

[andreyfe1]

Please be aware that using namespace experimental; is also used on line 201 of matrix-aot-amx.hpp, which could lead to similar problems in the future.

Would it be easier to fix the issue for matrix-aot-amx.hpp at this PR as well since the issue is the same? As I checked for whole repo only matrix-aot-amx.hpp and matrix-tensorcore.hpp have using namespace experimental

[bader]

@JackAKirk, please, open a PR in llvm-test-suite repository enabling Matrix/joint_matrix_tensorcore_double.cpp on CUDA.

[bader]

Please be aware that using namespace experimental; is also used on line 201 of matrix-aot-amx.hpp, which could lead to similar problems in the future.

@yubingex007-a11y, FYI.

[yubingex007-a11y]

Please be aware that using namespace experimental; is also used on line 201 of matrix-aot-amx.hpp, which could lead to similar problems in the future.

@yubingex007-a11y, FYI.

Sure I will create a PR to handle that. Sorry for inconvenience.

[JackAKirk]

@JackAKirk, please, open a PR in llvm-test-suite repository enabling Matrix/joint_matrix_tensorcore_double.cpp on CUDA.

Here it is, intel/llvm-test-suite#680, thanks.

[bader]

/verify with intel/llvm-test-suite#680