DeviceLib/built-ins/ext_native_math.cpp failed in pre-commit on CUDA backend

[steffenlarsen]

DeviceLib/built-ins/ext_native_math.cpp failed in pre-commit testing for #11559 on CUDA with the following error:

# .---command stderr------------
# | fatal error: error in backend: Cannot select: t154: v4i8 = bitcast t3
# |   t3: f[32](https://github.com/intel/llvm/actions/runs/6547288870/job/17779494384#step:21:33),ch = CopyFromReg t0, Register:f32 %7
# |     t2: f32 = Register %7
# | In function: _ZTSZZ18native_tanh_testerIN4sycl3_V16marrayIfLm3EEEEvNS1_5queueET_S5_S5_ENKUlRNS1_7handlerEE_clES7_EUlvE_
# | llvm-foreach: 
# | clang++: error: clang frontend command failed with exit code 70 (use -v to see invocation)
# | clang version 18.0.0 (https://github.com/intel/llvm 78[34](https://github.com/intel/llvm/actions/runs/6547288870/job/17779494384#step:21:35)1be24f512487961f4020b02b5bee06[37](https://github.com/intel/llvm/actions/runs/6547288870/job/17779494384#step:21:38)159d)
# | Target: x86_64-unknown-linux-gnu
# | Thread model: posix
# | InstalledDir: /__w/llvm/llvm/toolchain/bin
# | clang++: note: diagnostic msg: Error generating preprocessed source(s).
# `-----------------------------

See for example https://github.com/intel/llvm/actions/runs/6547288870/job/17779494384.

[JackAKirk]

Didn't reproduce it on A100 using cuda 11.7. CI is failing with A10 on 12.1

[JackAKirk]

We can xfail it before investigation to stop it causing issues with CI.

[pasaulais]

This now passes on tip sycl. Here is the reproducer:
oac-1586.ll.txt

The XFAIL has been reverted by: 49eb0e3

[npmiller]

Closing this as the test is now enabled