CTS group and nd_item category compfail 'Unresolved extern function '_Z22__spirv_GroupAsyncCopyjPU3AS1DF16_PU3AS3KDF16_mm9ocl_event'' with latest SYCL build and CUDA

[XunZhangIntel]

Describe the bug
CTS group and nd_item category compfail with CUDA backend on latest SYCL builds and latest CUDA version. See logs below:

FAILED: bin/test_nd_item
: && clang++ -ffp-model=precise  -fsycl -fsycl-unnamed-lambda -Wno-deprecated-declarations  -Wl,-no-relax    -fsycl -fsycl-targets=nvptx64-nvidia-cuda-sycldevice -sycl-std=2020 -Xsycl-target-backend --cuda-gpu-arch=sm_50 tests/nd_item/CMakeFiles/test_nd_item_objects.dir/nd_item_async_work_group_copy_fp16.cpp.o tests/common/CMakeFiles/main_function_object.dir/main.cpp.o -o bin/test_nd_item  -Wl,-rpath util/libutil.a oclmath/liboclmath.a libOpenCL.so && :
clang++: warning: argument 'nvptx64-nvidia-cuda-sycldevice' is deprecated, use 'nvptx64' instead [-Wdeprecated]
ptxas fatal   : Unresolved extern function '_Z22__spirv_GroupAsyncCopyjPU3AS1DF16_PU3AS3KDF16_mm9ocl_event'
llvm-foreach:
clang++: error: ptxas command failed with exit code 255 (use -v to see invocation)
clang version 14.0.0 (https://github.com/intel/llvm.git e61dcc6d1c11573b3631e7d184e43c60c1e6a0b2)
Target: x86_64-unknown-linux-gnu
Thread model: posix
InstalledDir:
clang++: note: diagnostic msg: Error generating preprocessed source(s).
ninja: build stopped: cannot make progress due to previous errors.

To Reproduce
Note: Due to a recent uplift of minimal supported C++ standard, one should modify this line to require C++17.

1. Get SYCL with CUDA support (GSG) and SYCL-CTS
2. Build CTS tests

cd SYCL-CTS
mkdir build && cd build
cmake -GNinja  -DSYCL_IMPLEMENTATION=Intel_SYCL -DINTEL_SYCL_ROOT=<path to built sycl> -Dopencl_platform_name=nvidia -Dopencl_device_name=opencl_gpu -DCMAKE_BUILD_TYPE=Release -DINTEL_SYCL_FLAGS="-Xsycl-target-backend;--cuda-gpu-arch=sm_50" -DCMAKE_EXE_LINKER_FLAGS=" -Wl,-no-relax " -DINTEL_SYCL_TRIPLE=nvptx64-nvidia-cuda-sycldevice -DSYCL_CTS_ENABLE_OPENCL_INTEROP_TESTS=Off -DSYCL_CTS_ENABLE_DOUBLE_TESTS=On -DSYCL_CTS_ENABLE_HALF_TESTS=On ..
ninja test_group
ninja test_nd_item

Environment (please complete the following information):
OS version: Ubuntu 20.04.3 LTS
CUDA version:

nvidia-smi

+-----------------------------------------------------------------------------+
| NVIDIA-SMI 470.57.02    Driver Version: 470.57.02    CUDA Version: 11.4     |
|-------------------------------+----------------------+----------------------+

[AidanBeltonS]

I have been able to replicate the error.
The unresolved symbol only exists for 16-bit floating points versions of GroupAsyncCopy.

The issue is linked to a mismatch between the mangled name generated from the application
and the name generated by libclc.

Application Name: _Z22__spirv_GroupAsyncCopyjPU3AS3DF16_PU3AS1KDF16_mm9ocl_event
Libclc Name: _Z22__spirv_GroupAsyncCopyjPU3AS3DF16_PU3AS1KDF16_ll9ocl_event

The difference in the name is because the functions fp16 definitions use __clc_size_t to indicate the number of
entries to copy, while all other defintions use size_t. __clc_size_t is a signed 64-bit int, size_t is unsigned.
This creates the difference in the functions mangled name.

I think the simple solution of changing __clc_size_t to unsigned is the best way to resolve this.