Wenju He

@AllanZyne which part of this PR requires gpu driver uplift as mentioned in https://github.com/intel/llvm/pull/14720#issuecomment-2249798071?

Failure in [Build - Adapters on HW (L0, Debug, clang, clang++)](https://github.com/oneapi-src/unified-runtime/actions/runs/8415343675/job/23047034417?pr=1470#logs) seems unrelated to this pr. ``` Run cmake -B/home/test-user/actions-runner/_work/unified-runtime/unified-runtime/build -DCMAKE_C_COMPILER=clang -DCMAKE_CXX_COMPILER=clang++ -DCMAKE_BUILD_TYPE=Debug -DUR_ENABLE_TRACING=ON -DUR_DEVELOPER_MODE=ON -DUR_BUILD_TESTS=ON -DUR_BUILD_ADAPTER_L0=ON -DUR_DPCXX=/home/test-user/actions-runner/_work/unified-runtime/unified-runtime/dpcpp_compiler/bin/clang++ -DUR_SYCL_LIBRARY_DIR=/home/test-user/actions-runner/_work/unified-runtime/unified-runtime/dpcpp_compiler/lib --...

@nrspruit could you please review? SYCL PR is at https://github.com/intel/llvm/pull/13926

@kbenzie can this PR be added to read-to-merge queue? thank you.

> > @kbenzie can this PR be added to read-to-merge queue? thank you. > > I'll add the _ready to merge_ label but please do fix the merge conflicts on...

Remove WIP from PR title since this PR is ready for review?

+1 OpenCL function clGetKernelSuggestedLocalWorkSizeKHR's input global_work_size and output suggested_local_work_size have type size_t.

> Maybe we need to wait for cpu&gpu driver upgrading to support "__devicelib_exit" builtin. Recently upgraded gpu driver in https://github.com/intel/llvm/pull/14838 has support of "__devicelib_exit" builtin

one motivation is intel gpu backend requires calling conv to be consistent.

> > one motivation is intel gpu backend requires calling conv to be consistent. > > Thanks for adding this support. Looks good. One question. Is there documentation anywhere that...