Basic/barrier_order.cpp hangs on opencl:gpu

[steffenlarsen]

The new SYCL/Basic/barrier_order.cpp (intel/llvm-test-suite#1122) test in intel/llvm-test-suite is failing sporadically in pre-commit CI. See for example https://github.com/intel/llvm/actions/runs/3427185876/jobs/5710264332.

******************** TEST 'SYCL :: Basic/barrier_order.cpp' FAILED ********************
Script:
--
: 'RUN: at line 1';    /__w/llvm/llvm/toolchain/bin/clang++     -Xsycl-target-backend=amdgcn-amd-amdhsa --offload-arch=gfx1031 -fsycl -fsycl-targets=amdgcn-amd-amdhsa /__w/llvm/llvm/llvm_test_suite/SYCL/Basic/barrier_order.cpp -o /__w/llvm/llvm/build/SYCL/Basic/Output/barrier_order.cpp.tmp.out
: 'RUN: at line 2';   true /__w/llvm/llvm/build/SYCL/Basic/Output/barrier_order.cpp.tmp.out
: 'RUN: at line 3';    env ONEAPI_DEVICE_SELECTOR=ext_oneapi_hip:gpu  /__w/llvm/llvm/build/SYCL/Basic/Output/barrier_order.cpp.tmp.out
--
Exit Code: 1

Command Output (stdout):
--
$ ":" "RUN: at line 1"
note: command had no output on stdout or stderr
$ "/__w/llvm/llvm/toolchain/bin/clang++" "-Xsycl-target-backend=amdgcn-amd-amdhsa" "--offload-arch=gfx1031" "-fsycl" "-fsycl-targets=amdgcn-amd-amdhsa" "/__w/llvm/llvm/llvm_test_suite/SYCL/Basic/barrier_order.cpp" "-o" "/__w/llvm/llvm/build/SYCL/Basic/Output/barrier_order.cpp.tmp.out"
note: command had no output on stdout or stderr
$ ":" "RUN: at line 2"
note: command had no output on stdout or stderr
$ "true" "/__w/llvm/llvm/build/SYCL/Basic/Output/barrier_order.cpp.tmp.out"
note: command had no output on stdout or stderr
$ ":" "RUN: at line 3"
note: command had no output on stdout or stderr
$ "env" "ONEAPI_DEVICE_SELECTOR=ext_oneapi_hip:gpu" "/__w/llvm/llvm/build/SYCL/Basic/Output/barrier_order.cpp.tmp.out"
# command output:
1111111111111111111111111111111111111111111111111111111111111111
0111111111111111111111111111111111111111111111111111111111111111
failed

error: command failed with exit status: 1

--

[steffenlarsen]

Temporarily disabled for HIP in intel/llvm-test-suite#1375.

[zjin-lcf]

1 std::cout << std::bitset<8 * sizeof(unsigned long long)>(*x) << std::endl;

2 q.wait_and_throw();

Is line2 in front of line1 in the test ?

[romanovvlad]

1 std::cout << std::bitset<8 * sizeof(unsigned long long)>(*x) << std::endl;

2 q.wait_and_throw();

Is line2 in front of line1 in the test ?

Right, the access to x should be after wait_and_throw. Could be the root cause.

[Fznamznon]

The test also fails on CUDA.
https://github.com/intel/llvm/actions/runs/3477880483/jobs/5814961418
https://github.com/intel/llvm/actions/runs/3480131323/jobs/5820955932