Repository navigation
Switch the disabled shared_memory_barrier tests to c2h - #12150
Conversation
|
Important Review skippedReview was skipped as selected files did not have any reviewable changes. ⚙️ Run configuration
You can disable this status message by setting the Use the checkbox below for a quick retry:
No actionable comments were generated in the recent review. 🎉 ℹ️ Recent review info
📝 Summary
Priority: ⬇️ Low Change: Other Merge Risk: 🟡 Moderate · up to
|
![coderabbitai[bot]](https://avatars.githubusercontent.com/in/347564?s=60&v=4){kind=small}
There was a problem hiding this comment.
Actionable comments posted: 1
ℹ️ Review info
⚙️ Run configuration
- Configuration used: Repository: NVIDIA/cccl/.coderabbit.yaml
- Review profile: CHILL
- Plan: Enterprise
- Run ID:
8d433a83-ef87-44a3-bc32-a8ebdf9b7036
📒 Files selected for processing: 1
libcudacxx/test/libcudacxx/cuda/barrier/shared_memory_barrier.cu
Included review availability: This review used your included allowance. Your plan provides up to 12 included reviews per hour; 11 remain after this review.
| NV_IF_TARGET(NV_PROVIDES_SM_90, | ||
| (test_shared_memory_barrier_choreography<Scope>(); test_shared_memory_barrier_common_extensions<Scope>(); | ||
| test_shared_memory_barrier_sm90_extensions<Scope>();)) |
There was a problem hiding this comment.
🎯 Functional Correctness | 🟠 Major | 🏗️ Heavy lift
🔎 Supported by static analysis
🏁 Script executed:
#!/bin/bash
# Inspect architecture settings for this test; also check its generated
# compilation command for an SM90-or-newer image.
rg -n -C 3 'CMAKE_CUDA_ARCHITECTURES|CUDA_ARCHITECTURES|gencode|compute_90|sm_90' libcudacxx/testRepository: NVIDIA/cccl
Length of output: 23631
🏁 Script executed:
set -u
printf '%s\n' '--- test references and registration ---'
rg -n -F --glob 'CMakeLists.txt' -e 'shared_memory_barrier' -e 'LIBCUDACXX_TEST' libcudacxx/test libcudacxx || test "$?" -eq 1
printf '%s\n' '--- libcudacxx test CMake context ---'
sed -n '1,180p' libcudacxx/test/libcudacxx/CMakeLists.txt
printf '%s\n' '--- relevant test source entrypoints and guards ---'
nl -ba libcudacxx/test/libcudacxx/cuda/barrier/shared_memory_barrier.cu | sed -n '1,90p;520,600p'
printf '%s\n' '--- architecture propagation and defaults ---'
rg -n -C 5 -e 'CMAKE_CUDA_ARCHITECTURES' -e 'LIBCUDACXX.*ARCH' -e 'compute_archs' libcudacxx/test cmake CMakeLists.txt || test "$?" -eq 1Repository: NVIDIA/cccl
Length of output: 34173
🏁 Script executed:
set -u
printf '%s\n' '--- cccl_add_executable definition and callers ---'
rg -n -C 8 -F -- 'function(cccl_add_executable' . || test "$?" -eq 1
rg -n -C 4 -F -- 'cccl_add_executable(' libcudacxx/test cmake CMakeLists.txt | head -240
printf '%s\n' '--- CUDA architecture defaults and target properties ---'
rg -n -C 6 -e 'CMAKE_CUDA_ARCHITECTURES' -e 'CUDA_ARCHITECTURES' cmake CMakeLists.txt libcudacxx/CMakeLists.txt libcudacxx/test/CMakeLists.txt | head -320Repository: NVIDIA/cccl
Length of output: 8895
🏁 Script executed:
set -u
printf '%s\n' '--- complete cccl_add_executable implementation ---'
nl -ba cmake/CCCLAddExecutable.cmake | sed -n '1,180p'
printf '%s\n' '--- supported architecture invocations and test build configuration ---'
rg -n -C 5 -e 'all-major-cccl' -e 'all-cccl' -e 'CMAKE_CUDA_ARCHITECTURES=' -e 'CUDA_ARCHITECTURES' .github ci cmake docs libcudacxx 2>/dev/null | head -400Repository: NVIDIA/cccl
Length of output: 31258
🏁 Script executed:
set -u
rg -n -F -- 'function(cccl_configure_target' cmake
file=$(rg -l -F -- 'function(cccl_configure_target' cmake | head -1)
if [ -n "$file" ]; then
nl -ba "$file" | sed -n '1,240p'
fiRepository: NVIDIA/cccl
Length of output: 2491
important: Ensure the shared-memory barrier target contains SM90 device code.
test_shared_memory_barrier_kernel removes all test calls when compiled below SM90. The C2H test checks only the host device capability. A lower-architecture PTX image can therefore run on an SM90 device while executing an empty kernel, and both sections can pass without testing the barrier.
Require an SM90-or-newer compilation target for this test, or skip configurations that cannot provide one. Do not rely on the host device check to select the device code.
Source: Path instructions
This comment has been minimized.
This comment has been minimized.
| assert(status.complete()); | ||
| assert(!status.has_report()); | ||
| REQUIRE_DEVICE(status.complete()); | ||
| REQUIRE_DEVICE(!status.has_report()); |
This comment has been minimized.
This comment has been minimized.
This comment has been minimized.
This comment has been minimized.
🥳 CI Workflow Results🟩 Finished in 1h 10m: Pass: 100%/147 | Total: 1d 13h | Max: 1h 10m | Hits: 99%/557909See results here. |
|
Successfully created backport PR for |
shared_memory_barrier lit test is currently disabled because of eager loading in lit and older drivers on some CI machines leading to failure before the main even runs. This PR switches the test to c2h so we can check the driver version and skip the test