Disable flaky Unit_NonHost_Printf hip-tests for compiler promotion (#4081)#4245
Disable flaky Unit_NonHost_Printf hip-tests for compiler promotion (#4081)#4245
Conversation
…4081) Signed-off-by: Lenine Ajagappane <Lenine.Ajagappane@amd.com>
lajagapp
requested review from
geomin12,
kiranzangam,
kzhuravl,
ronlieb and
searlmc1
searlmc1
requested review from
ScottTodd,
marbre and
stellaraccident
and removed request for
ScottTodd,
marbre and
stellaraccident
| "Unit_hipMemsetASyncMulti", | ||
| "Unit_hipHostAlloc_AllocateMoreThanAvailGPUMemory", | ||
| "Unit_hipStreamValue_Write - TestParams<uint32_t, PtrType::DevicePtrToHost>", | ||
| # TODO(#4081): Flaky with compiler submodule update — subprocess aborted. |
There was a problem hiding this comment.
I see this linked a PR. can we create a GitHub issue that has this info and issue? does the compiler team / hip test team know of these errors?
There was a problem hiding this comment.
Yes, updated github-issue (#4244) info now.
As Ron mentioned, Compiler team is already working on these failure tests.
|
yes we know of it, Sam and Jacob have PRs ready to start landing to resolve this. its a comgr and llvm-project fix |
Signed-off-by: Lenine Ajagappane <Lenine.Ajagappane@amd.com>
geomin12
left a comment
geomin12
left a comment
There was a problem hiding this comment.
Is this correct? Did we add it to the correct arch? I still see the error for gfx94X-dcgpu:
https://github.com/ROCm/TheRock/actions/runs/23900615264/job/70066496331?pr=4245#step:12:8248
Signed-off-by: Lenine Ajagappane <Lenine.Ajagappane@amd.com>
Yes, updated PR to disable these tests on gfx94X target family. |
5 participants
Summary
Disable three flaky
Unit_NonHost_Printfhip-tests that are intermittently failing (subprocess aborted) with the compiler submodule update in #4081. This unblocks the compiler promotion merge.Failing Tests
|
Unit_NonHost_Printf_loop| Subprocess aborted ||
Unit_NonHost_Printf_multiple_Threads| Subprocess aborted ||
Unit_NonHost_Printf_BufferAvailability| Subprocess aborted |Context
These tests fail intermittently with the compiler submodule update and are not related to a specific GPU architecture. Disabling them allows the compiler promotion to proceed while the root cause is investigated separately.
Related