This class of tests can fail randomnly since they assume that time to
create first set of stream should be greater than second set. This is
incorrect since there can be multiple reason why the time for second
batch can be greater than the first batch.
Change-Id: I5f91c2bab7119608014dead25fdeb38523c61682
Reasoning: The test assumes hipcc to be present, which might not be the
case on platforms running hip-tests. This test assumes only one type of
GPU present, iGPUs can mess the test up. It also assumes write access
inside the folder its executing, which might not be the case and lastly
it relies on a behavior of HIP which has changed a few times which is
how does a HIP application behave when code object in an exe does not
match the gpu isa. Also work on SWDEV-365241 might render this redundant,
incases where you try to run gfx1032 on gfx1030.
Change-Id: Ie99bdf7caef47e449e3d19c138af47ab82f73d2b
The corresponding change https://gerrit-git.amd.com/c/compute/ec/clr/+/940530
in clr is being reverted. This test can be enabled when this change
is reintroduced.
Change-Id: I7787d576b59b9000d43042ec5553219e748b0ae0
- Unit_hipLaunchParm fails in integrity test & in PSDB on MI100.
- Disable it until the failure is fixed.
Change-Id: Id055860e067189c0ef85c77583c73269db9d2b51
- Unit_hipHostRegister_Memcpy - int/float/double failures are fixed in latest mainline (SWDEV-393727).
- Remaining testcases were random failures seen only during stress test. Could be due to driver/hardware issue.
Jatin Chaudhary
Satyanvesh Dittakavi
Tao Sang
Rakesh Roy
Jaydeep Patel
Todd tiantuo Li
Ioannis Assiouras
kjayapra-amd
Rakesh Roy
Maneesh Gupta