* make sure symbolStatus after hipGetProcAddress() is the expected in the tests that are failing on Phoenix
* fix that calls to hipMemset2DAsync() and hipMemset2DAsync_spt() to set values on submatrices where not taking into account the pitch, causing test failures on Phoenix
* before testing whether hipMemset2DAsync() works, initialize the whole matrix to a known value. This makes sure the test fails even if the uninitialized memory returned after hipMallocPitch() happens to have the expected value
* Sbefore testing whether hipMemset2DAsync_spt() works, initialize the whole matrix to a known value. This makes sure the test fails even if the uninitialized memory returned after hipMallocPitch() happens to have the expected value
* fix up compiler error: dyn_hipMemset2DAsync_ptr should have been dyn_hipMemset2DAsync_spt_ptr
* when Unit_hipMalloc3D_Basic fails due to a potential memory leak, print the values that hipMemGetInfo() returns before and after the allocation/deallocation pair.
* Also print intermediate free memory
* Make sure Unit_hipMalloc3D_Basic allocates at least PalSettings::subAllocationChunkSize_ on Windows. Otherwise hipMemGetInfo() will not report an increase on available memory after hipFree() is called, as the minimum amount of memory that cause a call to the CoarseMemorySubAllocator::Create() is that chunk size
* Fix up previous commit; allocated too much memory
* skip some extent sizes in Unit_hipFreeMipmappedArrayImplicitSyncArray if allocating them would be require more memory than the actual totalGlobalMem of the device
* Do not expect an exact match when comparing the memory available memory before and after hipMalloc3D() + hipFree()
* Do not allocate more memory than the total GPU memory in Unit_hipFreeMipmappedArrayImplicitSyncArray
* fix expected available memory amount comparison in hipMalloc3D - Basic
* use SUCCEED() macro in Unit_hipFreeMipmappedArrayImplicitSyncArray to log more information when there is not enough memory for the mipmapped array to be allocated
* fix formatting
1.Remove clock functions from some tests that don't need them.
2.In some memory pool tests and coherency tests, timer-based kernel
delay isn't reliable, use pinned host based notification instead.
3.Add CHECK_PCIE_ATOMICS_SUPPORT before some tests.
4.catch/unit/memory/hipMemoryAllocateCoherent.cc is removed
as it is useless and originally excluded in building.
5.Some tests can still pass even if clock rate =0, thus they
will be kept as is.
6.Some logic and format improvement in some tests.
Change-Id: I6b3c6bf54c61cffd45cd6f17c75998f751b75725
Unit_hipMalloc_AllocateMoreThanTotalRAM is an invalid test. hipMalloc is able to allocate more than RAM size if VRam is larger, unless the test is meant to only allocate on host device.
Previously this test may have passed because of an incorrect check on hipMalloc limit: we had incorrectly limited allocation to physical ram size. This is no longer correct. We could always allocate Vram limit or sometimes system ram limit if Vram is full, fixed in https://github.com/AMD-ROCm-Internal/clr/pull/122
Co-authored-by: Jimbo Xie <jiabaxie@amd.com>
The test wrongly assumed that the reserved memory should always
be equal to the allocation size. However, suballocations are possible
Change-Id: I1d1f3c3cdb231188f8b35ed64ffe4e2a3a2f16f9
Unit_hipMemPoolApi_BasicAlloc expects to work on device 0, but other
tests will set not-0 devices in mgpu. This leads to hang of
Unit_hipMemPoolApi_BasicAlloc. Fix by set device 0 in head code
of Unit_hipMemPoolApi_BasicAlloc.
SWDEV-508872 - Fix Perf_hipPerfMemFill_test
When mem size is 2G, the test is so slow that it looks like stuckness.
Set top mem size to 1G can make the test pass in an acceptiable time.
Change-Id: Ie26dbf597e5ba8cb898d1aae5ed5ecf0267c3228
- For H2D transfer cases, we cannot enfore ErrorNotReady behavior if the
source is unpinned. In that case the driver can sync depending on if
it pins or stages the source buffer.
Change-Id: I436a4b667bff172cfbadc58631a46a215c40fcb1
Fix Unit_hipExtMallocWithFlags_CoherentTst hang on device fine
grain access memory because host builtin atomic functions malfunctions
on it.
Fix SWDEV-507938 also.
Change-Id: Ie26dbf6b7e5aa8cb898d11ae5ed5ecf1267c3228
- Since any-host-to-any-host as always synchronous, there's no need for
separate functions
- Removed 'return' statement as it skips the rest of the test.
Change-Id: If70e0b21ad8151311d5ac7c5de789f6265cf44b6
Fix memory leaks by adding missing destroy calls for
events, streams, and graphs at the end of tests.
Ensure that every test case executes destroy calls,
regardless of whether it passes or fails.
Change-Id: I814e35c528d90ed2abb34d77377f1a7fd3f1f11c
1.Test staging buffer copy when there is no P2P capabilities
2.Enable PeerAccess correctly for P2P copy
Change-Id: Ic0b75a60c8a763f3cfe68fbd92656fbfb4c207bc
Hernandez, Gerardo
Manocha, Rahul
Sang, Tao
Assiouras, Ioannis
Andryeyev, German
Arandjelovic, Marko
Xie, Pengda
Huang, AnZhong
Xie, Jiabao(Jimbo)
Brzak, Branislav
Saleel Kudchadker
Vladana Stojiljkovic
Marko Arandjelovic
Rambabu Swargam
Jaydeep Patel
SrinivasaRao
Satyanvesh Dittakavi
Pradeep Kumar Jain
Li, Todd tiantuo
Ajay
Rahul Manocha
Rupam Chetia