- Verify that memory stays mapped when reference count is non zero
- Verify that memory stays mapped if handle is closed in a second process
- Verify that attempting to close handle in the process that created it causes an error
- Correct header includes.
- Add additional pointer check to reference counting test.
- Disable Unit_hipIpcCloseMemHandle_Negative_Close_In_Originating_Process test for AMD.
[ROCm/hip commit: dc5d658843]
- Verify that memory stays mapped when reference count is non zero
- Verify that memory stays mapped if handle is closed in a second process
- Verify that attempting to close handle in the process that created it causes an error
- Correct header includes.
- Add additional pointer check to reference counting test.
- Disable Unit_hipIpcCloseMemHandle_Negative_Close_In_Originating_Process test for AMD.
- Verify that memory stays mapped when reference count is non zero
- Verify that memory stays mapped if handle is closed in a second process
- Verify that attempting to close handle in the process that created it causes an error
- Correct header includes.
- Add additional pointer check to reference counting test.
- Disable Unit_hipIpcCloseMemHandle_Negative_Close_In_Originating_Process test for AMD.
[ROCm/hip-tests commit: 86fa9725c4]
- Verify that memory stays mapped when reference count is non zero
- Verify that memory stays mapped if handle is closed in a second process
- Verify that attempting to close handle in the process that created it causes an error
- Correct header includes.
- Add additional pointer check to reference counting test.
- Disable Unit_hipIpcCloseMemHandle_Negative_Close_In_Originating_Process test for AMD.
- Validate that hipGetDeviceFlags returns flags with which the current context was created.
- Disable Unit_hipGetDeviceFlags_Positive_Context on AMD
[ROCm/hip commit: 7a3fecab50]
- Validate that hipGetDeviceFlags returns flags with which the current context was created.
- Disable Unit_hipGetDeviceFlags_Positive_Context on AMD
[ROCm/hip-tests commit: a6bbefbd95]
- mem pool should be from same device
Signed-off-by: sdashmiz <shadi.dashmiz@amd.com>
Change-Id: Id06a7f070c019548462c235bff3c0cd972e6eb8a
[ROCm/clr commit: 33f02abc41]
MES devices need GART mappings and therefore need non-paged memory. But
using non-paged memory introduces performance regression where it can
take over 80 ms to see the signal changes if the memory is in the wrong
NUMA node. Currently, we cannot control NUMA affinity when allocating
non-paged memory. Using non-paged memory allocation only on devices that
have MES scheduler
Change-Id: Ib27fb01d75247aa4f2bb2aa4503c6af5a98afda0
MES devices need GART mappings and therefore need non-paged memory. But
using non-paged memory introduces performance regression where it can
take over 80 ms to see the signal changes if the memory is in the wrong
NUMA node. Currently, we cannot control NUMA affinity when allocating
non-paged memory. Using non-paged memory allocation only on devices that
have MES scheduler
Change-Id: Ib27fb01d75247aa4f2bb2aa4503c6af5a98afda0
[ROCm/ROCR-Runtime commit: c1e836b6ab]
- Implemented P2P Device Access positive and negative tests for the following functions:
- hipDeviceCanAccessPeer
- hipDeviceEnablePeerAccess
- hipDeviceDisablePeerAccess
- Expand P2P Device Access negative tests
- Add hipDeviceEnablePeerAccess negative test when flag is invalid and access is already enabled
- Add hipDeviceDisablePeerAccess negative test when access is not enabled and when it is disabled twice
[ROCm/hip-tests commit: 79162275d9]
- Implemented P2P Device Access positive and negative tests for the following functions:
- hipDeviceCanAccessPeer
- hipDeviceEnablePeerAccess
- hipDeviceDisablePeerAccess
- Expand P2P Device Access negative tests
- Add hipDeviceEnablePeerAccess negative test when flag is invalid and access is already enabled
- Add hipDeviceDisablePeerAccess negative test when access is not enabled and when it is disabled twice
- Implemented P2P Device Access positive and negative tests for the following functions:
- hipDeviceCanAccessPeer
- hipDeviceEnablePeerAccess
- hipDeviceDisablePeerAccess
- Expand P2P Device Access negative tests
- Add hipDeviceEnablePeerAccess negative test when flag is invalid and access is already enabled
- Add hipDeviceDisablePeerAccess negative test when access is not enabled and when it is disabled twice
[ROCm/hip commit: cbd52de94b]
- Implemented P2P Device Access positive and negative tests for the following functions:
- hipDeviceCanAccessPeer
- hipDeviceEnablePeerAccess
- hipDeviceDisablePeerAccess
- Expand P2P Device Access negative tests
- Add hipDeviceEnablePeerAccess negative test when flag is invalid and access is already enabled
- Add hipDeviceDisablePeerAccess negative test when access is not enabled and when it is disabled twice
Using previous method of std::thread for SVM profiler task was causing
segfaults on thread launch on RHEL 8 if libhsa-runtime library is loaded
using dlopen.
Change-Id: Ic010cd6ae9bc6e6ed0605de02b93f6aae8ed3e97
Using previous method of std::thread for SVM profiler task was causing
segfaults on thread launch on RHEL 8 if libhsa-runtime library is loaded
using dlopen.
Change-Id: Ic010cd6ae9bc6e6ed0605de02b93f6aae8ed3e97
[ROCm/ROCR-Runtime commit: 0e4c7336ff]
Sync between compute and SDMA engines can be very expensive under Windows.
Use CP DMA for tiny transfers (< 1KiB) to avoid syncs and improve performance.
Change-Id: I9db39a2199f7b9e337ed08fd36d9cbc150502f1f
[ROCm/clr commit: 473621c008]
Sync between compute and SDMA engines can be very expensive under Windows.
Use CP DMA for tiny transfers (< 1KiB) to avoid syncs and improve performance.
Change-Id: I9db39a2199f7b9e337ed08fd36d9cbc150502f1f
KMD can enable hipHostRegister optimization with HMM path.
That will make CPU and GPU pointers matching.
Change-Id: Iad96ceada5cfa3bada20452b906f744f9dbaebbe
[ROCm/clr commit: 55d085ca8c]
KMD can enable hipHostRegister optimization with HMM path.
That will make CPU and GPU pointers matching.
Change-Id: Iad96ceada5cfa3bada20452b906f744f9dbaebbe
- Test behaviour with valid arguments for both APIs
- Test with invalid parameters.
- Default return value test
- Multithreaded test
- Enable tests for nvidia platform.
[ROCm/hip-tests commit: 44a48a387e]
- Test behaviour with valid arguments for both APIs
- Test with invalid parameters.
- Default return value test
- Multithreaded test
- Enable tests for nvidia platform.
- Test behaviour with valid arguments for both APIs
- Test with invalid parameters.
- Default return value test
- Multithreaded test
- Enable tests for nvidia platform.
[ROCm/hip commit: 2db7c8778b]
- Test behaviour with valid arguments for both APIs
- Test with invalid parameters.
- Default return value test
- Multithreaded test
- Enable tests for nvidia platform.