Added functional and negative scenarios for
hipMemcpyPeer and
hipMemcpyPeerAsync APIs
Change-Id: Ibaee0d84cb616727dc5db057bb9422702670b686
[ROCm/hip commit: 48c2a0fb27]
- when the first argument is null API should return hipErrorInvalidValue
- but when first and second argument both are null API was returning hipErrorInvalidHandle causing catch2 event tests to fail
Change-Id: I0978ce8b8462e4baa043be75a40b5bc45b036bb6
[ROCm/clr commit: 688288c2e3]
- when the first argument is null API should return hipErrorInvalidValue
- but when first and second argument both are null API was returning hipErrorInvalidHandle causing catch2 event tests to fail
Change-Id: I0978ce8b8462e4baa043be75a40b5bc45b036bb6
Add support for bfloat16.
Add ncclAvg reduction operation.
Improve performance for aggregated operations.
Improve performance for tree.
Improve network error reporting.
Add NCCL_NET parameter to force a specific network.
Add NCCL_IB_QPS_PER_CONNECTION parameter to split IB traffic onto multiple queue pairs.
Fix topology detection error in WSL2.
Fix proxy memory elements affinity (improve alltoall performance).
Fix graph search on cubemesh topologies.
Fix hang in cubemesh during NVB connections.
[ROCm/rccl commit: 7e51592129]
Add support for bfloat16.
Add ncclAvg reduction operation.
Improve performance for aggregated operations.
Improve performance for tree.
Improve network error reporting.
Add NCCL_NET parameter to force a specific network.
Add NCCL_IB_QPS_PER_CONNECTION parameter to split IB traffic onto multiple queue pairs.
Fix topology detection error in WSL2.
Fix proxy memory elements affinity (improve alltoall performance).
Fix graph search on cubemesh topologies.
Fix hang in cubemesh during NVB connections.
Add test on CXX and Fortran build in cmake.
Add test on hip::device interface linking in cmake.
Change-Id: I3fe6dba05a7a140a9a19df107b7885df83d28042
[ROCm/hip-tests commit: d5142afe58]
Add test on CXX and Fortran build in cmake.
Add test on hip::device interface linking in cmake.
Change-Id: I3fe6dba05a7a140a9a19df107b7885df83d28042
[ROCm/hip commit: 818aa18d59]
Re-enable __HIPCC_RTC__ in hip_vector_types.h which require
an upstream clang patch, 6823af0ca858b54e09e5be61a19d067ccd0bd6b7.
Once upstream patch has landed in mainline, merge this for
hiprtc functionality.
Change-Id: Ife884e0c3081b307bdadc8bec7804d1d7c60153b
[ROCm/clr commit: e5a1f25424]
Re-enable __HIPCC_RTC__ in hip_vector_types.h which require
an upstream clang patch, 6823af0ca858b54e09e5be61a19d067ccd0bd6b7.
Once upstream patch has landed in mainline, merge this for
hiprtc functionality.
Change-Id: Ife884e0c3081b307bdadc8bec7804d1d7c60153b
- Also revise the implementation of atomicInc/atomicDec based on clang
builtins.
Change-Id: Ia5e06e88c7be1da5f8fc7ac70037a120c604e343
[ROCm/clr commit: 45205c5e53]
Specify for packaging that clinfo is a program. This will set its
default permissions to include OWNER_EXECUTE, GROUP_EXECUTE and
WORLD_EXECUTE.
Change-Id: I986f6db73efa9a3f8ab5bedfb829adccad0759aa
[ROCm/clr commit: 31e688d8f0]
Specify for packaging that clinfo is a program. This will set its
default permissions to include OWNER_EXECUTE, GROUP_EXECUTE and
WORLD_EXECUTE.
Change-Id: I986f6db73efa9a3f8ab5bedfb829adccad0759aa
Added functional and negative scenarios for hipMemcpy2DToArray and
hipMemcpy2DToArrayAsync APIs
Change-Id: I97331a587fb1d51583faeec998cdb4ac09bfa564
[ROCm/hip-tests commit: 81f717774f]
Added functional and negative scenarios for hipMemcpy2DToArray and
hipMemcpy2DToArrayAsync APIs
Change-Id: I97331a587fb1d51583faeec998cdb4ac09bfa564
[ROCm/hip commit: edb1a96cdb]
Test Thunk multiple threads register and deregister same userptr race
condition, to emulate application register same userptr to multiple
GPUs using multiple threads.
Use thread barrier to sync the threads, to start register userptr at
same time.
Change-Id: I6723dc39f75908026fa14a490e39e1fe49a13a1b
Signed-off-by: Philip Yang <Philip.Yang@amd.com>
Test Thunk multiple threads register and deregister same userptr race
condition, to emulate application register same userptr to multiple
GPUs using multiple threads.
Use thread barrier to sync the threads, to start register userptr at
same time.
Change-Id: I6723dc39f75908026fa14a490e39e1fe49a13a1b
Signed-off-by: Philip Yang <Philip.Yang@amd.com>
[ROCm/ROCR-Runtime commit: 92076f6f1b]
- add std=c++11 to the g++ test hipMalloc.cpp to avoid failure on centos
- Also, enabled gcc test hipMalloc.c on amd platform
Change-Id: I3ff6dba05a70140a9a19df107b7885df83d28041
[ROCm/hip commit: 966741a0b3]
- add std=c++11 to the g++ test hipMalloc.cpp to avoid failure on centos
- Also, enabled gcc test hipMalloc.c on amd platform
Change-Id: I3ff6dba05a70140a9a19df107b7885df83d28041