To test if SVM support address mmap on file backed memory.
Change-Id: I227cc0b44d167c4fa2e63257e13a481aefa4f750
Signed-off-by: Philip Yang <Philip.Yang@amd.com>
To test if SVM support address mmap on file backed memory.
Change-Id: I227cc0b44d167c4fa2e63257e13a481aefa4f750
Signed-off-by: Philip Yang <Philip.Yang@amd.com>
[ROCm/ROCR-Runtime commit: d91f92669b]
1. Fix hipMallocManagedMultiChunkMultiDevice
Add device setting before kernel launching
Add stream sync
2. Fix hipMemcpyAtoH and hipMemcpyHtoA
Fix primary context initialization issue when device
is changed on cuda
Change-Id: I3fe6dbc35b7b24abb21ada297b7885df83d28152
1. Simply enable test on NV
Some need minor fix
performance/compute/hipPerfDotProduct.cpp
performance/dispatch/hipPerfDispatchSpeed.cpp
performance/memory/hipPerfBufferCopyRectSpeed.cpp
performance/memory/hipPerfBufferCopySpeed.cpp
performance/memory/hipPerfDevMemReadSpeed.cpp
performance/memory/hipPerfDevMemWriteSpeed.cpp
performance/memory/hipPerfMemcpy.cpp
performance/memory/hipPerfMemset.cpp
performance/memory/hipPerfSharedMemReadSpeed.cpp
performance/stream/hipPerfDeviceConcurrency.cpp
performance/stream/hipPerfStreamCreateCopyDestroy.cpp
2. Enable and fix on NV
performance/compute/hipPerfMandelbrot.cpp
Root cause: coordIdx is random
Solution: Initialize coordIdx correctly
performance/memory/hipPerfMemFill.cpp
Root cause: Hip ext Apis called.
Solution: Exclude case with Hip ext Apis involved
performance/memory/hipPerfMemMallocCpyFree.cpp
Root cause: Test allocates device memory more than GPU has.
Solution: Allocate device memory in terms of GPU capacity.
tests/performance/memory/hipPerfSampleRate.cpp
Root cause: Cuda has no operators += for float2 and float4.
Solution: Provide the operators.
performance/stream/hipPerfStreamConcurrency.cpp
Root cause:float4 format doesn't match cude.
operators are missing in cuda lib.
Solution: Use (x, y, z, w) format.
Add necessary float4 operatoris for cuda.
Change-Id: I5add29ebabcfb21fb3ef89d09004c5d13423a291
[ROCm/hip-tests commit: bfc2f4516a]
1. Simply enable test on NV
Some need minor fix
performance/compute/hipPerfDotProduct.cpp
performance/dispatch/hipPerfDispatchSpeed.cpp
performance/memory/hipPerfBufferCopyRectSpeed.cpp
performance/memory/hipPerfBufferCopySpeed.cpp
performance/memory/hipPerfDevMemReadSpeed.cpp
performance/memory/hipPerfDevMemWriteSpeed.cpp
performance/memory/hipPerfMemcpy.cpp
performance/memory/hipPerfMemset.cpp
performance/memory/hipPerfSharedMemReadSpeed.cpp
performance/stream/hipPerfDeviceConcurrency.cpp
performance/stream/hipPerfStreamCreateCopyDestroy.cpp
2. Enable and fix on NV
performance/compute/hipPerfMandelbrot.cpp
Root cause: coordIdx is random
Solution: Initialize coordIdx correctly
performance/memory/hipPerfMemFill.cpp
Root cause: Hip ext Apis called.
Solution: Exclude case with Hip ext Apis involved
performance/memory/hipPerfMemMallocCpyFree.cpp
Root cause: Test allocates device memory more than GPU has.
Solution: Allocate device memory in terms of GPU capacity.
tests/performance/memory/hipPerfSampleRate.cpp
Root cause: Cuda has no operators += for float2 and float4.
Solution: Provide the operators.
performance/stream/hipPerfStreamConcurrency.cpp
Root cause:float4 format doesn't match cude.
operators are missing in cuda lib.
Solution: Use (x, y, z, w) format.
Add necessary float4 operatoris for cuda.
Change-Id: I5add29ebabcfb21fb3ef89d09004c5d13423a291
1. Simply enable test on NV
Some need minor fix
performance/compute/hipPerfDotProduct.cpp
performance/dispatch/hipPerfDispatchSpeed.cpp
performance/memory/hipPerfBufferCopyRectSpeed.cpp
performance/memory/hipPerfBufferCopySpeed.cpp
performance/memory/hipPerfDevMemReadSpeed.cpp
performance/memory/hipPerfDevMemWriteSpeed.cpp
performance/memory/hipPerfMemcpy.cpp
performance/memory/hipPerfMemset.cpp
performance/memory/hipPerfSharedMemReadSpeed.cpp
performance/stream/hipPerfDeviceConcurrency.cpp
performance/stream/hipPerfStreamCreateCopyDestroy.cpp
2. Enable and fix on NV
performance/compute/hipPerfMandelbrot.cpp
Root cause: coordIdx is random
Solution: Initialize coordIdx correctly
performance/memory/hipPerfMemFill.cpp
Root cause: Hip ext Apis called.
Solution: Exclude case with Hip ext Apis involved
performance/memory/hipPerfMemMallocCpyFree.cpp
Root cause: Test allocates device memory more than GPU has.
Solution: Allocate device memory in terms of GPU capacity.
tests/performance/memory/hipPerfSampleRate.cpp
Root cause: Cuda has no operators += for float2 and float4.
Solution: Provide the operators.
performance/stream/hipPerfStreamConcurrency.cpp
Root cause:float4 format doesn't match cude.
operators are missing in cuda lib.
Solution: Use (x, y, z, w) format.
Add necessary float4 operatoris for cuda.
Change-Id: I5add29ebabcfb21fb3ef89d09004c5d13423a291
[ROCm/hip commit: 9035ae3154]
1. Simply enable test on NV
Some need minor fix
performance/compute/hipPerfDotProduct.cpp
performance/dispatch/hipPerfDispatchSpeed.cpp
performance/memory/hipPerfBufferCopyRectSpeed.cpp
performance/memory/hipPerfBufferCopySpeed.cpp
performance/memory/hipPerfDevMemReadSpeed.cpp
performance/memory/hipPerfDevMemWriteSpeed.cpp
performance/memory/hipPerfMemcpy.cpp
performance/memory/hipPerfMemset.cpp
performance/memory/hipPerfSharedMemReadSpeed.cpp
performance/stream/hipPerfDeviceConcurrency.cpp
performance/stream/hipPerfStreamCreateCopyDestroy.cpp
2. Enable and fix on NV
performance/compute/hipPerfMandelbrot.cpp
Root cause: coordIdx is random
Solution: Initialize coordIdx correctly
performance/memory/hipPerfMemFill.cpp
Root cause: Hip ext Apis called.
Solution: Exclude case with Hip ext Apis involved
performance/memory/hipPerfMemMallocCpyFree.cpp
Root cause: Test allocates device memory more than GPU has.
Solution: Allocate device memory in terms of GPU capacity.
tests/performance/memory/hipPerfSampleRate.cpp
Root cause: Cuda has no operators += for float2 and float4.
Solution: Provide the operators.
performance/stream/hipPerfStreamConcurrency.cpp
Root cause:float4 format doesn't match cude.
operators are missing in cuda lib.
Solution: Use (x, y, z, w) format.
Add necessary float4 operatoris for cuda.
Change-Id: I5add29ebabcfb21fb3ef89d09004c5d13423a291
* SWDEV-266829 - Enable more tests on AMD and NV devices
1. Enable tests on AMD and NV devices
tests/src/runtimeApi/event/hipEventMultiThreaded.cpp
Loops and threads per core are changed smaller so that test can
finish in a shorter time.
tests/src/runtimeApi/stream/hipStreamCreateWithPriority.cpp
Fix logic error on how to get priority_normal
2. Simply enable test on AMD device
tests/src/runtimeApi/memory/hipManagedKeyword.cpp
tests/src/runtimeApi/module/hipManagedKeyword.cpp
tests/src/runtimeApi/stream/hipStreamACb_MultiThread.cpp
tests/src/runtimeApi/memory/p2p_copy_coherency.cpp
3. Simply enable test on NV device
tests/src/runtimeApi/module/hipModuleLoadDataMultThreaded.cpp
4. Fix typo
tests/src/runtimeApi/stream/hipStreamAddCallbackCatch.cpp
5. Remove useless tests
tests/src/hipC.c
tests/src/hipHcc.cpp
Change-Id: Ia4406353e64d69bd34c58ebb56185701f7ce1caa
* Remove tests/src/runtimeApi/module/hipModuleLoadDataMultThreaded.cpp for cuda test
Co-authored-by: anusha GodavarthySurya <Anusha.GodavarthySurya@amd.com>
Co-authored-by: Jenkins <jenkins-compute@amd.com>
[ROCm/hip commit: 3fd16c0b5b]
* SWDEV-266829 - Enable more tests on AMD and NV devices
1. Enable tests on AMD and NV devices
tests/src/runtimeApi/event/hipEventMultiThreaded.cpp
Loops and threads per core are changed smaller so that test can
finish in a shorter time.
tests/src/runtimeApi/stream/hipStreamCreateWithPriority.cpp
Fix logic error on how to get priority_normal
2. Simply enable test on AMD device
tests/src/runtimeApi/memory/hipManagedKeyword.cpp
tests/src/runtimeApi/module/hipManagedKeyword.cpp
tests/src/runtimeApi/stream/hipStreamACb_MultiThread.cpp
tests/src/runtimeApi/memory/p2p_copy_coherency.cpp
3. Simply enable test on NV device
tests/src/runtimeApi/module/hipModuleLoadDataMultThreaded.cpp
4. Fix typo
tests/src/runtimeApi/stream/hipStreamAddCallbackCatch.cpp
5. Remove useless tests
tests/src/hipC.c
tests/src/hipHcc.cpp
Change-Id: Ia4406353e64d69bd34c58ebb56185701f7ce1caa
* Remove tests/src/runtimeApi/module/hipModuleLoadDataMultThreaded.cpp for cuda test
Co-authored-by: anusha GodavarthySurya <Anusha.GodavarthySurya@amd.com>
Co-authored-by: Jenkins <jenkins-compute@amd.com>
When unsetting runtime should use HSA_AMD_SVM_ATTRIB_AGENT_ACCESSIBLE
for the agent and not HSA_AMD_SVM_ATTRIB_AGENT_ACCESSIBLE_IN_PLACE
Change-Id: I3814802d1fb3b72c54e7566defafafed6b0d5cee
[ROCm/clr commit: d8a86e4870]
When unsetting runtime should use HSA_AMD_SVM_ATTRIB_AGENT_ACCESSIBLE
for the agent and not HSA_AMD_SVM_ATTRIB_AGENT_ACCESSIBLE_IN_PLACE
Change-Id: I3814802d1fb3b72c54e7566defafafed6b0d5cee
The original logic left only one slot for HW processing in the queue.
For some reason there is a race condition on CPU overwrite of the slot
before the current active. The workaround is to avoid the previous to
the current active slot for possible unfinished HW processing.
Change-Id: I565495a8feeaedffc9fc8a505edbee5ff5816975
[ROCm/clr commit: 65ddfcc6a8]
The original logic left only one slot for HW processing in the queue.
For some reason there is a race condition on CPU overwrite of the slot
before the current active. The workaround is to avoid the previous to
the current active slot for possible unfinished HW processing.
Change-Id: I565495a8feeaedffc9fc8a505edbee5ff5816975
hipGetLastError/hipPeekAtLastError should really return the last error
produced by any of the runtime calls that did not successfuly complete,
not just the error code of the last runtime call.
For example, in this snippet of code:
1: hipLaunchKernelGGL(MyKernel, dim3(2), dim3(128), 0, 0, 10000,x);
2: hipDeviceSynchronize();
3: hipError_t error = hipGetLastError();
If the call to hipLaunchKernelGGL fails, hipDeviceSynchronize (which
succeeds) should not reset the last error to hipSuccess. hipGetLastError
should still return the non-success error code returned by
hipLaunchKernelGGL.
The last error is reset to hipSuccess after calling hipGetLastError.
Change-Id: Ib7c039067c53c94c99c8ecd83f54212bcef06f81
[ROCm/clr commit: 3dc47bc839]
hipGetLastError/hipPeekAtLastError should really return the last error
produced by any of the runtime calls that did not successfuly complete,
not just the error code of the last runtime call.
For example, in this snippet of code:
1: hipLaunchKernelGGL(MyKernel, dim3(2), dim3(128), 0, 0, 10000,x);
2: hipDeviceSynchronize();
3: hipError_t error = hipGetLastError();
If the call to hipLaunchKernelGGL fails, hipDeviceSynchronize (which
succeeds) should not reset the last error to hipSuccess. hipGetLastError
should still return the non-success error code returned by
hipLaunchKernelGGL.
The last error is reset to hipSuccess after calling hipGetLastError.
Change-Id: Ib7c039067c53c94c99c8ecd83f54212bcef06f81
HIP_RETURN(x) is not a value, it is a statement ending with a return,
this commit replaces all instances of "return HIP_RETURN(x)" with simply
"HIP_RETURN(x)"
Change-Id: I03293b2684a65367ff55e02b3a71ea49ec7a517a
[ROCm/clr commit: 3911184607]
HIP_RETURN(x) is not a value, it is a statement ending with a return,
this commit replaces all instances of "return HIP_RETURN(x)" with simply
"HIP_RETURN(x)"
Change-Id: I03293b2684a65367ff55e02b3a71ea49ec7a517a
Philip Yang
sumanthtg
TomSang
Julia Jiang
Chauncey Hui
Satyanvesh Dittakavi
German Andryeyev
Jonathan R. Madsen
Laurent Morichetti
Jason Tang