SWDEV-145570 - [HIP] refactor hipMemcpy* functions to correctly handle copies using prepinned memory
The current implementation of hipMemcpy functions picks the copy type based on a flag that the user passes. However, one can use the hipMemcpyHostToDevice/hipMemcpyDeviceToHost flag in a combination with prepinned memory. By using the WriteMemoryCommand/ReadMemoyCommand in this case, we will pin the same host memory twice. This is fine on PAL/Linux, since pinning the same VA range is a noop, but this will start failing once we switch to using device memory with HIP/VDI/HSA.
The solution is to ignore the hipMemcpyKind flag and let the runtime decide what kind of copy is best to do. Except for the case when hipMemcpyHostToHost is passed, since both host pointers may be prepinned.
ReviewBoardURL = http://ocltc.amd.com/reviews/r/15482/diff/
Affected files ...
... //depot/stg/opencl/drivers/opencl/api/hip/hip_memory.cpp#37 edit
[ROCm/clr commit: b02040e934]
SWDEV-145570 - [HIP] refactor hipMemcpy* functions to correctly handle copies using prepinned memory
The current implementation of hipMemcpy functions picks the copy type based on a flag that the user passes. However, one can use the hipMemcpyHostToDevice/hipMemcpyDeviceToHost flag in a combination with prepinned memory. By using the WriteMemoryCommand/ReadMemoyCommand in this case, we will pin the same host memory twice. This is fine on PAL/Linux, since pinning the same VA range is a noop, but this will start failing once we switch to using device memory with HIP/VDI/HSA.
The solution is to ignore the hipMemcpyKind flag and let the runtime decide what kind of copy is best to do. Except for the case when hipMemcpyHostToHost is passed, since both host pointers may be prepinned.
ReviewBoardURL = http://ocltc.amd.com/reviews/r/15482/diff/
Affected files ...
... //depot/stg/opencl/drivers/opencl/api/hip/hip_memory.cpp#37 edit
SWDEV-145570 - [HIP] refactor hipMemcpy* functions to correctly handle copies using prepinned memory
The current implementation of hipMemcpy functions picks the copy type based on a flag that the user passes. However, one can use the hipMemcpyHostToDevice/hipMemcpyDeviceToHost flag in a combination with prepinned memory. By using the WriteMemoryCommand/ReadMemoyCommand in this case, we will pin the same host memory twice. This is fine on PAL/Linux, since pinning the same VA range is a noop, but this will start failing once we switch to using device memory with HIP/VDI/HSA.
The solution is to ignore the hipMemcpyKind flag and let the runtime decide what kind of copy is best to do. Except for the case when hipMemcpyHostToHost is passed, since both host pointers may be prepinned.
ReviewBoardURL = http://ocltc.amd.com/reviews/r/15482/diff/
Affected files ...
... //depot/stg/opencl/drivers/opencl/api/hip/hip_memory.cpp#37 edit
[ROCm/hip commit: 09b46cddf5]
SWDEV-145570 - [HIP] refactor hipMemcpy* functions to correctly handle copies using prepinned memory
The current implementation of hipMemcpy functions picks the copy type based on a flag that the user passes. However, one can use the hipMemcpyHostToDevice/hipMemcpyDeviceToHost flag in a combination with prepinned memory. By using the WriteMemoryCommand/ReadMemoyCommand in this case, we will pin the same host memory twice. This is fine on PAL/Linux, since pinning the same VA range is a noop, but this will start failing once we switch to using device memory with HIP/VDI/HSA.
The solution is to ignore the hipMemcpyKind flag and let the runtime decide what kind of copy is best to do. Except for the case when hipMemcpyHostToHost is passed, since both host pointers may be prepinned.
ReviewBoardURL = http://ocltc.amd.com/reviews/r/15482/diff/
Affected files ...
... //depot/stg/opencl/drivers/opencl/api/hip/hip_memory.cpp#37 edit
SWDEV-2 - Change OpenCL version number from 2705 to 2706.
Affected files ...
... //depot/stg/opencl/drivers/opencl/runtime/utils/versions.hpp#2453 edit
[ROCm/clr commit: ce73c861d6]
SWDEV-2 - Change OpenCL version number from 2705 to 2706.
Affected files ...
... //depot/stg/opencl/drivers/opencl/runtime/utils/versions.hpp#2453 edit
A README.txt file is added to help the opensource community to use kfdtest
effectively.
After building, run_kfdtest.sh in the building output folder can be used
to run the test.
Change-Id: I9612d9d5a63bd4cdc3a328efd9961d3cc92a6ba5
Signed-off-by: Yong Zhao <yong.zhao@amd.com>
A README.txt file is added to help the opensource community to use kfdtest
effectively.
After building, run_kfdtest.sh in the building output folder can be used
to run the test.
Change-Id: I9612d9d5a63bd4cdc3a328efd9961d3cc92a6ba5
Signed-off-by: Yong Zhao <yong.zhao@amd.com>
[ROCm/ROCR-Runtime commit: 1d43938ac7]
Thunk is an internal name and we'd better reference it using the library name.
Change-Id: I20042bda546e5249530311d3de30c71d99379033
Signed-off-by: Yong Zhao <yong.zhao@amd.com>
Thunk is an internal name and we'd better reference it using the library name.
Change-Id: I20042bda546e5249530311d3de30c71d99379033
Signed-off-by: Yong Zhao <yong.zhao@amd.com>
[ROCm/ROCR-Runtime commit: f8472a055c]
The code is a snapshot up to this commit around July 31 2018.
commit b00fadff36a3
Author: xinhui pan <xinhui.pan@amd.com>
Date: Mon Jul 30 09:53:03 2018 +0800
kfdtest: skip MMapLarge test on apu
Change-Id: I40e9a5a18e5c8f075e5290bb80532f1a3f689058
Signed-off-by: Yong Zhao <yong.zhao@amd.com>
The code is a snapshot up to this commit around July 31 2018.
commit b00fadff36a3
Author: xinhui pan <xinhui.pan@amd.com>
Date: Mon Jul 30 09:53:03 2018 +0800
kfdtest: skip MMapLarge test on apu
Change-Id: I40e9a5a18e5c8f075e5290bb80532f1a3f689058
Signed-off-by: Yong Zhao <yong.zhao@amd.com>
[ROCm/ROCR-Runtime commit: 6df62c78b8]
Implement the vector value types in terms of SCALAR_TYPE##SIZE_vector_value_type for float, int, uint and sizes 2, 3, 4, 8, and 16.
[ROCm/clr commit: ff1e49d191]
Implement the vector value types in terms of SCALAR_TYPE##SIZE_vector_value_type for float, int, uint and sizes 2, 3, 4, 8, and 16.
[ROCm/hip commit: e0e785a45d]
When shareWithAll memory (e.g., host memory) is allocated, set appId
in hc::AmPointerInfo to -1 to indicate that this memory is not mapped
to any device. Peer checking in ihipStream_t::canSeeMemory is not
necessary if memory is shared with all devices. Thus, it is skipped.
Note that earlier host memory is always mapped to device 0 and HIP
always performs peer checking for all kinds of hipMemcpy. Since the
peer checking process requires context locking, hipMemcpy from/to host
memory always grabs device 0's context lock. Therefore, if there is
another thread holding the context lock of device 0 (e.g.,
hipDeviceSynchronize on device 0), hipMemcpy will have to wait for the
lock until it can actually perform memcpy. This can significantly
deteriorate execution performance.
Signed-off-by: Sarunya Pumma <sarunya.pumma@amd.com>
[ROCm/clr commit: 84aadb9274]
When shareWithAll memory (e.g., host memory) is allocated, set appId
in hc::AmPointerInfo to -1 to indicate that this memory is not mapped
to any device. Peer checking in ihipStream_t::canSeeMemory is not
necessary if memory is shared with all devices. Thus, it is skipped.
Note that earlier host memory is always mapped to device 0 and HIP
always performs peer checking for all kinds of hipMemcpy. Since the
peer checking process requires context locking, hipMemcpy from/to host
memory always grabs device 0's context lock. Therefore, if there is
another thread holding the context lock of device 0 (e.g.,
hipDeviceSynchronize on device 0), hipMemcpy will have to wait for the
lock until it can actually perform memcpy. This can significantly
deteriorate execution performance.
Signed-off-by: Sarunya Pumma <sarunya.pumma@amd.com>
When shareWithAll memory (e.g., host memory) is allocated, set appId
in hc::AmPointerInfo to -1 to indicate that this memory is not mapped
to any device. Peer checking in ihipStream_t::canSeeMemory is not
necessary if memory is shared with all devices. Thus, it is skipped.
Note that earlier host memory is always mapped to device 0 and HIP
always performs peer checking for all kinds of hipMemcpy. Since the
peer checking process requires context locking, hipMemcpy from/to host
memory always grabs device 0's context lock. Therefore, if there is
another thread holding the context lock of device 0 (e.g.,
hipDeviceSynchronize on device 0), hipMemcpy will have to wait for the
lock until it can actually perform memcpy. This can significantly
deteriorate execution performance.
Signed-off-by: Sarunya Pumma <sarunya.pumma@amd.com>
[ROCm/hip commit: 8111fd3b8b]
When shareWithAll memory (e.g., host memory) is allocated, set appId
in hc::AmPointerInfo to -1 to indicate that this memory is not mapped
to any device. Peer checking in ihipStream_t::canSeeMemory is not
necessary if memory is shared with all devices. Thus, it is skipped.
Note that earlier host memory is always mapped to device 0 and HIP
always performs peer checking for all kinds of hipMemcpy. Since the
peer checking process requires context locking, hipMemcpy from/to host
memory always grabs device 0's context lock. Therefore, if there is
another thread holding the context lock of device 0 (e.g.,
hipDeviceSynchronize on device 0), hipMemcpy will have to wait for the
lock until it can actually perform memcpy. This can significantly
deteriorate execution performance.
Signed-off-by: Sarunya Pumma <sarunya.pumma@amd.com>
SWDEV-2 - Change OpenCL version number from 2704 to 2705.
Affected files ...
... //depot/stg/opencl/drivers/opencl/runtime/utils/versions.hpp#2452 edit
[ROCm/clr commit: 33c588c456]
SWDEV-2 - Change OpenCL version number from 2704 to 2705.
Affected files ...
... //depot/stg/opencl/drivers/opencl/runtime/utils/versions.hpp#2452 edit
xinhui pan
Aaron Enye Shi
Evgeny Mankov
foreman
Rahul Garg
Yong Zhao
Sarunya Pumma
Evgeny Mankov
Yaxun Sam Liu