SWDEV-189488 - [HIP] Caffe2 TensorTest.TensorSerializationMultiDevices fails
1. Make sure to set attributes->device to current device for host malloc'd
2. Return hipSuccess for hipDeviceCanAccessPeer
Affected files ...
... //depot/stg/opencl/drivers/opencl/api/hip/hip_memory.cpp#56 edit
... //depot/stg/opencl/drivers/opencl/api/hip/hip_peer.cpp#4 edit
[ROCm/clr commit: 61bacfd131]
+ Add missing cuda_profiler_api.h to hip/hip_profile.h transformation.
NOTE: HIP Profiler API is under development. This is NOT WORKING example.
TODO: Find out a way to generate HIP_SCOPED_MARKER, HIP_BEGIN_MARKER, HIP_END_MARKER, declared in hip/hip_profile.h in particular place (signatures are to obtain).
[ROCm/clr commit: 5e49c25faa]
* Revert "Revert "Use COMgr to read Kernel Args Metadata (#1006)""
This reverts commit f8d108a815.
* Revert "Use COMgr to read Kernel Args Metadata (#1006)"
This reverts commit 10048a5631.
* Revert "improve program state commentary"
This reverts commit 5233d41c6c.
* Revert "load program state once per agent"
This reverts commit 9cee2c5311.
* start moving function_names() into the hip shared lib
* start moving code_object_blobs to a new "state" object
* Consolidate various program state related static objects into a
single program_state object
* minor clean up
* move more stuffs from functional_grid_launch into program_state
* debug make_kernarg
* moving lookup for kernargs size_align into program_state
* clean up old code for kernarg size and alignment
* update hip_module to use newer api in program_state
* Create public member functions for program_state
* move most program state functions into shared library
* Pass the data buffer size to load_executable
Otherwise, it can't figure what the data size is
just from the char* (since the data is not really a string)
* turning free functions in program state into members of program_state_impl
* change the free function globals() into a member of program_state_impl
* replace the static mutex used for populating globals
* moving associate_code_object_symbols_with_host_allocation into
program_state_impl
* move load_code_object_and_freeze_executable into program_state_impl
* moving executables and functions_names into program_state_impl
* moving kernels() into program_state_impl
* moving functions() into program_state_impl
* move get_kernargs into program_state_impl
* moving kernel_descriptor into program_state_impl
* moving kernargs_size_align calculation into program_state_impl
* Changing the handle to program_state_impl to a pointer
* moving program_state_impl into a separate inline source file
* fixing/cleaning up some header file includes
* moving member function for kernargs_size_align into program_state.cpp
* moving Kernel_descriptor into program_state.inl
* add a new class to manage agent globals
* moving all agent globals processing functions into agent_globals_impl
* load program state once per agent
re-merging PR991 against other program state changes
* fix per-agent program state member initialization
* cache executables based on elf name, isa, and agent.
This avoids program state reloading executables after a shared library is dlopened.
re-merging PR1057 against other program state changes
* protect executables cache by a global mutex
* return ref to executables cache
* adapt PR#981 Make hipModuleGetGlobal be in HIP runtime
[ROCm/clr commit: 05a1b696da]
SWDEV-145570 - Fix device name mismatch for gfx906.
For now hip-clang can only emits gfx906 ISA with conservative configurations, i.e. with ecc on and xnack on, therefore it is always gfx906. It is still under discussion how to encode the target id for xnack off or ecc off.
Therefore, the reasonable solution for now is just allow code object marked as gfx906 to be loaded on any device name that starts with gfx906. We will have more detailed control once hip-clang is able to emit code object for gfx906 with ecc off or xnack off.
Affected files ...
... //depot/stg/opencl/drivers/opencl/api/hip/hip_platform.cpp#30 edit
[ROCm/clr commit: ed092053e1]
+ Only a generation of transformation map of CUDA entities is implemented.
+ 2 hipify-clang options are added: -python, -o-python-map-dir.
+ Explicitly set -roc option for cuda_to_hip_mappings.py generation.
+ Generated file already might be used by pytorch team.
[ROCm/clr commit: 0f4affde9c]
+ Only a generation of transformation map of CUDA entities supported by HIP is implemented.
+ 3 hipify-clang options are added: -perl, -o-perl-map, -o-perl-map-dir.
+ OptionsParser mode is changed from OneOrMore to Optional to support hipify-perl generation without actual hipification.
+ Add explicit control of source files specification absence in case of no perl generation.
[ROCm/clr commit: 849155d865]
Maneesh Gupta
Wenkai Du
Evgeny Mankov
foreman
Konstantin Pyzhov
Jenkins
Wen-Heng (Jack) Chung
Rahul Garg
Nick Curtis
Siu Chi Chan