Added command line option to specify which datatype(s) to capture into the roofline PDF(s).
All datatypes are still collected by roofline call if applicable, but only specific datatypes are plotted into PDF outputs. Will dump out all datatypes into one graph, but separate FP from Int into two graphs if needed. Will skip datatype and give error message if the datatype is not valid on a particular gpu arch.
Default is FP32
Reworked roofline calls and plotting to be general enough such that any new datatypes added into rocm-amdgpu-bench can easily be reflected in rocprof-compute with simple modifications in roofline_calc.py.
Adjusted ctest to reflect expected default pdf outputs from roofline.
---------
Signed-off-by: Carrie Fallows <Carrie.Fallows@amd.com>
[ROCm/rocprofiler-compute commit: a492e92034]
Added command line option to specify which datatype(s) to capture into the roofline PDF(s).
All datatypes are still collected by roofline call if applicable, but only specific datatypes are plotted into PDF outputs. Will dump out all datatypes into one graph, but separate FP from Int into two graphs if needed. Will skip datatype and give error message if the datatype is not valid on a particular gpu arch.
Default is FP32
Reworked roofline calls and plotting to be general enough such that any new datatypes added into rocm-amdgpu-bench can easily be reflected in rocprof-compute with simple modifications in roofline_calc.py.
Adjusted ctest to reflect expected default pdf outputs from roofline.
---------
Signed-off-by: Carrie Fallows <Carrie.Fallows@amd.com>
Adding a general stage for agents to release their resources on
shutdown. This avoids a circular dependency during shutdown because
we have to delete allocated resources before deleting memory pools, but
we also have to delete memory pools before destroying agents.
Adding a general stage for agents to release their resources on
shutdown. This avoids a circular dependency during shutdown because
we have to delete allocated resources before deleting memory pools, but
we also have to delete memory pools before destroying agents.
[ROCm/ROCR-Runtime commit: 947391deac]
The initial call to Refresh() in the constructor is
unnecessary as it's handled in Runtime::Load().
Signed-off-by: lyndonli <Lyndon.Li@amd.com>
[ROCm/ROCR-Runtime commit: c34a2798ce]
Also removes asserts in cooperative groups shfl functions since
__hip_bfloat16 shfl is present now
Change-Id: I57578b6e68dccc10c2ddcd194e9cc18bc7732ce1
[ROCm/clr commit: 376f23b86a]
Also removes asserts in cooperative groups shfl functions since
__hip_bfloat16 shfl is present now
Change-Id: I57578b6e68dccc10c2ddcd194e9cc18bc7732ce1
The debugger override will set the initial request mask to the
previously set request mask so use a different mask to assert
enablement.
Trap on wave start and end also run back to back, so fix the
previous override mask check as well.
In addition, unlike instruction traps, trap on wave start and end
will not require a rewind of the program counter on wave exit.
The debugger override will set the initial request mask to the
previously set request mask so use a different mask to assert
enablement.
Trap on wave start and end also run back to back, so fix the
previous override mask check as well.
In addition, unlike instruction traps, trap on wave start and end
will not require a rewind of the program counter on wave exit.
[ROCm/ROCR-Runtime commit: c710a06ee0]
* Delete static pmc files
* Counter parsing changes
- Move counter parsing logic to another function
- Fix counter parsing regex
- Log list of counters being collected
* Sanity check counters supported by rocprof
- Emit warning instead of error since rocprof support counters
list might be inaccurate
* Do not collect these counters
- TCP_TCP_LATENCY_sum (except for gfx908 and gfx90a)
- SQC_DCACHE_INFLIGHT_LEVEL
* Update logic of writing TCC channel counter definition yaml file
* Fix bug in capture_subprocess_output() utility function
- Make logging optional in capture_subprocess_output()
* Fix formatting and tests
* Update changelong
[ROCm/rocprofiler-compute commit: 58cf702d40]
* Delete static pmc files
* Counter parsing changes
- Move counter parsing logic to another function
- Fix counter parsing regex
- Log list of counters being collected
* Sanity check counters supported by rocprof
- Emit warning instead of error since rocprof support counters
list might be inaccurate
* Do not collect these counters
- TCP_TCP_LATENCY_sum (except for gfx908 and gfx90a)
- SQC_DCACHE_INFLIGHT_LEVEL
* Update logic of writing TCC channel counter definition yaml file
* Fix bug in capture_subprocess_output() utility function
- Make logging optional in capture_subprocess_output()
* Fix formatting and tests
* Update changelong
- Re-ran pip-compile on source/docs/sphinx/requirements.in
Co-authored-by: Jonathan R. Madsen <jonathanrmadsen@gmail.com>
[ROCm/rocprofiler-sdk commit: 2061c52817]
* Make sure all structs/enums can be forward declared
* Updates to counter collection
- consistency updates and cleanup
* Conversion of dimension information to info struct
* Added deprecated folder
* Testing changes
* merge changes
* Fix shadowed variable
* Source code formatting
* Fix shadowed variable
* Update rocprofiler_counter_info_v1_t member names
* Split version.h into version.h and ext_version.h
- ext_version.h contains external version info, e.g. ROCPROFILER_HSA_API_TABLE_MAJOR_VERSION, ROCPROFILER_HSA_RUNTIME_VERSION
- this reduces amount of recompilation after a commit since version.h gets updated with the git revision
* profile_config -> counter_config
* EOF new line
* [Samples] Reduce header includes + reorg counter collection samples
* Misc compilation fixes
- shadowed variables
- use of [[deprecated("...")]] in C code
- unused variables
* Minor misc modifications
- use common:: instead of rocprofiler::common:: when inside rocprofiler namespace
- counters.cpp
- move local anon namespace functions into rocprofiler::counters:: anon namespace
- use std::string_view for get_static_string
- const ref for get_static_ptr
- misc namespace shortening
* [Public API] rocprofiler_get_version_triplet + rocprofiler_version_triplet_t
- struct rocprofiler_version_triplet_t containing fields for the major, minor, and patch version
- public API function: rocprofiler_get_version_triplet
- define C++ operators for rocprofiler_version_triplet_t
- C++ function compute_version_triplet
* [Tests] Improve async-copy-testing test
- relax constraints
- improve logging
* Update counter_config.h doxygen docs
* ROCPROFILER_SDK_BETA_COMPAT
- ppdef which helps with renaming when set to 1
* Remove spurious include
* Fix includes for cxx/version.hpp
* Doxygen fixes for rocprofiler_get_version and rocprofiler_get_version_triplet
* Public API Experimental Designation
- ROCPROFILER_SDK_EXPERIMENTAL added to experimental function
- "(experimental)" added to doxygen @brief entries
* Fix use of assert instead of static_assert in hip/stream.cpp
* Use typedef instead of define for rocprofiler_profile_config_id_t
* Use inline rocprofiler_{create,destroy}_profile_config instead of ppdef
- added <rocprofiler-sdk/deprecated/profile_config.h>
* Doxygen for rocprofiler_{create,destroy}_profile_config
* ROCPROFILER_SDK_DEPRECATED_WARNINGS
* Temporarily comment out ROCPROFILER_SDK_DEPRECATED_WARNINGS=1
* cmake formatting
* Misc variable renaming in samples and tests
* Fix declarations of types
* Fix hip stream tracing service struct name
- rocprofiler_callback_tracing_stream_handle_data_t renamed to rocprofiler_callback_tracing_hip_stream_api_data_t
* Rename "HIP_STREAM_API" to "HIP_STREAM"
---------
Co-authored-by: Jonathan R. Madsen <jonathanrmadsen@gmail.com>
Co-authored-by: Benjamin Welton <bewelton@amd.com>
[ROCm/rocprofiler-sdk commit: 4cd121e27b]
* Make sure all structs/enums can be forward declared
* Updates to counter collection
- consistency updates and cleanup
* Conversion of dimension information to info struct
* Added deprecated folder
* Testing changes
* merge changes
* Fix shadowed variable
* Source code formatting
* Fix shadowed variable
* Update rocprofiler_counter_info_v1_t member names
* Split version.h into version.h and ext_version.h
- ext_version.h contains external version info, e.g. ROCPROFILER_HSA_API_TABLE_MAJOR_VERSION, ROCPROFILER_HSA_RUNTIME_VERSION
- this reduces amount of recompilation after a commit since version.h gets updated with the git revision
* profile_config -> counter_config
* EOF new line
* [Samples] Reduce header includes + reorg counter collection samples
* Misc compilation fixes
- shadowed variables
- use of [[deprecated("...")]] in C code
- unused variables
* Minor misc modifications
- use common:: instead of rocprofiler::common:: when inside rocprofiler namespace
- counters.cpp
- move local anon namespace functions into rocprofiler::counters:: anon namespace
- use std::string_view for get_static_string
- const ref for get_static_ptr
- misc namespace shortening
* [Public API] rocprofiler_get_version_triplet + rocprofiler_version_triplet_t
- struct rocprofiler_version_triplet_t containing fields for the major, minor, and patch version
- public API function: rocprofiler_get_version_triplet
- define C++ operators for rocprofiler_version_triplet_t
- C++ function compute_version_triplet
* [Tests] Improve async-copy-testing test
- relax constraints
- improve logging
* Update counter_config.h doxygen docs
* ROCPROFILER_SDK_BETA_COMPAT
- ppdef which helps with renaming when set to 1
* Remove spurious include
* Fix includes for cxx/version.hpp
* Doxygen fixes for rocprofiler_get_version and rocprofiler_get_version_triplet
* Public API Experimental Designation
- ROCPROFILER_SDK_EXPERIMENTAL added to experimental function
- "(experimental)" added to doxygen @brief entries
* Fix use of assert instead of static_assert in hip/stream.cpp
* Use typedef instead of define for rocprofiler_profile_config_id_t
* Use inline rocprofiler_{create,destroy}_profile_config instead of ppdef
- added <rocprofiler-sdk/deprecated/profile_config.h>
* Doxygen for rocprofiler_{create,destroy}_profile_config
* ROCPROFILER_SDK_DEPRECATED_WARNINGS
* Temporarily comment out ROCPROFILER_SDK_DEPRECATED_WARNINGS=1
* cmake formatting
* Misc variable renaming in samples and tests
* Fix declarations of types
* Fix hip stream tracing service struct name
- rocprofiler_callback_tracing_stream_handle_data_t renamed to rocprofiler_callback_tracing_hip_stream_api_data_t
* Rename "HIP_STREAM_API" to "HIP_STREAM"
---------
Co-authored-by: Jonathan R. Madsen <jonathanrmadsen@gmail.com>
Co-authored-by: Benjamin Welton <bewelton@amd.com>
* [SDK][rocprofv3] HIP API buffer records with args (ext)
- New buffer tracing domain(s) for HIP APIs which include the arguments and the return value in the buffer records
- Update HIP stream support for extended HIP buffer tracing
- Update rocprofv3 tool library and output library to use extended HIP buffer tracing recods
* Update stream.cpp
- handle hipStream_t address being reused for a new stream
* Update doxygen docs for rocprofiler_iterate_buffer_tracing_record_args
* Update rocprofv3 tool.cpp
- configure buffer tracing services with HIP_*_API_EXT variants
- tweak logging level for hip_stream_display_callback
* Fix validation tests
- add HIP_RUNTIME_API_EXT and HIP_COMPILER_API_EXT to valid domain names
* Serialization support for buffer tracing args
* Disable stream service for __hipPopCallConfiguration
- this is interpreted as a stream create but it doesn't create a stream
* Fix execute_buffer_record_emplace for HIP extended contexts
* Add uint64_t_retval to rocprofiler_hip_api_retval_t union
- reading in hipError_t_retval during serialization of pointer return value causes undefined behavior
* Fix compilation warning about unused but set parameter
- in hip/stream.cpp
* Add synchronization for async_copy_data
* Fix compilation error
* Fix compilation error
---------
Co-authored-by: Jonathan R. Madsen <jonathanrmadsen@gmail.com>
[ROCm/rocprofiler-sdk commit: e33dff7ad0]
* [SDK][rocprofv3] HIP API buffer records with args (ext)
- New buffer tracing domain(s) for HIP APIs which include the arguments and the return value in the buffer records
- Update HIP stream support for extended HIP buffer tracing
- Update rocprofv3 tool library and output library to use extended HIP buffer tracing recods
* Update stream.cpp
- handle hipStream_t address being reused for a new stream
* Update doxygen docs for rocprofiler_iterate_buffer_tracing_record_args
* Update rocprofv3 tool.cpp
- configure buffer tracing services with HIP_*_API_EXT variants
- tweak logging level for hip_stream_display_callback
* Fix validation tests
- add HIP_RUNTIME_API_EXT and HIP_COMPILER_API_EXT to valid domain names
* Serialization support for buffer tracing args
* Disable stream service for __hipPopCallConfiguration
- this is interpreted as a stream create but it doesn't create a stream
* Fix execute_buffer_record_emplace for HIP extended contexts
* Add uint64_t_retval to rocprofiler_hip_api_retval_t union
- reading in hipError_t_retval during serialization of pointer return value causes undefined behavior
* Fix compilation warning about unused but set parameter
- in hip/stream.cpp
* Add synchronization for async_copy_data
* Fix compilation error
* Fix compilation error
---------
Co-authored-by: Jonathan R. Madsen <jonathanrmadsen@gmail.com>
Timemory and Dyninst were updated to use BinUtil v2.42 by the default
Update Dyninst submodule
Update Timemory submodule
[ROCm/rocprofiler-systems commit: 1e9a82fe54]
* MI300 Stochastic PC sampling SDK API implementation
* ROCProfV3: Stochastic PC sampling Support (#94)
* ROCProfV3: MI300 Stochastic PC sampling initial draft
* ROCProfV3: Initial Stochastic PC sampling Tests (#95)
ROCProfV3: Initial Stochastic PC sampling tests
* Update rocprofiler_pc_sampling_record_stochastic_v0_t
- update doxygen docs for members
- replace rocprofiler_correlation_id_t with rocprofiler_async_correlation_id_t
* Relax the check in JSON tests
* drain PC sampling buffer during finalize_rocprofv3
* Increase timeout for "Test Install Build" step
- 10 minutes -> 20 minutes
- "Test Installed Packages" has 20 minutes so "Test Install Build" should also
---------
Co-authored-by: Jonathan R. Madsen <jonathanrmadsen@gmail.com>
[ROCm/rocprofiler-sdk commit: 49ce79a5b5]
* MI300 Stochastic PC sampling SDK API implementation
* ROCProfV3: Stochastic PC sampling Support (#94)
* ROCProfV3: MI300 Stochastic PC sampling initial draft
* ROCProfV3: Initial Stochastic PC sampling Tests (#95)
ROCProfV3: Initial Stochastic PC sampling tests
* Update rocprofiler_pc_sampling_record_stochastic_v0_t
- update doxygen docs for members
- replace rocprofiler_correlation_id_t with rocprofiler_async_correlation_id_t
* Relax the check in JSON tests
* drain PC sampling buffer during finalize_rocprofv3
* Increase timeout for "Test Install Build" step
- 10 minutes -> 20 minutes
- "Test Installed Packages" has 20 minutes so "Test Install Build" should also
---------
Co-authored-by: Jonathan R. Madsen <jonathanrmadsen@gmail.com>
* Provide rocprofiler_register_iterate_registration_info function
- stores the runtime arguments for later reference
* Fix compilation error
* Removed unused variable
* Store api tables in vector
* Update license
* Replace global_mutex usage with scoped_count
- in rocp_invoke_registrations
* Formatting
---------
Co-authored-by: Jonathan R. Madsen <jonathanrmadsen@gmail.com>
[ROCm/rocprofiler-register commit: fc712350ef]
* Provide rocprofiler_register_iterate_registration_info function
- stores the runtime arguments for later reference
* Fix compilation error
* Removed unused variable
* Store api tables in vector
* Update license
* Replace global_mutex usage with scoped_count
- in rocp_invoke_registrations
* Formatting
---------
Co-authored-by: Jonathan R. Madsen <jonathanrmadsen@gmail.com>
ywang103-amd
cfallows-amd
David Yat Sin
Yiannis Papadopoulos
Belton-Schure, Aidan
lyndonli
Dittakavi, Satyanvesh
Jonathan Kim
David Galiffi
vedithal-amd
Madsen, Jonathan
Welton, Benjamin
Trowbridge, Ian
dependabot[bot]
Pranjal Swarup
Indic, Vladimir