For large bar platform, it's not necessary to mmap
memory allocated on VRAM to cpu again
Change-Id: I0701680476829d4058b3e7b643e8df657d0c6168
Signed-off-by: Ruili Ji <ruiliji2@amd.com>
[ROCm/clr commit: 25fe45bb2a]
For large bar platform, it's not necessary to mmap
memory allocated on VRAM to cpu again
Change-Id: I0701680476829d4058b3e7b643e8df657d0c6168
Signed-off-by: Ruili Ji <ruiliji2@amd.com>
* Return the result by value and not through a pointer passed by
parameter
Change-Id: I8f872c95c4a330bebe299d486fb73f36660b469c
[ROCm/clr commit: 37849a0726]
The print statement had extra character 'f' resulting in syntax error and rocprofiler runtime package was not getting generated in centos
Change-Id: Ib4472f4879c4c50ffb00e8336ceef818021c965f
The print statement had extra character 'f' resulting in syntax error and rocprofiler runtime package was not getting generated in centos
Change-Id: Ib4472f4879c4c50ffb00e8336ceef818021c965f
[ROCm/rocprofiler commit: 231948235a]
* omnitrace-avail updates
- enables text wrapping for descriptions
- reworks the HW counters display layout
- added new column "Device" which has either "CPU" or "GPU"
- support sorting HW counters alphabetically
- fixed some minor csv issues
- reorganize the order of the argparse arguments
* Fix tests
[ROCm/rocprofiler-systems commit: b39a683eab]
* omnitrace-avail updates
- enables text wrapping for descriptions
- reworks the HW counters display layout
- added new column "Device" which has either "CPU" or "GPU"
- support sorting HW counters alphabetically
- fixed some minor csv issues
- reorganize the order of the argparse arguments
* Fix tests
Revert agent_info->dev_index in hsa_rsrc_factory.cpp
- disable using the driver_node_id and, instead, start at zero
because it breaks the lookup for rocprofiler_pool_fetch
lookup in rocprofiler.cpp. On my system (one AMD GPU and
one NVIDIA GPU), it has a value of 1, not 0 and the pool
size is 1 -- resulting in segfault
[ROCm/rocprofiler-systems commit: ab8894082b]
Revert agent_info->dev_index in hsa_rsrc_factory.cpp
- disable using the driver_node_id and, instead, start at zero
because it breaks the lookup for rocprofiler_pool_fetch
lookup in rocprofiler.cpp. On my system (one AMD GPU and
one NVIDIA GPU), it has a value of 1, not 0 and the pool
size is 1 -- resulting in segfault
* rocprofler_iterate_info workaround + omnitrace-avail update
- provides workaround for rocprofiler_iterate_info behavior change in ROCm 5.4.0-3
- update timemory submodule with argparse tweaks
- updates hsa_rsrc_factory.{hpp,cpp}
- colorized log in omnitrace-avail
- Bump version to 1.9.2
* Fix empty_base inheritance
- timemory's component::empty_base inherits from concepts::component so direct inheritance was removed
* Fix OMNITRACE_HIP_VERSION_COMPAT_STRING
- defined as "" when OMNITRACE_HIP_VERSION_MAJOR==0
* new defines + extra info
- define OMNITRACE_LIBRARY_ARCH (via CMAKE_LIBRARY_ARCHITECTURE)
- define OMNITRACE_SYSTEM_NAME (via CMAKE_SYSTEM_NAME)
- define OMNITRACE_SYSTEM_PROCESSOR (via CMAKE_SYSTEM_PROCESSOR)
- define OMNITRACE_SYSTEM_VERSION (via OMNITRACE_SYSTEM_VERSION)
- define OMNITRACE_COMPILER_ID (via CMAKE_CXX_COMPILER_ID)
- define OMNITRACE_COMPILER_VERSION (via CMAKE_CXX_COMPILER_VERSION)
- include this info in metadata
- include subset of this info in --version for bin tools
- tweak to perfetto verbose messages
[ROCm/rocprofiler-systems commit: 4ed5f3e67b]
* rocprofler_iterate_info workaround + omnitrace-avail update
- provides workaround for rocprofiler_iterate_info behavior change in ROCm 5.4.0-3
- update timemory submodule with argparse tweaks
- updates hsa_rsrc_factory.{hpp,cpp}
- colorized log in omnitrace-avail
- Bump version to 1.9.2
* Fix empty_base inheritance
- timemory's component::empty_base inherits from concepts::component so direct inheritance was removed
* Fix OMNITRACE_HIP_VERSION_COMPAT_STRING
- defined as "" when OMNITRACE_HIP_VERSION_MAJOR==0
* new defines + extra info
- define OMNITRACE_LIBRARY_ARCH (via CMAKE_LIBRARY_ARCHITECTURE)
- define OMNITRACE_SYSTEM_NAME (via CMAKE_SYSTEM_NAME)
- define OMNITRACE_SYSTEM_PROCESSOR (via CMAKE_SYSTEM_PROCESSOR)
- define OMNITRACE_SYSTEM_VERSION (via OMNITRACE_SYSTEM_VERSION)
- define OMNITRACE_COMPILER_ID (via CMAKE_CXX_COMPILER_ID)
- define OMNITRACE_COMPILER_VERSION (via CMAKE_CXX_COMPILER_VERSION)
- include this info in metadata
- include subset of this info in --version for bin tools
- tweak to perfetto verbose messages
Xie.AlexBin
Chauncey Hui
Ajay
Ruili Ji
Juan Manuel MARTINEZ CAAMAÑO
Icarus Sparry
Maneesh Gupta
Hao Zhou
Ranjith Ramakrishnan
Jonathan R. Madsen
Nirmal Unnikrishnan
AravindanC
Giovanni LB
gobhardw
Kiumars Sabeti