Unless SDMA blits have actually been used for copies, prevent the DMA
copy status from querying the blit's pending byte status to avoid
creating an unnecessary HW queue.
Change-Id: Ied1fbed73c08f0408f0e3583f9b56f2768c71708
Querying pending bytes on a blit kernel is unnecessary when runtime
runs out of SDMA resource since we are returning an SDMA availabilty
mask.
Change-Id: I347efba0c85b70ea3ba8749d76a499afc23909e8
Added HSA_AMD_MEMORY_POOL_GLOBAL_FLAG_EXT_SCOPE_FINE_GRAINED flag to enable extended scope memory region
where the device-scope atomics act as system-scope atomics
Change-Id: I79fc3207cb630dfc68bed2f8aabd75f35fe80b12
Enable sleep for all waiters with event age tracking support kernel.
Change-Id: Icd4e1e8d83b4a54e9f6aaa99691a6573211b3337
Signed-off-by: James Zhu <James.Zhu@amd.com>
KFD kernel version 1.13 starts to support event age
tracking which help elimating unncessary busy wait.
Change-Id: Ib447ed6e0350f3110a4d6b9b80a0388000dd0e72
Signed-off-by: James Zhu <James.Zhu@amd.com>
* Support ROCm 5.5 in docker
* Update containers workflow
- add Ubuntu and OpenSUSE container builds for ROCm 5.4 and 5.5
- add RHEL builds
* Update cpack workflow
- build on PR against main when cpack.yml or docker files updated
- removed packaging for ROCm < 5.2 for many OSes
- added packaging for ROCm 5.5
* Update OpenSUSE workflow
- add python 3.11 to OMNITRACE_PYTHON_ENVS
- upload-artifacts name includes strategy.job-index (prevent overwrite)
- only upload artifacts on failure
- continue on error if upload artifacts fails
* Update RedHat workflow
- provide run-name
- add python 3.11 to OMNITRACE_PYTHON_ENVS
- upload-artifacts name includes strategy.job-index (prevent overwrite)
- only upload artifacts on failure
- continue on error if upload artifacts fails
* Update Ubuntu (Bionic) workflow
- add python 3.11 to OMNITRACE_PYTHON_ENVS
- upload-artifacts name includes strategy.job-index (prevent overwrite)
- only upload artifacts on failure
- continue on error if upload artifacts fails
* Update Ubuntu (Focal) workflow
- add python 3.11 to OMNITRACE_PYTHON_ENVS
- upload-artifacts name includes strategy.job-index (prevent overwrite)
- only upload artifacts on failure
- continue on error if upload artifacts fails
- remove testing of ROCm 4.3, 5.0, 5.1
- add testing of ROCm 5.5
* Update Ubuntu (Jammy) workflow
- add python 3.11 to OMNITRACE_PYTHON_ENVS
- upload-artifacts name includes strategy.job-index (prevent overwrite)
- only upload artifacts on failure
- continue on error if upload artifacts fails
- add testing of ROCm latest
* Dockerfile.{rhel,opensuse} update
- remove use of amdgpu-install in favor of installing rocm-dev package
- In ROCm 5.5, amdgpu-install changed meaning of --usecase=rocm (added rocmdev use case)
* redhat workflow update
- remove use of amdgpu-install in favor of installing rocm-dev package
- In ROCm 5.5, amdgpu-install changed meaning of --usecase=rocm (added rocmdev use case)
* build-docker.sh update
- add '--progress plain' to docker build commands
* Ubuntu (jammy) workflow update
- fix rocm installation
* Update Dockerfile.rhel
- add LIBRARY_PATH for /opt/amdgpu/lib64 for redhat
* Update Dockerfile.rhel
- install libpciaccess for rocm
* Annotate perfetto with timemory component data
- support perfetto annotations via timemory component data, e.g. use PAPI component for exact HW counter annotations
* Tests for perfetto annotation via timemory data
* Update omnitrace-instrument
- remove --default-components argument as this overrides any components set in configuration file
- required by perfetto annotation via timemory data tests
* filter unavailable timemory components
- filter out unavailable timemory components before attempting to invoke the annotate operation on the bundle
* update annotate tests
- account for no PAPI support
* update lulesh-timemory test
- replace '-d wall_clock peak_rss' with '--env OMNITRACE_TIMEMORY_COMPONENTS="wall_clock peak_rss"'
* annotate tests update
- fix misnamed test
* annotate tests update
- restrict binary rewrite to run function to force instrumentation despite heuristics
* annotate tests update
- print {available,overlapping,excluded,instrumented} functions during binary rewrite
* annotate tests update
- add allow-overlapping flag
* Support PAPI with CAP_SYS_ADMIN
- do not disable PAPI if perf_event_paranoid > 2 but has CAP_SYS_ADMIN capability
Compiler behavior is undefined if the right operand is negative,
or greater than or equal to the width of the promoted left operand.
For release builds with address sanitizer enabled, this compiler
optimization behavior leads to unsupported queue size value since
current method shifts till 128 bits on a 64 bit value.
Change-Id: Iddcc15b43d2331bc8bf5fc3aa4725f76844655ec
Signed-off-by: Sreekant Somasekharan <sreekant.somasekharan@amd.com>
Earlier, hsa-runtime was unable to find symbols from a stripped ELF-image becasue
no support to find symbols from ".dynsym" section.
Looking for symbols in .dynsym is enabled by LOADER_USE_DYNSYM=1
environment variable
Change-Id: I4f0e8dd0eb053a6066d4d49b670c52e51149531a
Fix rocprofiler usage in ROCm >= 5.5.x
- starting with ROCm 5.5.0, rocprofiler throws exception if OnLoad + dlopen librocprofiler
- CI skipped for this PR since CI does not support GPU usage (tested locally)
kokkosp update
- disable tracing deep copies by default because inconsistent begin/end corrupts call-stack hierarchy
- tracing deep copy operations can be enabled via `OMNITRACE_KOKKOSP_DEEP_COPY`
This has slowly become less and less reliable on more and more ASICs,
so just blacklist it altogether. Using wall clock for performance
is not a reliable method for testing performance, so skip it to avoid
more failure reports on various systems.
Change-Id: I1a5744604e4620bc7675a629d146ba4ffba669d2
Remove docker hiplibsdk from amdgpu-install
- amdgpu-install use case hiplibsdk is not necessary and bloats the install
- same as above for package rocm-hip-sdk