Jonathan R. Madsen
4208b5654c
* Initial support for GPU hardware counters
* Update find modules for roctracer and rocprofiler
- /opt/rocm/{rocprofiler,roctracer} path is deprecated so tweak search procedure
* Improve ConfigCPack for MPI
* Update rocprofiler
- rocm_metrics()
- minor cleanup
* Update rocm find modules
* declare rocm_metrics + call in omnitrace-avail
* relocate omnitrace-launch-compiler
* REALPATH and find_modules
* Examples cmake (may drop)
* omnitrace-avail
- hw_counter categories
- init rocm
* setenv updates for rocprofiler in library.cpp and dl.cpp
* get_rocm_events config
* gpu::hip_device_count()
* rocm_metrics returns hardware_counters::info
* - relocated library/components/roctracer_callbacks.* to library/roctracer.*
- relocated library/components/rocprofiler.* to library/rocprofiler.*
- cleaned up rocprofiler.hpp
- added perfetto output of rocprofiler
- added timemory output of rocprofiler
- renamed omni.roctracer thread to roctracer.hip
- added roctracer.hsa thread name
- updated timemory submodule to support std::variant
- updated timemory submodule to support = in config value
- updated timemory submodule to support standalone storage
- updated timemory submodule to support new hw counter apis
- updated timemory submodule to prevent label/description caching in data_tracker
* update omnitrace-avail info_type generation
* Update timemory submodule
* rocprofiler component
* cmake formatting
* omnitrace-avail handle no GPUs
- Add -c command-line option for --categories
- support verbosity
* hsa_rsrc_factory throws exceptions
- throw exceptions to avoid aborting on HSA_STATUS_ERROR_NOT_INITIALIZED when advantageous
- removed duplicate specialization of is_available for component::rocprofiler
* rocprofiler symbols for when disabled
* Fix warning in omnitrace-avail
- std::stringstream from initializer list would use explicit constructor
* Fix finalization after settings are deleted
* Reorganized rocprofiler source
* Updated formatting
* Miscellaneous tweaks
- added using statements from timemory
- tweaked the main and thread bundle names
- fixed timemory header includes
100 líneas
3.2 KiB
Bash
Archivo Ejecutable
100 líneas
3.2 KiB
Bash
Archivo Ejecutable
#!/bin/bash -e
|
|
#
|
|
# This script allows CMAKE_CXX_COMPILER to be a standard
|
|
# C++ compiler and omnitrace sets RULE_LAUNCH_COMPILE and
|
|
# RULE_LAUNCH_LINK in CMake so that all compiler and link
|
|
# commands are prefixed with this script followed by the
|
|
# C++ compiler. Thus if $1 == $2 then we know the command
|
|
# was intended for the C++ compiler and we discard both
|
|
# $1 and $2 and redirect the command to linker.
|
|
# If $1 != $2 then we know that the command was not intended
|
|
# for the C++ compiler and we just discard $1 and launch
|
|
# the original command. Examples of when $2 will not equal
|
|
# $1 are 'ar', 'cmake', etc. during the linking phase
|
|
#
|
|
|
|
# emit a message about the underlying command executed
|
|
: ${DEBUG:=0}
|
|
: ${OMNITRACE_DEBUG_LAUNCH_COMPILER:=${DEBUG}}
|
|
|
|
debug-message()
|
|
{
|
|
if [ "${OMNITRACE_DEBUG_LAUNCH_COMPILER}" -ne 0 ]; then
|
|
echo -e "##### $(basename ${BASH_SOURCE[0]}) executing: \"$@\"... #####"
|
|
fi
|
|
}
|
|
|
|
# if omnitrace compiler is not passed, someone is probably trying to invoke it directly
|
|
if [ -z "${1}" ]; then
|
|
echo -e "\n${BASH_SOURCE[0]} was invoked without the omnitrace compiler as the first argument."
|
|
echo "This script is not indended to be directly invoked by any mechanism other"
|
|
echo -e "than through a RULE_LAUNCH_COMPILE or RULE_LAUNCH_LINK property set in CMake.\n"
|
|
exit 1
|
|
fi
|
|
|
|
# if omnitrace compiler is not passed, someone is probably trying to invoke it directly
|
|
if [ -z "${2}" ]; then
|
|
echo -e "\n${BASH_SOURCE[0]} was invoked without the C++ compiler as the second argument."
|
|
echo "This script is not indended to be directly invoked by any mechanism other"
|
|
echo -e "than through a RULE_LAUNCH_COMPILE or RULE_LAUNCH_LINK property set in CMake.\n"
|
|
exit 1
|
|
fi
|
|
|
|
# if there aren't two args, this isn't necessarily invalid, just a bit strange
|
|
if [ -z "${3}" ]; then exit 0; fi
|
|
|
|
# store the omnitrace compiler
|
|
OMNITRACE_COMPILER=${1}
|
|
|
|
# remove the omnitrace compiler from the arguments
|
|
shift
|
|
|
|
# store the expected C++ compiler
|
|
CXX_COMPILER=${1}
|
|
|
|
# remove the expected C++ compiler from the arguments
|
|
shift
|
|
|
|
# discards the clang-tidy arguments
|
|
if [ "$(basename ${1})" = "cmake" ] && [ "${2}" = "-E" ] && [ "${3}" = "__run_co_compile" ]; then
|
|
c=1
|
|
n=1
|
|
for i in "${@}"
|
|
do
|
|
if [ "${i}" = "--" ]; then
|
|
break;
|
|
fi
|
|
if [ "${c}" -gt 3 ]; then
|
|
n=$((${n} + 1))
|
|
fi
|
|
c=$((${c} + 1))
|
|
done
|
|
|
|
if [[ $# -gt ${c} ]]; then
|
|
n=$((${n} + 3)) # add three because of the first 3 args
|
|
for i in $(seq 1 1 ${n})
|
|
do
|
|
shift
|
|
done
|
|
fi
|
|
fi
|
|
|
|
if [[ "${CXX_COMPILER}" != "${1}" ]]; then
|
|
debug-message $@
|
|
# the command does not depend on omnitrace so just execute the command w/o re-directing to ${OMNITRACE_COMPILER}
|
|
eval $@
|
|
else
|
|
# the executable is the C++ compiler, so we need to re-direct to ${OMNITRACE_COMPILER}
|
|
if [ ! -f "${OMNITRACE_COMPILER}" ]; then
|
|
echo -e "\nError: the compiler redirect for omnitrace was not found at ${OMNITRACE_COMPILER}\n"
|
|
exit 1
|
|
fi
|
|
|
|
# discard the compiler from the command
|
|
shift
|
|
|
|
debug-message ${OMNITRACE_COMPILER} $@
|
|
# execute ${OMNITRACE_COMPILER} (again, usually nvcc_wrapper)
|
|
${OMNITRACE_COMPILER} $@
|
|
fi
|