Perfetto annotation from timemory components (#289)

* 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

[ROCm/rocprofiler-systems commit: 1aca8c177b]
This commit is contained in:
Jonathan R. Madsen
2023-06-19 19:18:04 -05:00
committed by GitHub
parent a0812bfa0b
commit c6929f545d
11 changed files with 377 additions and 101 deletions
@@ -21,5 +21,6 @@ include(${CMAKE_CURRENT_LIST_DIR}/omnitrace-critical-trace-tests.cmake)
include(${CMAKE_CURRENT_LIST_DIR}/omnitrace-attach-tests.cmake)
include(${CMAKE_CURRENT_LIST_DIR}/omnitrace-rccl-tests.cmake)
include(${CMAKE_CURRENT_LIST_DIR}/omnitrace-overflow-tests.cmake)
include(${CMAKE_CURRENT_LIST_DIR}/omnitrace-annotate-tests.cmake)
include(${CMAKE_CURRENT_LIST_DIR}/omnitrace-causal-tests.cmake)
include(${CMAKE_CURRENT_LIST_DIR}/omnitrace-python-tests.cmake)
@@ -0,0 +1,92 @@
# -------------------------------------------------------------------------------------- #
#
# papi tests
#
# -------------------------------------------------------------------------------------- #
if(OMNITRACE_USE_PAPI
AND (omnitrace_perf_event_paranoid LESS_EQUAL 3
OR omnitrace_cap_sys_admin EQUAL 0
OR omnitrace_cap_perfmon EQUAL 0))
set(_annotate_environment
"${_base_environment}"
"OMNITRACE_TIMEMORY_COMPONENTS=thread_cpu_clock papi_array"
"OMNITRACE_PAPI_EVENTS=perf::PERF_COUNT_SW_CPU_CLOCK"
"OMNITRACE_USE_SAMPLING=OFF")
omnitrace_add_test(
SKIP_BASELINE SKIP_RUNTIME
NAME annotate
TARGET parallel-overhead
RUN_ARGS 30 2 200
REWRITE_ARGS
-e
-v
2
-R
run
--allow-overlapping
--print-available
functions
--print-overlapping
functions
--print-excluded
functions
--print-instrumented
functions
--print-instructions
ENVIRONMENT "${_annotate_environment}"
LABELS "annotate;papi")
omnitrace_add_validation_test(
NAME annotate-binary-rewrite
PERFETTO_FILE "perfetto-trace.proto"
LABELS "annotate;papi"
ARGS --key-names perf::PERF_COUNT_SW_CPU_CLOCK thread_cpu_clock --key-counts 8 8)
omnitrace_add_validation_test(
NAME annotate-sampling
PERFETTO_FILE "perfetto-trace.proto"
LABELS "papi"
ARGS --key-names thread_cpu_clock --key-counts 6)
else()
set(_annotate_environment
"${_base_environment}" "OMNITRACE_TIMEMORY_COMPONENTS=thread_cpu_clock"
"OMNITRACE_USE_SAMPLING=OFF")
omnitrace_add_test(
SKIP_BASELINE SKIP_RUNTIME
NAME annotate
TARGET parallel-overhead
RUN_ARGS 30 2 200
REWRITE_ARGS
-e
-v
2
-R
run
--allow-overlapping
--print-available
functions
--print-overlapping
functions
--print-excluded
functions
--print-instrumented
functions
--print-instructions
ENVIRONMENT "${_annotate_environment}"
LABELS "annotate")
omnitrace_add_validation_test(
NAME annotate-binary-rewrite
PERFETTO_FILE "perfetto-trace.proto"
LABELS "annotate"
ARGS --key-names thread_cpu_clock --key-counts 8)
omnitrace_add_validation_test(
NAME annotate-sampling
PERFETTO_FILE "perfetto-trace.proto"
LABELS "annotate"
ARGS --key-names thread_cpu_clock --key-counts 6)
endif()
@@ -119,9 +119,8 @@ omnitrace_add_test(
--dynamic-callsites
-ME
[==[libgomp]==]
-d
wall_clock
peak_rss
--env
OMNITRACE_TIMEMORY_COMPONENTS="wall_clock peak_rss"
RUN_ARGS -i 10 -s 20 -p
ENVIRONMENT
"${_timemory_environment};OMNITRACE_CRITICAL_TRACE=OFF;OMNITRACE_USE_KOKKOSP=OFF"
@@ -995,22 +995,24 @@ function(OMNITRACE_ADD_VALIDATION_TEST)
)
endif()
add_test(
NAME validate-${TEST_NAME}-timemory
COMMAND
${OMNITRACE_VALIDATION_PYTHON}
${CMAKE_CURRENT_LIST_DIR}/validate-timemory-json.py -m ${TEST_TIMEMORY_METRIC}
${TEST_ARGS} -i
${PROJECT_BINARY_DIR}/omnitrace-tests-output/${TEST_NAME}/${TEST_TIMEMORY_FILE}
WORKING_DIRECTORY ${PROJECT_BINARY_DIR})
if(TEST_TIMEMORY_FILE)
add_test(
NAME validate-${TEST_NAME}-timemory
COMMAND
${OMNITRACE_VALIDATION_PYTHON}
${CMAKE_CURRENT_LIST_DIR}/validate-timemory-json.py -m
"${TEST_TIMEMORY_METRIC}" ${TEST_ARGS} -i
${PROJECT_BINARY_DIR}/omnitrace-tests-output/${TEST_NAME}/${TEST_TIMEMORY_FILE}
WORKING_DIRECTORY ${PROJECT_BINARY_DIR})
endif()
if(OMNITRACE_VALIDATION_PYTHON_PERFETTO EQUAL 0)
if(OMNITRACE_VALIDATION_PYTHON_PERFETTO EQUAL 0 AND TEST_PERFETTO_FILE)
add_test(
NAME validate-${TEST_NAME}-perfetto
COMMAND
${OMNITRACE_VALIDATION_PYTHON}
${CMAKE_CURRENT_LIST_DIR}/validate-perfetto-proto.py -m
${TEST_PERFETTO_METRIC} ${TEST_ARGS} -i
"${TEST_PERFETTO_METRIC}" ${TEST_ARGS} -i
${PROJECT_BINARY_DIR}/omnitrace-tests-output/${TEST_NAME}/${TEST_PERFETTO_FILE}
WORKING_DIRECTORY ${PROJECT_BINARY_DIR})
endif()
@@ -71,6 +71,20 @@ if __name__ == "__main__":
"-p", "--print", action="store_true", help="Print the processed perfetto data"
)
parser.add_argument("-i", "--input", type=str, help="Input file", required=True)
parser.add_argument(
"--key-names",
type=str,
help="Require debug args contain a specific key",
default=[],
nargs="*",
)
parser.add_argument(
"--key-counts",
type=int,
help="Required number of debug args",
default=[],
nargs="*",
)
args = parser.parse_args()
@@ -128,6 +142,26 @@ if __name__ == "__main__":
except RuntimeError as e:
print(f"{e}")
ret = 1
for key_name, key_count in zip(args.key_names, args.key_counts):
slice_args = tp.query(
f"select * from slice join args using (arg_set_id) where key='debug.{key_name}'"
)
count = 0
if args.print:
print(f"{key_name} (expected: {key_count}):")
for row in slice_args:
count += 1
if args.print:
for key, val in row.__dict__.items():
print(f" - {key:20} :: {val}")
print(f"Number of entries with {key_name} = {count} (expected: {key_count})")
if key_count != count:
ret = 1
if ret == 0:
print(f"{args.input} validated")
else:
print(f"Failure validating {args.input}")
sys.exit(ret)