extend benchmark to make the db-sync mode and node-profiler configurable #11647
Conversation
This comment was marked as outdated.
This comment was marked as outdated.
arvidn
force-pushed
the
node-benchmark
branch
2 times, most recently
from
f230a3f to
4684702
Compare
tools/run_benchmark.sh
Outdated
| @@ -2,13 +2,15 @@ | |||
|
|
|||
| run_benchmark() { | |||
| # shellcheck disable=SC2086 | |||
| python ./tools/test_full_sync.py run $3 --profile --test-constants "$1" & | |||
| python ./tools/test_full_sync.py run $3 --profile --node-profiler --test-constants "$1" & | |||
There was a problem hiding this comment.
it just occurred to me that combining --profile and --node-profiler here will cause two cProfile instances to be active at the same time. Is that supported by python? I don't recall anything stating it isn't..
There was a problem hiding this comment.
it does look like they clobber each other. I don't think it's useful to enable both.
tools/run_benchmark.sh
Outdated
| @@ -2,13 +2,15 @@ | |||
|
|
|||
| run_benchmark() { | |||
| # shellcheck disable=SC2086 | |||
| python ./tools/test_full_sync.py run $3 --profile --test-constants "$1" & | |||
| python ./tools/test_full_sync.py run $3 --profile --node-profiler --test-constants "$1" & | |||
There was a problem hiding this comment.
Is there any reduced performance from this?
There was a problem hiding this comment.
pretty minor impact on wall-time:
without --node-profiler: 116.05 s
with --node-profiler: 116.72 s
|
I removed |
example output from the node profiler: