[BOLT] Add itrace aggregation for AUX data #70426
Conversation
|
Question for reviewers: Do any Bolt tests run Bolt and pass in a perf.data file and check the .fdata is correct? |
|
@maksfb @rafaelauler Please can I have a review. |
yota9
changed the title
|
Please use [BOLT] in commit message |
jonathandavies-arm
force-pushed
the
itrace-aggregation
branch
from
fe96ac5 to
f87cb30
Compare
We have several internal tests for |
bolt/lib/Profile/DataAggregator.cpp
Outdated
| @@ -155,6 +160,8 @@ void DataAggregator::findPerfExecutable() { | |||
| } | |||
|
|
|||
| void DataAggregator::start() { | |||
| std::string ItracePerfScriptArgs; | |||
There was a problem hiding this comment.
Is there a reason to declare ItracePerfScriptArgs in this scope?
There was a problem hiding this comment.
I just saw other std::string being declared at the beginning of functions. I've moved the declaration into the if statement scope.
Noted. I didn't port those as we had to sanitize inputs and perf data is tricky since it's opaque and captures a lot of system information by default. We do have some perf2bolt tests in rafaelauler/bolt-tests repo, but not a comprehensive suite. CC @lanza |
If you have a perf.data with Arm ETM data the only way to use perf2bolt with Branch Aggregation is to first run
perf inject --itrace=l64i1us -o perf-brstack.dataand then pass the new perf-brstack.data into perf2bolt. perf2bolt then runsperf script -F pid,ip,brstackto produce the brstacks.This PR adds
--itracearg to perf2bolt to enable Itrace Aggregation. It takes a string which is what is passed to theperf script -F pid,ip,brstack --itrace={0}. This command produces the brstacks without having to run perf inject and creating a new perf.data file.