Open arpaterson opened 2 years ago
Thanks for letting me know. I haven't tested this recently, but I can see why multiple threads might be blocked writing to the same log file. There might be some options to make
or tee
that can resolve this.
Possibly ninja
build could do a better job, although I don't know if cmake -G ninja
works with OpenCV yet. Thanks for filing the issue. I'll look into this if I get a chance. Otherwise, feel free to submit a PR disabling logging for the moment.
No worries!
(On a Jetson Orin)
This script does not use more than one core, regardless of the value in
$CPUS
or$JOBS
.I have been having issues building opencv and getting it working with realsense-ros, so I've been reconfiguring and building using my own script, and the one in this git.
I noticed mine is dramatically faster, despite being almost identical aside from a few flags.
Changing nano_build_opencv's make command
make -j${JOBS} 2>&1 | tee -a build.log
tomake -j$(($(nproc)-1))
, removes the redirection of the output to the log file but gives a dramatic speed up, with jtop showing multiple or all 12 cores pegged at near 100%.I think the cmake step is also affected.
Suspect this affects all systems.