[IEEE ICRA'23] A new lightweight LiDAR-inertial odometry algorithm with a novel coarse-to-fine approach in constructing continuous-time trajectories for precise motion correction.
Hey there,
when I compile dlio (and dlo) with clang, the average computation time more than triples in comparison to using gcc ( ~100ms instead of ~30ms for an Ouster OS0 128).
I tried this on a AMD Ryzen 7 5800H with ubuntu 20.04 using clang 12.0 and gcc 9.4.
With clang, the CPU was running at full core utilization (x16) and a CPU load around 99% while dropping around 30% of the scans.
With gcc, the CPU runs at around 8% and slightly above 1 CPU utilized.
Any idea were this problem may come from?
Hey there, when I compile dlio (and dlo) with clang, the average computation time more than triples in comparison to using gcc ( ~100ms instead of ~30ms for an Ouster OS0 128). I tried this on a AMD Ryzen 7 5800H with ubuntu 20.04 using clang 12.0 and gcc 9.4. With clang, the CPU was running at full core utilization (x16) and a CPU load around 99% while dropping around 30% of the scans. With gcc, the CPU runs at around 8% and slightly above 1 CPU utilized. Any idea were this problem may come from?
Best regards, Jan