I tried to use the following samples.
My environment is as follows.
Jetson Orin NX 16GB
CPU arch: aarch64
Jetson Linux 35.4.1
CUDA 11.4
Compute Sanitizer 2021.2.2
And, I add my account to debug group.
https://docs.nvidia.com/compute-sanitizer/ComputeSanitizer/index.html#tegra-setup
cd Memcheck
dbg=1 make
make run_leakcheck
As a result, Compute Sanitizer could not detect memory leak.
Running out_of_bounds_kernel: no error
========= LEAK SUMMARY: 0 bytes leaked in 0 allocations
========= ERROR SUMMARY: 2 errors
aladram
September 5, 2023, 11:33pm
2
This sample requires CUDA 12.1 or newer (missing cudaDeviceReset
, see release notes ), which is currently not available on Jetson. Editing file memcheck_demo.cu
by adding cudaDeviceReset();
before return 0;
in main
should address the issue.
Thank you for your reply.
I add cudaDeviceReset
in main().
As a result, Compute Sanitizer could detect memory leak.
========= Leaked 1024 bytes at 0x2028ce000
========= Saved host backtrace up to driver entry point at cudaMalloc time
========= Host Frame: [0x1dc980]
========= in /usr/lib/aarch64-linux-gnu/tegra/libcuda.so.1
========= Host Frame: [0x42be0]
========= in /home/dandelion/compute-sanitizer-samples/Memcheck/memcheck_demo
========= Host Frame: [0xf854]
========= in /home/dandelion/compute-sanitizer-samples/Memcheck/memcheck_demo
========= Host Frame: [0x4fe2c]
========= in /home/dandelion/compute-sanitizer-samples/Memcheck/memcheck_demo
========= Host Frame: [0x97d8]
========= in /home/dandelion/compute-sanitizer-samples/Memcheck/memcheck_demo
========= Host Frame:__libc_start_main [0x20e10]
========= in /lib/aarch64-linux-gnu/libc.so.6
========= Host Frame: [0x9490]
========= in /home/dandelion/compute-sanitizer-samples/Memcheck/memcheck_demo
=========
========= LEAK SUMMARY: 1024 bytes leaked in 1 allocations
system
Closed
September 20, 2023, 2:08am
4
This topic was automatically closed 14 days after the last reply. New replies are no longer allowed.