I create an nvjpegencoder instance with fd mode, and then use it for continuous encoding.After running for a while, the coding will be particularly slow. If a nvjpegencoder instance is created for each frame of encoding, and destroy it after the completion of encoding, there is no such problem. Is there any internal resource leakage?
The hardware platform used is Xavier, and the system version is 32.3.1
Hi @hongliuzhu,
Could you upgrade to JP4.4DP release?
This issue is not reproduciable on JP4.4DP release with below test command.
// could change “#define PERF_LOOP 300” to more loops,e.g. 30000
$ ./jpeg_encode …/…/data/Picture/nvidia-logo.yuv 1920 1080 test.jpg --encode-fd --perf