2

I am trying to compare at least 40000 screenshots from videos with Imagick command "/dev/shm/convert s1 s2 -metric RMSE -format %[distortion] -compare info:". I have succeded to get an average speed of ~ 0.04 seconds.

If I am using in 3 threads the convert command CPU rises to 80-90% an all cores(witch it should be), but also the load on server increase exponentially and affects the average speed to ~0.09 seconds

I am using s1 and s2 image paths from /dev/shm to lower the impact on loading server and SSD. Also I am using the convert command from /dev/shm, is not enough , because it uses other libraries from the SSD.

I dont know what are the dependencies library to copy them to /dev/shm ,or any other approach would be good to achieve the same result.

B.Petrus
  • 51
  • 1
  • 6
  • How are you specifying 3 threads? The start of your question implies you are running `convert` from `/dev/shm` successfully, yet at the end you imply you don't know how to do that - which is it? Are you using the Q8 version of IM - it is often faster - check with `identify -version`? – Mark Setchell Aug 29 '17 at 18:19
  • I am referring to 3 paralel scripts with convert comand inside. Imagick with OpenMP scales to all cpu cores ,using on my server 30% on any core. Convert file from /dev/shm or /usr/bin(default path) uses other libraries like /lib64/ld-linux-x86-64.so.2 ..... (I opened the file and I saw some paths) and thats why I said that is not enough to have convert copied to /dev/shm – B.Petrus Aug 29 '17 at 18:38

0 Answers0