Bug #1396
CWC - high cpu load
0%
Description
The qnap use a Marvell 2.0 GHz CPU. Model Typ: TS-219P II
CPU load for one stream (HD 1080) with CWC
PID USER STATUS RSS PPID %CPU %MEM COMMAND 30941 admin S 70M 1 34.1 14.0 ld-linux.so.3 11293 admin R 960 11265 0.5 0.1 top
CPU load for two streams (HD 1080) with CWC
PID USER STATUS RSS PPID %CPU %MEM COMMAND 30941 admin S 74M 1 94.7 14.7 ld-linux.so.3 11293 admin R 1032 11265 0.7 0.2 top
CPU load for two streams (HD 1080) without CWC
PID USER STATUS RSS PPID %CPU %MEM COMMAND 30941 admin S 74M 1 1.9 14.7 ld-linux.so.3 5833 admin S 14M 1 0.9 2.9 mediasrv
related stream https://www.lonelycoder.com/redmine/boards/5/topics/6331
History
Updated by Adam Sutton almost 12 years ago
- Status changed from New to Need feedback
- Affected Versions 3.3 added
Have you tried running from the stable 3.2 tree (release/3.2). It may well be that the problem exists in both, but knowing whether or not this is the case could be useful input.
Adam
Updated by Adam Sutton almost 12 years ago
- Affected Versions 3.2 added
Thanks, that at least rules out all of the recent mods. I'll have to give some thought to what might be wrong.
Adam
Updated by John Törnblom almost 12 years ago
I'm guessing this is an arm cpu. Without mmx, sse etc, decrypting video streams with ffdecsa can be very demanding.
Updated by Adam Sutton almost 12 years ago
I think we need to figure out if this is simply a limitation of the hardware (or at least the code we're using on it etc..) I know there was talk of another lib, but although a PR was submitted there has been no follow up on it.
Adam
Updated by Robert Mabel almost 12 years ago
I test the same version on intel-based system without any problems. so i think it has to do with the arm cpu.
Updated by Adam Sutton almost 12 years ago
I'm inclined to say this is not a bug and its simply a limitation of the hardware (or at least the lack of optimisation of the software for that hardware).
Adam
Updated by Adam Sutton almost 12 years ago
I have just pushed a very old PR that adds support for dvbcsa (an alternative descrambling lib), it has been suggested this will perform better on non x86 processors since we have no optimisations for these archs.
It might be worth trying this, you'll probably need to get libdvbcsa and compile for your ARM arch etc... but might be worth a shot? I can't offer any promises though as I'm merely the messenger
Adam
Updated by Adam Sutton almost 12 years ago
- Status changed from Need feedback to Rejected
I'm closing this one, there is enough evidence to suggest the reason for the poor performance problems are lack of optimisation for descrambling on ARM platforms.
Please try building and linking libdvbcsa which may perform better on such hardware.
Adam