Analysis & Opinion

[Feature Request] Benchmarking

70 posts
Thanks Meter: 25
By vivan000, Member on 20th November 2014, 01:58 PM
Post Reply Subscribe to Thread Email Thread
Once upon a time there was TCPMP (CorePlayer) for WM. It had one interesting feature - benchmarking. This options forces player to play video as fast as possible and measures FPS (until you hit stop or video ends).

Such feature is really useful for testing overall perfomance and perfomance hit/gain of different options (how much "speed-up tricks" help, how much subtitle rendering consumes, new decoder optimisations, etc.).
21st November 2014, 07:49 AM |#2  
CDB-Man's Avatar
Senior Member
Thanks Meter: 193
Interesting thought. Though, you could always use something like Antutu, or get an FPS meter app?
21st November 2014, 10:51 AM |#3  
vivan000's Avatar
OP Member
Flag Saint-Petersburg
Thanks Meter: 25
Originally Posted by CDB-Man

Interesting thought. Though, you could always use something like Antutu, or get an FPS meter app?

It's a bit different - it measures real playback perfomance, not some abstract number. Even if it was measuring pure cpu perfomance - different archictures have different efficiency at video decoding (think about extentions like MMX/SSE/AVX on x86), plus decoder gets better over time (you get more fps for same cpu perfomance).

Is 40k in antutu enough to play 720p hi10p flawlessy? "It depends".
Fpsmeter will (at best) show only frame drops - when player was not fast enough to draw a frame. If you play 30 fps video and it will say that it plays at 20 fps - it doesn't mean that you can play similar video at 20 fps or that you need to get 50% faster. And if it plays without frame drops - you'll never know how much extra perfomance you have.

But that way it would be possible to do such things:
1) Run video and say:
- "hey, it runs at >120%, I don't need to touch anything to be happy".
- "it runs at 100%, which means that it barely could play it - I need to do something".
- "it runs at <80%, nothing will help so it's better to give up".
2) Change settings and say:
- "switching to yuv/rgb32/rgb16 made it 10% faster, so I should probably use it if I'm happy with quality"
- "I needed some extra perfomance and speed-up tricks got me extra 30% - just what I needed"
3) Give video and ask to benchmark it and then judge how capable the device it (I've seen people that say "flawlessly"/"watchable"/"playable" at 15 fps).

For example I've wasted hours testing hi10p perfomance on my Z3c - sometimes it plays flawlessely, sometimes overheats (drops cpu freq), sometimes lags... and there're different setting to play with, let alone videos with different complexity (and subtiles).
Last edited by vivan000; 21st November 2014 at 10:57 AM.
22nd November 2014, 01:28 PM |#4  
sumantaadk's Avatar
Flag Narayangarh
Thanks Meter: 54
Mx is a media player not a benchmarking tool. I think this feature will only hog unnecessary space for thousands of people.
The Following 2 Users Say Thank You to sumantaadk For This Useful Post: [ View ]
6th December 2014, 12:43 PM |#5  
Junior Member
Thanks Meter: 2
I partially agree with the OP.
Benchmarking would help with identifying how fast the decoding/rendering is on a certain device.

However, I think there's more value doing this for the ffmpeg team
Post Reply Subscribe to Thread
Previous Thread Next Thread
Thread Tools Search this Thread
Search this Thread:

Advanced Search
Display Modes