And how they relate to cook time?
For example for a quicktime animation movie which is roughly the size of 3 full HD I get 12.2 for last_frame_decode_time and 0.1 for last_frame_hd_read_time, but cpu cook time is only 0.1 ms
Also a related question :
Same quicktime encoded as png, file size is 10 times smaller but last_frame_decode_time is now 76 and touch can’t keep up 60 fps.
Is png decoding not multithreaded and any advantage except for file size?
Both of those occur in a differnet thread, so they wont show up in the cook time. As long as the sum of those two is below your target FPSs frame time, then they wont impact your fps. If they are two large itll cause you to drop frames.
Smaller files often have a larger decode time since its more work to decompress a heavily compressed file vs a lightly compressed one.
Great, thanks for the quick answer.
Since the decoding occurs in a different thread, I guess it’s a different thread per TOP, which is why it makes sense to crop larger res in smaller movies and stitch after, when decode time is pretty big?
Hi Malcolm, I have one more question :
If I have a few moviein TOPs reading a quicktime animation, as I keep adding moviein TOPS, at some point overall frame time rises and fps starts to drop, but I’m not sure what’s the bottleneck.
For each of the moviein, last_frame_decode_time is still below 20, and last_frame_hd_read_time still around 0.2.
Not much shows up in the performance monitor, all the time is in “rendering a window”
I have enough gpu ram as well.
It doesn’t look like it’s hard drive speed issue, but looking at the performance in the windows task manager, I’m only at 14%.
Do the threads for the different moviein TOPs run on the same processor/core ?
Thanks for clearing things up!
ps : Same test with hapq, first bottleneck seems to be the hard drive read, which makes sense.
Shows up as “last_pre_read_miss_wait” increasing, and "waiting for…"in the performance monitor.
Also what are cpu_movie_cache_hits and misses ? misses keeps increasing event when everything is smooth.
Yes, every TOP has its own threads. Ignore the CPU cache stuff, its unused in most cases and not used in this one too. How many TOPs are you playing? At somempoint the GPU upload. an be the bottleneck, which isnt visible anywhere really
Thanks Malcolm. It starts to slow down with 6-7 moviein, animation codec.
I notice gpu upload time is between 0.05 and 0.1, when for hap q it’s much lower, 0.01.
So could be it. I guess for hap q the compressed frame is uploaded whereas for cpu based codecs the uncompressed frame is uploaded.