v4l2-ctl: Question on last_ts

v4l2-ctl: Question on last_ts

[Max Schulze]

Hello Hans,


I do not understand the concept on last_ts.

> cap dqbuf: 0 seq:      0 bytesused: 512000 ts: 137342.420951 delta 
> last: 137342420.951 ms (ts-monotonic, ts-src-soe)
> cap dqbuf: 1 seq:      2 bytesused: 512000 ts: 137342.484954 dropped: 
> 1 (ts-monotonic, ts-src-soe)

It seems to print the timestamp on the first frame, but then never again 
(how useful is it, to print the same timestamp twice?).


> if (last_ts <= 0.0)
>             fprintf(f, " delta last: %.03f ms", (ts - last_ts) * 1000.0);
>         last_ts = ts;

Might it be, that commit 222e1760c7 "utils: fix implicit float 
conversion" got the float conversion wrong and it should rather be

>         if (last_ts > 0.0)

To my understanding, this would yield a much-more useful output of 
printing the diff to the last frame _in every frame_


> cap dqbuf: 0 seq:      0 bytesused: 512000 ts: 143215.916971 
> (ts-monotonic, ts-src-soe)
> cap dqbuf: 1 seq:      2 bytesused: 512000 ts: 143215.984995 delta 
> last: 68.024 ms dropped: 1 (ts-monotonic, ts-src-soe)
> cap dqbuf: 2 seq:      3 bytesused: 512000 ts: 143216.016956 delta 
> last: 31.961 ms (ts-monotonic, ts-src-soe)
> cap dqbuf: 3 seq:      4 bytesused: 512000 ts: 143216.052981 delta 
> last: 36.025 ms (ts-monotonic, ts-src-soe)
> cap dqbuf: 0 seq:      5 bytesused: 512000 ts: 143216.088920 delta 
> last: 35.939 ms fps: 29.08 (ts-monotonic, ts-src-soe)

Re: v4l2-ctl: Question on last_ts

[Hans Verkuil]

On 27/01/2021 11:44, Max Schulze wrote:
> Hello Hans,
> 
> 
> I do not understand the concept on last_ts.
> 
>> cap dqbuf: 0 seq:      0 bytesused: 512000 ts: 137342.420951 delta 
>> last: 137342420.951 ms (ts-monotonic, ts-src-soe)
>> cap dqbuf: 1 seq:      2 bytesused: 512000 ts: 137342.484954 dropped: 
>> 1 (ts-monotonic, ts-src-soe)
> 
> It seems to print the timestamp on the first frame, but then never again 
> (how useful is it, to print the same timestamp twice?).
> 
> 
>> if (last_ts <= 0.0)
>>             fprintf(f, " delta last: %.03f ms", (ts - last_ts) * 1000.0);
>>         last_ts = ts;
> 
> Might it be, that commit 222e1760c7 "utils: fix implicit float 
> conversion" got the float conversion wrong and it should rather be
> 
>>         if (last_ts > 0.0)
> 
> To my understanding, this would yield a much-more useful output of 
> printing the diff to the last frame _in every frame_

Yep, you are correct. It's now fixed. I was wondering about that as well
when I looked at the code.

Thanks for reporting this!

	Hans

> 
> 
>> cap dqbuf: 0 seq:      0 bytesused: 512000 ts: 143215.916971 
>> (ts-monotonic, ts-src-soe)
>> cap dqbuf: 1 seq:      2 bytesused: 512000 ts: 143215.984995 delta 
>> last: 68.024 ms dropped: 1 (ts-monotonic, ts-src-soe)
>> cap dqbuf: 2 seq:      3 bytesused: 512000 ts: 143216.016956 delta 
>> last: 31.961 ms (ts-monotonic, ts-src-soe)
>> cap dqbuf: 3 seq:      4 bytesused: 512000 ts: 143216.052981 delta 
>> last: 36.025 ms (ts-monotonic, ts-src-soe)
>> cap dqbuf: 0 seq:      5 bytesused: 512000 ts: 143216.088920 delta 
>> last: 35.939 ms fps: 29.08 (ts-monotonic, ts-src-soe)
>