Current Server Time: 23:29 (Central Europe)

#apertus IRC Channel Logs

2019/06/13

Timezone: UTC


01:04
illwieckz
joined the channel
01:13
Y_G
left the channel
02:28
futarisIRCcloud
joined the channel
02:58
apurvanandan[m]
Hello Bertl, are you available?
04:30
Bertl
kind of ... almost off to bed now
04:31
Bertl
apurvanandan[m]: what can I do for you?
04:33
apurvanandan[m]
I was stuck somewhere, but I found the way out. And yes I am going to test my first ft601 confroller just now :)
04:33
Bertl
excellent!
04:34
apurvanandan[m]
Good night :)
04:41
Bertl
have a good one!
04:41
Bertl
changed nick to: Bertl_zZ
05:57
BAndiT1983|away
changed nick to: BAndiT1983
06:17
BAndiT1983
changed nick to: BAndiT1983|away
07:53
se6astian|away
changed nick to: se6astian
08:16
se6astian
left the channel
08:16
apurvanandan[m]
<Bertl "have a good one!"> Hey Bertl , The ft601 controller worked :D . I am able to receive the count from the counter on the connected PC at 335 MBps. I tried at +1 increment and +2 increment and both worked. Very low resources of MachXO2 are being used by ft601 controller. Yet I will optimise the code and push on github :)
08:16
se6astian
joined the channel
08:17
se6astian
left the channel
08:17
se6astian
joined the channel
08:18
se6astian
left the channel
08:18
apurvanandan[m]
But currently I used the ft245 mode, will write for ft600 mode also easily :)
08:18
se6astian
joined the channel
11:32
se6astian
left the channel
13:42
Bertl_zZ
changed nick to: Bertl
13:42
Bertl
morning folks!
13:42
Bertl
apurvanandan[m]: great news!
15:54
dev__
joined the channel
15:58
BAndiT1983|away
changed nick to: BAndiT1983
16:05
dev__
Hello BAndiT1983 , Can I get some Info related to "Evaluate how we could implement a video clip" which describes itself as to make a class having reference to frames.
16:05
BAndiT1983
hi dev__
16:06
BAndiT1983
what's your understanding of the task?
16:07
dev__
We currenly have OCImage which have parameters related to frames
16:08
BAndiT1983
not really, as OCImage is the frame at the moment
16:08
dev__
Now, May be , we want to a class which gives us access to all frames position in RAW input
16:08
BAndiT1983
so in order to be able to keep data small and divided between responsible classes, we should add a class which presents a video clip
16:10
dev__
a video clip contain more n number of frames and we are trying to store there information , right ?
16:10
dev__
(more than one)*
16:11
Bertl
off for now ... bbl
16:11
Bertl
changed nick to: Bertl_oO
16:11
BAndiT1983
frames would only contain necessary info, but video clip would hold such things as resolution
16:12
BAndiT1983
as it doesn't really change, partially it also relates to the other item in trello, about the order of loader process, as downscaler returns data which is half as big, because it grabs only known frames and we would lose a lot of space otherwise
16:13
BAndiT1983
also DNG loading is a bit broken at the moment, later we wouldn't really load single DNG files, but folders full of DNG, and this also requires video clip class
16:13
BAndiT1983
for single DNGs there are better options, like RawTherapee or Darktable, but we are targeting moving pictures
16:16
dev__
So this class will contain functionalities by which we can read files present in a directory and keep there information
16:19
BAndiT1983
yes, as most infos don't change, like resolution or frame rate
16:20
BAndiT1983
this would also allow to do check for playback, like total frame count and so on, but you should also evaluate how to read big files, without decreasing performance
16:20
BAndiT1983
maybe memory mapping or similar would help
16:20
BAndiT1983
regarding frames, they should contain only a reference to the bucket in the pool allocator, like red -> bucket 34
16:21
BAndiT1983
frames will occupy 3 buckets of uint16_t, for RGB
16:22
BAndiT1983
later we can try to allocate multiple pages in static allocator, if we need more amount and OS would limit us, e.g. only 512mb per block/page
16:26
BAndiT1983
is it clear so far?
16:26
dev__
does Each single file would represent a frames ?
16:28
BAndiT1983
in case of DNG yes
16:28
dev__
Do we read MLV files like this , I guess no
16:28
BAndiT1983
but if we load first frame and store the offset of image data, then we can grab the data of every other frame at same offset
16:29
BAndiT1983
like how?
16:29
dev__
Each MLV file already contain more than one frame
16:30
dev__
we can grab the data of every other frame at same offset : yes , its clear , as all files have same parameters
16:31
BAndiT1983
i was talking about DNG, MLV doesn't need special handling, but it would also benefit from faster loading if we can check what the frameSpace amount is, and if it's not different just load the data, as the resolution wouldn'T change, so we don't need to read headers over and over
16:31
BAndiT1983
later it has to be extended, as DNG files contain more data, like white balance, but this is for later point in time
16:32
BAndiT1983
supragya has evaluated MLV last year, so he knows the details, in case we want to do smarter loader
16:33
dev__
okay , First I will focus on DNG only
16:35
BAndiT1983
stay with MLV first, as i expect extended allocator first, no need to jump between tasks, as DNG loader is implemented and just needs some adjustments
16:35
BAndiT1983
first allocator, then video clip class to track frames and hold general infos like resolution
16:35
BAndiT1983
afterwards we can evaluate further
16:35
BAndiT1983
have you done first refactoring? can i check the code?
16:36
dev__
yes, I tried to make MLVLoader small
16:36
dev__
https://github.com/kakashi-Of-Saringan/opencine/commits/dev
16:37
BAndiT1983
the formatter settings are not good at the moment, will check on them, as i don't like the indentation of =
16:38
BAndiT1983
maybe we can keep it, as for the loader it's not that bad, but looks a lot like old C stuff
16:38
BAndiT1983
will review later, as i have to accomplish other stuff which is not related to apertus or computers at all
16:39
dev__
okay , please see Pool allocator code also , I have updated it
16:40
BAndiT1983
will do
16:40
BAndiT1983
have you fixed GetFrameCount() and other things?
16:40
dev__
Nope, I just pushed changes related to MLVloader.cpp
16:40
BAndiT1983
frames should not reference memory, but rather buckets, hope that it will be implemented, if it wasn't yet
16:40
dev__
I will do that
16:41
BAndiT1983
GetFrameCount() which gives amount of channels back is very confusing
16:41
BAndiT1983
that's why video clip class is required and frame class
16:41
BAndiT1983
OCImage will be renamed later and will contain only frame related data
16:42
BAndiT1983
like i've mentioned before, probably OC::Image::Frame
16:42
dev__
okay ,
16:46
BAndiT1983
in the future it shouldn't be handled by allocator, but by playback, which would request to load more frames, so you would use a bit different request like videoClip->GetTotalFrameCount()
16:47
BAndiT1983
this is just an example of simple playback, usually this would be done through events and whatnot
16:48
dev__
left the channel
16:55
sebix
left the channel
17:08
dev__
joined the channel
17:11
Dev
joined the channel
17:11
Dev
changed nick to: Guest42896
17:21
se6astian
joined the channel
17:39
danieel
left the channel
17:42
Y_G
joined the channel
18:12
Guest42896
left the channel
18:12
dev__
left the channel
18:16
BAndiT1983
changed nick to: BAndiT1983|away
18:16
Y_|G
joined the channel
18:17
Y_G
left the channel
18:19
BAndiT1983|away
changed nick to: BAndiT1983
18:24
Y_|G
left the channel
18:43
danieel
joined the channel
21:08
niemand
joined the channel
21:10
Y_G
joined the channel
21:24
BAndiT1983
changed nick to: BAndiT1983|away
21:37
BAndiT1983|away
changed nick to: BAndiT1983
21:59
BAndiT1983
Y_G, added some comments to trello regarding the code
22:00
BAndiT1983
off for today
22:00
BAndiT1983
changed nick to: BAndiT1983|away
22:07
Y_G
Hi BAndiT1983|away, had a look .Will incorporate the changes
22:15
se6astian
off to bed, good night
22:16
se6astian
changed nick to: se6astian|away
22:16
Y_G
left the channel
22:28
niemand
left the channel