#apertus IRC Channel Logs

And I didnt find any solution to the weird behaviour yet :/

hi

10:27

I have a question

as AVFS does not work on linux

why does it not work in linxu?

Read somewhere.. let me find the link

Your link: https://github.com/vapoursynth/vapoursynth/issues/93

another alternative

some guys on magiclantern use fuse, but we wanted to use existing stuff, which also would allow to do adjustments to format in between

https://www.magiclantern.fm/forum/index.php?topic=13152.0

we could ask g3gg0, when he is online, as he is mentioned there

Let me have a look

haven't seen him online for quite some time

[thinking about fixes for VS] - ?

FUSE is also focused on linux mostly, so it's also a bit of hassle to use possilby

nowadays there should be not that much problems

don't know if performance is degraded much

You know better than me :)

good question, maybe first a general research about frame servers, so we know if there is an alternative

supragya, C++17 gets filesystem lib, maybe it will be not that difficult to use it in the future to provide frame server for multiple systems

[maybe it will be not that difficult to use it in the future to provide frame server for multiple systems] - it being MLV Fuse or VS

just general thought

don't know how much effort it requires to get stable linux version

when with VS or MLV?

C++17 is just partially implemented in compilers, as far as i know, so it would be a lot of hassle to track down possible features at the moment

https://en.cppreference.com/w/cpp/compiler_support

okay...

if we know what frame server does, then we can estimate what it takes to fix VS or try FUSE or think about some own implementation, maybe also with C++17, as the compilers seems to be rather complete

some pages are not really conclusive about the state of filesystem in compilers

VLC specific?

frame server should avoid that

anyhow, to accept the streams from the frameserver, there is a requirement of plugins for frameserver anyhow

?

the general goal of the task is to have a dummy video file, which is operated through frame server

BAndiT1983: supragya: using fuse on windows shouoldnt be too hard, as there are multiple projects, that implement userspace filesystemdriver libs with fuse warappers

anuejn, do you know the general process to get frame server through it?

well... i would create a virtual file and deliver correct frame data in the read callback

and with constant bitrate and constant framerate

but having almost uncompressed video wont hurt that much in this case, i think

but with those assumptions it should be quite doable

but i would focus mainly on one wildly supported format, as it is has almost no benefit to be able to support multiple formats

I guess it will

supragya, just another thing, as i have to leave in a moment

i think some quite agressive caching of frames in ram would be helpful for performance

*embarassed

TofuLynx didn't know, can you explain it to hitm, please?

once he comes online

I will tell him

that's why i'm asking as i have to pack my furniture etc., so i won't be online for some time

okay

he should just commit it to his fork first and then i can review it with him, as he had trouble with dual images, something is off in the buffers

see you

https://github.com/apertus-open-source-cinema/beta-software/tree/master/software/cmv_tools/cmv_hist3

In task it says "First steps are understanding and improving this tool with small new features (crop image and extract histogram only from selection), etc." does that mean I have to improve cmv_hist3 source code?

Hi Iti nmdis1999 and Bertl_oO !

for extracting histogram?

Hi Supragya!

Cool :D

We have to read image data from raw12 file in T734?

hey!

I click the fork button?

TofuLynx, sorry was away

TofuLynx, how can I help?

okay..

Yes, just click that fork button, it will be forked to your account

Just simply as that?

Or if there were no changes made to OC after you cloned it locally

From what BAndiT1983 told me, you should then move your changes to the forked repo

He will then have a look at that

this is similar to the raw format but not identical

1.Read Image type (I still don't know what should be the image type)

3.After reading image I'll convert it to grayscale (I want to use Luminosity method)

What do you think?

it is basically 4 sensel R,G1,G2,B packed together with overlay information in each 64bit word

for performance reasons, I would avoid going through the memory more than once (memory access is the main limiting factor here)

cropping an image (as suggested in the task) is best done by leaving out the memory access for those pixels completely

btw, I just realized, that we did not explicitely mention applying for more than one task, but as we have quite a number of tasks and are unlikely to get slots for each task, it makes a lot of sense to apply for more than one task to increase your chances to get a GSoC spot

Yes, indeed. But considering last year scenarios (you guys got 3 slots) this year it's probable they'll give at least 3 slots, isn't it?

and of course, you can still work on your tasks when you are not selected for GSoC and we _will_ mentor you as time permits

True, I can.

Have you guys recieved any application in VHDL task?

In memory pattern.

What kind of feedback do you guys give about the applications?

http://www.magiclantern.fm/modules/modules/mlv_structure.zip/mlv_structure.zip

so the in memory representation (as for T34) is basically an address in memory where you have a number of 64bit words which contain R/G/G/B values each

Doesn't raw12 have sometimes some info appended at the end?

Understood!

(they are a register dump of the CMV12k sensor registers)

regarding feedback to applications: you should make a good first impression there (you only get one chance to do that :) so it should already be 'as good as possible' (layout, style, content, etc)

Before submitting the application I think it's better if get it reviewed , right?

Roger that! :)

How's your work going @TofuLynx?

Noted, thanks for this :D

TikZ and PGF for diagrams and illustrations is a good idea too :)

good afternoon

Hello Davelister!

@TofuLynx: Google Docs is good as well. I'm decent in LaTeX but haven't written any article or stuff in it. Just used it casually.

I haven't tried LaTeX yet, although I heard great things about it

otherwise, that's great :-)

(I can't count how many hours I spent to fix just a page rendering into my dissertation)

But yeah overall it's very good

When I resize the image to half the width and Height, the output becomes weird, with the correct width and height, but the image gets scaled to twice it's height

I took some screenshots

https://i.imgur.com/oJBE0ki.png - normal one, without resizing the OCimage

after resizing the OCimage values to half: https://i.imgur.com/LzzVVVL.png

I think there are a few culprits here..

What are your suspicions?

so this means that there is error in debayering code

For most part... I am more than sure it's that

Not LodePNG

This happened to me during the challenge

So you're suggesting that the debayering class has hardcoded size ?

If you look in the second image, even the aspect is incorrect

Yeah, it's vertically scaled wrongly

Can you explain why this behaviour happens?

I have a few theories as to why this happens but am not really sure

Thanks :)

The image is cropped to 1/4

Point me to the specific files, will you TofuLynx ?

downscaler class?

That extracts only a quarter of the pixels

To improve the performance

Check the TIFFloader

It is responsible for controlling the extraction and the process

his class should replace bayerframepreprocessor if we need fast preview

So I have to remove from TIFFloader the debayer function?

BAndiT1983, Then I am more than sure that if his implementation is right, this problem will not occur

Ok! Thanks for the productive discussion! :)

i hoped not

It will show grayscale

supragya, why should it be grayscale?

something made it RRRRRRR...GGGGGG....BBBBB

but is rendered as LCT_GREY or similar

BAndiT1983, we should clearly look at the values that are fed in

It should show full image color

TofuLynx has just to walk through every second line and get values from 2x2 blocks

R, B and (G1 + G2) / 2

for(unsigned int i = 0; i<width*height; i++){

img[3*i + 1] = ((int*)ocimg.GreenChannel())[i]<<4;

}

so color data is there, will have green tint etc., but sufficient

the same error we see here

*?

Even without it, we have error

as the shift right was used to get 8bit per channel

processingtest has something similar for conversion for thumbnail test ->

{

interleavedArray[i * 3 + 1] = (greenArray[i] >> 4) * 1.0;

}

It maybe the case

you want to shift pointer, if i see correctly

This happened earlier to me

Well I just have to remove the debayer process I guess

Only the thumbnail image will get wrong, rift?

replace bayerframeprocessor with your downscaler

It's just it doesn't circumvent debayering

debayering is done in presenter, just comment out process()

i call it downsclaer, because the image will be half the resolution, it just grabs filled pixels

BAndiT1983, FUSE is perhaps what we need.

I forgot it.

supragya, you just have to skip some pixels when extracting from memory :)

libfuse is there for OSX

frameserving using libfuse will be much easier than VS..

many years ago i had to build blender almost every day, as the fire/smoke branch was not included, it was done while gsoc by the way

ubuntu gave me almost no problems, although a bit cumbersome to get all libs, but windows was more awkward to get to build

Wow you contributed to the particles system of Blender?

was just building so i could do special effects for some hobby film group

a year later, i've joined apertus, so had no time to enter blender development

You joined apertus in 2015?

joined 4,5 years ago

BAndiT1983, https://www.cs.nmsu.edu/~pfeiffer/fuse-tutorial/src/bbfs.c

Ok :)

I guess that is it

Much easier and useful that something cumbersome like VS

BAndiT1983, haven't gone much deep into libfuse but last GSoC, I had to read about VFS-FUSE coupling for microkernel systems

{

interleavedArray[i * 3 + 1] = (greenArray[i] >> 4) * 1.0;

}

So, I know much of FS is easily done through FUSE

However, this specific problem needs an eye

main advantage when doing multi-platform stuff, different bugs and erros are reported on different systems

?

DNG image is imported into OCimage, right?

but only certain image is placed there at the momen, like width and height plus RGB data arrays

Uh

So it's always changing OCimage for each frame?

Several OCimages?

you have to check it, i don't have an overview right now, was doing many other things in apertus for last year

Ok!

could you upload your code soon?

https://www.ibm.com/developerworks/library/l-fuse/

i suppose, but i don't know the details yet, general outline is known, but i would like to know on which events we have to react and so on

i have bookmarked your links for later, cannot dive into that stuff currently, other things have more priority in my life at the moment

If I get home soon (I'm shopping at the moment), I think I can upload today

TofuLynx, no problem, take your time, will take a look into it when i'm in the train tomorrow morning

:D

was glad that we have included some selenium UI tests

supragya, first the study, it's more important

BAndiT1983, It's Heritage and Culture exam. No one studies

and this is more interesting

reminds me of big bang theory when Sheldon describes to Raj indian rituals and Raj asks why he explains his own culture to him :D

not tomorrow, today.. it's midnight

how is Gandhi seen in india?

usually apertus stuff is taking my time at the moment

what is the exam exactly about? his life? influence?

about british part of it also?

Low battery, see you soon :)

BAndiT1983, where are you staying btw, which country, state? and the rest of apertus community?

i'm leaving in hessen/germany

Wu bist du, Guten Tag, Guten Morgan, handynummer - that's my german

always get confused in s, z and beta

and german is not my native language

AndroUser connects from android?

this s, z, and beta is why I got [B] in german

I'm Iti nmdis1999, just download an app for my android phone. Don't know how to change nickname her

AndroIRC is it?

XD

have only used andchat

It's late and I didn't wanted to use laptop rn

http://wiki.androirc.com/preferences

waiting for qt install in VM

AndroUser, have you tried to open options?

This app isn't taking my commands -. -

Will do later... I don't have energy to waste on this xD

nope, russian

Changing name is easy

now BAndiT1983 on android

yeah, tested name changing for Iti

I know it's easy but this won't change it

App*

Iti, I am laughing too hard.... nice one BAndiT1983

just wanted to check if i can connect if i'M not at home

You watch Bing Bang Theory?

everybody loves sheldon, no? :)

Yes!!!

new sheldon?

haven't seen last seasons

Lol xD

just watching first season with my girlfriend, as she didn't knew this series

Well truth be told... never watched BBT

o. O

? why is to be behind with movies and series not hip today?

Is it? I mean I watch less series and movies compared to my friends but that's because I'm busy working on different things. Although, I do love playing games.

some people looked in awe at me, that i don't know certain series or movies

*shocked

well I am behind this because I like work and then in spare time connecting with people, not machines.

have gathered enough hobbies in my life, that i can easily bypass watching tv

?

Well, I personally don't like FB

*when they hear

it'S not aobut open source, it's generally about technical stuff and software development as hobby

Because you work in your extra time, it's not common as I said. Most people work for money, status etc etc. You are devoting time open source which means you are giving all your time to work.

(BAndiT1983) [have just transformed my hobby into my profession] : that's why :)

crytek, money not appealing ?

game company sound funny, but you have to work a lot and compared to what you get paid, it's not that much

once i saw crysis and cryengine, and then frostbite.... I was dumbstuck by visuals

pffff, when i've seen prince of persia 1 and f-19 i've been struck by that and started to learn qbasic :P

ahahahahahhhahahahah

https://www.apertus.org/node/324

Bertl_oO, totally forgot about that!

f22 was awesome for that time, especially after beatiful comanche and other engines based on their game

I personally feel it's a great game.

what are you doing to me? I am lauging so hard

Bertl_oO, really, CEO of well known Japanese company

supragya, without 4k display you don't see all the details of pong ;)

Even my calculator is good enough for that

liquid/smoke for pong?

not same but similar

Really cool :)

so, OC is finally working in VM, can see DNG, now i just need code from TofuLynx to help with his downscaler

good luck and good night

such laugh

AndroUser: nn

with a dynamic wallpaper and it is actually running on the Beta FPGA

Bertl_oO, is it possible to do purely in FPGA?

i know that logic will be harder to implement, but it's an interesting challenge to use minimal number of slices

is full one required or are just parts sufficient

you can do the logic in pure HDL as well, but it is probably not worth the efford

but pong in an FPGA is not a big problem even without soft cores

yup

still can't wrap my head around the topic, but it's just a matter of time

matter of time

i like the approach of the software, but some things are a bit hard to use

nothing more

just remember, that it has almost single contributor

Hey

hi TofuLynx_

for VS?

ah, haven't done pull request, but asked them to take a look if it's ok to submit, not important at the moment as golang is not used for apertus firmware currently

Hmm, btw what is golang good at... why to go for it?

was suggested by some other guys and i had to learn it, to be able to help the student last year

started with go 1.6, which had large compiled files, after short time seen that 1.9 is out and switched, files got smaller

no real advantages for me at least

good laughs xD

see the april fool video

I saw the april fools video xD

supragya BAndiT1983

please upload you code to forked repo, my OC works now and i can take a look as long as i have time

have 3,5 hours tomorrow, when riding train, so then it will be latest i can response to you about it, hope this time wifi works

But mark my word, the patch is in the arithmetic

but my fork isn't ready yet

i know, but it takes quite some time, so doing it already, afterwards i can pull your commit

just commit your code and i can try it

it shows inline if there are errors or warnings for questionable code

?

TofuLynx: did you cause it?

then nothing to be sorry about:)

(dont know the correct name in english)

magic wand is ok i think

depends on what you mean ... google for 'magic wand' and pick one :)

Bertl_oO, just google for "Zauberstab englisch"

No I am asking why I did so

oh no xD

https://static.fnac-static.com/multimedia/Images/PT/NR/71/a3/0c/828273/1540-1/tsp20160812194238/Bosch-Varinha-Magica-MSM88110.jpg

looks reasonably similar, but I guess it's getting worse *G*

magic

https://www.youtube.com/watch?v=N-bjiCSnDug

;D

I can't commit or push using Qt Creator

haven't used that for commit

have you set up your git, e.g. name, email etc.?

let me check

sebastian says: we want to explore the emotional depths of these ____ players locked in an eternal battle for balls and glory XD

xD

but GUI is very important when doing bigger software, can't get enough overview in the terminal

Going to dinner now, see you :)

BAndiT1983, It's a heritage exam with open book system. So I am just copying notes

supragya, know such exams, some even allowed laptops, one guy has got math software for an exam with him :D

i know, just looks funny, if you know the smilie

this -> o_O

my fav smilie... also doesnot need any fancy system to show and is not tilted like :)

captures human emotion better

see you

good night

nn

Bertl_oO

nothing yet :)

hmm

is getzi still around here?

goodnight raj!

TofuLynx_, have adjusted a bit, seems to work now, but haven'T tested thoruroughly

{

_dataGreen[dataIndex] = (_outputData[index + 1] + _outputData[index + _width]) >> 1;

dataIndex++;

I saw that G3ggo just published the new task

index += _width;

}

dataSize is width * size, just for performance reason, so CPU does not re-calculate with each iteration

i haven'T compared changed code yet, let us discuss that tomorrow, it'S late here and i have to catch the train in the morning

Ok :) We shall discuss tomorrow

main change was the loop

Good night everyone!