#apertus IRC Channel Logs

off to bed now ... have a good one everyone!

good morning to all

https://github.com/gr8geek/apertus-challenge

also https://github.com/gr8geek/apertus-challenge/blob/master/characterdevice/test.c has long functions, which is also not good for proper maintenance, software developmemnt is not about "it works", but about "it works and other developers can improve or extend it, even after long time"

BandiT1983: I made the changes 🙂

by the way, formatting is still gruesome

I follwed kernel.org guidelines

and what do they say about empty lines between methods?

I will fix it

Well I jusst realized that the technique I use to calculate checksum is tough sensitive to character position but still is not appropriate

Bertl: didn't know that english queen has made you a Sir

which is technically wrong but still has become a habit and passed from generation to generation because of Once we were part of British colony

A more appropriate address for Herbert would be Mage. He likes that.

i hope a mage which is not like Rincewind

98%'ish.

RexOrCine: is it news power board?

That's the ironic thing about it. No, it's not PBV2X.

yep, preparing the base and then you or whoever feels like can continue, am at home mostly, because of restrictions, so looking at interesting topics to work on and visual stuff is one of them

would like to know for rendering, if it is possible to let it look alike

can imagine

ok, thasnks, will look for BRD data

I just now saw the axiom micro project on apertus team talk on youtube

Good to see you exploring AXIOM's world fully pratyush.

Your account should be activated in the Labs now.

I am an AI and electronics enthuaist

Definitely.

Long live apertus

whats the password

i shall reset

morning folks!

BAndiT1983|away: I'm okay with Rincewind ;)

Bertl: Is it okay?

at line 240 -242

https://github.com/axiom-micro/software/tree/master/camera_control

pratyush: so line 240-242 is okay to look at?

everything?

well, looking at chardvr.c, line 15 is the first formatting error

fixing it

so what was the problem there?

well, not so much about 'alignment' more about having a space between #include and <>

oh noo #define parts

can you fix it?

nah, doesn't look like it

tab removed

good start, still no banana there

Bertl: Please do build and see the output

pratyush: I'm not going to download and test the code before all the formatting is correct

so you can do better than what you did so far?

pratyush: let me know when the formatting is perfect!

Bertl: to find out my shortcomings

(kernel coding style guide)

yep, but looking at existing code should be quite helpful as well

yes, that looks okay

is it parallel clock or embedded clock? or 8b/10b serdes?

hmm? context?

at line 20

please try to rephrase the question

Heading 5

is dev_t a struct or a pointer?

okay

Also we are currently working on the third revision of the hardware, that will most likely include significant changes many core components of the current micro.

vup:is there any project related to computational photography

vup : I have something on my mind http://image-sensors-world.blogspot.com/2019/02/adobe-unveils-ai-based-demosaicing.html

feel free to suggest your own idea (T727) ... note: please do not change the task itself, add your own task and link it there

okay

Well there is no fixed plan but anuejn and we are quite interested in such projects. I think anuejn already looked a bit into AI based debayering.

pratyush: you also need to find a mentor able and willing to mentor your idea

Bertl : I have fixed formatting plese review

spacing error

pratyush, isn't it easier to get a formatter and not guess around every time?

axiom remote bootloader

anuejn : I have a project idea related to AI based debayering algorithim

pratyush, just by posting PDFs it isn't done, it requires more informative document, but from your side, which shows how it should be done and how to implement it for the apertus base

Off to bed now ... good night

Bertl: Hello! I have completed the task1... i.e. added ioctl read... please have a look: https://github.com/Swaraj1998/apertus-kernel-challenge/tree/master/task1

Okay!

any idea why that could be so?

precisely

first, csum_arg_t is not something you do in the kernel

and then, using a string to transfer an uint32_t via IOCTL is rather weird

actually i did do a dynamic one before, where I used char* as an argument, the issue with it was that there was no size information... easy to screw up for the user... so i thought to wrap it up in a struct... for that I had to fix a size... and hence I did stack allocation of the struct (is 32 bytes an issue for stack?)... But I could allocate the struct in heap....

btw, the copy_to_user() line is over the line length limit :)

yes, there it is just fine, although I would probably drop the prefix as you only have _one_ value there

actually you should try to fit it into less than 80 characters

<Bertl "yes, there it is just fine, alth"> I can drop the prefix everywhere....

for proc it makes working with proc values simpler

(in a bash script)

the integer?

oh okk

i should drop the prefix then

if you have several values like for example in /proc/meminfo

gotcha

will make the changes then

yes I am on it... bdw proposal and CV will just be the one proposal document right? Or a different cv is required?

bluez_[m]: you can make it one, but it wouldn't hurt to have a separate CV

<Bertl "bluez_: you can make it one, but"> I don't yet have a seperate cv for now... was thinking to make it one... I

bertl: I would like to work with the pixel remapper project. Alongside I am thinking to take upon on the post processing tasks after acquisition of frames and build another module to perform writes to sd card in order to store the resulting video? Is this already done? If not, do you think that this would be useful?

and where can I know the current status of the project as a whole (Regarding functionalities)

we are dealing with a lot of usually raw data here and the SD card is magnitudes slower

certainly

please elaborate!

This makes sure that the quality of video remains same even if the area is overlit or underlit

I do not think that is particularily useful for the typical use case, i.e. making movies

and you definitely do not want to mess with the raw data on the camera to hide over or under exposure

We can provide an additional feature of controlling contrast which is seen in usual cameras also

they also have the big advantage that they do not mess up the raw data

hey max_bxl!

Do you think lossless compression of raw images can be done? If not on FPGA but with using cpu so that it even reduce the processing latency as a whole?

on teh CPU side, the performance is too low to keep up with the data

Is it the lossless JPEG encoder?

about updating the wiki with firmware 2.0 commands : should I keep the "old" command under my changes (if someone still uses the old firmware) or should I make it super clear ?

IMHO it would be best to have two pages, one for the old firmware and another one for the new

sounds like a good idea

or just add a section at the beginning

(at the top)

yes

my impression was, that the algorithm was invented for ML

you can work on whatever you like, but if you want to apply for GSoC 2020, you need to pick an existing task or propose your own, which needs to be convincing, useful and you need to find a mentor for it as well

Hello

Wanderer_: one big problem of jpeg2000 appears to be, that you need quite a lot of memory bandwidth to calculate it

any idea where the problem come from ?

probably the wrong png library

Requirement already satisfied: pypng in /usr/lib/python3.8/site-packages (0.0.20)

check with the author of the python script?

max_bxl: is your png.py similar to this? -> https://github.com/google/AppSpeedIndex/blob/master/png.py

where can I find png.py in the Beta ?

called pypng

but its already installed in this firmware I assume as you are not getting the error message "ImportError: No module named 'png' "

anuejn:

ok I find where png.py is

pip show --files pypng

great

reference: https://pip.pypa.io/en/stable/quickstart/

so this can be fixed in the stock firmware already

can you check the file itself?

how to check with bash ?

or nano

i don't mean the file date, but there is a history table inside the png.py

yes version 0.0.18 was the last one that had it

it seems it was a helper for some cli functionality, that was now moved into its own file: pripamtopng, I think downgrading to version 0.0.18 is a ok workaround for now

0.0.19 hand plenty of changes, according to the logs

here is the changelog in the middle -> https://github.com/drj11/pypng

i've got this for png.py __version__ = "0.0.20"

yeah just try pip install --force "pypng==0.0.18" for now

seems to work but I forgot to do axiom-start.sh beforehand so I have to wait for Bertrand to power off/power on the camera...

and it's done !

thank for everything, will go to bed now and update the wiki tomorrow !

bye everyone!

bye