#apertus IRC Channel Logs

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

Bertl : I'd like to get started on T731 and it'd be great if you could provide resources for the same.

(but he checks chat logs)

hi [simon], can you elaborate? do you want to propose something from Qualification tasks?

ah yes, you can choose from one of the related tasks, for which you have done the qualification challenge. (make sure you don't do VHDL challenge when applying for kernel btw) :)

(if I want to see the history of the channel is there a way to display it in my irc client?)

I cannot see other way of doing so... until you are from apertus :)

a ok

[simon], yes weekly schedule are a part of GSoC proposal

if you are starting a new project... you should understand the task thoroughly(or atleast have a general sense of what is expected), figure out how difficult it is and then put forth your proposal. Note that weekly schedules are great way to keep track of time and deliverables, but you will not be failed in evaluations just because things moved ahead or before your expectations. Still, keep it as close to what you plan to do in summer :)

?

rahul_: the focus peaking should probably go in the middle of the image pipeline (beginning of the output pipeline), but we will decide where to put it exactly

okay, In my proposal should I add the AXI Stream and AXi lite interface as well? or just pixel data_in (RGB+EOL+EOF) and Pixel clock.

so basically we will tap off between when the memory is fetched (via AXI) and before the data is sent to TMDS encoding

Bertl, I didn't add it for reviewing yet any help will be appreciated also can i send my proposal draft to you on mail to check it out?

an yes, of you have some challenge code or proposal to review, I'll gladly do it

Bertl, is there any specific template format for Apertus, that I need to follow.

it also is a good idea to make it look nice and appealing (think CV) and thus avoid things like Microsoft Word or similar (go for TeX, LaTeX or so)

Bertl: are you also avoiding libreoffice and such?

BAndiT1983: well, libreoffice basically tries to do what the Microsoft Office package does ... it does it a little better but it is still no comparison to proper type setting (as with LaTeX)

(after all it is Google SoC :)

Bertl: met a lot of people who are know nothing about software development etc., so latex is a problem, as they also don't know HTML and similar

well, there is Lyx for those who need/want a GUI for LaTeX

still like the feature to add custom tags to latex a lot, we used it for ABM already

would like to use latex in the company, but i'm afraid that my colleagues would block it right away, although it would get rid of badly formatted word docs when several people adjusted them already

it depends how it was done, also know several people who have done their diploma/master thesis in word, switched to latex for mine back when, as word broke my layout

*back then

you forget about people with little technical knowledge, a secretary wouldn't hack text into LyX

fear of computer technology

he or she would not have to 'learn' how to 'work' with the text processor but could focus on entering text

MK: checking the updated challenge task now ...

quite nicely formatted now ... only a tiny number of missing spaces

on the semantic part, I wonder, have you tested this code with a real FPGA toolchain (i.e. further than synthesis)?

okay, because there are some constructs in your VHDL which have a good chance to fail on almost any real world toolchain/FPGA

i know what you are about should i change them ?

I'm fine if you're aware of the problems (maybe add a note to the code what the problem is and how you could/would work around it)

Bertl, sure i will add a comment

hi BAndiT1983

it takes the size from the OCimage it gets

Bertl, did you see my question?

Bertl saw your question, however maked it with TD-Linux rather than TofuLynx :)

ah i suspected that

supragyaraj, not sure if I understood you, but you want the frame to resize automatically?

yeah, continue

so when bytes are requested, they are provided

there is an external app that will control the quality (downscale)

Oh I see

what if...

the downscaler resizes the result?

basically, subsampling

Yeah I would suggest subsampling

currently it's resizing

as per your needs

TofuLynx, do that please! :)

I will have to edit my proposal

Yeah, it's probably just simple maths

is it Only me or apertus lab is not working

lab is down

this too anuejn ? -> https://lab.apertus.org/tag/google_summer_of_code_2018/

seen some trouble today at lab however... few hours ago wasn't working for me

my friend's phone work but mine not lol

hi TofuLynx

can you look it and perhaps run it at your end?

I am beginning to couple the AVIEncoder with libfuse now... finally will couple with OCcore when it is done

your loops for synthetic frames can be made simpler and faster

width / 3 repeats itself a couple of times, this can be extracted, otherwise the software would calculate it at every iteration, possibly also multiple times

this would hold the value and no recalculation required

okay

i'm not a fan of C style, but it will work for now

your code looks more like C and not C++

later

sure :)

for the headers?

ah, i've thought about other calcs

okay

you should use -> using namespace ... in .cpp files, to avoid cluttering it up with namespace::

does using namespace span accross files or is file specific?

using namespace ... is only for the file

for this reason only, I have put it all into AviContainer::calculateSize(sizeType type)

looked at it, so the question came up

you can also avoid setting struct members to 0

if you want to ensure that all of them are zero at the beginning, then try memset

usually yes, just wanted to ensure

if you do new ..., then it should be zeroed

I am not looking into pool allocator for now... would like to just frameserve this avi first before anything

is it okay that way?

don't worry, was just some consideration for later

subsampling vs resizing while downscaling

all the offsets change too...

BAndiT1983, if we have 100x100 image and quality set to 1/2, the resulting frame is 100x100 or 50x50?

for 1:1 we need some interpolation, but quick one would be sufficient for preview

?

but it was never considered to do it on the fly

more important is the "on the fly" thing for frame serving, evrything else can be set when servinf was paused

if however we fill every alternate pixel with nearest value, we can sample it up to constant framesize and then it becomes same "on the fly"

BAndiT1983, can you explain this? [more important is the "on the fly" thing for frame serving, evrything else can be set when servinf was paused]

[simon], can you give the github link?

https://github.com/Symon97x/GoogleSummerOfCode.git

it should be up now.. kindly check

I have an account in lab aperture...I'll try to add it

ok I'll remove them

use .gitignore feature Simon :)

hi TofuLynx

hmm Andrej, are you occupied right now?

regarding errorhandling, here three examples:

file without permissions: Couldn't open specified file!

Hmm?

I see the point

the third one is just misleading because you probably start to fiddle with args 2 and 3

what do you mean by fiddling with args 2 and 3?

the error suggests that something is wrong with them

btw, I will be doing in depth reviews/testing this weekend (on the GSoC challenge tasks) so that's a good chance for folks to improve ...

because it should say that the file is empty or doesn't contain enough data

Ok! Roger!

yep

[simon], by binary executables, BAndiT1983 did not mean to say the raw12 file, he meant a.out file :)

aaaaa ok lol

found it, checking now ...

I agree with process labels, but I don't understand the comment there: "If you can't it is because it either do too much work or sometimes too little"

generics on the other hand can (and probably should) be treated like constants

_d, _dd, _d<n> for delayed versions of a signal

_i, _o and _t for input, output, tristate of one signal

_a for an array type

_clk for a clock signal

main : process (all) :)

well, sometimes that's fine, but usually not, I agree, but I wouldn't have interpreted your comment as that

and I think it would be a good idea to make some examples regarding spaces around operators and indentation for various statements

how can i rephrase this comment? what do you think?

like keep processes and functions small and specialized with a meaningful label

MK: almost everything ... it's hard to read and it opens a can of work regarding how to handle lowerCaseNames vs UpperCaseNames

Exactly that is why i add it some people like and the other not

Off for today! Goodnight!

let's stick with snake_case for VHDL :)