Where communities thrive


  • Join over 1.5M+ people
  • Join over 100K+ communities
  • Free without limits
  • Create your own community
People
Repo info
Activity
  • Apr 10 15:31
    codecov[bot] commented #573
  • Apr 10 14:36
    codecov[bot] commented #573
  • Apr 10 14:36
    lpranam synchronize #573
  • Apr 10 14:06
    codecov[bot] commented #573
  • Apr 10 14:06
    Build #147 passed
  • Apr 10 13:11
    lpranam synchronize #573
  • Apr 10 11:29
    mloskot commented #583
  • Apr 10 07:59
    meshtag converted_to_draft #583
  • Apr 10 07:59
    meshtag commented #583
  • Apr 10 07:20
    Scramjet911 edited #588
  • Apr 10 07:17
    Sayan-Chaudhuri commented #392
  • Apr 10 07:12
    Scramjet911 edited #576
  • Apr 10 06:08
    codecov[bot] commented #583
  • Apr 10 06:01
    meshtag synchronize #583
  • Apr 10 06:00
    codecov[bot] commented #582
  • Apr 10 05:07
    codecov[bot] commented #581
  • Apr 10 04:16
    codecov[bot] commented #582
  • Apr 10 04:16
    meshtag synchronize #582
  • Apr 10 03:43
    meshtag synchronize #581
  • Apr 10 03:17
    codecov[bot] commented #581
Mateusz Łoskot
@mloskot
From my own fairly long experience with GIL, I can only confirm that returning back to the docs is worth it. They may feel overwhelming at first but that is why reaching them frequently pays off
Mateusz Łoskot
@mloskot
Speaking of docs, I need to get back to structuring the image processing.
I want to create chapters almost directly based on structuring proposed in "Principles of Digital Image Processing" by Burger & Burge
Mateusz Łoskot
@mloskot
@miralshah365 I'm looking at your PR. Please, take the nitpicking comments as just comments for some future consideration, not as request for changes now
Mateusz Łoskot
@mloskot
This seems like a cool idea for a pet project for GIL user
https://www.pyimagesearch.com/2015/09/07/blur-detection-with-opencv/
Olzhas Zhumabek
@simmplecoder
I'm writing senior project regarding 3d gaze estimation. I'm still unsure though how to integrate GIL with CUDA. The simplest approach is I believe just creating some sort of a buffer that would hold the content part of image, then just add ability to create image_views from it. I have no idea though how to initialize image_view with just a pointer, or if it is even possible. Too many abstraction layers to peel off
Mateusz Łoskot
@mloskot
Look at interleaved_view and other view creating functions.
If not inside GIL source tree, web search for interleaved_view and you will find lots of examples. I think it's one of most popular GIL function :-)
Olzhas Zhumabek
@simmplecoder
nvcc seems to choke on GIL real hard. I believe I'll need to setup some sort of incremental compilation or something like that
I believe it is because it tries to compile every function, and it needs to do some register allocation stuff which slows it down
Mateusz Łoskot
@mloskot
I wish I could help, but zero experience with CUDA here.
You may have much better luck on channels of cpplang.slack.com, #cuda, #gpu etc
Olzhas Zhumabek
@simmplecoder
thanks! I believe I'll be able to take it from here. May be create a cpp file that I could just link to, and compile only once
Olzhas Zhumabek
@simmplecoder
I have a question: is there any reason why add_deref functions need type specified in the type itself? I believe making the function a template would make it so much easier to use, because I found that I have to write auto deref_fn = [](...){...}; auto deref_view = ...::add_deref<decltype(deref_fn)>(view, deref_fn); instead of just writing the lambda inline
it might be a breaking change though, and I believe it will break ABI as template parameter will get mangled into the name
I believe one solution would be to just write a global function that would just place the right types. May be there is already something similar?
Mateusz Łoskot
@mloskot

@simmplecoder

is there any reason why add_deref functions need type specified in the type itself?

I have never analysed / considered that myself. Possibly C++98 kind of legacy.
I'm not aware of any 'factory' function deducing the types.

IMHO, breaking API/ABI is fine, especially if it will help to modernise GIL as well as GIL users codebase.
If it is not possible to extend/overload existing functions, then breaking is the only way.

What I'd suggest is: get your idea implemented, submit PR for review, and let's see.

Mateusz Łoskot
@mloskot
@miralshah365 & @simmplecoder :point_right: https://lists.boost.org/boost-gil/2019/10/0314.php
Olzhas Zhumabek
@simmplecoder
roger that
Mateusz Łoskot
@mloskot
It's certainly not a request, rather FYI :)
Pranam Lashkari
@lpranam
are we going to release any GSoC work with this release?
Mateusz Łoskot
@mloskot
Yes, as stated in the thread
Pranam Lashkari
@lpranam
I have a few comments I'll share in the mail
BTW have we decided till wich commit we are going to release?
Mateusz Łoskot
@mloskot
I think @stefanseefeld will do the grand merge of everything in develop
@lpranam If you want to share any feedback, objection, suggestion, please do it ASAP following up to the boost-gil thread
Pranam Lashkari
@lpranam
@mloskot just sent the mail
Pranam Lashkari
@lpranam
it awaits approval
Mateusz Łoskot
@mloskot
Looks like you are not subscribed. I will accept your post this time
Pranam Lashkari
@lpranam
This is my first email to GIL-list that's why
Mateusz Łoskot
@mloskot
Indeed. I did not realise the moderation flag is on by default for new posters
Mateusz Łoskot
@mloskot

@stefanseefeld Since time will run short soon, I have to nudge you to ensure you got this on time. Please, have a look at

Thanks @lpranam for heads up!

Stefan Seefeld
@stefanseefeld
@mloskot Thanks. https://lists.boost.org/boost-gil/2019/10/0316.php is referring to a commit to develop, no ? Is there anything left to do ? (other than merge develop to master, I mean ?)
I can certainly do the merge once you confirm that develop is complete, i.e. there are no other expected commits targeted for 1.72.
Mateusz Łoskot
@mloskot
Yes, I made those adjustments in develop only.
Outstanding is this one PR by @simmplecoder waiting for your review https://github.com/boostorg/gil/milestones/Boost%201.72
Once that is merged to develop
Or postponed after 1.72 (milestone removed)
Then develop is ready for merge into master, I think
Mateusz Łoskot
@mloskot
@stefanseefeld :point_up:
Pranam Lashkari
@lpranam
Wait I'll send a PR to close #391 in a few hours
Almost ready just a minor bug needa to be resolved
Pranam Lashkari
@lpranam
@mloskot just created a new PR for the same
Mateusz Łoskot
@mloskot
Approved. Thanks for speedy PR!
Pranam Lashkari
@lpranam
:)
Olzhas Zhumabek
@simmplecoder
should I hide my code under detail too, since it depends on the convolve_2d?
Olzhas Zhumabek
@simmplecoder
it is weird that in my PR the public interface returns class from detail
Pranam Lashkari
@lpranam
@simmplecoder your argument makes sense to me..
Mateusz Łoskot
@mloskot
@simmplecoder It all depends on what "since it depends on class from detail" mean, I think
Since it is about returning, then you and @lpranam rightly observe it smells
But, it also can be fine and indicate user that type is internal and avoid using the type in own interfaces
While using instances of the type in user's code is perfectly fine