Episode 007 / Nathan Letwory, 2.5 Development Retrospective

This episode we look back on 2.5x development where we discuss with Nathan Letwory how it went, the people invoked, the problems that had to be resolved and some of the remaining issues nobody got around to fixing yet :S.

*** Disclaimer ***
This is a really geeky show and while we try to explain the topics for non techies, listeners who are not so interested in a few nerds reminiscing about the past 4 years might want to skip this episode :)

Rest assured we will keep up the interviews and less esoteric topics.

– Campbell

Length: 01:40:31
Recorded: 26. August 2011

Play

Download: OGG MP3

Links:
Winter Camp
Blender 2.5 Early Days

21 thoughts on “Episode 007 / Nathan Letwory, 2.5 Development Retrospective

    • The link is dead, but what is Hyper-mode? Have you just speeded up the audio file and lowered the pitch? Or have you in some way managed to delete dead air from the file?

      • I’m glad you asked, my friend! :D
        It is a relatively simple process that you yourself can do at home on Audacity:

        1. Truncate silence 4:1 at -25 dB threshold with a minimum silence of 100 ms and max 300 ms.
        2. Change tempo +10%.

        And you’re done!

  1. Hey! Glad to see you running advertisements! Guess you’ve seen the Blender Artists thread. But you might want to put up a warning message about repetitively clicking the ads, as people will probably click the ads just to help out the BF, and that could lead to trouble with Google.

  2. Great episode! It is very interesting to hear the views of the developers on areas of Blender, and also the difficulties that arise in such a big project in day to day coding work. You all are very much appreciated for doing this, thank you!

    So Campbell, Thomas and Nathan, thanks again for another informative episode. One slight suggestion though: I find I can’t understand Thomas very well at the beginning, the music is a tad bit too loud.

  3. Awesome show, I got into blender just before 2.5 came out. I started to learn on 2.49 but when 2.5 came out with its new UI … I was SO happy that I dont have to search for obscure tutorials to learn how to do things. The UI is very intuitive and easy to navigate through.
    Just wanted to say thanks to campbell, thomas and nathan on your guys work on blender, and also on the show.

  4. Really enjoy the development talk.
    Would be great to see a picture of that 4-monitor setup!

    It could be good (for a future podcast episode maybe?) to talk more on the topic of what things in blender eat up a lot of memory. I have found that no matter how much RAM an artist has, it will not be enough if he does not understand how to use it sensibly. I have 24gb RAM on one machine, and I can still max it out just as easily as my old 512mb.

    Also, perhaps it could be good to discuss the difference between what things max out the RAM (subsurf level 6, large image textures, etc), and what things simply take forever to render (SSS, Global Illumination, Full Oversampling?).

    I find the more that I, as an artist, can come to understand how Blender works, the more efficiently I can make it use my computer’s resources.

    • Strange. I Recently upgraded from 2 GB to 8GB, and I’m having trouble utilizing it. My computer simply refuses to use more than 1.2 GB for regular tasks, and Blender typically uses only about 800 megs. Even after cranking the cache limit to a ridiculously high value- 6 GB, it still doesn’t use much.

      • To use more than 4gb of memory per process you need a 64bit OS, even so, windows XP has some limit of… ~1.8gig per process. 32bit linux has a PAE extension which is mostly enabled by distro makers and lets 32bit systems make use of more then 4gig of ram but the per-process limit remains.

        Also note that memory can have fragmentation issues – so even if you are supposed to have free ram, an allocation may fail if it tries to get a large contiguous chunk of memory.

        See:
        http://blog.mikepan.com/blender-memory-issues-and-64bit/

  5. Very nice podcast and interesting topics – I enjoyed listening to all the episodes :-)

    Two bits that stuck out to me that you might want to look into:
    * Echo – quite a lot of it in some cases. Not sure how easy it is for you to avoid this?
    * Talking over the intro music. That was a single case I think. I don’t remember in what episode it happened, might’ve been one of the earlier ones. In that case the music was pretty loud and drowned out the voice.

    Nothing major, just wanted to point these out as friendly criticism. :-)

    Cheers and thanks for the poscast,
    Martin

  6. I was glad to hear Campbell acknowledge that we’ve lost some functionality with the (otherwise vastly superior) new Python API. There’s a few operations that aren’t just harder in 2.5, they’re completely impossible. Even if we never get back any of that stuff, it does my heart good just to know that some of the developers aren’t completely satisfied with it.

  7. @mbue, I didn’t notice the echo, The problem is only *one* of us has to have our headphones turned up too loud, or some bad setup and we end up with some echo, Thomas and myself can sort this out since we do this often enough but for guests we cant expect too much in the way of a great recording setup – we could try do basic sound tests beforehand but think it gets too much of a hassle if we get really picky with these things and I don’t want it to be off-putting to people have on the show.

    Point taken for intro vs Thoma’s volume levels, will take more care next time.

    @Lee, The 2.4x Python API had so many functions added here and there for special cases and just because it seemed like a good idea at the time, With 2.5x in general we’re aware there are many areas missing but when people don’t notice them – we tend to assume they were not used much anyway. – Example EKey in file selected in 2.4x opened the file with an editor! I found this while reading the source code but nobody missed this AFAIK :-)

    I’d be interested to know what areas of the api are missing for you, even if we cant add right away. they could be added to the TODO’s or perhaps theres some other way to do the same thing.

  8. Wow, I didn’t mean to be so catty. I really shouldn’t listen to the podcast while frustrated. Sorry about that.

    But since you asked, I ran into two missing functions while trying to get fur to an external renderer. The big one is that there doesn’t seem to be a way to export child particles anymore, short of converting the particle system to a temporary mesh. In 2.4x, Python could read the path cache (including child particles) but not the hair keys; now it’s the opposite. It would be great if we could have both.

    The other missing bit is the ability to evaluate a texture. That function did seem like a hack, though, and it never really did what I wanted. I can’t say I’m sad to see it go.

  9. Hi Lee, thanks for spotting texture.evaluate(), I used this in 2.4x on drivers so we could use textures for animation (fun!) and it never gave me any issues.

    Added back into blender, r39963, will be in 2.60 release.
    http://markmail.org/message/4noodprqilhvoglm

    Child particles issue is more complicated, at least I’d have to investigate further before saying anything useful about how hard it is or how long it would take.

  10. It would be really nice if there would be an option to subscribe to this podcast by email.
    Then I would not have to keep coming here every day just to see if there is a new podcast.

    :)