Garbage collection

[Toralion]

Hi there,

first... thx for that great peace of code! Well done!

But... ;-)

Have you tried to load more than one soundfile at once? Say:
context.loadSound("depeche_mode_-a_photograph_of_you1.mp3", "One");
context.loadSound("depeche_mode-a_question_of_lust1.mp3", "two");
context.loadSound("depeche_mode-a_question_of_time1.mp3", "three");
context.loadSound("depeche_mode-agent_orange1.mp3", "four");
context.loadSound("depeche_mode-and_then...1.mp3", "five");
context.loadSound("depeche_mode-black_celebration1.mp3", "six");
context.loadSound("depeche_mode-blasphemous_rumours-12_b1.mp3", "seven");
context.loadSound("depeche_mode_-breathe1.mp3", "eight");
context.loadSound("depeche_mode-but_not_tonight-extended1.mp3", "nine");
context.loadSound("depeche_mode_-_dream_on1.mp3", "ten");

That are about 60MB of mp3 files. Seems not so much, but look into the Task-Manager what happens with your Memory used by the Chrome-Task ... long story short... Chrome crashes around the seventh or eighth sound.

Don't know if it's a problem of that js code, or if it's a memory leak in the browser.

Btw. I'm using Google Chrome 23.0.1271.91 m on WIndows 7(64bit).
Hardware: Intel i7 -2600k @3.4GHz, 16 GB RAM

So long,
Toralion

[TheNotary]

thanks for the report, i'll look into this now.

[TheNotary]

Yeah, I'm loading up a muse album (122mb) and chrome's process bloats up to 1.2gb. I didn't see any crashing on my two systems, but something's definitely not right. I'm going to try manually loading them with just the API before handing this off to a chrome engineer, you probably found a spot where chrome can be improved, congrats! =D

[TheNotary]

Yeah, I get the same memory behavior when i load things up just using the bare API. I can't reproduce any crashes, so if you're interested you can report this issue to the chromium community and they will likely be able to straighten things out.
http://chromiumbugs.appspot.com/

Also, the memory consumption is probably normal (though I only eyeballed it) because the function decodeAudioData is essentially converting the MP3 files into WAV files. This WAV/PCM data has to sit in memory until you delete the audio.

It might therefore be a good idea to separate the downloading process from the decoding process so that MP3 data can be stored in the browser, and then when it comes time to play the sound, decoding would take place. You'd have to work outside this framework to do that though. Or if you fork this code and come up with a smooth way to handle this, I'd accept a pull request.

[Toralion]

Hello I'm back again.
Sorry for the delay! To be honest... I forgot about my post in here.. because we avoided our problem by not using the webaudioapi at last.
Instead we used the old -Dom Element for playing mp3-files.

I then stumbled upon your project while trying to find a solution for playing mp3-files on the iPad without delay and exact timing.
It worked great using the webaudioapi with small sound pieces but crashes soon as I used some bigger files as described above.
So I searched around the net for people with similar experiences and hit your project. :)

In the end I think that the whole problem is as you mentioned in your last comment a problem of the chromium.

I'm gone open a ticket at chromium.

So long,
Toralion

[TheNotary]

Hey again, welcome back =)

First of all, I really appreciate your post, you bring in more data that allows me to think of the project in new ways that I wouldn't have originally and that's really cool.

About your needs for playing long audio though, I think with a little tweaking, you might be able to get something to your liking. On iPad, you're probably under a memory constraint right?

I imagined it might be cool to implement a new feature into this project where there is a special "channel" for it to play "background/ one-at-a-time" style audio files. So for instance, the code to use this feature of the API might be something like this (be careful though, I haven't thought of this library in a while since I ran out of free time to work on the underlining project it was being developed for):

var context = initializeNewWebAudioContext();

var soundBuffer1 context.precacheSound('urlToSound1.mp3', 'mySoundName1')  // This method DL's compressed
                                                                                                                    // audio (.mp3) but doesn't
                                                                                                                    // decompress it so it doesn't
                                                                                                                    // fill up the system memory
                                                                                                                    // like loadSound() does...
                                                                                                                    // It might be a very tiny bit laggier
                                                                                                                    // but maybe not; haven't tested

var soundBuffer2 context.precacheSound('urlToSound2.mp3', 'mySoundName2')

context.commenceBgSound(soundBuffer1)   // The background sound would be converted into PCM (which 
                                                                 //  takes up lots of memory)
                                                                 //  and then is played at the same time!  

context.commenceBgSound(soundBuffer2)   //  Now when this happens, the soundBuffer1 PCM data is
                                                                 //   explicitly deleted so the system's memory doesn't get
                                                                 //   all gummed up with sounds you don't even want playing.
                                                                 //   anymore.    

I'm focusing on Ruby and Rails development right now, and I don't find working with js to be very fulfilling, but if the project you're working on is open source then by all means share the link and maybe we can implement some missing functionality within this project together. It would be nice for the internet to have an all-in-one sound kind of library to work with, although I find it annoying and overly time consuming writing all the feature/browser checks and fallback routines (the main reason I don't enjoy .js these days).

In a couple weeks I'll have some spare time but right now I'm a bit 'todo' saturated. If you're down to help write some of the code than I'll look through the project and point you at where to start and provide you with some stubs for you to flesh out while I'm indisposed.