Re: Web Audio API ambient drone

Actually, no, I didn't mean that Math.random was a bottleneck; just that
this approach was probably more costly than it needs to be.  Having to
generate samples from Math.random() and write them into the output buffer
in a JSnode (plus the cost of calling the JSnode) would be, I expect, more
costly than just looping a precalculated buffer.  But it was just a side
comment.

On Mon, Jul 16, 2012 at 12:10 AM, Marcus Geelnard <mage@opera.com> wrote:

> Den 2012-07-15 17:03:08 skrev Chris Wilson <cwilso@google.com>:
>
>
>  On Sat, Jul 14, 2012 at 12:26 PM, Matt Diamond <mdiamond@jhu.edu> wrote:
>>
>>  Hey all,
>>>
>>> Put together a little experiment in synthesizing an ambient sound texture
>>> using the Web Audio API:
>>>
>>> http://matt-diamond.com/drone.**html<http://matt-diamond.com/drone.html>
>>>
>>> Matt
>>>
>>
> Matt, nice demo!
>
>
>  Hey Matt-
>>
>> Neat!  I like the apparent pitch generated by bandpass filter.
>>
>> One comment - instead of continually generating noise via
>> JavaScriptAudioNodes, you'd probably find it much more performant to
>> pre-generate a couple of seconds (to avoid any detectable periodicity) of
>> noise in a Buffer and playing it as an AudioNode.  That would probably be
>> a
>> bit more performant than continually calling Math.random for each sample.
>>
>> var lengthInSamples =  2 * audioContext.sampleRate;
>> var noiseBuffer = audioContext.createBuffer(1, lengthInSamples,
>> audioContext.sampleRate);
>> var bufferData = noiseBuffer.getChannelData(0);
>> for (var i = 0; i < lengthInSamples; ++i) {
>>     bufferData[i] = (2*Math.random() - 1);
>> }
>>
>
> Surely, Math.random() can't be the bottleneck here. On my computer
> Math.random() can generate 48000 samples in less than one millisecond. In
> other words, it should consume less than 4% CPU if 40 such sources are used
> - yet Chrome consumes 45% CPU when running the demo.
>
> /Marcus
>
>
> --
> Marcus Geelnard
> Core Graphics Developer
> Opera Software ASA
>