Just an update to what was going on, this might help others, because what I was
doing in the first place seemed a sleek solution, that alas proved to be
counterproductive.
I actually built a structure containing an image destination coordinates, its
size, and the image itself with a javascript patch.
It went like this, without surprise:
function (__structure controlStruct) main (__number Destination_X,
__number Destination_Y,
__number Size,
__image
Image)
{
var result = new Object();
__structure params = new Object();
if(!_testMode) {
params["Destination_X"]=Destination_X;
params["Destination_Y"]=Destination_Y;
params["Size"]= Size;
params["Image"]= Image;
}
result.controlStruct = params;
return result;
}
This way of building a compact structure works, but changing the inputs every 4
seconds lead to a crash after several hours, because the textures were never
flushed and kept piling on in the VRAM. I finally ended pushing and sampling
the image separately, in parallel, and things went well. This is still is a
valuable solution for an init sequence, or for a short presentation, but the
VRAM gets dirty, and it is in no way the best practice for ambient media design.
So that was to say thanks to the list and Christopher who helped me.
dimitri
On 1 mars 2012, at 22:35, Dimitri Delcourt wrote:
> Using the OpenGL Driver Monitor, I can see that the textures graph is
> inexorably growing.
> Investigating image structures now.
>
> Thank you for heading me in the right direction!
>
> dimitri
>
>
>
>
> On 29 févr. 2012, at 17:51, Christopher Wright wrote:
>
>>> 21 com.apple.QuartzComposer 0x00007fff871ebd90
>>> -[QCProvider_CoreImage
>>> createTextureBufferForManager:withFormat:target:transformation:bounds:colorSpace:options:]
>>> + 1368
>>
>> ... who in turn was called by QC to create a texture from a CIImage. This
>> is a normal operation, but it's possible that textures are leaking. Use
>> OpenGL Driver Monitor to look at the number of textures, and see if it grows
>> without bound.
>
>
>
>
> On 1 mars 2012, at 18:21, Christopher Wright wrote:
>
>> Hi Dimitri,
>>
>>> <Quartz Composer_2012-02-27-110318_dimitri-i7.crash>
>>
>>
>> That crash is in the exact same place. So it's probably not a system memory
>> leak. It still looks like a GPU resource leak though.
>> Passing images through JS is relatively dangerous - that might be leaking
>> vram/other GPU resources.
>>
>> --
>> Christopher Wright
>> christopher_wri...@apple.com
_______________________________________________
Do not post admin requests to the list. They will be ignored.
Quartzcomposer-dev mailing list (Quartzcomposer-dev@lists.apple.com)
Help/Unsubscribe/Update your Subscription:
https://lists.apple.com/mailman/options/quartzcomposer-dev/archive%40mail-archive.com
This email sent to arch...@mail-archive.com
