Re: [osg-users] [Compute Program executed on useprogram :/?!]

2017-11-27 Thread Julien Valentin
Hi Robert That's sound more reasonnable than what i was affraid to interpret... I'll check on my own, because I'm sure I've something that might help us here...(A thing we discuss long time ago about making VAS a property) I'll make a PR to know your feeling about it Cheers robertosfield

Re: [osg-users] [Compute Program executed on useprogram :/?!]

2017-11-27 Thread Robert Osfield
HI Julien, On 25 November 2017 at 19:02, Julien Valentin wrote: > If I don't misunderstood: > Using a dedicated class for CS would introducing a new base class > "Dispatchable" for Drawable(vas) and DispatchCompute(no vas) that would > replace Drawable as

Re: [osg-users] [Compute Program executed on useprogram :/?!]

2017-11-25 Thread Julien Valentin
Hi Robert If I don't misunderstood: Using a dedicated class for CS would introducing a new base class "Dispatchable" for Drawable(vas) and DispatchCompute(no vas) that would replace Drawable as Renderleaf. Is that what you think about? Cheers robertosfield wrote: > Hi Julien, > > The

Re: [osg-users] [Compute Program executed on useprogram :/?!]

2017-11-18 Thread Robert Osfield
Hi Julien, The Compute shader work is a submission I merged from the community, I clearly didn't think deeply enough about the submission when I merged it though, as in hindisight having the glDispatchCompute directly coupled with the useProgram is wrong - on several different counts. The

Re: [osg-users] [Compute Program executed on useprogram :/?!]

2017-11-17 Thread Julien Valentin
...Perhaps a flexier solution would be to let dispatchcompute feature be a drawable::drawcallback ... mp3butcher wrote: > Hi all > I ran into some strange behavior debuging an app: It appears that compute > shader are executed at each pcp.useprogram() call... > I don't think it's a correct

[osg-users] [Compute Program executed on useprogram :/?!]

2017-11-17 Thread Julien Valentin
Hi all I ran into some strange behavior debuging an app: It appears that compute shader are executed at each pcp.useprogram() call... I don't think it's a correct behavior since uniforms requires the po to be bound before to be set... It yield an undefined behavior in GLObjectVisitor when