more slabby questions- the td.rota.jxs slab


    Jun 23 2006 | 9:41 pm
    first off, before i start whining- i want to say that the feedback transformations i usually work with in performance with the mop operators consistently result in like 23fps at 320x240. and now that i'm finally moving things to the slab world, a first effort resulted in 15fps at 1280x854. staggering. thank you for the shader helper files- they go a long way towards breaking the mental block i've had with shaders.
    now, to whining:
    i use jit.rota @interp 1 a lot. i've gotten very good at knowing exactly what settings to feed jit.brcosa to make up for the corresponding decrease in matrix values that comes with interpolation in a feedback system. but now that i'm moving this chain
    source-->jit.xfade-->jit.brcosa-->jit.rota-->jit.xfade 2nd inlet
    to this one:
    640x480 source-->jit.gl.slab bork @file co.average.jxs-->jit.gl.slab bork @file cc.brightness.ip.jxs-->jit.gl.slab bork @file cc.contrast.ip.jxs-->jit.gl.slab bork @file cc.saturation.ip.jxs-- >jit.gl.slab bork @file td.rota.jxs-->averaging slab (in render context bork)
    of course, everything has changed. it's MEGA FAST, but it's changed.
    is there a shader equivalent to the @interp attribute on jit.rota? and also: how can i alter the td.rota.jxs file to include the weird trig stuff- the cosoffsets and sinscales and thetasinscales of jit.rota? i'm not going to go so far as to ask that i be given a new td.rota.jxs, but can you at least point me in the right direction on how to do it myself?
    many thanks again for the upgrade. this is great.
    j

    • Jun 23 2006 | 10:11 pm
      Hi Joshua, Thanks for looking at this. First off, textures will automatically use linear interpolation unless specified otherwise using the @filter attribute to jit.gl.texture.
      The jit.rota helpfile gives the equation used to calculate locations, so you could try making a shader based on that if you want all the trig params. I opted for a linear algebra approach in this shader.
      Developing this shader was a real headache on a Mobility card. We ended up moving a lot of the code over to the vertex program because we were hitting the instruction limit on AluBooks. For some reason, the calculations necessary for "boundmode" make it difficult to add much other stuff. If you were to remove the boundmode code, you would have a lot more headroom for other parameters and calculation.
      Arguably, it's probably way easier to simply rotate a videoplane in GL space.
      It's cool that you're getting into slab stuff.
      Enjoy! Andrew B.
    • Jun 25 2006 | 5:09 pm
      the boundmodes are important to me, so i'm not going to mess with those.
      another question now- i see from the archives that there is a performance benefit to combining slabs.
      if i'm trying to use shaders to imitate jit.brcosa, in what order should i do brightness/contrast/saturation?
    • Jun 25 2006 | 5:16 pm
      Try saturation - contrast - brightness.
      jb
      Am 25.06.2006 um 19:09 schrieb joshua goldberg:
      > if i'm trying to use shaders to imitate jit.brcosa, in what order > should i do brightness/contrast/saturation?
    • Jun 25 2006 | 8:12 pm
      Hi,
      I just noticed this thread that deals with td.rota.jxs. This shader isn't part of the bundle I'm using (11/NOV/05). I couldn't see a new shader bundle distributed on the cycling74 website. Is the td.rota.jxs a free-floating shader that is privately distributed or part of a distribution that I somehow missed?
      thx, E.
    • Jun 26 2006 | 10:53 pm
      This is one of many shaders that were added to the public UB beta of Jitter 1.6 for intel mac. This will be included in the next release version of Jitter for all platforms.
      Cheers, Andrew B.
    • Jun 26 2006 | 11:26 pm
      >This is one of many shaders that were added to the public UB beta of Jitter >1.6 for intel mac. This will be included in the next release version of >Jitter for all platforms.
      thx
      E.