Long term direction

cwright's picture

So, smokris and I spent some time tonight trolling through the latest Snow Leopard seed, to see what we could find (we were trying to figure out how some wwdc demos were accomplished, after doing some low-level tests that indicated impossibility without some magic going on behind the scenes). What we found has shaken us quite a bit.

Basically, in QC-SnowLeopard, there's a supplement to Structures called "Streams" (they're used just like structures, and end-users won't know the difference). What they are is basically this:

  • C Arrays
  • GPU 'Textures' (blocks of vram for gpu processing via opencl)

So, In one fell swoop, QC Snow Leopard will destroy two technologies we've had slated for development: Spreads, and Kineme3D (longer term).

So, assuming you could move tons of data (dozens of megabytes) around very quickly (C array speed, or gpu vram speed) in QC, what would you do? What would be left to solve (and, by extension, what should kineme focus on to stay useful?)

Comment viewing options

Select your preferred way to display the comments and click "Save settings" to activate your changes.

tobyspark's picture
...just catching up with

...just catching up with kineme.

i'd think there's still a ton of stuff, and at root better enabled by built-in tech such as this.

the two obvious directions for "hard" stuff that needs to be solved, and solved well, for me would be - physics, flocking, etc on large object data sets. - computer-vision

but this is off the top of my head, having been deep in application development land rather than cunning qc land.

toby

gtoledo3's picture
Re: Long term direction

This is kind of interesting in retrospect!

usefuldesign.au's picture
Re: Long term direction

A hit, a palpable hit.