Support OpenCL / task parallelism / computations on many-core architectures / scalable memory allocation

[kowpa1990]

When I use QTGMC on Avisynth on my dual-core Pentium performance is low; when I use MT version, efficiency increases only about 5fps but still my Radeon HD7770 remains unused. To encourage people to move to Vapoursynth that they saw apparent advantages it is necessary to introduce innovative solutions.

It may be helpful:

STL-like libraries:
AMD's Bolt / NVIDIA's Thrust / Boost.Compute

expression or math-libaries:
VexCL / ViennaCL etc.

Discussion about the differences between libraries.

One more comparison.

[myrsloik]

Not a specific enough issue. There already is support for multiple cpus and gpu support is very far off into the future.