Development of VP9 encoder still focused on CPU

[GoogleCodeExporter]

My AMD 5500 CPU encodes into vp9 format very well, but on speed 1 fpm only. I 
think it will be much faster if encoder would use its integrated GPU. And very 
much faster with discrete GPU which I also have. The problem is: now, having 
amazing OpenCL technology, encoder still developing manually by hand for each 
kind of supported SIMD instrictions for each CPU architecture. Same time many 
of these architectures already supports OpenCL (by pocl, for example) and, of 
course OpenCL gives access to incredible power of GPU devices (which present 
even on smartphones). It would me more clever to spend time for optimizing 
existing OpenCL frameworks instead of rewriting low-level code each time when 
Intel releases new SIMD instructions.

Original issue reported on code.google.com by infer...@gmail.com on 22 Jun 2015 at 8:17

[GoogleCodeExporter]

There has been some progress on making use of GPU via opencl here: 
https://github.com/ittiamvpx/libvpx

Original comment by ya...@google.com on 16 Jul 2015 at 10:24

• Changed state: Started

[GoogleCodeExporter]

We are also planning to do more on vp10 development

Original comment by ya...@google.com on 16 Jul 2015 at 10:25

• Changed state: WontFix
• Added labels: Type-Enhancement