1

I am working on evaluating a embedded processor architecture which offers below features:

  • 8 SIMD co-processing DSP kind of cores,
  • Each core can do a 8 way SIMD
  • Each core is a 8 execution slot VLIW as well.

I want to use high video encoder(H.264, 1080p, 60fps) or 3D Video encoder to run on this processor/hardware. I am trying to perform architectural exploration and find

  • What are the good featrues of a processor should have which help in carrying out multimedia(Video/Image) Signal processing applications in power/cycle/memory efficient way.

  • What peripherals,memory structure, either cache memory or internal memory;additional assembly instructions help in efficient execution of code for multimedia applications

  • What are most power efficient and fast processor architectures for Multimedia(Video/Image) processing applications

PS: It has to be low power as it is for portable applications.

Any pointers(papers/blogs) would be helpful.

thank you.

-AD.

osgx
  • 90,338
  • 53
  • 357
  • 513
goldenmean
  • 18,376
  • 54
  • 154
  • 211
  • what a cores do you talk about? It is a processor with 8-way SIMD or it is a processor with a dsp coprocessor? – osgx Nov 12 '10 at 15:35

1 Answers1

0

I think that "most power efficient and fast processor architectures for Multimedia(Video/Image) processing" is a special hardware cores to do a specific video/image encoding operation. E.g. the fastest mpeg4 avc encoder will be a hardware encoder, isnt it?

For example, there is scheme of VP9 hardware decoder: http://www.webmproject.org/hardware/vp9/ Scheme of Google's VP9 hardware decoder

It even has no any big CPU/MCU. The encoder is more complex and less public thing... But there is some project: http://hardh264.sourceforge.net/H264-encoder-manual.html with scheme:

Zexia video H.264 Hardware Encoder scheme

or VP8 encoder from google: http://www.webmproject.org/hardware/vp8/

WebM Video Encoder Hardware IP from google - scheme

Again, no big CPU. And you can check what stages was separated by designer, and think about how you will implement them in CPU: DCT, quantisation, prediction, entrophy coding, YUV/RBG conversion, etc

osgx
  • 90,338
  • 53
  • 357
  • 513