ThinkMeta LogoDeutsch
English
LAME: Quantization

Quantization is the most extensive und thus the most CPU-bound stage of the encoding process. Here the compression is made.

The possibilities to use asynchronous programming are very limited in this stage. In the case study this was done only in two places in VBR encoding:

  • New VBR: The encoding function for one frame could be executed in parallel for each channel and each granule.
  • Old VBR: The encoding function for one frame could be executed in parallel for each channel.

Furthermore, SSE instructions were used where possible.

All classes are located in the "Quantization" folder of the project workspace.