Context model cache-management in a dual-pipeline CABAC architecture

Patent metadata

Field	Value
Publication number	US-9258565-B1
Application number	US-201113172775-A
Country	US
Kind code	B1
Filing date	Jun 29, 2011
Priority date	Jun 29, 2011
Publication date	Feb 9, 2016
Grant date	Feb 9, 2016

Abstract

A method and system are disclosed for managing cache memory in a dual-pipelined CABAC encoder. A request for a context model is received from both encoder pipelines. If the requested context model is not stored in cache, the requested context model is retrieved from a context table. At least one context model stored in cache is written to the context table. The retrieved context model is updated and written to the cache. If the requested context model is stored in cache, and if the requested context model was updated in the previous clock cycle, the requested context model is retrieved from the pipeline, updated, and written to cache. If the requested context model is not stored in cache, and if the requested contest model was not updated in the previous clock cycle, the requested context model retrieved from cache, updated, and written back to cache.

First claim

What is claimed is:

1. A method for managing cache memory in context-adaptive binary arithmetic coding (CABAC) implemented in a synchronous dual-pipelined environment, the method comprising:
receiving, at a CABAC encoder cache memory in a first clock cycle, a first request for a first context model from a first context lookup module communicatively coupled to a first encoder and a second request for a second context model from a second context lookup module communicatively coupled to a second encoder;
determining if the first context model is stored in cache memory, wherein the cache memory is configured to store at least one previously updated context model;
responsive to a determination that a cached version of the first context model is stored in the cache memory, determining whether an updated version of the first context model is available at a respective output of either the first encoder or the second encoder;
responsive to a determination that the updated version of the first context model is available at the respective output, retrieving the updated version of the first context model from the respective output;
providing the updated version of the first context model to the first encoder, wherein the first encoder is configured to
utilize the updated version of the first context model to encode a first bin, and
update the updated version of the first context model to produce a subsequent version of the first context model;

receiving, at the cache memory in a second clock cycle, the subsequent version of the first context model from the output of the first encoder; and
writing the subsequent version of the first context model in the cache memory in response to the receiving the subsequent version at the cache memory.

2. The method of claim 1 , further comprising:
writing the subsequent version of the first context model in a context table, wherein
the cache memory is communicatively coupled to the context table,
the context table stores an older version of the first context model, and
the subsequent version of the first context model overwrites the older version of the first context model in the context table.

3. The method of claim 2 , further comprising:
determining which of the at least one previously updated context model stored in the cache memory was least recently updated; and
writing the determined previously updated context model to the context table.

4. The method of claim 2 , further comprising:
determining which of the at least one previously updated context model stored in the cache memory was most recently requested; and
writing the determined previously updated context model to the context table.

5. The method of claim 2 , further comprising:
determining which of the at least one previously updated context model stored in the cache memory is least likely to be requested in the future; and
writing the determined previously updated context model to the context table.

6. The method of claim 1 , further comprising:
writing the updated version of the first context model in the cache memory, wherein
the cached version of the first context model is stored at a particular location in the cache memory, and
the updated version of the first context model is written to the particular location in the cache memory.

7. The method of claim 1 , further comprising:
responsive to a determination that the first context model is not stored in cache memory,
retrieving a stored version of the first context model from a context table, wherein
the context table is configured to store a plurality of context models utilized in the synchronous dual-pipelined environment; and

providing the stored version of the first context model to the first encoder, wherein the first encoder is configured to
utilize the stored version of the first context model to encode the first bin, and
update the stored version of the first context model to produce the subsequent version of the first context model.

8. The method of claim 1 , further comprising:
responsive to a determination that the updated version of the first context model is not available at the respective output, retrieving the cached version of the first context model from the cache memory; and
providing the cached version of the first context model to the first encoder, wherein the first encoder is configured to
utilize the cached version of the first context model to encode the first bin, and
update the cached version of the first context model to produce the subsequent version of the first context model.

9. The method of claim 1 , wherein the method is implemented in an H.264 video encoder.

10. The method of claim 1 , wherein the requested first context model and the requested second context model comprise the same context model.

11. The method of claim 1 , wherein the first context lookup module is configured to:
receive the first bin;
issue the first request for the first context module based on the first bin; and
output the first bin to the first encoder; and

wherein the second context lookup module is configured to:
receive a second bin;
issue the second request for the second context module based on the second bin; and
output the second bin to the second encoder.

12. The method of claim 1 , the method further comprising:
determining if the second context model is stored in the cache memory;
responsive to a determination that a cached version of the second context model is stored in the cache memory, determining whether an updated version of the second context model is available at the respective output of either the first encoder or the second encoder;
responsive to a determination that the updated version of the second context model is available at the respective output, retrieving the updated version of the second context model from the respective output;
providing the updated version of the second context model to the second encoder, wherein the second encoder is configured to
utilize the updated version of the second context model to encode a second bin, and
update the updated version of the second context model to produce a subsequent version of the second context model;

receiving, at the cache memory in the second clock cycle, the subsequent version of the second context model from the output of the second encoder; and
writing the subsequent version of the second context model in the cache memory in response to the receiving the subsequent version of the second context model at the cache memory.

13. A system for managing cache memory in context-adaptive binary arithmetic coding (CABAC) implemented in a synchronous dual-pipelined environment, the system comprising:
a cache memory manager configured to:
receive, in a first clock cycle, a first request for a first context model from a first context lookup module communicatively coupled to a first encoder and a second request for a second context model from a second context lookup module communicatively coupled to a second encoder;
determine if the first context model is stored in cache memory, wherein the cache memory is configured to store at least one previously updated context model;
responsive to a determination that a cached version of the first context model is stored in the cache memory, determine whether an updated version of the context model is available at a respective output of either the first encoder or the second encoder;
responsive to a determination that the updated version of the first context model is available at the respective output, retrieve the updated version of the first context model from the respective output;
provide the updated versi

Assignees

• Jacob Rojit
• Freescale Semiconductor Inc

Inventors

• Jacob Rojit

Classifications

• H04N19/91 Primary

  Entropy coding, e.g. variable length coding [VLC] or arithmetic coding · CPC title

• H04N19/102

  characterised by the element, parameter or selection affected or controlled by the adaptive coding · CPC title

• H04N19/13 Primary

  Adaptive entropy coding, e.g. adaptive variable length coding [AVLC] or context adaptive binary arithmetic coding [CABAC] · CPC title

• H04N19/423

  characterised by memory arrangements (H04N19/433 takes precedence) · CPC title