Digital signal processor having inverse discrete cosine transform engine for video decoding and partitioned distributed arithmetic multiply/accumulate unit therefor

• US 7,574,468 B1
• Filed: 03/18/2005
• Issued: 08/11/2009
• Est. Priority Date: 03/18/2005
• Status: Active Grant

• Current Assignee: VeriSilicon Holdings Co., Ltd. (Verisilicon Microelectronics (Shanghai) Co. Ltd.)
• Original Assignee: VeriSilicon Holdings Co., Ltd. (Verisilicon Microelectronics (Shanghai) Co. Ltd.)
• Inventors: Rayala, Jitendra
• Primary Examiner(s): Malzahn; David H
• Application Number:
  US11/083,575
• Time in Patent Office: 1,607 Days
• Field of Search: 708/402
• US Class Current: 708/402
• CPC Class Codes: G06F 17/147 Discrete orthonormal transforms, e.g. discrete cosine transform, discrete sine transform, and variations therefrom, e.g. modified discrete cosine transform, integer transforms approximating the discrete cosine transform G06F17/145 takes precedence

Abstract

A distributed arithmetic multiply/accumulate (MAC) unit for computing inverse discrete cosine transformations (IDCTs). In one embodiment, the distributed arithmetic MAC unit includes: (1) a first pipeline stage configured to perform dot products on received sequential input data and (2) a second pipeline stage coupled to the first pipeline stage and configured to compute additions and subtractions of the dot products to yield sequential output data.

Claims

1. A distributed arithmetic multiply/accumulate (MAC) unit for computing inverse discrete cosine transformations, comprising:

a first pipeline stage of said distributed arithmetic MAC configured to perform dot products on received sequential input data, wherein row transformations are performed initially and column transformations are performed subsequent to said row transformations; and

a second pipeline stage of said distributed arithmetic MAC coupled to said first pipeline stage and configured to compute additions and subtractions of said dot products to yield sequential output data,

wherein said distributed arithmetic MAC unit has a 12-bit precision.

2. The distributed arithmetic MAC unit as recited in claim 1 wherein said first pipeline stage comprises at least one look-up table and is configured to perform look-up table operations and a shifting and accumulation to perform said dot products.

3. The distributed arithmetic MAC unit as recited in claim 1 wherein said dot products are vectors of length four.

4. The distributed arithmetic MAC unit as recited in claim 1 wherein appropriate bits of said additions and subtractions are extracted to yield said output data.

5. The distributed arithmetic MAC unit as recited in claim 1 wherein said first pipeline stage is configured to compute eight 1D IDCTs in a pipelined fashion.

6. The distributed arithmetic MAC unit as recited in claim 1 wherein said output data is rounded.

7. The distributed arithmetic MAC unit as recited in claim 1 wherein said first pipeline stage has sub-units configured to operate on four bits of said input data.

8. The distributed arithmetic MAC unit as recited in claim 1 wherein said input data corresponds to a selected one of MPEG-2 profiles from MP@LL to MP@HL and MPEG-4 profiles from SP@L3 to Main@L4.

9. A digital signal processor (DSP), comprising:

a core containing a pipeline control unit having a co-processor interface; and

a co-processor coupled to said core and containing an distributed arithmetic multiply/accumulate (MAC) unit for computing inverse discrete cosine transformations, including;

a first pipeline stage of said distributed arithmetic MAC configured to perform dot products based on sequential input data received from said co-processor interface, wherein row transformations are performed initially and column transformations are performed subsequent to said row transformations, and

a second pipeline stage of said distributed arithmetic MAC coupled to said first pipeline stage and configured to compute additions and subtractions of said dot products to yield sequential output data for transmission toward said co-processor interface,

wherein said distributed arithmetic MAC unit has a 12-bit precision.

10. The distributed arithmetic MAC unit as recited in claim 9 wherein said first pipeline stage comprises at least one look-up table and is configured to perform look-up table operations and a shifting and accumulation to perform said dot products.

11. The distributed arithmetic MAC unit as recited in claim 9 wherein said dot products are vectors of length four.

12. The distributed arithmetic MAC unit as recited in claim 9 wherein appropriate bits of said additions and subtractions are extracted to yield said output data.

13. The distributed arithmetic MAC unit as recited in claim 9 wherein said first pipeline stage is configured to compute eight 1D IDCTs in a pipelined fashion.

14. The distributed arithmetic MAC unit as recited in claim 9 wherein said output data is rounded.

15. The distributed arithmetic MAC unit as recited in claim 9 wherein said first pipeline stage has sub-units configured to operate on four bits of said input data.

16. The distributed arithmetic MAC unit as recited in claim 9 wherein said input data corresponds to a selected one of MPEG-2 profiles from MP@LL to MP@HL and MPEG-4 profiles from SP@L3 to Main@L4.