Method and apparatus for efficient addressing of DRAM in a video decompression processor

An apparatus for storing .and retrieving prediction area pixel data of successive video frames from a random access memory (RAM) to minimize the number of times that different rows of said RAM must be addressed, comprising:

• means for allocating the pixel data representing said video frames into a plurality of tiles;
• a memory manager for generating write addresses for writing said tiles into said RAM on a one tile per row basis, wherein each tile has a height of M pixels and a width of N pixels and substantially fills a different row of said RAM;
• a decoder, coupled to receive motion vectors associated with said video frames, for decoding motion vector data to determine said prediction areas, each having a height equivalent to a vertical space occupied by H pixels and a width equivalent to a horizontal space occupied by W pixels;
• ,
• a prediction address generator which is responsive to said decoder for generating read addresses for said video frames for retrieving the prediction area pixel data thereof from said RAM;
• timing means for providing a timing signal which defines consecutive access periods in which said prediction area pixel data may be retrieved from said RAM;
• wherein:
• (a) when a selected video frame prediction area encompasses only one of said rows, said prediction address generator generates read addresses to allow retrieval of the corresponding pixel data in a first of said access periods;
• said decoder decodes motion vector data to determine another prediction area;
• and said prediction address generator generates read addresses to allow retrieval of the corresponding pixel data from one row of said RAM in a second of said access periods which directly follows said first access period;
• and
• (b) when a selected video frame prediction area encompasses only two of said rows, said prediction address generator generates read addresses to allow retrieval of the corresponding pixel data of the first of said rows in a first of said access periods, and to retrieve the corresponding pixel data of the second of said rows in a second of said access periods which directly follows said first access period;
• said decoder decodes motion vector data to determine another prediction area, and said prediction address generator generates read addresses to allow retrieval of the corresponding pixel data from one row of said RAM in a third of said access periods which directly follows said second access period.

An apparatus for storing and retrieving prediction area pixel data of successive video frames from a random access memory (RAM) to minimize the number of times that different rows of said RAM must be addressed, wherein said video frames are provided in a field mode format, comprising:

• means for allocating the pixel data representing said video frames into a plurality of tiles, each tile including at least one macroblock;
• wherein each macroblock is divided into first and second sub-macroblocks;
• a memory manager for generating write addresses for writing said tiles into said RAM on a one tile per row basis, wherein each tile has a height of M pixels and a width of N pixels and substantially fills a different row of said RAM;
• a decoder, coupled to receive motion vectors associated with said video frames, for decoding motion vector data to determine said prediction areas, each having a height equivalent to a vertical space occupied by H pixels and a width equivalent to a horizontal space occupied by W pixels;
• wherein said prediction areas are at least as large as said sub-macroblocks;
• a prediction address generator which is responsive to said decoder for generating read addresses for said video frames for retrieving the prediction area pixel data thereof from said RAM;
• timing means for providing a timing signal which defines consecutive access periods in which said prediction area pixel data may be retrieved from said RAM;
• wherein
• (a) when a selected video frame prediction area of said first field of one of said video frames encompasses only one of said rows, said prediction address generator generates read addresses to allow retrieval of the corresponding pixel data in a first of said access periods;
• said decoder decodes motion vector data to determine another prediction area of said first field of said one of said video frames;
• and said prediction address generator generates read addresses to allow retrieval of the corresponding pixel data from one row of said RAM in a second of said access periods which directly follows said first access period;
• and
• (b) when a selected video frame prediction area of said first field of one of said video frames encompasses only two of said rows, said prediction address generator generates read addresses to allow retrieval of the corresponding pixel data of the first of said rows in a first of said access periods, and to retrieve the corresponding pixel data of the second of said rows in a second of said access periods which directly follows said first access period;
• said decoder decodes motion vector data to determine another prediction area of said first field of said one of said video frames, and said prediction address generator generates read addresses to allow retrieval of the corresponding pixel data from one row of said RAM in a third of said access periods which directly follows said second access period.