Method and system for improved memory interface during image rendering

• US 6,104,418 A
• Filed: 04/06/1998
• Issued: 08/15/2000
• Est. Priority Date: 04/06/1998
• Status: Expired due to Term

• Current Assignee: InnoLux Corporation
• Original Assignee: Silicon Magic Corp.
• Inventors: Kuo, Sean Hsi-an, Choy, Kenneth, Stefanek, Gregory M., Tanaka, Greg L., Lee, Michael M.
• Primary Examiner(s): Tung, Kee M.
• Application Number:
  US09/055,566
• Time in Patent Office: 862 Days
• Field of Search: 345/507-510, 345/430, 345/521, 345/503, 345/515, 711/5, 711/118
• US Class Current: 345/531
• CPC Class Codes: G06T 1/60 Memory management

Abstract

Aspects for increasing efficiency of memory accesses during graphics rendering are provided. A preferred method aspect includes providing a plurality of memory banks for data, and decoding input signals that indicate accessing of at least one of the plurality of memory banks for a desired plurality of words of data. The method further includes splitting data access across the plurality of memory banks to allow parallel selection of an output from at least one of the plurality of memory banks as the desired plurality of words of the data, wherein latency of data access is amortized. A system aspect for improving data transfer from memory to a texture mapping unit includes a plurality of cache banks for storing texel data, a bank decode unit coupled to the plurality of cache banks for decoding a plurality of input signals indicative of selection of texel data from one or more of the plurality of banks, and selection control logic for receiving decoded address data from the bank decode unit and controlling selection of the plurality of cache banks to retrieve the texel data for output to the texture mapping unit.

Claims

1. A method for achieving independent access to memory that stores data in a graphics processing system, the method comprising:

providing a plurality of memory banks for data;

decoding signals that indicate accessing of at least one of the plurality of memory banks for a desired plurality of words of the data;

splitting data access across the plurality of memory banks to allow parallel selection of an output from at least one of the plurality of memory banks as the desired plurality of words of the data, wherein latency of data access is amortized, including multiplexing first, second, third and fourth words of data in pairs from the plurality of memory banks, multiplexing the resulting selected first, second, third and fourth words from each of the pairs to provide the desired plurality of words of the data, and byte-swapping the first, second, third and fourth words before transferring the desired plurality of words of the data to a texture mapping unit.

3. The method of claim 1 wherein providing further comprises providing four memory banks.

4. The method of claim 3 wherein decoding input signals further comprises decoding a top left, top right, bottom left and bottom right signals.

2. A method for achieving independent access to memory that stores data in a graphics processing system, the method comprising:

providing a plurality of memory banks for data,

decoding signals that indicate accessing of at least one of the plurality of memory banks for a desired plurality of words of the data;

splitting data access across the plurality of memory banks to allow parallel selection of an output from at least one of the plurality of memory banks as the desired plurality of words of the data, wherein latency of data access is amortized; and

remapping data in the plurality of memory banks by remapping most frequently changing bits to part of memory incurring a minimal penalty, wherein the remapping maintains locality of the data to provide efficient accessing of the plurality of memory banks.

5. The method of claim 2 further comprising transferring the desired plurality of words of the data to a texture mapping unit.

6. A system for improving data transfer from memory to a texture mapping unit in a graphics rendering system, the system comprising:

a plurality of cache banks for storing texel data;

a bank decode unit coupled to the plurality of cache banks for decoding a plurality of signals indicative of selection of the texel data from one or more of the plurality of banks; and

selection control logic for receiving decoded address data from the bank decode unit and controlling selection of the plurality of cache banks to retrieve the texel data in parallel for output to the texture mapping unit, including a fetch state machine to receive decoded input signals from the bank decode unit and identify when one or more of the plurality of cache banks store the selected texel data.

7. The system of claim 6 wherein the plurality of cache banks comprise four cache banks.

8. The system of claim 6 wherein the fetch state machine further controls syncing up the selected texel data from the plurality of cache banks to account for varying retrieval times due to varying latencies among the plurality of cache banks.

9. The system of claim 6 wherein the selection control logic further comprises a plurality of multiplexor means for routing the selected texel data in a predetermined order to the texture mapping unit.

10. A computer system with an improved memory interface during texture mapping in image processing, the system comprising:

a central processing unit (CPU) for providing rendering commands and data;

core logic coupled to the CPU for transferring the rendering commands and data; and

a graphics processing means coupled to the core logic, the graphics processing means comprising;

a texture mapping unit for receiving and processing texture elements; and

a memory unit coupled to the texture mapping unit, the memory unit providing a plurality of texture elements to the texture mapping unit in a single cycle, and including a plurality of cache banks for storing the texture elements with most frequently changing bits remapped to a part of memory incurring a minimal penalty, wherein locality of the data is maintained to provide efficient accessing of the plurality of memory banks.

11. The computer system of claim 10 wherein the graphics processing means further comprises a 2D GUI and a 3D accelerator.

12. The computer system of claim 11 wherein the 3D accelerator further comprises a rasterization pipeline, the rasterization pipeline including the texture mapping unit.

13. The computer system of claim 10 wherein the memory unit further comprises a bank decode unit coupled to the plurality of cache banks for decoding a plurality of input signals indicative of selection of texture elements from one or more of the plurality of cache banks.

14. The computer system of claim 13 wherein the memory unit further comprises selection control logic for receiving decoded address data from the bank decode unit and controlling selection of the plurality of cache banks to retrieve the texel data in parallel for output to the texture mapping unit.

15. The computer system of claim 13 wherein the bank decode unit decodes input signals comprising a top left, top right, bottom left and bottom right signals.

16. The computer system of claim 10 wherein each of the plurality of cache banks further comprises a memory bank.

17. A memory device capable of effectively increasing bandwidth for memory accesses during graphics rendering, the memory device comprising:

a plurality of memory banks, each of the plurality of memory banks divided into a plurality of memory blocks; and

decoding logic coupled to the plurality of memory banks, the decoding logic supporting parallel selection of data from at least one memory block of at least one memory bank in a single cycle, and including a fetch state machine to receive decoded input signals and identify when one or more of the plurality of memory banks store the selected texel data.

18. The memory device of claim 17 wherein the plurality of memory banks further comprises four memory banks.

19. The memory device of claim 18 wherein the plurality of blocks further comprises eight memory blocks for each memory bank.