Abstract: Various embodiments include techniques for accessing extended memory in a parallel processing system via a high-bandwidth path to extended memory residing on a central processing unit. The disclosed extended memory system extends the directly addressable high-bandwidth memory local to a parallel processing system and avoids the performance penalties associated with low-bandwidth system memory. As a result, execution threads that are highly parallelizable and access a large memory space execute with increased performance on a parallel processing system relative to prior approaches.

Abstract: One embodiment of the present invention includes a memory management unit (MMU) that is configured to efficiently process requests to access memory that includes protected regions. Upon receiving an initial request via a virtual address (VA), the MMU translates the VA to a physical address (PA) based on page table entries (PTEs) and gates the response based on page-specific secure state information. To thwart software-based attempts to illicitly access the protected regions, the secure state information is not stored in page tables. However, to expedite subsequent requests, after the MMU identifies the PTE and the corresponding secure state information, the MMU stores both the PTE and the secure state information as a cache line in a translation lookaside buffer.

Abstract: A system and method are provided for implementing multi-stage translation of virtual addresses. The method includes the steps of receiving, at a first memory management unit, a memory request including a virtual address in a first address space, translating the virtual address to generate a second virtual address in a second address space, and transmitting a modified memory request including the second virtual address to a second memory management unit. The second memory management unit is configured to translate the second virtual address to generate a physical address in a third address space. The physical address is associated with a location in a memory.

Abstract: A system and method for decompressing compressed data (e.g., in a frame buffer) and optionally recompressing the data. The method includes determining a portion of an image to be accessed from a memory and sending a conditional read corresponding to the portion of the image. In response to the conditional read, an indicator operable to indicate that the portion of the image is uncompressed may be received. If the portion of the image is compressed, in response to the conditional read, compressed data corresponding to the portion of the image is received. In response to receiving the compressed data, the compressed data is uncompressed into uncompressed data. The uncompressed data may then be written to the memory corresponding to the portion of the image. The uncompressed data may then be in-place compressed for or during subsequent processing.

Abstract: One embodiment of the present invention sets forth a technique for converting alpha values into pixel coverage masks. Geometric coverage is sampled at a number of “real” sample positions within each pixel. Color and depth values are computed for each of these real samples. Fragment alpha values are used to determine an alpha coverage mask for the real samples and additional “virtual” samples, in which the number of bits set in the mask bits is proportional to the alpha value. An alpha-to-coverage mode uses the virtual samples to increase the number of transparency levels for each pixel compared with using only real samples. The alpha-to-coverage mode may be used in conjunction with virtual coverage anti-aliasing to provide higher-quality transparency for rendering anti-aliased images.

Abstract: One embodiment of the present invention includes a memory management unit (MMU) that is configured to efficiently process requests to access memory that includes protected regions. Upon receiving an initial request via a virtual address (VA), the MMU translates the VA to a physical address (PA) based on page table entries (PTEs) and gates the response based on page-specific secure state information. To thwart software-based attempts to illicitly access the protected regions, the secure state information is not stored in page tables. However, to expedite subsequent requests, after the MMU identifies the PTE and the corresponding secure state information, the MMU stores both the PTE and the secure state information as a cache line in a translation lookaside buffer.

Abstract: A method for compressing framebuffer data is presented. The method includes determining a reduction ratio for framebuffer data in a tile including multiple samples. The reduction ratio determined is independent of the sampling mode, where the sampling mode is the number of samples within each pixel in the tile. The method further includes comparing a first portion of the framebuffer data for each of the multiple samples to determine an equality comparison result and also comparing a second portion of the framebuffer data for each one of the multiple samples to compute per-channel differences for each one of the multiple samples and testing the per-channel differences against a threshold value to determine a threshold comparison result. Finally, the method comprises compressing the framebuffer data for the tile based on the reduction ratio, the equality comparison result and the threshold comparison result to produce output framebuffer data for the tile.

Abstract: A system and method are described for compressing image data using a combination of compression methods. Compression method combinations are provided to compress image data of a particular frame buffer format and antialiasing mode. Each method in the compression method combination is tried in turn to compress the image data in a tile. The best method that succeeded in compressing the image data is encoded in the compression bit state associated with the tile. Together, the compression bits, the compression method combination, and the frame buffer format provide sufficient information to decompress a tile.

Abstract: A system and method are provided for protecting data. In operation, a request to read data from memory is received. Additionally, it is determined whether the data is stored in a predetermined portion of the memory. If it is determined that the data is stored in the predetermined portion of the memory, the data and a protect signal are returned for use in protecting the data. In certain embodiments of the invention, data stored in the predetermined portion of the memory may be further processed and written hack to the predetermined portion of the memory. In other embodiments of the invention, such processing may involve unprotected data stored outside the predetermined portion of the memory.

Abstract: A system and method are provided for implementing multi-stage translation of virtual addresses. The method includes the steps of receiving, at a first memory management unit, a memory request including a virtual address in a first address space, translating the virtual address to generate a second virtual address in a second address space, and transmitting a modified memory request including the second virtual address to a second memory management unit. The second memory management unit is configured to translate the second virtual address to generate a physical address in a third address space. The physical address is associated with a location in a memory.

Abstract: Sequential write operations to a unit of compressed memory, known as a compression tile, are examined to see if the same compression tile is being written. If the same compression tile is being written, the sequential write operations are coalesced into a single write operation and the entire compression tile is overwritten with the new data. Coalescing multiple write operations into a single write operation improves performance, because it avoids the read-modify-write operations that would otherwise be needed.

Abstract: One embodiment of the present invention sets forth a technique for rendering graphics primitives in parallel while maintaining the API primitive ordering. Multiple, independent geometry units perform geometry processing concurrently on different graphics primitives. A primitive distribution scheme delivers primitives concurrently to multiple rasterizers at rates of multiple primitives per clock while maintaining the primitive ordering for each pixel. The multiple, independent rasterizer units perform rasterization concurrently on one or more graphics primitives, enabling the rendering of multiple primitives per system clock.

Abstract: One embodiment of the present invention sets forth an architecture for advancing the Z-test operation prior to pixel shading whenever possible. The current rendering state, as maintained by the setup engine, determines whether advancing the Z-test function above the shader engine for “early” Z-testing is possible or whether the Z-test function should be deferred until after shading operations for “late” Z-testing. Data is dynamically routed to each processing engine in the pipeline, so that the appropriate data flow for either early Z or late Z is dynamically constructed, as determined by the current rendering state. The same functional units are utilized in both early Z and late Z configurations.

Abstract: Sequential write operations to a unit of compressed memory, known as a compression tile, are examined to see if the same compression tile is being written. If the same compression tile is being written, the sequential write operations are coalesced into a single write operation and the entire compression tile is overwritten with the new data. Coalescing multiple write operations into a single write operation improves performance, because it avoids the read-modify-write operations that would otherwise be needed.

Abstract: One embodiment of the present invention sets forth a compression status cache configured to store compression information for blocks of memory stored within an external memory. A data cache unit is configured to request, in response to a cache miss, compressed data from the external memory based on compression information stored in the compression status bit cache. The compression status for active buffers is dynamically swapped into the compression status cache as needed. Different compression formats may be specified for one or more tiles within an active buffer. One advantage of the disclosed compression status cache is that a lame amount of attached memory may be allocated as compressible memory blocks, without incurring a corresponding die area cost because a portion of the compression status stored off chip in attached memory is cached in the compression status cache.

Abstract: A system and method are provided for protecting data. In operation, a request to read data from memory is received. Additionally, it is determined whether the data is stored in a predetermined portion of the memory. If it is determined that the data is stored in the predetermined portion of the memory, the data and a protect signal are returned for use in protecting the data. In certain embodiments of the invention, data stored in the predetermined portion of the memory may be further processed and written hack to the predetermined portion of the memory. In other embodiments of the invention, such processing may involve unprotected data stored outside the predetermined portion of the memory.

Abstract: A system and method for decompressing compressed data (e.g., in a frame buffer) and optionally recompressing the data. The method includes determining a portion of an image to be accessed from a memory and sending a conditional read corresponding to the portion of the image. In response to the conditional read, an indicator operable to indicate that the portion of the image is uncompressed may be received. If the portion of the image is compressed, in response to the conditional read, compressed data corresponding to the portion of the image is received. In response to receiving the compressed data, the compressed data is uncompressed into uncompressed data. The uncompressed data may then be written to the memory corresponding to the portion of the image. The uncompressed data may then be in-place compressed for or during subsequent processing.

Abstract: One embodiment of the present invention sets forth a technique for using a shared memory to store hardware-managed virtual buffers. A circular buffer is allocated within a general-purpose multi-use cache for storage of primitive attribute data rather than having a dedicated buffer for the storage of the primitive attribute data. The general-purpose multi-use cache is also configured to store other graphics data sinces the space requirement for primitive attribute data storage is highly variable, depending on the number of attributes and the size of primitives. Entries in the circular buffer are allocated as needed and released and invalidated after the primitive attribute data has been consumed. An address to the circular buffer entry is transmitted along with primitive descriptors from object-space processing to the distributed processing in screen-space.

Abstract: One embodiment of the present invention sets forth a technique for rendering graphics primitives in parallel while maintaining the API primitive ordering. Multiple, independent geometry units perform geometry processing concurrently on different graphics primitives. A primitive distribution scheme delivers primitives concurrently to multiple rasterizers at rates of multiple primitives per clock while maintaining the primitive ordering for each pixel. The multiple, independent rasterizer units perform rasterization concurrently on one or more graphics primitives, enabling the rendering of multiple primitives per system clock.

Abstract: A parallel array architecture for a graphics processor includes a multithreaded core array including a plurality of processing clusters, each processing cluster including at least one processing core operable to execute a pixel shader program that generates pixel data from coverage data; a rasterizer configured to generate coverage data for each of a plurality of pixels; and pixel distribution logic configured to deliver the coverage data from the rasterizer to one of the processing clusters in the multithreaded core array. A crossbar coupled to each of the processing clusters is configured to deliver pixel data from the processing clusters to a frame buffer having a plurality of partitions.