Abstract: A memory management system and method are described. In one embodiment, a memory management system includes a memory management unit for virtualizing context memory storage and independently controlling access to the context memory without interference from other engine activities. The shared resource management unit overrides a stream of access denials (e.g., NACKs) associated with an access problem. The memory management system and method facilitate access to memory while controlling translation between virtual and physical memory “spaces”. In one embodiment the memory management system includes a translation lookaside buffer and a fill component. The translation lookaside buffer tracks information associating a virtual memory space with a physical memory space. The fill component tracks the status of an access request progress from a plurality of engines independently and faults that occur in attempting to access a memory space.

Abstract: A system and method for buffering intermediate data in a processing pipeline architecture stores the intermediate data in a shared cache that is coupled between one or more pipeline processing units and an external memory. The shared cache provides storage that is used by multiple pipeline processing units. The storage capacity of the shared cache is dynamically allocated to the different pipeline processing units as needed, to avoid stalling the upstream units, thereby improving overall system throughput.

Abstract: Systems and methods are disclosed for managing the number of affirmatively associated cache lines related to the different sets of a data cache unit. A tag look-up unit implements two thresholds, which may be configurable thresholds, to manage the number of cache lines related to a given set that store dirty data or are reserved for in-flight read requests. If the number of affirmatively associated cache lines in a given set is equal to a maximum threshold, the tag look-up unit stalls future requests that require an available cache line within that set to be affirmatively associated. To reduce the number of stalled requests, the tag look-up unit transmits a high priority clean notification to a frame buffer logic when the number of affirmatively associated cache lines in a given set approaches the maximum threshold. The frame buffer logic then processes requests associated with that set preemptively.

Abstract: A method for cleaning dirty data in an intermediate cache is disclosed. A dirty data notification, including a memory address and a data class, is transmitted by a level 2 (L2) cache to frame buffer logic when dirty data is stored in the L2 cache. The data classes may include evict first, evict normal and evict last. In one embodiment, data belonging to the evict first data class is raster operations data with little reuse potential. The frame buffer logic uses a notification sorter to organize dirty data notifications, where an entry in the notification sorter stores the DRAM bank page number, a first count of cache lines that have resident dirty data and a second count of cache lines that have resident evict_first dirty data associated with that DRAM bank. The frame buffer logic transmits dirty data associated with an entry when the first count reaches a threshold.

Abstract: Deadlocks are avoided by marking read requests issued by a parallel processor to system memory as “special.” Read completions associated with read requests marked as special are routed on virtual channel 1 of the PCIe bus. Data returning on virtual channel 1 cannot become stalled by write requests in virtual channel 0, thus avoiding a potential deadlock.

Abstract: A memory access technique, in accordance with one embodiment of the present invention, includes selectively overriding attributes contained in a translation lookaside buffer or page table data structure with attributes contained in a context specifier.

Abstract: A memory access technique, in accordance with one embodiment of the present invention, includes caching segmentation data. The technique utilizes a separate memory for storing a plurality of context specifiers and an MMU. The MMU includes an on-chip cache and a segmentation unit. The MMU receives a location of a particular context specifier and a corresponding context index for each of one or more of the plurality of context specifiers stored in the separate memory. The segmentation unit retrieves the particular context specifier and caches it locally. The segmentation unit also binds the cache location of the particular context specifier to the corresponding context index. After caching one or more context specifiers and generating a corresponding binding, the segmentation unit may receive a memory access request that includes a given context index. A given context specifier that is cached locally is accessed by the segmentation unit using the context index to get a base address.

Abstract: A shared memory management system and method are described. In one embodiment, a memory management system includes a memory management unit for concurrently managing memory access requests from a plurality of engines. The shared memory management system independently controls access to the context memory without interference from other engine activities. In one exemplary implementation, the memory management unit tracks an identifier for each of the plurality of engines making a memory access request. The memory management unit associates each of the plurality of engines with particular translation information respectively. This translation information is specified by a block bind operation. In one embodiment the translation information is stored in a portion of instance memory. A memory management unit can be non-blocking and can also permit a hit under miss.

Abstract: A method of accessing memory, in accordance with one embodiment, includes receiving a memory access request that includes a virtual address. An address of a given page table is determined utilizing a page directory stored in a particular one of a plurality of computing device-readable media. A given one of the plurality of computing device-readable media that stores the given page table is determined from a table aperture attribute in the page directory. A given physical address of a page is determined utilizing the given page table stored in the given computing device-readable media. A corresponding one of the plurality of computing device-readable media that stores the page is determined from a page aperture attribute in the given page table. The corresponding computing device-readable media at the given physical address is then accessed.

Abstract: In a multiprocessor system level 2 caches are positioned on the memory side of a routing crossbar rather than on the processor side of the routing crossbar. This configuration permits the processors to store messages directly into each other's caches rather than into system memory or their own coherent caches. Therefore, inter-processor communication latency is reduced.

Abstract: Techniques are disclosed for maintaining cache coherency across a serial interface bus such as a Peripheral Component Interconnect Express (PCIe) bus. The techniques include generating a snoop request (SNP) to determine whether first data stored in a local memory is coherent relative to second data stored in a data cache, the snoop request including destination information that identifies the data cache on the serial interface bus and causing the snoop request to be transmitted over the serial interface bus to a second processor. The techniques further include extracting a cache line address from the snoop request, determining whether the second data is coherent, generating a complete message (CPL) indicating that the first data is coherent with the second data, and causing the complete message to be transmitted over the bus to the first processor. The snoop request and complete messages may be vendor defined messages.

Abstract: A system and method for converting data from one format to another in a processing pipeline architecture. Data is stored in a shared cache that is coupled between one or more clients and an external memory. The shared cache provides storage that is used by multiple clients rather than being dedicated to separately convert the data format for each client. Each client may interface with the memory using a different format, such as a compressed data format. Data is converted to the format expected by the particular client as it is read from the cache and output to the client during a read operation. Bytes of a cache line may be remapped to bytes of an unpack register for output to a naïve client, which may be configured to perform texture mapping operations. Data is converted from the client format to the memory format as it is stored into the cache during a write operation.

Abstract: In one embodiment, a method for managing information related to dirty data stored in an intermediate cache coupled to one or more clients and to an external memory includes receiving a dirty data notification related to dirty data residing in the intermediate cache, the dirty data notification including a memory address indicating a location in the external memory where the dirty data should be stored and a data type associated with the dirty data, and extracting a bank page number from the memory address that identifies a bank page within the external memory where the dirty data should be stored. The embodiment also includes incrementing a first count associated with a first entry in a notification sorter that is affirmatively associated with the bank page, determining that the dirty data has a first data type, and incrementing a second count associated with the first entry.

Abstract: A method and system for maintaining cache coherency across a serial interface bus such as a Peripheral Component Interconnect Express (PCIe) bus. The method includes generating a snoop request (SNP) to determine whether first data stored in a local memory is coherent relative to second data stored in a data cache, the snoop request including destination information that identifies the data cache on the serial interface bus, and causing the snoop request to be transmitted over the serial interface bus to a second processor. The method further includes extracting a cache line address from the snoop request, determining whether the second data is coherent, generating a complete message (CPL) indicating that the first data is coherent with the second data, and causing the complete message to be transmitted over the bus to the first processor. The snoop request and complete messages may be vendor defined messages.

Abstract: One embodiment of the invention sets forth a mechanism for using the L2 cache as a buffer for data associated with read/write commands that are processed by the frame buffer logic. A tag look-up unit tracks the availability of each cache line in the L2 cache, reserves necessary cache lines for the read/write operations and transmits read commands to the frame buffer logic for processing. A data slice scheduler transmits a dirty data notification to the frame buffer logic when data associated with a write command is stored in an SRAM bank. The data slice scheduler schedules accesses to the SRAM banks and gives priority to accesses requested by the frame buffer logic to store or retrieve data associated with read/write commands. This feature allows cache lines reserved for read/write commands that are processed by the frame buffer logic to be made available at the earliest clock cycle.

Abstract: One embodiment of the present invention sets forth a method for implementing ECC protection in an on-chip L2 cache. When data is written to or read from an external memory, logic within the L2 cache is configured to generate ECC check bits and store the ECC check bits in the L2 cache in space typically allocated for storing byte enables. As a result, data stored in the L2 cache may be protected against bit errors without incurring the costs of providing additional storage or complex hardware for the ECC check bits.

Abstract: One embodiment of the invention sets forth a mechanism for efficiently processing atomic operations transmitted from multiple general processing clusters to an L2 cache. A tag look-up unit tracks the availability of each cache line in the L2 cache, reserves the necessary cache lines for the atomic operations and transmits the atomic operations to an ALU for processing. The tag look-up unit also increments a reference counter associated with a reserved cache line each time an atomic operation associated with that cache line is received. This feature allows multiple atomic operations associated with the same cache line to be pipelined to the ALU. A ROP unit that includes the ALU may request additional data necessary to process an atomic operation from the L2 cache. Result data is stored in the L2 cache and may also be returned to the general processing clusters.

Abstract: One embodiment of the invention is a method for evicting data from an intermediary cache that includes the steps of receiving a command from a client, determining that there is a cache miss relative to the intermediary cache, identifying one or more cache lines within the intermediary cache to store data associated with the command, determining whether any of data residing in the one or more cache lines includes raster operations data or normal data, and causing the data residing in the one or more cache lines to be evicted or stalling the command based, at least in part, on whether the data includes raster operations data or normal data. Advantageously, the method allows a series of cache eviction policies based on how cached data is categorized and the eviction classes of the data. Consequently, more optimized eviction decisions may be made, leading to fewer command stalls and improved performance.

Abstract: One embodiment of the invention sets forth a mechanism for efficiently processing atomic operations transmitted from multiple general processing clusters to an L2 cache. A tag look-up unit tracks the availability of each cache line in the L2 cache, reserves the necessary cache lines for the atomic operations and transmits the atomic operations to an ALU for processing. The tag look-up unit also increments a reference counter associated with a reserved cache line each time an atomic operation associated with that cache line is received. This feature allows multiple atomic operations associated with the same cache line to be pipelined to the ALU. A ROP unit that includes the ALU may request additional data necessary to process an atomic operation from the L2 cache. Result data is stored in the L2 cache and may also be returned to the general processing clusters.

Abstract: One embodiment of the present invention sets forth a method for implementing ECC protection in an on-chip L2 cache. When data is written to or read from an external memory, logic within the L2 cache is configured to generate ECC check bits and store the ECC check bits in the L2 cache in space typically allocated for storing byte enables. As a result, data stored in the L2 cache may be protected against bit errors without incurring the costs of providing additional storage or complex hardware for the ECC check bits.