Abstract: A system and method are described that manage incremental state updates in such a way that multiple threads within a processor can each operate, in effect, on their own set of state data. The system and method are applicable to any processor in which multiple threads require access to sets of state information which differ from one another by a relatively small number of state changes.

Abstract: The present invention provides a system for handling extra contexts for shader constants, and applications thereof. In an embodiment there is provided a computer-based method for executing a series of compute packets in an execution pipeline. The execution pipeline includes a first plurality of registers configured to store state-updates of a first type and a second plurality of registers configured to store state-updates of a second type. A first number of state-updates of the first type and a second number of state-updates of the second type are respectively identified and stored in the first and second plurality of registers. A compute packet is sent to the execution pipeline responsive to the first number and the second number. Then, the compute packet is executed by the execution pipeline.

Abstract: A cache memory with improved cache-miss performance is implemented by providing cache-miss data from system memory directly to its requester. One embodiment of the invention operates as a texture cache in a graphics system. The graphics system comprises a system memory that stores texture data, coupled to a texture cache memory, which is coupled to at least one requester. The texture cache memory is divided into a cache tags unit and a data cache unit. The data cache unit is configured to receive at least two cache address inputs, and has at least two data output ports each coupled to a respective first input of a respective multiplexer. A respective second input of each multiplexer is configured to receive cache-miss data from the system memory. The select input of each multiplexer is configured to receive a respective hit/miss indicator signal associated with the respective cache address input. In case of a cache-miss, cache-miss data from system memory bypasses the data cache unit and is output directly.

Abstract: Data, such as data received by a memory I/O from a memory unit in a DDR SDRAM system, is captured using a trigger signal, which may be a non free-running clock signal such as a DQS signal in a DDR SDRAM system, and is transferred to a host system, which may be part of an ASIC, using the host system's clock. The memory I/O includes a data capture register that latches the data received from the memory unit using DQS. The memory I/O also includes a FIFO buffer that latches the data output by the data capture register using a delayed version of DQS. A single edge of the delayed DQS is available to the FIFO for latching each set of data that corresponds to a single pulse of DQS. The FIFO transfers the data to the host system using the host system's clock, which represents a different clock domain than DQS.

Abstract: Cache access is optimized through identifying redundant accesses (read-requests made to identical system memory addresses), and issuing a single cache data request for each group of redundant accesses. One embodiment of the invention is a graphics system comprising a system memory that stores texture data, coupled to a texture cache that is coupled to one or more texture pipes. Each pipe processes information for a respective spatial bin. A cache preprocessor receives read-requests for texels from the texture pipes and generates a control code corresponding to each read-request, indicating whether the read-request is a redundant access, and linking redundant accesses to a single cache data request. The cache preprocessor provides the control codes and the read-requests to a cache arbiter, which issues the codes and the cache data requests to the texture cache.

Abstract: A system and a method for improving magnified texture-mapped pixel performance in a single-pixel pipeline. A plurality of textured pixel addresses corresponding to a plurality of pixels may be generated. A FIFO or other memory unit may be used to linearly order the plurality of textured pixel addresses. Two consecutive textured pixel addresses out of the plurality of textured pixel addresses may be examined if they map to a common set of texels in texture space. The two consecutive textured pixel addresses may be merged together and propagated down the pipeline if they map to the common set of texels. However, only a first of the two consecutive textured pixel addresses may be propagated down the pipeline if the two consecutive textured pixel addresses do not map to a common set of texels. Texel data may be generated in response to receiving either the combined texel structure or the first of the two textured pixel addresses.

Abstract: A predictive memory performance optimizing unit for use with an interleaved memory, for example a DDR SDRAM memory, and suitable for use in a computer graphics system, among others, is described. The unit maintains a queue of pending requests for data from the memory, and prioritizes precharging and activating interleaves with pending requests. Interleaves which are in a ready state may be accessed independently of the precharging and activation of non-ready interleaves. The unit utilizes idle cycles occurring between consecutive requests to activate interleaves with pending requests.

Abstract: The propagation of a feedback signal, such as a DQS signal generated in response to a read request in a Double Data Rate (DDR) memory system, into a digital host system, such as an ASIC, is controlled by using a programmable delay circuit and detection sequence to compensate for variable I/O delay. The memory system includes a controller and an interface, both on the ASIC, and memory units coupled to the controller through the interface. The interface uses the read request signal, sent by the controller to initiate read operations, to generate a select signal. A programmable delay element inside the interface unit is programmed using a delay value generated by a delay manager unit inside the controller. The programmable delay element delays the select signal, and an enable signal is generated from the delayed select signal, using DQS. The propagation of DQS is controlled by the enable signal.

Abstract: A method for optimizing a cache memory used for multitexturing in a graphics system is implemented. The graphics system comprises a texture memory, which stores texture data comprised in texture maps, coupled to a texture cache memory. Active texture maps for an individual primitive, for example a triangle, are identified, and the texture cache memory is divided into partitions. In one embodiment, the number of texture cache memory partitions equals the number of active texture maps. Each texture cache memory partition corresponds to a respective single active texture map, and is operated as a direct mapped cache for its corresponding respective single active texture map. In one embodiment, each texture cache memory partition is further operated as an associative cache for the texture data comprised in the partition's corresponding respective single active texture map. The cache memory is dynamically re-configured for each primitive.

Abstract: A rendering unit positions a supertile so that it intersects a primitive. The rendering unit repeatedly walks over bins of the supertile, applying a layer of texture to the bins of the supertile in each iteration of said repeated walking. The rendering unit advances to the next texture layer after having applied the current texture layer to each candidate bin of the supertile. The results of each texture layer application to the bins may be stored in a texture accumulation buffer. The size of the supertile corresponds to the size of the texture accumulation buffer. After applying a last layer of texture to the bins of the supertile, the supertile may be advanced to a new position. The rendering unit traverses the primitive with the supertile so that the union of areas visited by the supertile covers the primitive.

Abstract: The propagation of a feedback signal, such as a DQS signal generated in response to a read request in a Double Data Rate (DDR) memory system, into a digital host system, such as an ASIC, is controlled by using delay tracking to compensate for variable I/O delay. The memory system includes a controller and an interface, both on the ASIC, and memory units coupled to the controller through the interface, all configured on a printed circuit board (PCB). The interface uses the read request signal, sent by the controller to initiate read operations, to generate a read-enable signal, which is transmitted to a trace on the PCB one-half cycle of the system clock before DQS is expected to reach the interface. The trace tracks the total delay encountered by the system clock and DQS between the interface unit and memory units, and is routed back to the interface unit, where read-enable is used to generate an enable signal that allows DQS to propagate into the ASIC only when DQS is a valid digital signal.

Abstract: A multi-chip system and method are disclosed for incorporating a primitive assembler in each of one or more geometry chips and one or more rasterization chips. This system may allow per-primitive operations to be performed in the geometry chips, and also allow use of a vertex data interface for sending vertex data to the rasterization chips. The primitive assemblers in the geometry chips may assemble vertices into primitives for clipping tests. The geometry chips may also test an assembled primitive against the projected boundaries of a set of screen space regions, where each region is assigned to one of the rasterization chips. Those primitives residing in more than one region may be sub-divided into two or more new primitives so that each new primitive resides in only one screen space region. The geometry chip may then send the vertex data for each primitive to the corresponding rasterization chip.

Abstract: A non-blocking cache for texture mapping is implemented by separating Cache Tags from Cache Data. Multiple requests for data may be processed in parallel without strict ordering or synchronization. Separating Cache Tags and Cache Data results in a texture memory cache design that preempts stalling which would otherwise occur in case of cache-misses. Multiple Cache Tags with corresponding respective system memory controllers and Data Cache units allow for simultaneous processing of multiple requests without strict ordering. In preferred embodiments the texture memory cache may also be configured to predict cache misses and merge with burst reads from memory, and may equally be configured to minimize memory read-requests necessary during multitexturing, thus maximizing bandwidth.

Abstract: A graphics system configured to apply multiple layers of texture information to primitives. The graphics system receives parameters defining a primitive and performs a size test on the primitive. If the size test cannot guarantee that a fragment size of the primitive is less than or equal to a fragment capacity of a texture accumulation buffer, the primitive is divided into subprimitives, and the graphics system applies the multiple layers of texture to fragments which intersect the primitive. The graphics system switches from a current layer to the layer next when it has applied textures corresponding to the current layer to all the fragments intersecting the primitive. The graphics system stores color values associated with the primitive fragments in the texture accumulation buffer between the application of successive texture layers.

Abstract: A graphics system may include a frame buffer, a processing device coupled to output data, a multipurpose memory device that includes a plurality of storage locations and is coupled to store data output from the processing device, and a multipurpose memory controller coupled to the multipurpose memory device. The multipurpose memory controller may be configured to allocate a first plurality of the storage locations to a first image buffer configured to store image data, a second plurality of the storage locations to a first texture buffer configured to store texture data, and a third plurality of the storage locations to a first accumulation buffer configured to store accumulation buffer data. The multipurpose memory device may be configured to include a first image buffer, a first texture buffer, and a first accumulation buffer at the same time.

Abstract: A method and a system for stalling large pipelined designs. A computational pipeline may comprise a first module and a second module coupled together. The first module may propagate one or more signals to the second module. A stall-signal may be asserted in order to stall the computational pipeline if the second module is not ready to receive the one or more signals from the first module. The one or more signals propagated from the first module and the asserted stall-signal may be buffered in a stall-buffer. The asserted stall-signal may be propagated to the first module in a next cycle. The first module may be stalled in response to the first module receiving the propagated asserted stall-signal. Next, the asserted stall-signal may be propagated up the computational pipeline.

Abstract: A system and method for preserving the order of data items through a divergence-and-reconvergence of two or more paths in a hardware device. A host processor may write a first token to a first path in the hardware device. A convergence unit in the hardware device may receive and store the first token in a synchronization register. The host processor may poll the synchronization register to determine when the first token arrives in the synchronization register. In response to determining that the first token has arrived in the synchronization register, the host processor may safely write a sequence of one or more data items to a second path in the hardware device.

Abstract: A graphics system applies multiple layers of texture information to triangles. The graphics system includes a hardware accelerator, a frame buffer and a video output processor. The hardware accelerator receives vertices of a triangle, identifies fragments of a sampling space which intersect the triangle, and applies the multiple layers of texture to the intersecting fragments. The multiple layers of textures may be stored in a texture memory external to the hardware accelerator. The hardware accelerator switches to a next texture layer after applying the textures of a current layer to all the fragments of the triangle. The hardware accelerator includes (or couples to) a texture accumulation buffer which stores color values associated with the triangle fragments between the application of successive texture layers. The frame buffer stores the samples and pixels generated from the samples by filtration. The video output processor transforms the pixels into a video signal.

Abstract: A system and a method for improving magnified texture-mapped pixel performance in a single-pixel pipeline. Two textured pixel addresses corresponding to two pixels may be generated. The two textured pixel addresses may then be passed to the next unit in the pipeline, where the two textured pixel addresses can be examined if the corresponding two pixels correspond to a common set of texels in texture space. The two textured pixel addresses may be merged together if the two pixels correspond to the common set of texels. Merging may operate to create a combined texel structure. Texel data may be generated in response to receiving the combined texel structure. The texel data may be filtered using one or more texture filters in order to generate texture values.

Abstract: A system and method for reading register contents from a computational pipeline having a plurality of computational units. The system includes a readback bus and a read control unit. The readback bus has a plurality of logic units coupled in a series. Each logic unit couples to a corresponding one of the computational units. The read control unit couples to each of the computational units through a corresponding load line, and is configured to assert a load signal on one of the load lines in response to a register read request. Each of the computational units is configured to transmit a data value from a selected register onto the readback bus in response to detecting an assertion of the load signal on its corresponding load line.