Abstract: Methods and systems are disclosed for executing operations on single-instruction-multiple-data (SIMD) units. Techniques disclosed perform a dot product operation on input data during one computer cycle, including convolving the input data, generating intermediate data, and applying one or more transitional operations to the intermediate data to generate output data. Aspects described, wherein the input data is an input to a layer of a convolutional neural network and the generated output data is the output of the layer.

Abstract: A first workload is executed in a first subset of pipelines of a processing unit. A second workload is executed in a second subset of the pipelines of the processing unit. The second workload is dependent upon the first workload. The first and second workloads are suspended and state information for the first and second workloads is stored in a first memory in response to suspending the first and second workloads. In some cases, a third workload executes in a third subset of the pipelines of the processing unit concurrently with executing the first and second workloads. In some cases, a fourth workload is executed in the first and second pipelines after suspending the first and second workloads. The first and second pipelines are resumed on the basis of the stored state information in response to completion or suspension of the fourth workload.

Abstract: Embodiments include methods, systems and non-transitory computer-readable computer readable media including instructions for executing a prefetch kernel that includes memory accesses for prefetching data for a processing kernel into a memory, and, subsequent to executing at least a portion of the prefetch kernel, executing the processing kernel where the processing kernel includes accesses to data that is stored into the memory resulting from execution of the prefetch kernel.

Abstract: Embodiments include methods, systems and non-transitory computer-readable computer readable media including instructions for executing a prefetch kernel with reduced intermediate state storage resource requirements. These include executing a prefetch kernel on a graphics processing unit (GPU), such that the prefetch kernel begins executing before a processing kernel. The prefetch kernel performs memory operations that are based upon at least a subset of memory operations in the processing kernel.

Abstract: A processing element is implemented in a stage of a pipeline and configured to execute an instruction. A first array of multiplexers is to provide information associated with the instruction to the processing element in response to the instruction being in a first set of instructions. A second array of multiplexers is to provide information associated with the instruction to the first processing element in response to the instruction being in a second set of instructions. A control unit is to gate at least one of power or a clock signal provided to the first array of multiplexers in response to the instruction being in the second set.

Abstract: A multi-die parallel processor semiconductor package includes a first base IC die including a first plurality of virtual compute dies 3D stacked on top of the first base IC die. A first subset of a parallel processing pipeline logic is positioned at the first plurality of virtual compute dies. Additionally, a second subset of the parallel processing pipeline logic is positioned at the first base IC die. The multi-die parallel processor semiconductor package also includes a second base IC die including a second plurality of virtual compute dies 3D stacked on top of the second base IC die. An active bridge chip communicably couples a first interconnect structure of the first base IC die to a first interconnect structure of the second base IC die.

Abstract: A method, computer system, and a non-transitory computer-readable storage medium for performing primitive batch binning are disclosed. The method, computer system, and non-transitory computer-readable storage medium include techniques for generating a primitive batch from a plurality of primitives, computing respective bin intercepts for each of the plurality of primitives in the primitive batch, and shading the primitive batch by iteratively processing each of the respective bin intercepts computed until all of the respective bin intercepts are processed.

Abstract: A processor maintains an access log indicating a stream of cache misses at a cache of the processor. In response to each of at least a subset of cache misses at the cache, the processor records a corresponding entry in the access log, indicating a physical memory address of the memory access request that resulted in the corresponding miss. In addition, the processor maintains an address translation log that indicates a mapping of physical memory addresses to virtual memory addresses. In response to an address translation (e.g., a page walk) that translates a virtual address to a physical address, the processor stores a mapping of the physical address to the corresponding virtual address at an entry of the address translation log. Software executing at the processor can use the two logs for memory management.

Abstract: An array processor includes processor element arrays distributed in rows and columns. The processor element arrays perform operations on parameter values. The array processor also includes memory interfaces that are dynamically mapped to mutually exclusive subsets of the rows and columns of the processor element arrays based on dimensions of matrices that provide the parameter values to the processor element arrays. In some cases, the processor element arrays are vector arithmetic logic unit (ALU) processors and the memory interfaces are direct memory access (DMA) engines. The rows of the processor element arrays in the subsets are mutually exclusive to the rows in the other subsets and the columns of the processor element arrays in the subsets are mutually exclusive to the columns in the other subsets. The matrices can be symmetric or asymmetric, e.g., one of the matrices can be a vector having a single column.

Abstract: A method and apparatus for performing a multi-precision computation in a plurality of arithmetic logic units (ALUs) includes pairing a first Single Instruction/Multiple Data (SIMD) block channel device with a second SIMD block channel device to create a first block pair having one-level staggering between the first and second channel devices. A third SIMD block channel device is paired with a fourth SIMD block channel device to create a second block pair having one-level staggering between the third and fourth channel devices. A plurality of source inputs are received at the first block pair and the second block pair. The first block pair computes a first result, and the second block pair computes a second result.

Abstract: A graphics processing unit (GPU) or other apparatus includes a plurality of shader engines. The apparatus also includes a first front end (FE) circuit and one or more second FE circuits. The first FE circuit is configured to schedule geometry workloads for the plurality of shader engines in a first mode. The first FE circuit is configured to schedule geometry workloads for a first subset of the plurality of shader engines and the one or more second FE circuits are configured to schedule geometry workloads for a second subset of the plurality of shader engines in a second mode. In some cases, a partition switch is configured to selectively connect the first FE circuit or the one or more second FE circuits to the second subset of the plurality of shader engines depending on whether the apparatus is in the first mode or the second mode.

Abstract: An apparatus such as a graphics processing unit (GPU) includes a plurality of processing elements configured to concurrently execute a plurality of first waves and accumulators associated with the plurality of processing elements. The accumulators are configured to store accumulated values representative of behavioral characteristics of the plurality of first waves that are concurrently executing on the plurality of processing elements. The apparatus also includes a dispatcher configured to dispatch second waves to the plurality of processing elements based on comparisons of values representative of behavioral characteristics of the second waves and the accumulated values stored in the accumulators. In some cases, the behavioral characteristics of the plurality of first waves comprise at least one of fetch bandwidths, usage of an arithmetic logic unit (ALU), and number of export operations.

Abstract: The address of the draw or dispatch packet responsible for creating an exception is tied to a shader/wavefront back to the draw command from which it originated. In various embodiments, a method of operating a graphics pipeline and exception handling includes receiving, at a command processor of a graphics processing unit (GPU), an exception signal indicating an occurrence of a pipeline exception at a shader stage of a graphics pipeline. The shader stage generates an exception signal in response to a pipeline exception and transmits the exception signal to the command processor. The command processor determines, based on the exception signal, an address of a command packet responsible for the occurrence of the pipeline exception.

Abstract: Improvements in the graphics processing pipeline are disclosed. More specifically, a new primitive shader stage performs tasks of the vertex shader stage or a domain shader stage if tessellation is enabled, a geometry shader if enabled, and a fixed function primitive assembler. The primitive shader stage is compiled by a driver from user-provided vertex or domain shader code, geometry shader code, and from code that performs functions of the primitive assembler. Moving tasks of the fixed function primitive assembler to a primitive shader that executes in programmable hardware provides many benefits, such as removal of a fixed function crossbar, removal of dedicated parameter and position buffers that are unusable in general compute mode, and other benefits.

Abstract: An array processor includes processor element arrays (PEAs) distributed in rows and columns. The PEAs are configured to perform operations on parameter values. A first sequencer received a first direct memory access (DMA) instruction that includes a request to read data from at least one address in memory. A texture address (TA) engine requests the data from the memory based on the at least one address and a texture data (TD) engine provides the data to the PEAs. The PEAs provide first synchronization signals to the TD engine to indicate availability of registers for receiving the data. The TD engine provides second synchronization signals to the first sequencer in response to receiving acknowledgments that the PEAs have consumed the data.

Abstract: A processing system includes a first set and a second set of general-purpose registers (GPRs) and memory access circuitry that fetches nonzero values of a sparse matrix into consecutive slots in the first set. The memory access circuitry also fetches values of an expanded matrix into consecutive slots in the second set of GPRs. The expanded matrix is formed based on values of a vector and locations of the nonzero values in the sparse matrix. The processing system also includes a set of multipliers that concurrently perform multiplication of the nonzero values in slots of the first set of GPRs with the values of the vector in corresponding slots of the second set. Reduced sum circuitry accumulates results from the set of multipliers for rows of the sparse matrix.

Abstract: A processing system executes wavefronts at multiple arithmetic logic unit (ALU) pipelines of a single instruction multiple data (SIMD) unit in a single execution cycle. The ALU pipelines each include a number of ALUs that execute instructions on wavefront operands that are collected from vector general process register (VGPR) banks at a cache and output results of the instructions executed on the wavefronts at a buffer. By storing wavefronts supplied by the VGPR banks at the cache, a greater number of wavefronts can be made available to the SIMD unit without increasing the VGPR bandwidth, enabling multiple ALU pipelines to execute instructions during a single execution cycle.

Abstract: A system, method and a non-transitory computer readable storage medium are provided for hybrid rendering with deferred primitive batch binning. A primitive batch is generated from one or more primitives. A bin is identified for processing the primitive batch. At least a portion of each primitive intersecting the identified bin is processed and a next bin for processing the primitive batch is identified based on an intercept walk order. The processing is iteratively repeated for the one or more primitives in the primitive batch for successive bins until all primitives of the primitive batch are completely processed. Then, the one or more primitives in the primitive batch are further processed.

Abstract: A graphics processing unit (GPU) or other apparatus includes a plurality of shader engines. The apparatus also includes a first front end (FE) circuit and one or more second FE circuits. The first FE circuit is configured to schedule geometry workloads for the plurality of shader engines in a first mode. The first FE circuit is configured to schedule geometry workloads for a first subset of the plurality of shader engines and the one or more second FE circuits are configured to schedule geometry workloads for a second subset of the plurality of shader engines in a second mode. In some cases, a partition switch is configured to selectively connect the first FE circuit or the one or more second FE circuits to the second subset of the plurality of shader engines depending on whether the apparatus is in the first mode or the second mode.

Abstract: An array processor includes processor element arrays distributed in rows and columns. The processor element arrays perform operations on parameter values. The array processor also includes memory interfaces that are dynamically mapped to mutually exclusive subsets of the rows and columns of the processor element arrays based on dimensions of matrices that provide the parameter values to the processor element arrays. In some cases, the processor element arrays are vector arithmetic logic unit (ALU) processors and the memory interfaces are direct memory access (DMA) engines. The rows of the processor element arrays in the subsets are mutually exclusive to the rows in the other subsets and the columns of the processor element arrays in the subsets are mutually exclusive to the columns in the other subsets. The matrices can be symmetric or asymmetric, e.g., one of the matrices can be a vector having a single column.