Abstract: An approach is provided for processing near-memory processing commands, e.g., PIM commands, using PIM register definition data that defines multiple combinations of source and/or destination registers to be used to process PIM commands. A particular combination of source and/or destination registers to be used to process a PIM command is specified by the PIM command or determined by a near-memory processing element processing the PIM command. According to another implementation, the PIM register definition data specifies an initial combination of source and/or destination registers and one or more update functions for each PIM command. A near-memory processing element processes a PIM command using the initial combination of source and/or destination registers and uses the one or more update functions to update the combination of source and/or destination registers to be used the next time the PIM command is processed.

Abstract: Systems and methods related to priority-based and performance-based selection of a render mode, such as a two-level binning mode, in which to execute workloads with a graphics processing unit (GPU) of a system are provided. A user mode driver (UMD) or kernel mode driver (KMD) executed at a central processing unit (CPU) configures low and medium priority workloads to be executed in a two-level binning mode and selects a binning mode for high priority workloads based on whether performance heuristics indicate that one or more binning conditions or override conditions have been met. High priority workloads are maintained in a high priority queue, while low and medium priority workloads are maintained in a low/medium priority queue, such that execution of low and medium priority workloads at the GPU can be preempted in favor of executing high priority workloads.

Abstract: Graphics processing unit (GPU) selection based on a utilized power source, including: determining that an apparatus is using a direct current (DC) power source instead of an Alternating Current (AC) power source; and causing, in response to the apparatus using the DC power source, the apparatus to preferentially utilize an integrated graphics processing unit (iGPU) over a discrete graphics processing unit (dGPU) while using the DC power source.

Abstract: Enhanced methods for memory context restore are described. A device may include a physical layer (PHY) having an interface to support communication of command signals and data with a physical memory. The PHY implements a training mode to train the interface, detect values of a plurality of parameters as part of training the interface, and store the detected values as initial training data. The PHY also implements a retraining mode to use the initial training data as seed data to retrain the interface.

Abstract: An accelerated processing device is provided which comprises a plurality of compute units each including a plurality of SIMD units, and each SIMD unit comprises a register file. The accelerated processing device also comprises LDS in communication with each of the SIMD units. The accelerated processing device also comprises a first portion of cache memory, in communication with each of the SIMD units and a second cache portion of memory shared by the compute units. The compute units are configured to execute a program in which a storage portion of at least one of the register file of a SIMD unit, the first portion of cache memory and the LDS is reserved as part of another of the register file, the first portion of cache memory and the LDS.

Abstract: A method and apparatus for integrating data compression in a computer system includes receiving first data at a first system level. Based upon a number of planes of the first data being less than or equal to a threshold, the data is compressed with a first data compression scheme, and transferred to a second system level for processing. Based upon the number of planes of the first data exceeding the threshold, the first data is transferred uncompressed to the second system level for processing. Based upon the received data at the second system level being compressed with the first compression scheme, the data is transferred to a third system level, and based upon the received data at the second system level being uncompressed with the first compression scheme, compressing the data with a second compression scheme, and transferring the compressed data to the third system level.

Abstract: Selectively bypassing cache directory lookups for processing-in-memory instructions is described. In one example, a system maintains information describing a status—clean or dirty—of a memory address, where a dirty status indicates that the memory address is modified in a cache and thus different than the memory address as represented in system memory. A processing-in-memory request involving the memory address is assigned a cache directory bypass bit based on the status of the memory address. The cache directory bypass bit for a processing-in-memory request controls whether a cache directory lookup is performed after the processing-in-memory request is issued by a processor core and before the processing-in-memory request is executed by a processing-in-memory component.

Abstract: A semiconductor module includes two or more semiconductor dies and an interconnect structure coupled to the two or more semiconductor dies. The interconnect structure implements a plurality of die-to-die connection pathways having a first density and a plurality of fan-out redistribution pathways having a second density that is different from the first density.

Abstract: Systems, apparatuses, and methods for achieving higher cache hit rates for machine learning models are disclosed. When a processor executes a given layer of a machine learning model, the processor generates and stores activation data in a cache subsystem a forward or reverse manner. Typically, the entirety of the activation data does not fit in the cache subsystem. The processor records the order in which activation data is generated for the given layer. Next, when the processor initiates execution of a subsequent layer of the machine learning model, the processor processes the previous layer's activation data in a reverse order from how the activation data was generated. In this way, the processor alternates how the layers of the machine learning model process data by either starting from the front end or starting from the back end of the array.

Abstract: A technique for manipulating a generic tensor is provided. The technique includes receiving a first request to perform a first operation on a generic tensor descriptor associated with the generic tensor, responsive to the first request, performing the first operation on the generic tensor descriptor, receiving a second request to perform a second operation on generic tensor raw data associated with the generic tensor, and responsive to the second request, performing the second operation on the generic tensor raw data, the performing the second operation including mapping a tensor coordinate specified by the second request to a memory address, the mapping including evaluating a delta function to determine an address delta value to add to a previously determined address for a previously processed tensor coordinate.

Abstract: A data processor that is operable to be coupled to a memory includes a memory operation array, a controller, a refresh logic circuit, and a selector. The memory operation array is for storing memory operations for a first power state of the memory. The controller is responsive to a power state change request to execute a plurality of memory operations from the memory operation array when the first power state is selected. The refresh logic circuit generates refresh cycles periodically for the memory. The selector is for multiplexing the refresh cycles with the memory operations during a power state change to the first power state.

Abstract: A method, apparatus and computer readable medium that use of a lightweight finite state machine (FSM) control flow block to enable limited execution of data-dependent control flow, thereby enhancing the control flow flexibility of array scale SIMD processors. In certain cases, the FSM block contains registers responsible for decoding and managing single global instructions into multiple local instructions that can incorporate data-dependent control flow.

Abstract: The disclosed computer-implemented method for video encoding rate control can include governing, by at least one processor, a video encoding rate at least partly in response to video encoding quality information. The method can additionally include generating, by the at least one processor, an encoded video data bitstream based on input pixel data and according to the video encoding rate. The method can also include determining, by the at least one processor, the video encoding quality information based on reconstructed pixel data. Various other methods, systems, and computer-readable media are also disclosed.

Abstract: Devices and methods for rendering objects using ray tracing are provided which include generating a low resolution version of a high resolution mesh representing objects in the scene, determining points on curved surfaces of curved surface patches defined for one of triangles and bi-linear quadrangles of the low resolution version of the high resolution mesh, performing ray intersection testing by casting rays toward surfaces of the high resolution mesh which are approximated from new points calculated by offset values along interpolated normals from the points on the curved surfaces of the curved surface patches and rendering the objects in the scene based on the ray intersection testing.

Abstract: A technique for generating video is provided. The technique includes obtaining a plurality of source frames with a wide-angle camera and a narrow-angle camera; identifying a plurality of central portions and a plurality of peripheral portions of the plurality of source frames based on image stabilization; and combining the plurality of central portions and the plurality of peripheral portions to generate a plurality of resulting frames of an output video.

Abstract: The disclosed computing device can include host circuitry configured to provide a physical function and guest circuitry configured to provide a virtual function. The host circuitry is configured to dynamically assign request identifiers for accessing at least the host circuitry in a manner that allows the request identifiers to change on a command-to-command basis instead of a time-to-time basis that uses fixed value request identifiers in time slices. Various other methods, systems, and computer-readable media are also disclosed.

Abstract: Graph analytics system are described. In accordance with the described techniques, a graph having vertices that include a first vertex and a second vertex that are associated with access control metadata are received. An updated graph is output based on a merging of the first vertex and the second vertex into a merged vertex of a group of vertices based on the first vertex and the second vertex being associated with access control metadata common to the first vertex and the second vertex and based on a reordering technique. A single copy of the access control metadata is stored for the first vertex and the second vertex.

Abstract: A computer-implemented method for data communication bus address sharing can include selecting, by at least one processor, one of two or more peripheral devices sharing an address of a data communication bus. The method can additionally include modulating, by the at least one processor, a duty cycle of a clock signal transmitted over the data communication bus to the two or more peripheral devices, wherein the modulating causes a low period of the clock signal to satisfy a threshold condition for indicating selection of the selected one of the two or more peripheral devices. The method can also include performing, by the at least one processor using the address, data communication over the data communication bus with the selected one of the two or more peripheral devices. Various other methods, systems, and computer-readable media are also disclosed.

Abstract: A computer-implemented method for resynchronization at execution time can include detecting, by at least one processor and during an execution time of an instruction, a resynchronization. The method can additionally include regenerating, by the at least one processor and in response to the detection, an instruction pointer. The method can also include performing, by the at least one processor and during the execution time of the instruction, the resynchronization by using the regenerated instruction pointer. Various other methods and systems are also disclosed.

Abstract: A disclosed technique includes based on a clock pattern, determining an enable configuration for setting enable signals for one or more multi-cycle paths of a hardware logic network; controlling a selector to set the enable configuration for the one or more multi-cycle paths; and executing testing operations for the hardware logic network with the one or more multi-cycle paths enabled according to the enable configuration.