Abstract: A computing system may implement a low priority arbitration interrupt method that includes receiving a message signaled interrupt (MSI) message from an input output hub (I/O hub) transmitted over an interconnect fabric, selecting a processor to interrupt from a cluster of processors based on arbitration parameters, and communicating an interrupt service routine to the selected processor, wherein the I/O hub and the cluster of processors are located within a common domain.

Abstract: A monitoring system predicts voltage droops at a processor by monitoring one or more performance characteristics of the processor, selecting a response policy based on the prediction, and adjusting a parameter of the processor. Multiple predictions of voltage droop conditions at different locations of the processor are made simultaneously, with the processor generating one or more responses and resulting in adjusting one or more parameters of the processor. By predicting voltage droop conditions before they occur, the deleterious effects of such droop conditions can be minimized or avoided.

Abstract: A processing system selectively enables and disables a result lookaside buffer (RLB) based on a hit rate tracked by a counter, thereby reducing power consumption for lookups at the result lookaside buffer during periods of low hit rates and improving the overall hit rate for the result lookaside buffer. A controller increments the counter in the event of a hit at the RLB and decrements the counter in the event of a miss at the RLB. If the value of the counter falls below a threshold value, the processing system temporarily disables the RLB for a programmable period of time. After the period of time, the processing system re-enables the RLB and resets the counter to an initial value.

Abstract: Systems and methods for allocating computing resources within a distributed computing system are disclosed. Computing resources such as CPUs, GPUs, network cards, and memory are allocated to jobs submitted to the system by a scheduler. System configuration and interconnectivity information is gathered by a mapper and used to create a graph. Resource allocation is optimized based on one or more quality of service (QoS) levels determined for the job. Job performance characterization, affinity models, computer resource power consumption, and policies may also be used to optimize the allocation of computing resources.

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 server includes a plurality of nodes that are connected by a network that includes an on-chip network or an inter-chip network that connects the nodes. The server also includes a controller to configure the network based on relative priorities of workloads that are executing on the nodes. Configuring the network can include allocating buffers to virtual channels supported by the network based on the relative priorities of the workloads associated with the virtual channels, configuring routing tables that route the packets over the network based on the relative priorities of the workloads that generate the packets, or modifying arbitration weights to favor granting access to the virtual channels to packets generated by higher priority workloads.

Abstract: A processing system includes an interconnect fabric coupleable to a local memory and at least one compute cluster coupled to the interconnect fabric. The compute cluster includes a processor core and a cache hierarchy. The cache hierarchy has a plurality of caches and a throttle controller configured to throttle a rate of memory requests issuable by the processor core based on at least one of an access latency metric and a prefetch accuracy metric. The access latency metric represents an average access latency for memory requests for the processor core and the prefetch accuracy metric represents an accuracy of a prefetcher of a cache of the cache hierarchy.

Abstract: Various semiconductor chips and packages are disclosed. In one aspect, an apparatus is provided that includes a semiconductor chip that has a side, and plural conductive pillars on the side. Each of the conductive pillars includes a pillar portion that has an exposed shoulder facing away from the semiconductor chip. The shoulder provides a wetting surface to attract melted solder. The pillar portion has a first lateral dimension at the shoulder. A solder cap is positioned on the pillar portion. The solder cap has a second lateral dimension smaller than the first lateral dimension.

Abstract: In one form, a memory controller includes a command queue, an arbiter, and a replay queue. The command queue receives and stores memory access requests. The arbiter is coupled to the command queue for providing a sequence of memory commands to a memory channel. The replay queue stores the sequence of memory commands to the memory channel, and continues to store memory access commands that have not yet received responses from the memory channel. When a response indicates a completion of a corresponding memory command without any error, the replay queue removes the corresponding memory command without taking further action. When a response indicates a completion of the corresponding memory command with an error, the replay queue replays at least the corresponding memory command. In another form, a data processing system includes the memory controller, a memory accessing agent, and a memory system to which the memory controller is coupled.

Abstract: A method and system for performing graphics processing is provided. The method and system includes storing stencil buffer values in a stencil buffer; generating either or both of a reference value and a source value in a fragment shader; comparing the stencil buffer values against the reference value; and processing a fragment based on the comparing the stencil buffer values against the reference value.

Abstract: A read clock circuit selectively provides a read clock signal from a memory to a memory controller over a memory bus. A pulse-amplitude modulation (PAM) driver including an input and an output capable of driving at least three levels indicating respective digital values. A digital control circuit is coupled to the PAM driver and operable to cause the PAM driver to provide a preamble signal before the read clock signal, the preamble signal including an initial toggling state in which the PAM driver toggles between two selected levels at a first rate, and a final toggling state in which the PAM driver toggles between two selected levels at a second rate higher than the first rate, with a length of the initial toggling state and a length of the final toggling state are based on values in a mode register.

Abstract: Shader source code performance prediction is described. In accordance with the described techniques, an update to shader source code for implementing a shader is received. A prediction of performance of the shader on a processing unit is generated based on the update to the shader source code. Feedback about the update is output. The feedback includes the prediction of performance of the shader. In one or more implementations, generating the prediction of performance of the shader includes compiling the shader source code with the update to generate a representation of the shader, inputting the representation of the shader to one or more machine learning models, and receiving the prediction of performance of the shader as an output from the one or more machine learning models.

Abstract: A memory includes a read clock state machine and a read clock driver circuit. The read clock state machine has a first input for receiving a read command signal, a second input for receiving a read clock mode signal, and an output for providing a drive enable signal. The read clock driver circuit has an output for providing a read clock signal in response to a clock signal when the drive enable signal is active. When the read clock mode signal indicates a read-only mode, the read clock state machine starts toggling the read clock signal during a read preamble period before a data transmission of a first read command, and continues toggling the read clock signal for at least a read postamble period following the data transmission of the first read command.

Abstract: A memory controller monitors memory command selected for dispatch to the memory and sends commands controlling a read clock state. A memory includes a read clock circuit and a mode register. The read clock circuit has an output for providing a hybrid read clock signal in response to a clock signal and a read clock mode signal. The mode register provides the read clock mode signal in response to a read clock mode, wherein the read clock circuit provides the hybrid read clock signal as a free-running clock signal that toggles continuously when the read clock mode is a first mode, and as a strobe signal that is active only in response to the memory receiving a read command when the read clock mode is a second mode.

Abstract: A memory includes a read clock circuit and a mode register. The read clock circuit has an output for providing a hybrid read clock signal in response to a clock signal and a read clock mode signal. The mode register provides the read clock mode signal in response to a read clock mode, wherein the read clock circuit provides the hybrid read clock signal as a free-running clock signal that toggles continuously when the read clock mode is a first mode, and as a strobe signal that is active only in response to the memory receiving a read command when the read clock mode is a second mode.

Abstract: A high-bandwidth dual-inline memory module (HB-DIMM) includes a plurality of memory chips, a plurality of data buffer chips, and a register clock driver (RCD) circuit. The data buffer chips are coupled to respective sets of the memory chips and transmit data from the memory chips over a host bus at a data rate twice that of the memory chips. The RCD circuit includes a host bus interface and a memory interface coupled to the plurality of memory chips. The RCD circuit implements commands received over the host bus by routing command/address (C/A) signals to the memory chips for providing at least two independently addressable pseudo-channels, the RCD circuit addressing each respective pseudo-channel based on a chip identifier (CID) bit derived from the C/A signals.

Abstract: An integrated circuit product includes a redistribution layer, an integrated circuit die disposed above the redistribution layer, a row of discrete devices disposed laterally with respect to the integrated circuit die, and encapsulant mechanically coupling the redistribution layer, integrated circuit die, and the row of discrete devices. In at least one embodiment, the row of discrete devices is a row of decoupling capacitors disposed proximate to the integrated circuit die and coupled to the integrated circuit die and a power distribution network. In at least one embodiment, a second integrated circuit die is disposed above the redistribution layer and disposed laterally with respect to the integrated circuit die and the row of discrete devices. The second integrated circuit die is mechanically coupled to the redistribution layer, integrated circuit die, and the row of discrete devices and is partially surrounded by the row of discrete devices.

Abstract: A memory controller includes an arbiter for selecting memory requests from a command queue for transmission to a dynamic random access memory (DRAM) memory. The arbiter includes a bank group tracking circuit that tracks bank group numbers of three or more prior write requests selected by the arbiter. The arbiter also includes a selection circuit that selects requests to be issued from the command queue, and prevents selection of write requests and associated activate commands to the tracked bank group numbers unless no other write request is eligible in the command queue. The bank group tracking circuit indicates that a prior write request and the associated activate commands are eligible to be issued after a number of clock cycles has passed corresponding to a minimum write-to-write timing period for a bank group of the prior write request.

Abstract: Memory operations using compound memory commands, including: receiving, by a memory module, a compound memory command indicating one or more operations to be applied to each portion of a plurality of portions of contiguous memory in the memory module; generating, based on the compound memory command, a plurality of memory commands to apply the one or more operations to each portion of the plurality of portions of contiguous memory; and executing the plurality of memory commands.

Abstract: Methods, devices, and systems for prefetching data. First data is loaded from a first memory location. The first data in cached in a cache memory. Other data is prefetched to the cache memory based on a compression of the first data and a compression of the other data. In some implementations, the compression of the first data and the compression of the other data are determined based on metadata associated with the first data and metadata associated with the other data. In some implementations, the other data is prefetched to the cache memory based on a total of a compressed size of the first data and a compressed size of the other data being less than a threshold size. In some implementations, the other data is not prefetched to the cache memory based on the other data being uncompressed.