Methods and systems are disclosed to aggregate traffic from multiple server devices through a peripheral component interconnect (PCI) hosting device. In one embodiment, the PCI hosting device comprises a network interface to couple the PCI hosting device to a network, a plurality of PCI interfaces, a processing circuit to forward packets, and a power supply to supply power to the PCI interfaces independently from the plurality of server devices. Each of the PCI interfaces is designed to be coupled to one server device to the PCI hosting device, which is registered as a first PCI board of a first server device through a first PCI interface and as a second PCI board of a second server device through a second PCI interface, and the PCI hosting device is designed to forward packets between the network interface and the first server device, and the network interface and the second server device.

A computer storage device having a host interface, a controller, non-volatile storage media, and firmware. The firmware instructs the controller to: receive, via the host interface, a request from a host to allocate a namespace of a quantity of non-volatile memory; generate, in response to the request, a namespace map identifying a plurality of blocks of addresses having a same predetermined block size, and a partial block of addresses having a size smaller than the predetermined block size; and convert, using the namespace map, logical addresses in the namespace communicated from the host to physical addresses for the quantity of the non-volatile memory. For example, the request for allocating the namespace can be in accordance with an NVMe protocol.

High level synthesis (HLS) begins with high-level specification of a problem, where behavior is generally decoupled from e.g., clock-level timing. Programming code can be run and debugged during functional simulation using debugging techniques. However, it is not possible to understand execution flow of register transfer level instructions (RTL) generated during RTL debug. Conventionally, it is challenging and not possible due to nature of debugging techniques which ignore printf statements in code for invocation. Systems and methods of present disclosure synthesize printf and/or scanf statements for generating debug messages in HLS code, wherein printf and/or scanf statements is/are included before/after function(s) in sections comprising instructions in code and synthesized as a block during run-time which communicate with host system and debug messages are generated for display on screen. This enables traceability of the code execution on the screen and printf/scanf statements output can be observed without any challenges.

Aspects of the present disclosure are directed to performing varying frequency memory sub-system background scans using either or both a timer and an I/O event limit. This can be accomplished by identifying a background scan trigger event from one of multiple possible types of background scan trigger events, such as a timer expiration or reaching an event count limit. In response to the background scan trigger event, a background scan can be initiated on a memory portion. The background scan can produce results, such as CDF-based data. When a metric based on the results exceeds a background scan limit, a refresh relocation can be performed and logged. A metric can be generated based on the CDF-based data, obtained error recovery depth data, or refresh relocation event data. When the metric is above or below corresponding background scan thresholds, a background scan frequency can be adjusted.

Aspects of the embodiments are directed to systems, methods, and computer program products that facilitate a downstream port to operate in Separate Reference Clocks with Independent Spread Spectrum Clocking (SSC) (SRIS) mode. The system can determine that the downstream port supports one or more SRIS selection mechanisms; determine a system clock configuration from the downstream port to a corresponding upstream port connected to the downstream port by the PCIe-compliant link; set an SRIS mode in the downstream port; and transmit data across the link from the downstream port using the determined system clock configuration.

In example implementations, an apparatus is provided. The apparatus includes a riser card body, a first interface, a first 2×8 slot on a surface of the riser card body, and a second 2×8 slot on a same side of the surface of the riser card body as the first 2×8 slot. The first interface includes a first set of fingers and a second set of fingers at an end of the riser card body to connect to a peripheral component interconnect e×press (PC1e) slot of a motherboard. The first 2×8 slot and the second 2×8 slot are positioned perpendicular to the PCIe slot of the motherboard.

The present disclosure provides a processing device including: a coarse-grained pruning unit configured to perform coarse-grained pruning on a weight of a neural network to obtain a pruned weight, an operation unit configured to train the neural network according to the pruned weight. The coarse-grained pruning unit is specifically configured to select M weights from the weights of the neural network through a sliding window, and when the M weights meet a preset condition, all or part of the M weights may be set to 0. The processing device can reduce the memory access while reducing the amount of computation, thereby obtaining an acceleration ratio and reducing energy consumption.

The invention provides a transaction layer circuit of a PCIe. The transaction layer circuit includes transaction layer processing channels, a channel selection circuit, and a merge circuit. The transaction layer processing channels are coupled to a data bus transmitting at least one packet data output by a data link layer circuit of the PCIe. The channel selection circuit receives packet start/end location information in a current clock cycle from the data link layer circuit, and distributes at least one packet data in the current clock cycle to at least one transaction layer processing channel according to the packet start/end location information. The merge circuit is coupled to the transaction layer processing channels and selectively merges transaction layer processing results output by the transaction layer processing channels based on the distribution of the packet data in the current clock cycle to the transaction layer processing channels via the channel selection circuit.

An apparatus is provided. The apparatus comprises interface circuitry, machine-readable instructions, and processing circuitry to execute the machine-readable instructions to determine that a first composite link of a plurality of composite PCIe links terminating at the same PCIe root port lacks support for enabling a desired power saving state or an exit latency for the first composite link is above a first latency threshold. The processing circuitry is further configured to determine whether an exit latency for a second composite link of the plurality of composite PCIe links is below a second latency threshold and selectively trigger at least one sub-link of the second composite link to enable the desired power saving state if the exit latency for the second composite link is below the second latency threshold.

The present disclosure is directed to enabling operation of a field programmable gate array (FPGA) while preventing application quiescence during FPGA reconfiguration. In embodiments of the disclosure, proxy agent firmware may enable downstream transactions (e.g., PCIe transactions) to be serviced during reconfiguration of the FPGA. Programmable logic states (e.g., PCIe configuration states or memory-mapped input/output (MMIO) states) are saved in memory and maintained by the proxy agent (via a management controller running the proxy agent). Once the FPGA is reconfigured, the state may be restored to the FPGA's programmable logic, and the FPGA may operate on the current state of the transactions.