An example programmable integrated circuit (IC) includes a processor, a plurality of endpoint circuits, a network-on-chip (NoC) having NoC master units (NMUs), NoC slave units (NSUs), NoC programmable switches (NPSs), a plurality of registers, and a NoC programming interface (NPI). The processor is coupled to the NPI and is configured to program the NPSs by loading an image to the registers through the NPI for providing physical channels between NMUs to the NSUs and providing data paths between the plurality of endpoint circuits.

Methods and systems for facilitating an equitable bandwidth distribution across downstream devices in asymmetrical switch topologies, and in particular asymmetrical PCIe switch topologies. The equitable distribution of bandwidth is achieved in asymmetrical topologies using virtual switch partitioning. An upstream switch that is connected to the root complex via an upstream port and that receives bandwidth B from the upstream port, is virtualized into two or more virtual switches. Each virtual switch equally shares the bandwidth. Each virtual switch is allocated to downstream devices that are connected to the upstream switch as well as to one or more downstream switches that are connected to the upstream switch. Each downstream switch may be connected to one or more additional downstream devices.

In one embodiment, an apparatus includes: a transmitter to send a first plurality of flits to a second device coupled to the apparatus via a link; and a control circuit coupled to the transmitter to change a configuration of the link from a flit-based encoding to a packet-based encoding. In response to the configuration change, the transmitter is to send a first plurality of packets to the second device via the link. Other embodiments are described and claimed.

In example implementations, an apparatus is provided. The apparatus includes a riser card body, a first interface, a first 28 slot on a surface of the riser card body, and a second 28 slot on a same side of the surface of the riser card body as the first 28 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 express (PCIe) slot of a motherboard. The first 28 slot and the second 28 slot are positioned perpendicular to the PCIe slot of the motherboard.

A method for stack timing adjustment for serial communications is provided. The method includes receiving a USB communication, decoding the USB communication into UART frames, and adjusting the timing of the UART frames according to a serial protocol.

A data transmission system includes a first apparatus that is connected to a first external apparatus using the first transmission scheme and a second apparatus that is connected to a second external apparatus using the first transmission scheme. The first apparatus is connected to the second apparatus using a second transmission scheme that is different from the first transmission scheme. When the operation mode of the first apparatus in the first transmission scheme and the operation mode of the second apparatus in the first transmission scheme are different, data in the first external apparatus or the second external apparatus is transmitted to the second external apparatus or the first external apparatus, respectively, via the first apparatus and the second apparatus.

Methods, systems, and computer program products are described herein an extensible input stack for processing input device data received from a plurality of different input devices attached to a computing device. The extensible input stack comprises a plurality of stack layers. Each of the plurality of stack layers performs a particular set of processing with respect to the input device data, among other operations. Each of the plurality of stack layers comprises a code interface, which is used to provide and/or or receive data from the input device and/or other stack layers. Each of the stack layers is extensible to include additional functionality to support new input devices. By separating out the functionality performed by the input stack into separate stack layers, and having each layer accessible via a code interface, the functionality of each of stack layers may be easily extended to support any type of input device.

An example programmable integrated circuit (IC) includes a processor, a plurality of endpoint circuits, a network-on-chip (NoC) having NoC master units (NMUs), NoC slave units (NSUs), NoC programmable switches (NPSs), a plurality of registers, and a NoC programming interface (NPI). The processor is coupled to the NPI and is configured to program the NPSs by loading an image to the registers through the NPI for providing physical channels between NMUs to the NSUs and providing data paths between the plurality of endpoint circuits.

A solid-state drive (SSD) for handling an Asynchronous Event Request (AER) command includes a command receiving circuit and a command management circuit. The command receiving circuit is configured to receive at least one command from at least one host. The command management circuit is configured to determine if the received at least one command from the at least one host is an AER command, store the AER command into an AER queue reserved for deferred AER command handling, if the received at least one command is the AER command, and generate a dummy response for the AER command and release resources occupied by the AER command.

This disclosure relates to methods, non-transitory computer readable media, and systems that asynchronously train a machine learning model across client devices that implement local versions of the model while preserving client data privacy. To train the model across devices, in some embodiments, the disclosed systems send global parameters for a global machine learning model from a server device to client devices. A subset of the client devices uses local machine learning models corresponding to the global model and client training data to modify the global parameters. Based on those modifications, the subset of client devices sends modified parameter indicators to the server device for the server device to use in adjusting the global parameters. By utilizing the modified parameter indicators (and not client training data), in certain implementations, the disclosed systems accurately train a machine learning model without exposing training data from the client device.