Abstract: Examples described herein relate to a network interface device. In some examples, the network interface device includes a network interface, a direct memory access (DMA) circuitry, a host interface, memory, one or more processors, and circuitry to: based on a configuration of operation specifying a standalone operation, cause the network interface device to operate in standalone to execute one or more applications and based on a configuration of operation specifying a companion operation, cause the network interface device to operate in companion to provide at least one host system with access to one or more hardware resources accessible by the network interface device.

Abstract: Embodiments described herein are generally directed to improving performance of a transactional API protocol by scheduling function calls based on data dependencies. In an example, a function associated with the transactional API is received that is to be carried out by an executer on behalf of an application. It is determined whether the function has a dependency on a value that is invalid. If so, execution of the function is delayed by causing a function ID of the function to be queued for a global memory reference associated with the value. After the value becomes valid, the function is caused to be executed by the executer. When the first function is determined to have no such dependency, the function may be immediately scheduled for execution by the executer without delay.

Abstract: Various examples relate to apparatuses, devices, methods, computer systems and computer programs for handling remote procedure calls. A non-transitory, computer-readable medium comprises machine-readable instructions that, when the program code is executed on a processor of a requesting host, causes the processor to provide an interface for locally receiving remote procedure calls from a plurality of threads of a computer program, and forward, upon receiving a remote procedure call from one of the threads of the computer program, the remote procedure call to a providing host that provides the functionality associated with the remote procedure call, wherein the remote procedure call is forwarded together with information on the thread having issued the remote procedure call.

Abstract: A method of offloading performance of a workload includes receiving, on a first computing system acting as an initiator, a first function call from a caller, the first function call to be executed by an accelerator on a second computing system acting as a target, the first computing system coupled to the second computing system by a network; determining a type of the first function call; and generating a list of parameter values of the first function call.

Abstract: Embodiments described herein are generally directed to improving performance of a transactional API protocol by batch scheduling data dependent functions. In an example, a prescribed sequence of function calls associated with a transactional application programming interface (API) is received that is to be carried out by an executer (e.g., a compute service or a second processing resource remote from a first processing resource with which an application is associated) to perform an atomic unit of work on behalf of the application. Transport latency over an interconnect between the application and the executer is reduced by: (i) creating a batch representing the prescribed sequence of function calls in a form of a list of function descriptors in which variable arguments of the prescribed sequence of function calls are replaced with corresponding global memory references; and (ii) transmitting the batch via the interconnect as a single message.

Abstract: A method of offloading performance of a workload includes receiving, on a first computing system acting as an initiator, a first function call from a caller, the first function call to be executed by an accelerator on a second computing system acting as a target, the first computing system coupled to the second computing system by a network; determining a type of the first function call; and generating a list of parameter values of the first function call.

Abstract: A method of offloading performance of a workload includes receiving, on a first computing system acting as an initiator, a first function call from a caller, the first function call to be executed by an accelerator on a second computing system acting as a target, the first computing system coupled to the second computing system by a network; determining a type of the first function call; and generating a list of parameter values of the first function call.