An apparatus includes a first processor to execute a user-level application to operate in a virtual address, and a co-processor to execute a computing kernel associated with user-level application elements to be performed on the co-processor. The computing kernel is to operate in the virtual address. A memory includes physical addresses, and a partition used to map the virtual address associated with the first processor and to map the virtual address associated with the co-processor. A packet processor manages communications between the first processor and the co-processor. The packet processor receives packets from the first processor, with the packets including memory addresses identifying code and data of the computing kernel. The packet processor stores the packets in a queue associated with the user-level application, and outputs the packets to the co-processor, such that the co-processor is enabled to execute the computing kernel.