Offloading computations to multiple GPUs is not an easy task. It requires decomposing data, distributing computations and handling communication manually. GPU drop-in libraries (which require no program rewrite) have made it easy to offload computations to multiple GPUs by hiding this complexity inside library calls. Such encapsulation prevents the reuse of data between successive kernel invocations resulting in redundant communication. This limitation exists in multi-GPU libraries like CUBLASXT.

Offloading computations to multiple GPUs is not an easy task. It requires decomposing data, distributing computations and handling communication manually. Drop-in GPU libraries have made it easy to offload computations to multiple GPUs by hiding this complexity inside library calls. Such encapsulation prevents the reuse of the data between successive kernel invocations resulting in redundant communication. This limitation exists in multi-GPU libraries like CUBLASXT. In this paper, we introduce SemCache++, a semantics-aware GPU cache that automatically manages communication between the CPU and multiple GPUs in addition to optimizing communication by eliminating redundant transfers using caching. SemCache++ is used to build the first multi-GPU drop-in replacement library that (a) uses the virtual memory to automatically manage and optimize multi-GPU communication and (b) requires no program rewriting or annotations. Our caching technique is efficient; it uses a two level caching directory to track matrices and sub-matrices. Experimental results show that our system can eliminate redundant communication and deliver significant performance improvements over multi-GPU libraries like CUBLASXT.

Recently, GPU libraries have made it easy to improve application performance by offloading computation to the GPU. However, using such libraries introduces the complexity of manually handling explicit data movements between GPU and CPU memory spaces. Unfortunately, when using these libraries with complex applications, it is very difficult to optimize CPU-GPU communication between multiple kernel invocations to avoid redundant communication.

This paper presents a distributed system for remote monitoring of vehicle diagnostics and geographical position. This is achieved by using on-board microcomputer system, called on-board smart box (OBSB), general packet radio service (GPRS) and a remote server. The OBSB which is equipped with an integrated global positioning system (GPS) receiver is empowered by a software application that manages the processes of local data acquisition and transmission of the acquired data to the remote server via GPRS. When programmed with speed limits in a certain geographical region, the OBSB allows the traffic control authority to supervise violations of speed limits from inside vehicles rather than outside supervision via certain check points. Appropriate vocal and text warning messages are issued when a vehicle exceeds the permitted speed limit at a certain location. A prototype system is designed and implemented with a small number of sensors. On-road experiments have demonstrated the robustness, efficiency and applicability of the proposed system.