Altis

Altis是由德克萨斯大学奥斯汀分校开发的现代GPGPU计算基准套件，旨在解决现有基准套件如Rodinia和SHOC未能跟上硬件和编程框架发展的问题。该数据集专注于CUDA平台，通过整合和扩展Rodinia和SHOC中的应用，并添加新的应用，以更好地代表现代GPGPU工作负载，支持GPGPU在架构和系统软件方面的研究。Altis特别强调了神经网络等新兴应用领域，通过引入常用DNN内核来丰富其内容。此外，Altis还支持最新的CUDA特性，如统一内存、HyperQ和动态并行，以适应不断发展的GPU计算需求。

Altis is a modern general-purpose GPU (GPGPU) computing benchmark suite developed by The University of Texas at Austin, designed to address the shortcomings of existing benchmark suites such as Rodinia and SHOC that fail to keep pace with the evolution of hardware and programming frameworks. Focused on the CUDA platform, this suite integrates and extends applications from Rodinia and SHOC while adding new applications to better represent modern GPGPU workloads, thereby supporting research on GPGPU architecture and system software. Altis places particular emphasis on emerging application domains such as neural networks, enriching its content by incorporating commonly used DNN kernels. Furthermore, Altis supports the latest CUDA features including Unified Memory, HyperQ, and Dynamic Parallelism to accommodate the evolving demands of GPU computing.