Search for: All records

In recent years, we have been enhancing and updating gem5’s GPU support, including enhanced gem5’s GPU support to enable running ML workloads. Moreover, we created, validated, and released a Docker image with the proper software and libraries needed to run AMD’s GCN3 and Vega GPU models in gem5. With this container, users can run the gem5 GPU model, as well as build the ROCm applications that they want to run in the GPU model, out of the box without needing to properly install the appropriate ROCm software and libraries. Additionally, we updated gem5 to make it easier to reproduce results, including releasing support for a number of GPU workloads in gem5-resources and enabling continuous integration testing for a variety of GPU workloads. Current gem5 support focuses on Carrizo- and Vega-class GPUs. Unfortunately, these models do not always provide high accuracy relative to the equivalent ”real” GPUs. This leads to a mismatch in expectations: when prototyping new optimizations in gem5 users may draw the wrong conclusions about the efficacy of proposed optimizations if gem5’s GPU models do not provide high fidelity. Accordingly, to help bridge this divide, we design a series of micro-benchmarks designed expose the latencies, bandwidths, and sizes of a variety of GPU components on real GPUs. By iteratively applying fixes and improvements to gem’s GPU model, we significantly improve its fidelity relative to real AMD GPUs. 

With the waning of Moore’s Law and the end of Dennard’s Scaling, systems are turning towards heterogeneity, mixing conventional cores and specialized accelerators to continue scaling performance and energy efficiency. Specialized accelerators are frequently used to improve the efficiency of computations that run inefficiently on conventional, general-purpose processors. As a result, systems ranging from smartphones to data-centers, hyper-scalars, and supercomputers are increasingly using large numbers of accelerators to provide better efficiency than CPU-based solutions. However, heterogeneous systems face key challenges: changes to the underlying technology which threaten continued scaling, as well as the voracious scaling from applications, which require additional research to address. Traditionally, simulators could be used to perform early exploration for this research. However, existing simulators lack important support for these key challenges. Detailed simulation of modern systems can take extremely long times in existing tools and infrastructure. Furthermore, prototyping optimizations at scale can also be challenging, especially for newly proposed accelerators. Although other simulators such as Accel-Sim, SCALE-Sim, and Gemmini enable some early experiments, they are limited in their ability to target a wide variety of accelerators. In comparison, gem5 has support for various CPUs, GPUs, DSPs, and many other important accelerators. However, efficiently simulating large-scale workloads on gem5’s cycle-level models requires prohibitively long times. We aim to enhance gem5’s support to make running these workloads practical while retaining accuracy.