skip to main content
US FlagAn official website of the United States government
dot gov icon
Official websites use .gov
A .gov website belongs to an official government organization in the United States.
https lock icon
Secure .gov websites use HTTPS
A lock ( lock ) or https:// means you've safely connected to the .gov website. Share sensitive information only on official, secure websites.

Explore Research Products in the PAR
It may take a few hours for recently added research products to appear in PAR search results.


Title: Design, Optimization, and Benchmarking of Dense Linear Algebra Algorithms on AMD GPUs
Dense linear algebra (DLA) has historically been in the vanguard of software that must be adapted first to hardware changes. This is because DLA is both critical to the accuracy and performance of so many different types of applications, and because they have proved to be outstanding vehicles for finding and implementing solutions to the problems that novel architectures pose. Therefore, in this paper we investigate the portability of the MAGMA DLA library to the latest AMD GPUs.We use auto tools to convert the CUDA code in MAGMA to the Heterogeneous-Computing Interface for Portability (HIP) language. MAGMA provides LAPACK for GPUs and benchmarks for fundamental DLA routines ranging from BLAS to dense factorizations, linear systems and eigen-problem solvers. We port these routines to HIP and quantify currently achievable performance through the MAGMA benchmarks for the main workload algorithms on MI25 and MI50 AMD GPUs. Comparison with performance roofline models and theoretical expectations are used to identify current limitations and directions for future improvements.  more » « less
Award ID(s):
1740250
PAR ID:
10289436
Author(s) / Creator(s):
; ; ;
Date Published:
Journal Name:
2020 IEEE High Performance Extreme Computing Conference (HPEC)
Page Range / eLocation ID:
1 to 7
Format(s):
Medium: X
Sponsoring Org:
National Science Foundation
More Like this
  1. ; ; ; ; ; (, ACM Transactions on Parallel Computing)
    Due to the recent announcement of the Frontier supercomputer, many scientific application developers are working to make their applications compatible with AMD (CPU-GPU) architectures, which means moving away from the traditional CPU and NVIDIA-GPU systems. Due to the current limitations of profiling tools for AMD GPUs, this shift leaves a void in how to measure application performance on AMD GPUs. In this article, we design an instruction roofline model for AMD GPUs using AMD’s ROCProfiler and a benchmarking tool, BabelStream (the HIP implementation), as a way to measure an application’s performance in instructions and memory transactions on new AMD hardware. Specifically, we create instruction roofline models for a case study scientific application, PIConGPU, an open source particle-in-cell simulations application used for plasma and laser-plasma physics on the NVIDIA V100, AMD Radeon Instinct MI60, and AMD Instinct MI100 GPUs. When looking at the performance of multiple kernels of interest in PIConGPU we find that although the AMD MI100 GPU achieves a similar, or better, execution time compared to the NVIDIA V100 GPU, profiling tool differences make comparing performance of these two architectures hard. When looking at execution time, GIPS, and instruction intensity, the AMD MI60 achieves the worst performance out of the three GPUs used in this work. 
    more » « less
  2. ; ; ; ; ; (, Annual MVAPICH User Group (MUG) Conference)
    AWP-ODC is a 4th-order finite difference code used for linear wave propagation, Iwan-type nonlinear dynamic rupture and wave propagation, and Strain Green Tensor simulation2. We have ported and verified the linear and topography version of AWP-ODC, with discontinuous mesh as well as topography, to HIP so that it can also run on AMD GPUs. The topography code achieved a 99.6% parallel efficiency on 4,096 nodes on Frontier, a Leadership Computing Facility at Oak Ridge National Laboratory. We have also implemented CUDA-aware features and on-the-fly GDR compression in the linear version of the ported HIP code. These enhancements significantly improve data transfer efficiency between GPUs, reducing communication overhead and boosting overall performance. We have also extended CUDA-aware features to the topography version and are actively working on incorporating GDR compression into this version as well. We see 154% benefits over IMPI in MVAPICH2-GDR with CUDA-aware support and on-the-fly compression for linear AWP-ODC on Lonestar-6 A100 nodes. Furthermore, we have successfully integrated a checkpointing feature into the nonlinear IWAN version of AWP-ODC, prepared for future extreme-scale simulation during Texascale Days of Frontera at TACC. 
    more » « less
  3. ; ; ; ; ; ; (, SCEC Publications)
    We have ported and verified the topography version of AWP-ODC, with discontinuous mesh feature enabled, to HIP so that it runs on AMD MI250X GPUs. 103.3% parallel efficiency was benchmarked on Frontier between 8 and 4,096 nodes or up to 32,768 GCDs. Frontier is a two exaflop/s computing system based on the AMD Radeon Instinct GPUs and EPYC CPUs, a Leadership Computing Facility at Oak Ridge National Laboratory (ORNL). This HIP topography code has been used in the production runs on Frontier, a primary computing engine currently utilizing the 2024 SCEC INCITE allocation, a 700K node-hours supercomputing time award. Furthermore, we implemented ROCm-Aware GPU direct support in the topo code, and demonstrated 14% additional reduction in time-to-solution up to 4,096 nodes. The AWP-ODC-Topo code is also tuned on TACC Vista, an Arm-based NVIDIA GH200 Grace Hopper Superchip, with excellent performance demonstrated. This poster will demonstrate the studies of weak scaling and the performance characteristics on GPUs. We discuss the efforts of verifying the ROCm-Aware development, and utilizing high-performance MVAPICH libraries with the on-the-fly compression on modern GPU clusters. 
    more » « less
  4. ; (, Annual MVAPICH USER Group (MUG) Meeting)
    We integrate GPU-aware MVAPICH2 in AWP-ODC, a scalable finite difference code for wave propagation in nonlinear media. On OLCF Frontier, HIP-aware MVAPICH2 yields a 17.8% T2S improvement over the non-GPU-aware version and achieves 95% parallel efficiency on 65,536 AMD MI250X GCDs. On TACC Vista, CUDA-aware MVAPICH2 delivers a 3.5% performance gain across 2-256 Nvidia GH200 GPUs, with parallel efficiencies of 82% in the linear case and 92% in the computationally more intense nonlinear case. We deploy the code for production-scale earthquake simulations on leadership-class systems 
    more » « less
  5. ; ; ; ; ; (, Proceedings International Conference on High Performance Computing)
    In this paper, we present work towards the development of a new data analytics and machine learning (ML) framework, called MagmaDNN. Our main goal is to provide scalable, high-performance data analytics and ML solutions for scientific applications running on current and upcoming heterogeneous many-core GPU-accelerated architectures. To this end, since many of the functionalities needed are based on standard linear algebra (LA) routines, we designed MagmaDNN to derive its performance power from the MAGMA library. The close integration provides the fundamental (scalable high-performance) LA routines available in MAGMA as a backend to MagmaDNN. We present some design issues for performance and scalability that are specific to ML using Deep Neural Networks (DNN), as well as the MagmaDNN designs towards overcoming them. In particular, MagmaDNN uses well established HPC techniques from the area of dense LA, including task-based parallelization, DAG representations, scheduling, mixed-precision algorithms, asynchronous solvers, and autotuned hyperparameter optimization. We illustrate these techniques and their incorporation and use to outperform other frameworks, currently available. 
    more » « less
  • Citation Formats
  • MLA
  • APA
  • Chicago
  • BibTeX
  • Save / Share this Record
  • Send to Email