Engineering Papers⌕ Search

Engineering topics

Wong, Daniel

Publications and source records attributed to Wong, Daniel.

LC-MEMENTO: A Memory Model for Accelerated Architectures

With the advent of heterogeneous architectures, in particular, with the ubiquity of multi-GPU systems, it is becoming increasingly important to manage device memory efficiently in order to reap the benefits of the additional core count. To date, such responsibility mainly falls on the programmer where device-to-host data communication (and vice versa), if not done properly, may incur costly memory transfer operations and synchronization. The problem may be compounded by additional requirement to maintain system-wide memory consistency that may involve expensive synchronization overhead. In this paper, we present Location Consistency Memory Model for Enhanced Transfer Operations (LC-MEMENTO). This framework considers incorporating runtime techniques for multi-GPU memory management to support relaxed synchronization semantics and memory transfer operations automatically. Specifically, we implement a relaxed form of a memory consistency model based on the Location Consistency (LC) in an Asynchronous Many-Task Runtime (ARTS) and demonstrate that, this memory model enables additional optimization opportunities for the three representative applications encompassing different computational patterns (scientific computation, graphs, data streaming, etc.).

Memory Models, Accelerators, Adaptive Optimization↗

MAPA: Multi-Accelerator Pattern Allocation Policy for Multi-Tenant GPU Servers

Multi-accelerator servers are increasingly being deployed in shared multi-tenant environments (such as in cloud data centers) in order to meet the demands of large-scale compute-intensive workloads. In addition, these accelerators are increasingly being inter-connected in complex topologies and workloads are exhibiting a wider variety of inter-accelerator communication patterns. However, existing al-location policies are ill-suited for these emerging use-cases. Specifically, this work identifies that multi-accelerator workloads are commonly fragmented leading to reduced bandwidth and increased latency for inter-accelerator communication. We propose Multi-Accelerator Pattern Allocation (MAPA), a graph pattern mining approach towards providing generalized allocation support for allocating multi-accelerator workloads on multi-accelerator servers. We demonstrate that MAPA is able to improve the execution time of multi-accelerator workloads and that MAPA is able to provide generalized benefits across various accelerator topologies. Finally, we demonstrate a speedup of 12.4% for75th percentile of jobs with the worst case execution time reduced by up to 35% against baseline policy using MAPA.

Ranganath, Kiran↗

WIREFRAME: Supporting Data-dependent Parallelism through Dependency Graph Execution in GPUs

GPUs lack fundamental support for data-dependent parallelism and synchronization. While CUDA Dynamic Parallelism signals progress in this direction, many limitations and challenges still re-main. This paper introducesWireframe, a hardware-software solution that enables generalized support for data-dependent parallelism and synchronization. Wireframe enables applications to naturally express execution dependencies across different thread blocks through a dependency graph abstraction at run-time, which is sent to the GPU hardware at kernel launch. At run-time, the hardware enforces the dependencies specified in the dependency graph through a dependency-aware thread block scheduler. Overall, Wireframe is able to improve total execution time up to 65.20% with an average of 45.07%.

Abdolrashidi, AmirAli↗

Sci-Thur PM - Colourful Interactions: Highlights 06: Patient-specific calibration of cone-beam computed tomographic (CBCT) image sets for on-line dose tracking and plan assessment

Purpose: To develop a radiotherapy dose tracking and plan evaluation technique using cone-beam computed tomography (CBCT) images. Methods: We developed a patient-specific method of calibrating CBCT image sets for dose calculation. The planning CT was first registered with the CBCT using deformable image registration (DIR). A scatter plot was generated between the CT numbers of the planning CT and CBCT for each slice. The CBCT calibration curve was obtained by least-square fitting of the data, and applied to each CBCT slice. The calibrated CBCT was then merged with original planning CT to extend the small field of view of CBCT. Finally, the treatment plan was copied to the merged CT for dose tracking and plan evaluation. The proposed patient-specific calibration method was also compared to two methods proposed in literature. To evaluate the accuracy of each technique, 15 head-and-neck patients requiring plan adaptation were arbitrarily selected from our institution. The original plan was calculated on each method’s data set, including a second planning CT acquired within 48 hours of the CBCT (serving as gold standard). Clinically relevant dose metrics and 3D gamma analysis of dose distributions were compared between the different techniques. Results: Compared to the gold standard of using planning CTs, the patient-specific CBCT calibration method was shown to provide promising results with gamma pass rates above 95% and average dose metric agreement within 2.5%. Conclusions: The patient-specific CBCT calibration method could potentially be used for on-line dose tracking and plan evaluation, without requiring a re-planning CT session.

60 APPLIED LIFE SCIENCES↗

LULESH Programming Model and Performance Ports Overview

This document gives a description of various versions of the LULESH proxy application described in [1]. All the the codes described in this document are available at https://codesign.llnl. gov/lulesh.php. This report describes versions that fall into three categories: performance tuning exploration, programming model exploration and other versions. In this document we focus on giving a brief overview of what makes each version different from the others. The purpose is to allow others to reuse our work as a starting point for their exploration. Therefore, references to publications and presentations are provided to allow those who want more detailed information to be able to find it.

97 MATHEMATICS AND COMPUTING↗