Abstract
The management of separate memory spaces of CPUs and GPUs brings an additional burden to the development of software for GPUs. To help with this, CUDA unified memory provides a single address space that can be accessed from both CPU and GPU. The automatic data transfer mechanism is based on page faults generated by the memory accesses. This mechanism has a performance cost, that can be with explicit memory prefetch requests. Various hints on the inteded usage of the memory regions can also be given to further improve the performance. The overall effect of unified memory compared to an explicit memory management can depend heavily on the application. In this paper we evaluate the performance impact of CUDA unified memory using the heterogeneous pixel reconstruction code from the CMS experiment as a realistic use case of a GPU-targeting HEP reconstruction software. We also compare the programming model using CUDA unified memory to the explicit management of separate CPU and GPU memory spaces.
Highlights
Graphics Processing Units (GPUs) are commonly used to accelerate scientific computing because of their cost and power efficiency in solving many data-parallel problems
The algorithms are organized in five CMS data processing software (CMSSW) framework modules, depicted in Figure 1 as a directed acyclic graph (DAG) by their data dependencies, that communicate the intermediate data in the device memory through the CMSSW event data
We identify two use cases where programming would be simpler than with explicit memory: for data to be transferred to many devices, and for data structures that heavily use pointers to refer to other locations in the unified memory space
Summary
Graphics Processing Units (GPUs) are commonly used to accelerate scientific computing because of their cost and power efficiency in solving many data-parallel problems. Their programming model introduces a concept of separate memory spaces between the host (CPU) and devices (GPUs). We evaluate the performance impact of the CUDA unified memory compared to manage the separate host and device memory spaces explicitly.
Sign-in/Register to view highlights & summary Sign-in/Register to access full text options 
Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have
Schedule a callDisclaimer: All third-party content on this website/platform is and will remain the property of their respective owners and is provided on "as is" basis without any warranties, express or implied. Use of third-party content does not indicate any affiliation, sponsorship with or endorsement by them. Any references to third-party content is to identify the corresponding services and shall be considered fair use under The CopyrightLaw.
