TU Darmstadt / ULB / TUprints

In-Situ Profiling Feedback for GPGPU Code

Pauli, Armin ; Buelow, Max von ; Ströter, Daniel (2024)
In-Situ Profiling Feedback for GPGPU Code.
doi: 10.26083/tuprints-00027344
Report, Primary publication, Preprint

[img] Text
Copyright Information: CC BY 4.0 International - Creative Commons, Attribution.

Download (176kB)
Item Type: Report
Type of entry: Primary publication
Title: In-Situ Profiling Feedback for GPGPU Code
Language: English
Date: 10 May 2024
Place of Publication: Darmstadt
Collation: 2 ungezählte Seiten
DOI: 10.26083/tuprints-00027344

The evolving landscape of software development increasingly prioritizes functionality, maintainability, and developer productivity. This typically comes hand in hand with the shortcoming that less focus is invested on optimizing for runtime performance of programs. However, optimizing for performance is an important task in time-critical domains. Additionally, optimizing for performance can be an important way of reducing actual hardware requirements and achieving a better ecological footprint. So, why not bringing program optimization closer to the software engineer and reducing the disconnect between profiling results and their interpretability? This poster presents a GPU-focused in-situ profiling approach that visualizes memory profiling metrics directly inside the source code and gives the software engineer an direct hint for identifying inefficient parts during development. Performance metrics evaluated on each line are highlighted in the source code.

Status: Preprint
URN: urn:nbn:de:tuda-tuprints-273445
Classification DDC: 000 Generalities, computers, information > 004 Computer science
Divisions: 20 Department of Computer Science > Interactive Graphics Systems
Date Deposited: 10 May 2024 12:51
Last Modified: 17 May 2024 07:54
URI: https://tuprints.ulb.tu-darmstadt.de/id/eprint/27344
PPN: 518191664
Actions (login required)
View Item View Item