LPGPU.org

GLOVE™ (GL Over Vulkan^®), a cross-platform software library that acts as an intermediate layer between an OpenGL^® ES application and Vulkan and optimized by the LPGPU2 Profiler tool, now supports Android. GLOVE™ is focused towards embedded systems and is comprised of OpenGL ES and EGL^® implementations, which translate at runtime all OpenGL ES / EGL calls & ESSL shaders to Vulkan commands & SPIR-V shader respectively and finally relays them to the underlying Vulkan driver. GLOVE™ has been designed towards facilitating developers to easily build and integrate new features, allowing at the same time its further extension, portability and interoperability. Initially, GLOVE™ supported OpenGL ES and EGL^® on Linux and now it is further extended to support Android as well; however, its modular design can be easily extended to encompass implementations of other client APIs as well. GLOVE™ is considered as a work-in-progress and is open-sourced under the LGPL v3 license through which it is provided as free software with unlimited use for educational and research purposes. Future planned extensions of GLOVE™ include the support for OpenGL ES 3.x and OpenGL applications.

GLOVE™ is available for download from https://github.com/Think-Silicon/GLOVE. Please follow the link to view our demo and try it out on your phones by following the guidelines. We are thrilled to have your feedback!

Even though the LPGPU2 project has concluded the work continues…

Thanks to the data driven nature of the DC API and the Remote Protocol designed by Samsung for the LPGPU2 project, moving to new hardware or even a new operating system becomes much easier.

After a few short weeks of work, using the source code from the public LPGPU2 repo ( https://github.com/codeplaysoftware/LPGPU2-CodeXL ) we are able to successfully connect to a Tizen mobile device, read back counters (via DC API), and intercept API calls (thanks to the shim).

The screen shots below show captured API and counter data: (click to view larger images)

Why don’t you check out the LPGPU2 repo and see what new devices and operating systems you can enable today?

Ben Juurlink will attend the 16th Workshop on Scalable Approaches to High Performance and High Productivity Computing (ScalPerf’18: http://www.dei.unipd.it/~versacif/scalperf18) to present LPGPU2 research results. The title of his presentation is “Power Modeling of Heterogeneous Mobile SoCs using Machine Learning”. It describes a Neural Network model that predicts the power consumption of an application running on a mobile SoC from CPU and GPU performance counters. The Neural Network model achieves a mean relative error of about 4.85%, which is twice as accurate as the state of the art. The ScalPerf workshop aims at taking an integrated look at the opportunities and constraints on the road to ever higher performance and productivity of computing systems. Distinguished researchers are invited to exchange their perspectives on different areas that can contribute to scalable computing.

Prof. Ben Juurlink, the coordinator of the LPGPU2 project, has been invited to join the program committees of the International Parallel and Distributing Processing Symposium (IPDPS 2019 http://www.ipdps.org) to be held in Rio de Janeiro, Brazil and the International Conference on Architecture of Computing Systems (ARCS 2019 http://arcs2018.itec.kit.edu) to be held in Braunschweig, Germany.

Over the past weeks and months, we’ve been busy working with the good people at euresearcher.com to produce an article all about LPGPU2.

You can read the article (and find out more about the LPGPU2 project) here: The LPGPU2 Project

Today the LPGPU2 consortium issue the final press release for the project. Find out about our work on Data collection, analysis and automated suggestions for @OpenCLorg @opengl @VulkanAPI and OpenGL ES for mobile devices (mainly Android) here: https://lpgpu.org/wp/publications/press-release-2/

New metrics to identify Regions of Interests are added

Think Silicon developed new ways to identify potential Regions of Interests (RoIs) and make the feedback engine of LPGPU2 tool more effective. Through the new functionalities of the LPGPU2 tool, it is now possible to identify RoIs based on the average value of either a hardware counter or a software performance counter within a frame. The user defines the desired threshold, which is then compared with the average value of the selected counter for each frame. This offers the opportunity firstly to identify the frames that do not have the expected performance behavior and then to proceed with the analysis of the highlighted frames in greater detail, detecting any specific performance problems. The tool also enables the user to easily locate frames with interesting features, e.g., the frames with the most memory accesses, with the most pipeline stalls, with the most z-buffer traffic, etc.

The LPGPU2 framework, extended by Think Silicon within the LPGPU2 project, enabled the optimization of GLOVE™, an open source middleware, which allows developers for Android™, Linux^® and Windows^® operating systems to run seamlessly OpenGL^® ES on supported hardware by translating at runtime OpenGL ES API calls to Vulkan^® API commands for that platform. The GLOVE™ demo was running on a ZC706 Evaluation Board using NEMA^®|t, the smallest Internet-of-Things (IoT) Graphics Processor Unit (GPU) with 3D functionality. In depth analysis of the GLOVE™ software stack was performed by using the LPGPU2 tools CodeXL, SHIM and DC API on a Linux platform. The LPGPU2 tool performed a performance analysis, identifying bottlenecks and guided Think Silicon’s developers to achieve the best power-performance balance of the GPU, by statistically analyzing how the code runs inside the GPU. GLOVE™ is available for download from https://github.com/Think-Silicon/GLOVE.

The paper “VComputeBench: A Vulkan Benchmark Suite for GPGPU on Mobile and Embedded GPUs” by Nadjib Mammeri and Ben Juurlink has been accepted at IISWC 2018. The paper proposes VComputeBench, a set of benchmarks that help developers understand the differences in performance and portability of Vulkan and evaluates its suitability as an emerging cross-platform GPGPU framework by conducting a thorough analysis of its performance compared to CUDA and OpenCL on mobile as well as on desktop platforms.

The IEEE International Symposium on Workload Characterization is a well-established symposium for state-of-the-art research dedicated to the understanding and characterization of workloads that run on all types of computing systems. The 2018 edition of this conference will take place from September 30-October 2, 2018 in Raleigh, North Carolina, USA .

More information can be found at: http://www.iiswc.org/iiswc2018/index.html