Juan Fumero<p>Hardware <a href="https://mastodon.online/tags/Acceleration" class="mention hashtag" rel="tag">#<span>Acceleration</span></a> for <a href="https://mastodon.online/tags/Java" class="mention hashtag" rel="tag">#<span>Java</span></a> Ray Tracing: <a href="https://mastodon.online/tags/TornadoVM" class="mention hashtag" rel="tag">#<span>TornadoVM</span></a> in Action</p><p><a href="https://www.youtube.com/watch?v=7q9AGvpZ4Hw" target="_blank" rel="nofollow noopener" translate="no"><span class="invisible">https://www.</span><span class="">youtube.com/watch?v=7q9AGvpZ4Hw</span><span class="invisible"></span></a></p>
mastodon.online is one of the many independent Mastodon servers you can use to participate in the fediverse.
A newer server operated by the Mastodon gGmbH non-profit
Administered by:
Server stats:
11Kactive users
mastodon.online: About · Status · Profiles directory · Privacy policy
Mastodon: About · Get the app · Keyboard shortcuts · View source code · v4.4.0-nightly.2025-03-25
Juan Fumero@snatverk@mastodon.online
Research Fellow at the University of Manchester.
I work on Java, VMs, and Heterogeneous Computing. I am also an Intel Innovator, and TornadoVM lead architect. Formerly at the University of Edinburgh, Oracle Labs and CERN OpenLabs.
Juan Fumero<p>New blogpost: Learn how to build JDK 21 and JDK 25 with the HotSpot Disassembler (HSDIS) plugin enabled for Linux. `hsdis` is a tool to inspect the JVM’s JIT-compiled assembly code, and this post explains how to configure it:</p><p>🔗 <a href="https://jjfumero.github.io/posts/2025/02/14/jdk-hsdis-build" target="_blank" rel="nofollow noopener" translate="no"><span class="invisible">https://</span><span class="ellipsis">jjfumero.github.io/posts/2025/</span><span class="invisible">02/14/jdk-hsdis-build</span></a> </p><p><a href="https://mastodon.online/tags/java" class="mention hashtag" rel="tag">#<span>java</span></a> <a href="https://mastodon.online/tags/hsdis" class="mention hashtag" rel="tag">#<span>hsdis</span></a> <a href="https://mastodon.online/tags/openjdk" class="mention hashtag" rel="tag">#<span>openjdk</span></a></p>
Juan Fumero<p>New Article: <a href="https://mastodon.online/tags/Babylon" class="mention hashtag" rel="tag">#<span>Babylon</span></a> OpenJDK - A Guide for Beginners and Comparisons with <a href="https://mastodon.online/tags/TornadoVM" class="mention hashtag" rel="tag">#<span>TornadoVM</span></a></p><p>🔗<a href="https://jjfumero.github.io/posts/2025/02/07/babylon-and-tornadovm" target="_blank" rel="nofollow noopener" translate="no"><span class="invisible">https://</span><span class="ellipsis">jjfumero.github.io/posts/2025/</span><span class="invisible">02/07/babylon-and-tornadovm</span></a></p><p><a href="https://mastodon.online/tags/java" class="mention hashtag" rel="tag">#<span>java</span></a> <a href="https://mastodon.online/tags/ai" class="mention hashtag" rel="tag">#<span>ai</span></a> <a href="https://mastodon.online/tags/gpus" class="mention hashtag" rel="tag">#<span>gpus</span></a> <a href="https://mastodon.online/tags/openjdk" class="mention hashtag" rel="tag">#<span>openjdk</span></a></p>
Juan Fumero<p>How to Fix CUDA GCC Unsuported Versions on Linux</p><p>🔗 <a href="https://jjfumero.github.io/posts/2025/01/16/cuda-gcc-versions" target="_blank" rel="nofollow noopener" translate="no"><span class="invisible">https://</span><span class="ellipsis">jjfumero.github.io/posts/2025/</span><span class="invisible">01/16/cuda-gcc-versions</span></a></p>
Juan Fumero<p>New post: Learn how to setup WSL for GPU compute, including TornadoVM</p><p>🔗<a href="https://jjfumero.github.io/posts/2025/01/14/gpu-wsl" target="_blank" rel="nofollow noopener" translate="no"><span class="invisible">https://</span><span class="ellipsis">jjfumero.github.io/posts/2025/</span><span class="invisible">01/14/gpu-wsl</span></a></p>
Ian Brown :verified:<p>It has been a busy year, but I was able to finally get some downtime over the holidays to finish this book. </p><p>It is a great introduction to the core concepts of programming with GPUs and other co-processors, and surveys the future directions that managed runtimes like the <a href="https://mastodon.hccp.org/tags/JVM" class="mention hashtag" rel="nofollow noopener" target="_blank">#<span>JVM</span></a> will take in order to simplify developing software on heterogeneous hardware environments.</p><p>It was super accessible, and I'd recommend this for anyone interested in the subject. Even managers! </p><p><a href="https://mastodon.hccp.org/@igb/112386562916088384" rel="nofollow noopener" translate="no" target="_blank"><span class="invisible">https://</span><span class="ellipsis">mastodon.hccp.org/@igb/1123865</span><span class="invisible">62916088384</span></a></p>
Ian Brown<p>Finally found the downtime to complete this fantastic survey of managed runtimes (e.g. the JVM) and heterogeneous hardware (e.g. CPUs and GPUs or FPGAs) by <a href="https://mastodon.online/users/snatverk" rel="nofollow noopener" target="_blank">@snatverk@mastodon.online</a>, <a href="https://mastodon.sdf.org/users/thanos_str" rel="nofollow noopener" target="_blank">@thanos_str@mastodon.sdf.org</a>, and <a href="https://mastodon.online/users/kotselidis" rel="nofollow noopener" target="_blank">@kotselidis@mastodon.online</a>. </p>
<p>Required reading for those who want a look at the future of software development.</p>
<p><a href="https://books.hccp.org/hashtag/177" rel="nofollow noopener" target="_blank">#TornadoVM</a> <a href="https://books.hccp.org/hashtag/184" rel="nofollow noopener" target="_blank">#JOCL</a> <a href="https://books.hccp.org/hashtag/175" rel="nofollow noopener" target="_blank">#OpenCL</a> <a href="https://books.hccp.org/hashtag/176" rel="nofollow noopener" target="_blank">#CUDA</a></p><p>(comment on <a href="https://books.hccp.org/book/31088" rel="nofollow noopener" target="_blank">"Programming Heterogeneous Hardware via Managed Runtime Systems"</a>)</p>
Juan Fumero<p>TornadoVM Performance vs OpenCL for Matrix Multiplication on NVIDIA RTX 4090.</p><p>I have been working on this video for a while, and now it is here! Performance analysis of <a href="https://mastodon.online/tags/TornadoVM" class="mention hashtag" rel="tag">#<span>TornadoVM</span></a> vs <a href="https://mastodon.online/tags/OpenCL" class="mention hashtag" rel="tag">#<span>OpenCL</span></a> native. Plenty of Java, parallelism, OpenCL, and GPU programming.</p><p><a href="https://www.youtube.com/watch?v=xj8Te517Wtc" target="_blank" rel="nofollow noopener" translate="no"><span class="invisible">https://www.</span><span class="">youtube.com/watch?v=xj8Te517Wtc</span><span class="invisible"></span></a></p>
Juan Fumero<p>Want to know the capabilities of <a href="https://mastodon.online/tags/TornadoVM" class="mention hashtag" rel="tag">#<span>TornadoVM</span></a> on RTX 4090? Do you know that for some applications, TornadoVM is faster than native OpenCL? Is that even possible? If you are curious, I explore all of these questions in my latest blog article. </p><p>Using Matrix Multiplication as an example, I dive in into all optimisations that the TornadoVM JIT compiler and runtime performs to make the code fast on NVIDIA GPUs. How easy/hard is to beat <a href="https://mastodon.online/tags/TornadoVM" class="mention hashtag" rel="tag">#<span>TornadoVM</span></a>? Check it out! </p><p><a href="https://jjfumero.github.io/posts/2024/12/17/tornadovm-vs-opencl" target="_blank" rel="nofollow noopener" translate="no"><span class="invisible">https://</span><span class="ellipsis">jjfumero.github.io/posts/2024/</span><span class="invisible">12/17/tornadovm-vs-opencl</span></a></p>
Juan Fumero<p>Interesting Read: "Pascal: The Underrated Gem Among Programming Languages" </p><p><a href="https://simplifycpp.org/?id=a0534" target="_blank" rel="nofollow noopener" translate="no"><span class="invisible">https://</span><span class="">simplifycpp.org/?id=a0534</span><span class="invisible"></span></a></p><p>Pascal was my first programming language. I even implemented a mini-pascal compiler in Pascal during my Compilers course in the University.</p>
Juan Fumero<p>Running benchmarks with <a href="https://mastodon.online/tags/TornadoVM" class="mention hashtag" rel="tag">#<span>TornadoVM</span></a> on a <a href="https://mastodon.online/tags/RISCV" class="mention hashtag" rel="tag">#<span>RISCV</span></a> SBC CPU with vector units target hardware accelerators via <a href="https://mastodon.online/tags/OCK" class="mention hashtag" rel="tag">#<span>OCK</span></a>. Very promising results -> 11x over Java Sequential and 4.5x over Java Parallel Streams on the same CPU. More info very soon!</p>
Juan Fumero<p>The Level Zero JNI library is open sourced, and we implemented this library to dispatch SPIR-V kernels in TornadoVM. However, TornadoVM could implement shared memory buffers, and get extra performance on shared memory systems.</p><p>So, what's the catch? The Level-Zero JNI contains hand-written kernels with explicit runtime calls to manage and dispatch the code on GPUs, while in TornadoVM, well, it's parallelized and automatically accelerated from the Java sequential code.</p>
Juan Fumero<p>Java <a href="https://mastodon.online/tags/Llama2" class="mention hashtag" rel="tag">#<span>Llama2</span></a> fork extended with GPU support using Level Zero JNI lib to run on Intel ARC and Integrated GPUs. The initial version from my colleague Michalis Papadimitriou includes <a href="https://mastodon.online/tags/TornadoVM" class="mention hashtag" rel="tag">#<span>TornadoVM</span></a>.</p><p>The level-zero port achieves higher tok/s vs TornadoVM on Integrated GPUs.</p><p>🔗 <a href="https://github.com/jjfumero/llama2.tornadovm.java" target="_blank" rel="nofollow noopener" translate="no"><span class="invisible">https://</span><span class="ellipsis">github.com/jjfumero/llama2.tor</span><span class="invisible">nadovm.java</span></a></p>
Juan Fumero<p>We just drop a new TornadoVM version, 1.0.8 with improvements and many fixes. This version expands the profiler of the LevelZero/SPIRV backend with power metrics, new API calls for log and debug execution plans, fixes for running on OSx 14.6 and more! </p><p>Check it out!<br />🔗 <a href="https://github.com/beehive-lab/TornadoVM/releases/tag/v1.0.8" target="_blank" rel="nofollow noopener" translate="no"><span class="invisible">https://</span><span class="ellipsis">github.com/beehive-lab/Tornado</span><span class="invisible">VM/releases/tag/v1.0.8</span></a></p><p><a href="https://mastodon.online/tags/java" class="mention hashtag" rel="tag">#<span>java</span></a> <a href="https://mastodon.online/tags/ai" class="mention hashtag" rel="tag">#<span>ai</span></a> <a href="https://mastodon.online/tags/gpus" class="mention hashtag" rel="tag">#<span>gpus</span></a> <a href="https://mastodon.online/tags/fpgas" class="mention hashtag" rel="tag">#<span>fpgas</span></a> <a href="https://mastodon.online/tags/accelerators" class="mention hashtag" rel="tag">#<span>accelerators</span></a> <a href="https://mastodon.online/tags/graalvm" class="mention hashtag" rel="tag">#<span>graalvm</span></a></p>
Juan Fumero<p>Join me at the UXL oneAPI DevSummit on the 10th October 2024 </p><p>🔗 <a href="https://www.oneapi.io/events/oneapi-devsummit-hosted-by-uxl-foundation/" target="_blank" rel="nofollow noopener" translate="no"><span class="invisible">https://www.</span><span class="ellipsis">oneapi.io/events/oneapi-devsum</span><span class="invisible">mit-hosted-by-uxl-foundation/</span></a></p>
Juan Fumero<p>[Blog] Using oneAPI Construction Kit and hashtag#TornadoVM to accelerate Java Programs on x86, ARM and RISC-V CPUs. In this blog, we explore how OCK can be used by TornadoVM to run on different CPU architectures.</p><p>This post shows how to setup OCK to be used with TornadoVM for Intel, ARM Neoverse V2 (NVIDIA Grace Hopper Superchip) and RISC-V CPUs. Besides, it shows a performance evaluation of TornadoVM/OCK compared to Java Parallel Streams running on the same CPU. </p><p>🔗 <a href="https://jjfumero.github.io/posts/2024/09/10/tornadovm-ock" target="_blank" rel="nofollow noopener" translate="no"><span class="invisible">https://</span><span class="ellipsis">jjfumero.github.io/posts/2024/</span><span class="invisible">09/10/tornadovm-ock</span></a></p>
Juan Fumero<p>We are improving the driver support for new hardware accelerators for <a href="https://mastodon.online/tags/TornadoVM" class="mention hashtag" rel="tag">#<span>TornadoVM</span></a>. With the help of <br />Codeplay and the AERO EU Project, we are now able to run Java programs on <a href="https://mastodon.online/tags/RISCV" class="mention hashtag" rel="tag">#<span>RISCV</span></a> simulators using a fully open source software stack:</p><p>1️⃣ oneAPI Construction Kit<br />2⃣TornadoVM</p>
Juan Fumero<p>The good thing about the OCK is that it is a potential CPU implementation of OpenCL, not just for x86 CPUs, but also for ARM and RISC-V accelerators, and even custom accelerators. Stay tuned!</p>
Juan Fumero<p>Thanks to Codeplay Software and the AERO Project, we are able to accelerate Java programs on CPUs via the <a href="https://mastodon.online/tags/TornadoVM" class="mention hashtag" rel="tag">#<span>TornadoVM</span></a> framework and the oneAPI Construction Kit (OCK).</p><p>Acceleration of 3.4x over the execution with Java Streams on the same CPU (32 cores) for running Matrix Multiplications of size 1024x1024. </p><p>I will probably write a detailed blog post about how to use it and some performance numbers. </p><p><a href="https://mastodon.online/tags/java" class="mention hashtag" rel="tag">#<span>java</span></a> <a href="https://mastodon.online/tags/acceleration" class="mention hashtag" rel="tag">#<span>acceleration</span></a> <a href="https://mastodon.online/tags/ai" class="mention hashtag" rel="tag">#<span>ai</span></a> <a href="https://mastodon.online/tags/cpus" class="mention hashtag" rel="tag">#<span>cpus</span></a> <a href="https://mastodon.online/tags/gpus" class="mention hashtag" rel="tag">#<span>gpus</span></a></p>
Lindsey Kuper<p>We're seeking speakers for this year's Languages, Systems, and Data Seminar! 🎉</p><p>We meet on Fridays in fall, winter, and spring, in person and on Zoom. Most of our speakers are PhD students working in the areas of programming languages, systems, databases, etc.</p><p>If you've never given an LSD talk and you'd like to, nominate yourself: <a href="https://forms.gle/jGN6qUqKrQbABJvo8" rel="nofollow noopener" translate="no" target="_blank"><span class="invisible">https://</span><span class="">forms.gle/jGN6qUqKrQbABJvo8</span><span class="invisible"></span></a></p><p>We try not to have repeats, so if you *have* given an LSD talk before, pass the form along to a friend who you think would give a good talk!</p>
ExploreLive feeds
Mastodon is the best way to keep up with what's happening.
Follow anyone across the fediverse and see it all in chronological order. No algorithms, ads, or clickbait in sight.
LoginDrag & drop to upload