A great milestone achieved by our team. Just released #GPULlama3.java, the first Java implementation of Llama3 inference accelerated on GPUs via #TornadoVM.
GitHub v0.1.0-beta: https://github.com/beehive-lab/GPULlama3.java
Looking forward to your feedback.
#opensource #Java #AI #LLM #GPUs
🚀 Learn how to accelerate performance on RISC-V CPUs using TornadoVM & vector instructions via the oneAPI Construction Kit in my latest article.
In this article I explore the #RISCV architecture, talk briefly about the RISC-V extensions, some real hardware and the status of #Java for RISC-V. Then, I explore how we are using #TornadoVM and the oneAPI Construction Kit (#OCK) to accelerate Java workloads on RISC-V CPUs with vector extensions.
https://jjfumero.github.io/posts/2025/05/05/riscv-java-acceleration
New #TornadoVM v1.1.0 release!
👉 Improved profiling and logging.
👉 New memory and buffer management features to persist and reuse data.
👉 Support for more atomic operations.
Work done under #horizoneurope and UK Research and Innovation funded projects: AERO Project, TANGO EU, P2CODE project, ENCRYPT Project!
hanks to all our contributors! 👏
Hardware #Acceleration for #Java Ray Tracing: #TornadoVM in Action
New Article: #Babylon OpenJDK - A Guide for Beginners and Comparisons with #TornadoVM
🔗https://jjfumero.github.io/posts/2025/02/07/babylon-and-tornadovm
TornadoVM Performance vs OpenCL for Matrix Multiplication on NVIDIA RTX 4090.
I have been working on this video for a while, and now it is here! Performance analysis of #TornadoVM vs #OpenCL native. Plenty of Java, parallelism, OpenCL, and GPU programming.
Want to know the capabilities of #TornadoVM 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.
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 #TornadoVM? Check it out!
https://jjfumero.github.io/posts/2024/12/17/tornadovm-vs-opencl
Running benchmarks with #TornadoVM on a #RISCV SBC CPU with vector units target hardware accelerators via #OCK. Very promising results -> 11x over Java Sequential and 4.5x over Java Parallel Streams on the same CPU. More info very soon!
Java #Llama2 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 #TornadoVM.
The level-zero port achieves higher tok/s vs TornadoVM on Integrated GPUs.
We are improving the driver support for new hardware accelerators for #TornadoVM. With the help of
Codeplay and the AERO EU Project, we are now able to run Java programs on #RISCV simulators using a fully open source software stack:
1️⃣ oneAPI Construction Kit
2⃣TornadoVM
Thanks to Codeplay Software and the AERO Project, we are able to accelerate Java programs on CPUs via the #TornadoVM framework and the oneAPI Construction Kit (OCK).
Acceleration of 3.4x over the execution with Java Streams on the same CPU (32 cores) for running Matrix Multiplications of size 1024x1024.
I will probably write a detailed blog post about how to use it and some performance numbers.
🎉So exited, my #TornadoVM talk got accepted at the Intel Innovation 2024 in San Jose, CA. If someone in Santa Clara/San Jose area ~[20-27]th September wants to meet and discuss about #GPUs, #Java, and #AI, just ping me!
#santaclara #tornadovmteam
Wednesday Links - Edition 2024-05-29
https://dev.to/0xkkocel/wednesday-links-edition-2024-05-29-dbm
#java #jvm #sbom #spring #springboot #bug #jobrunr #tornadovm #microservices #ci #kotlin #asserj #gradle
What's the #TornadoVM programming model? From a #Java API for parallel programming to the Task-Graphs and Execution Plans building blocks, @snatverk explains it all on Foojay :foojay:!
https://foojay.io/today/the-tornadovm-programming-model-explained/
@petrillic @igb@books.hccp.org #TornadoVM is an interesting approach to the above (at least for the JVM) https://www.tornadovm.org/
My copy of @snatverk, @thanos_str, and @kotselidis's new book "Programming Heterogeneous Hardware via Managed Runtime Systems" just arrived! Can't wait to read it! #TornadoVM
TornadoVM reaching 8500 GLOPS on NVIDIA #GH200 #GraceHopper GPU for Matrix Multiplication of size 2048x2048. This is the naive matrix multiplication offloaded and optimised by the #TornadoVM + #GraalVM JIT compilers and the TornadoVM runtime. Stay tuned for more experiments on this platform. 🚀
Thank you to the University of Pisa for giving us access to this system via the AERO Project 🙏
Selbstgemacht und Handcompiled: Wir freuen uns über die neue Version von #TornadoVM 1.0(.3) Release erlaubt #Java auf der #GPU
https://theradio.cc/blog/2024/04/14/ll280-maintainers-chained-by-supply/
Finally, our book is available! We have been working on this for the past year and it shows an introduction to Programming Heterogeneous Hardware from Managed Runtime Systems.
We include an overview of GPU architectures, an introduction of parallel programming languages, a discussion about the challenges and current solutions of implementing and deploying data parallel applications from within managed runtime programming systems, and #TornadoVM too! Spread the word!
Discover the innovative Dynamic Reconfiguration feature of TornadoVM, a groundbreaking technology crafted by the The University of Manchester.
Learn how this cutting-edge feature will be integrated into #TANGO to significantly enhance energy efficiency, marking a leap forward in green computing. 🔽 🔽 🔽
https://tango-project.eu/articles/dynamic-reconfiguration-tornadovm-v10
👨🔬 Thanos Stratikopoulos, Juan Fumero, Christos Kotselidis
#TornadoVM #DynamicReconfiguration #GreenComputing #HorizonEU
