Foojay Podcast #17: Execute Java Code with TornadoVM on CPUs, GPUs, and FPGAs
Running Java on a CPU is the default, but what happens when you point the same code at a GPU or an FPGA? TornadoVM does exactly that. It offloads JVM workloads to accelerators and can speed up parts of an existing program by orders of magnitude. In this Foojay Podcast #17, host Erik Costlow talks with Juan Fumero, Christos Kotselidis, Thanos Stratikopoulos, and Jakob Jenkov about how the project works and where it fits.
What we talked about
- What TornadoVM is and how it fits in the JVM landscape
- How applications leverage TornadoVM acceleration
- Differences between CPU threads and GPU instruction chains
- Concrete use cases for TornadoVM
- Results on Apple M1 hardware
- Running TornadoVM in cloud environments
- The API and how developers use it
- AWS Lambda integration potential
- Programming complexity for GPU and FPGA targets
- Building heating, cost reduction, and ChatGPT-style applications
- The relationship with Project Panama
- Getting started guidance for new users
Why it matters
Most Java developers never touch a GPU, yet the hardware sits in almost every machine. TornadoVM lowers that bar and keeps the code in plain Java. The conversation shows where the speedups come from and which problems suit this approach.
See the Foojay Podcast #17 for all info, shownotes, links, etc.