Java on the GPU Where are we now?

on YouTube ↗51m

" summary

History plus hands‑on: GPUs move from fixed shaders to OpenGL bindings in Java (JOGL) and full GPGPU via OpenCL and CUDA; Java bindings available include JOCL/JavaCL (many abandoned) and JCuda. Measured example: 10 million vector adds run on a GPU kernel in ~800 ms with large setup and host/device copy overhead dominating; integrated GPUs avoid copies. Higher‑level routes (Aparapi/bytecode→GPU, IBM GPU enabler for Stream.forEach) auto‑offload lambdas; a Java EE heat‑map rendered on an AWS Tesla in the demo.

AI-written summary · about

" chapters (39)

1. 0:00Start
2. 1:13What is a video card?
3. 2:22What is a GPU?
4. 3:55GPGPU
5. 4:35The CPU in general looks like this
6. 5:41Let's simplify!
7. 6:53So we come to SIMD paradigm
8. 7:33And here we start to talk about vectors..
9. 8:23It all started with a shader
10. 10:24There are several shaders languages
11. 11:35Several abstractions were created
12. 12:56OpenGL in Java
13. 16:07BrookGPU
14. 17:00Why should we ever use GPU on Java
15. 18:37For OpenCL
16. 19:31Disclaimer: its hard to work with GPU!
17. 21:13Typical lifecycle of an OpenCL app
18. 25:49Datatypes: scalars
19. 26:16Datatypes:vectors
20. 27:06Memory Model
21. 27:59Execution model
22. 28:50ND Range - what is that?
23. 30:02Typical GPU
24. 31:23CUDA kernel
25. 31:45CUDA setup
26. 31:55CUDA copy to memory and run
27. 32:23But CUDA has some other superpowers
28. 34:35nVidia memory looks like this
29. 36:34Optimizations..
30. 36:46-Optimizations
31. 37:19Project Sumatra
32. 37:50AMD HSAIL
33. 38:52AMD APU tries to solve the main issue..
34. 39:45IBM patched JVM for GPU
35. 45:02nVidia GRID
36. 48:08AMD is a bit behind...
37. 48:20Its here: Nvidia GPU
38. 48:34Its here: AMD APU
39. 49:23Intel with VEGA??

" related talks

• 47mJava and GPU … are we nearly there yet?Gary Frost · JVM Language Summit 2023
• 40mJava GPGPU Enablement: Are We There Yet?Gary Frost · JavaOne 2025
• 49mMassive Parallelism with GPUs in JavaAdam Roberts · Devoxx US 2017
• 43mHow to make your electrons do more in Java?Konrad Szałkowski · JDD 2023
• 50mUnder the HAT: Empowering GPU Acceleration for JavaJuan Fumero Alfonso · JavaOne 2026
• 40mHow to make your electrons do more in JavaKonrad Szałkowski · GeeCON 2023
• 44mWrite Once, Accelerate Everywhere: GPU-Ready Java with TornadoVMThanos Stratikopoulos · Voxxed Days Ticino 2026
• 43mFrom CPU to GPU and FPGAs: Supercharging Java Applications with TornadoVMJuan Fumero · JVM Language Summit 2023