Docker for TornadoVM

We have two docker configurations for TornadoVM using 2 different JDKs:

TornadoVM Docker for NVIDIA GPUs: See instructions
- JDKs supported:
  - TornadoVM with OpenJDK 17
  - TornadoVM with GraalVM 22.3.1 and JDK 17
TornadoVM Docker for Intel Integrated Graphics, Intel CPUs, and Intel FPGAs (Emulated Mode): See instructions
- JDKs supported:
  - TornadoVM with OpenJDK 17
  - TornadoVM with GraalVM 22.3.1 and JDK 17

Nvidia GPUs

Prerequisites

The tornadovm-nvidia-openjdk docker image needs the docker nvidia daemon. More info here: https://github.com/NVIDIA/nvidia-docker.

How to run?

Pull the image

For the tornadovm-nvidia-openjdk image:

$ docker pull beehivelab/tornadovm-nvidia-openjdk:latest

This image uses the latest TornadoVM for NVIDIA GPUs and OpenJDK 17.

Run an experiment

We provide a runner script that compiles and run your Java programs with Tornado. Here's an example:

$ git clone https://github.com/beehive-lab/docker-tornado
$ cd docker-tornado

## Run Matrix Multiplication - provided in the docker-tornado repositoryu
$ ./run_nvidia_openjdk.sh tornado -cp example/target/example-1.0-SNAPSHOT.jar example.MatrixMultiplication

Computing MxM of 2048x2048
	CPU Execution: 0.36 GFlops, Total time = 48254 ms
	GPU Execution: 277.09 GFlops, Total Time = 62 ms
	Speedup: 778x

Using TornadoVM with GraalVM for NVIDIA GPUs

With JDK 17:

$ docker pull beehivelab/tornadovm-nvidia-graalvm:latest

Some options

# To see the generated OpenCL kernel
$ ./run_nvidia.sh tornado --printKernel example/MatrixMultiplication

# To check some runtime info about the kernel execution and device
$ ./run_nvidia.sh tornado --debug example/MatrixMultiplication

The tornado command is just an alias to the java command with all the parameters for TornadoVM execution. So you can pass any Java (OpenJDK or Hotspot) parameter.

$ ./run_nvidia.sh tornado --jvm="-Xmx16g -Xms16g" example/MatrixMultiplication

Intel Integrated Graphics

Prerequisites

The beehivelab/tornadovm-intel-openjdk docker image Intel OpenCL driver for the integrated GPU installed. More info here: https://github.com/intel/compute-runtime.

How to run?

Pull the image

For the beehivelab/tornadovm-intel-openjdk image:

$ docker pull beehivelab/tornadovm-intel-openjdk:latest

This image uses the latest TornadoVM for Intel integrated graphics and OpenJDK 17.

Run an experiment

We provide a runner script that compiles and run your Java programs with Tornado. Here's an example:

$ git clone https://github.com/beehive-lab/docker-tornado
$ cd docker-tornado

## Run Matrix Multiplication - provided in the docker-tornado repository
$ ./run_intel_openjdk.sh tornado -cp example/target/example-1.0-SNAPSHOT.jar example.MatrixMultiplication --params "256"

Computing MxM of 256x256
	CPU Execution: 1.53 GFlops, Total time = 22 ms
	GPU Execution: 8.39 GFlops, Total Time = 4 ms
	Speedup: 5x

Running on FPGAs (Emulation mode)?

The TornadoVM docker image for the Intel platforms contain the FPGA in device 1:0. To offload a Java application onto an FPGA, you can use the following command (example running the DFT application).

$ ./run_intel_openjdk.sh tornado --threadInfo  --jvm="-Ds0.t0.device=1:0" -m tornado.examples/uk.ac.manchester.tornado.examples.dynamic.DFTDynamic --params="256 default 1"
WARNING: Using incubator modules: jdk.incubator.foreign, jdk.incubator.vector
Initialization time:  1066024424 ns
 
Task info: s0.t0
        Backend           : OPENCL
        Device            : Intel(R) FPGA Emulation Device CL_DEVICE_TYPE_ACCELERATOR (available)
        Dims              : 1
        Global work offset: [0]
        Global work size  : [256]
        Local  work size  : [64, 1, 1]
        Number of workgroups  : [4]
 
Total time:  276927741 ns 
 
Is valid?: true
 
Validation: SUCCESS

Using TornadoVM with GraalVM for Intel Integrated Graphics

With JDK 17:

$ docker pull beehivelab/tornadovm-intel-graalvm:latest

Enjoy Tornado!!

Docker scripts have been inspired by blang/latex-docker

License

This project is developed at The University of Manchester, and it is fully open source under the Apache 2 license.