Skip to content
/ tornadovm-examples Public

Set of examples written for hardware acceleration via TornadoVM

License

Apache-2.0 license
15 stars 2 forks Branches Tags Activity
Star
Notifications You must be signed in to change notification settings

jjfumero/tornadovm-examples

BranchesTags

Repository files navigation

TornadoVM Examples

TornadoVM is a Java library for hardware acceleration of Java and JVM applications. It has a JIT compiler, a runtime system, and several backends that offload, manage memory and handle execution on GPUs, FPGAs, and multicore CPUs transparently.

This repository contains a few examples for demonstration purposes.

Note: Examples using TornadoVM v1.0.9-dev

Outline:

Topic Link
Install TornadoVM link
Setup the examples link
Mandelbrot demo link
Blur Filter demo link
Multi-Image Processing demo link
KMeans Clustering demo link
Live Task Migration demo link
Matrix Multiplication link

1. Build TornadoVM

To run the examples, first build TornadoVM with any backend (OpenCL, PTX and/or SPIR-V).

Important: If you do not have an NVIDIA GPU and the CUDA SDK installed, do not use the flag --ptx in the following command. Similarly, if your device/system does not support SPIR-V, do not use the --spirv flag.

To install TornadoVM, it requires as prerequisite:

  1. The driver installed (e.g., NVIDIA + CUDA Driver for NVIDIA GPUs, or oneAPI for Intel platforms).
  2. Maven

TornadoVM includes an easy installer script for Linux and OSx:

$ git clone https://github.com/beehive-lab/TornadoVM
$ cd TornadoVM
## Choose the backend/s that applies to your system. You can install multiple ones
$ ./bin/tornadovm-installer --jdk jdk21 --backend=opencl,ptx,spirv  
$ source setvars.sh

Check installation:

tornado --devices

Number of Tornado drivers: 2
Total number of PTX devices  : 1
Tornado device=0:0
	PTX -- NVIDIA GeForce RTX 2060 with Max-Q Design
		Global Memory Size: 5.8 GB
		Local Memory Size: 48.0 KB
		Workgroup Dimensions: 3
		Max WorkGroup Configuration: [1024, 1024, 64]
		Device OpenCL C version: N/A

Total number of OpenCL devices  : 3
Tornado device=1:0
	NVIDIA CUDA -- NVIDIA GeForce RTX 2060 with Max-Q Design
		Global Memory Size: 5.8 GB
		Local Memory Size: 48.0 KB
		Workgroup Dimensions: 3
		Max WorkGroup Configuration: [1024, 1024, 64]
		Device OpenCL C version: OpenCL C 1.2

Tornado device=1:1
	Intel(R) OpenCL HD Graphics -- Intel(R) UHD Graphics [0x9bc4]
		Global Memory Size: 24.9 GB
		Local Memory Size: 64.0 KB
		Workgroup Dimensions: 3
		Max WorkGroup Configuration: [256, 256, 256]
		Device OpenCL C version: OpenCL C 3.0

Tornado device=1:2
	Intel(R) CPU Runtime for OpenCL(TM) Applications -- Intel(R) Core(TM) i9-10885H CPU @ 2.40GHz
		Global Memory Size: 31.1 GB
		Local Memory Size: 32.0 KB
		Workgroup Dimensions: 3
		Max WorkGroup Configuration: [8192, 8192, 8192]
		Device OpenCL C version: OpenCL C 2.0

Note that, depending on the devices you have and the drivers installed (e.g., NVIDIA CUDA, OpenCL, SPIR-V), you will see different implementations.

2. Setup the examples

git clone https://github.com/jjfumero/tornadovm-examples
cd tornadovm-examples
source /path-to-your-Tornado-DIR/source.sh
export TORNADO_SDK=/path-to-your-Tornado-DIR/bin/sdk
mvn clean package

Running demos

Mandelbrot

## List all devices and backends available 
tornado --devices 

## Run the multi-thread version for reference 
tornado -cp target/tornadovm-examples-1.0-SNAPSHOT.jar io.github.jjfumero.Mandelbrot mt

## Run the TornadoVM Version (it will select the device 0:0 by default)
tornado -cp target/tornadovm-examples-1.0-SNAPSHOT.jar io.github.jjfumero.Mandelbrot tornado

## Print the device and thread information in which the application is running
tornado --threadInfo -cp target/tornadovm-examples-1.0-SNAPSHOT.jar io.github.jjfumero.Mandelbrot tornado

## Get the SPIRV code (assuming the SPIRV backend is installed in the device 0:0)
tornado --debug --threadInfo -cp target/tornadovm-examples-1.0-SNAPSHOT.jar io.github.jjfumero.Mandelbrot tornado
spirv-dis <spirv-binary> 

## Change the device
tornado --threadInfo --jvm="-Dfractal.mandelbrot.device=1:1" -cp target/tornadovm-examples-1.0-SNAPSHOT.jar io.github.jjfumero.Mandelbrot tornado

## Run with the tornadoVM profiler
tornado --enableProfiler console --threadInfo --jvm="-Dfractal.mandelbrot.device=0:0" -cp target/tornadovm-examples-1.0-SNAPSHOT.jar io.github.jjfumero.Mandelbrot tornado

Blur Filter

This examples shows a blur effect in a photo. Example of computational photography. Place an JPEG image in ./image.jpg or feel free to change the path your images.

## List all devices and backends available 
tornado --devices 

## Run the Java Parallel Stream Version on CPU for reference 
tornado -cp target/tornadovm-examples-1.0-SNAPSHOT.jar io.github.jjfumero.BlurFilter mt

## Run the Accelerated Version on the default device 
tornado -cp target/tornadovm-examples-1.0-SNAPSHOT.jar io.github.jjfumero.BlurFilter tornado

## Run the Accelerated Version on the default device with info about the accelerator 
tornado --threadInfo -cp target/tornadovm-examples-1.0-SNAPSHOT.jar io.github.jjfumero.BlurFilter tornado

## Print the generated kernel
tornado --printKernel --threadInfo -cp target/tornadovm-examples-1.0-SNAPSHOT.jar io.github.jjfumero.BlurFilter tornado

## Run in another device
tornado --threadInfo -cp target/tornadovm-examples-1.0-SNAPSHOT.jar io.github.jjfumero.BlurFilter tornado device=1:1

## Run with concurrent multi-devices
tornado --threadInfo --enableConcurrentDevices --jvm=" -Dblur.red.device=1:0 -Dblur.green.device=2:0 -Dblur.blue.device=1:2" -cp target/tornadovm-examples-1.0-SNAPSHOT.jar io.github.jjfumero.BlurFilter tornado

Multi-Image Processing

Demonstration of a Task-Graph to compute:

  • Black and White filter
  • Blur Filer

using multiple GPUs (or accelerators) at the same time for each task.

## Run the Java Parallel Stream Version on CPU for reference 
tornado -cp target/tornadovm-examples-1.0-SNAPSHOT.jar io.github.jjfumero.MultiImageProcessor mt

## Run the Accelerated Version on the default device 
tornado -cp target/tornadovm-examples-1.0-SNAPSHOT.jar io.github.jjfumero.MultiImageProcessor tornado

The examples is created based on the following device setup:

## List all devices and backends available 
$ tornado --devices 

Number of Tornado drivers: 3
Driver: SPIRV
  Total number of SPIRV devices  : 1
  Tornado device=0:0  (DEFAULT)
	SPIRV -- SPIRV LevelZero - Intel(R) UHD Graphics 770
		Global Memory Size: 24.9 GB
		Local Memory Size: 64.0 KB
		Workgroup Dimensions: 3
		Total Number of Block Threads: [512]
		Max WorkGroup Configuration: [512, 512, 512]
		Device OpenCL C version:  (LEVEL ZERO) 1.3

Driver: OpenCL
  Total number of OpenCL devices  : 4
  Tornado device=1:0
	OPENCL --  [NVIDIA CUDA] -- NVIDIA GeForce RTX 3070
		Global Memory Size: 7.8 GB
		Local Memory Size: 48.0 KB
		Workgroup Dimensions: 3
		Total Number of Block Threads: [1024]
		Max WorkGroup Configuration: [1024, 1024, 64]
		Device OpenCL C version: OpenCL C 1.2

  Tornado device=1:1
	OPENCL --  [Intel(R) OpenCL Graphics] -- Intel(R) UHD Graphics 770
		Global Memory Size: 24.9 GB
		Local Memory Size: 64.0 KB
		Workgroup Dimensions: 3
		Total Number of Block Threads: [512]
		Max WorkGroup Configuration: [512, 512, 512]
		Device OpenCL C version: OpenCL C 1.2

  Tornado device=1:2
	OPENCL --  [Intel(R) OpenCL] -- 12th Gen Intel(R) Core(TM) i7-12700K
		Global Memory Size: 31.1 GB
		Local Memory Size: 32.0 KB
		Workgroup Dimensions: 3
		Total Number of Block Threads: [8192]
		Max WorkGroup Configuration: [8192, 8192, 8192]
		Device OpenCL C version: OpenCL C 3.0

  Tornado device=1:3
	OPENCL --  [Intel(R) FPGA Emulation Platform for OpenCL(TM)] -- Intel(R) FPGA Emulation Device
		Global Memory Size: 31.1 GB
		Local Memory Size: 256.0 KB
		Workgroup Dimensions: 3
		Total Number of Block Threads: [67108864]
		Max WorkGroup Configuration: [67108864, 67108864, 67108864]
		Device OpenCL C version: OpenCL C 1.2

Driver: PTX
  Total number of PTX devices  : 1
  Tornado device=2:0
	PTX -- PTX -- NVIDIA GeForce RTX 3070
		Global Memory Size: 7.8 GB
		Local Memory Size: 48.0 KB
		Workgroup Dimensions: 3
		Total Number of Block Threads: [2147483647, 65535, 65535]
		Max WorkGroup Configuration: [1024, 1024, 64]
		Device OpenCL C version: N/A

To change the accelerator, use the following instructions:

TornadoDevice device0 = TornadoExecutionPlan.getDevice(0, 0);
TornadoDevice device1 = TornadoExecutionPlan.getDevice(1, 0);
TornadoDevice device2 = TornadoExecutionPlan.getDevice(1, 1);
TornadoDevice device3 = TornadoExecutionPlan.getDevice(1, 2);
TornadoDevice device4 = TornadoExecutionPlan.getDevice(2, 0);

executionPlan.withConcurrentDevices() //
	.withDevice("imageProcessor.blackAndWhite", device0) //
	.withDevice("imageProcessor.blurRed", device1) //
	.withDevice("imageProcessor.blurGreen", device2) //
	.withDevice("imageProcessor.blurBlue", device4);

KMeans Clustering

Full KMeans in which the assign-cluster function is expressed with TornadoVM.

# For example use 1000000 data points and classify them into 10 clusters.
# Points are selected randomly. This is just for quick experiments 

## Sequential
tornado -cp target/tornadovm-examples-1.0-SNAPSHOT.jar io.github.jjfumero.KMeans seq 1000000 10


# TornadoVM version 
tornado -cp target/tornadovm-examples-1.0-SNAPSHOT.jar io.github.jjfumero.KMeans tornado 1000000 10

Live Task Migration (Client-Server App)

## Run Server in one terminal
./runServer.sh

## Client in another terminal
./runClient.sh  ## Change device during runtime 

## Note: the application selects the backend 0:0 (default backend)

# type different <backend:deviceNumber> version from the client. 
# Examples:
# 0:1 
# 1:0 
## etc

Matrix Multiplication

This application shows different implementation for Matrix Multiplications for:

  1. Single-core Sequential Implementation using Panama Segments
  2. Multi-core Parallel Streams Using Panama Segments
  3. Multi-core Java Threads using Panama Segments
  4. Single-core Sequential implementation using Panama Vector Types for the reduction part (dotProduct)
  5. Multi-core Parallel Stream implementation using Panama Vector Types for the reduction part (dotProduct)
  6. TornadoVM to run on the default device (usually a GPU if a GPU backends is available and installed)

How to run?

mvn clean pacakge
./runMxM.sh

About

Set of examples written for hardware acceleration via TornadoVM

Topics

java fpga jvm gpu opencl parallel-computing spirv ptx fpga-programming tornadovm gpuprogramming

Resources

Readme

License

Apache-2.0 license
Activity

Stars

15 stars

Watchers

3 watching

Forks

2 forks
Report repository

Releases

No releases published

Packages

No packages published

Languages