Special coloring of Mandelbrot example output

[otabuzzman]

Output of Mandelbrot example looks somewhat inverse compared to well-known images. Changed int arithmetic to float and scaling to get color gradients.

Source file with changes provided below. Wasn't sure if you prefer issues or pull requests. Code changes in Mandelbrot.java affected two lines, each containing

short r = (short) ((y * 255) / iterations);

replaced by

short r = (short) ((float)(y * 255 * 255) / iterations);

I wrote a tutorial about TornadoVM for a German computer magazine (iX) to be published in January or February and used the Mandelbrot example to add some color to the text (appreciated by the audience) of part 1. Part 2 will cover the ray tracer use case on multiple accelerators.

Mandelbrot.java

/*
 * Copyright (c) 2013-2023, APT Group, Department of Computer Science,
 * The University of Manchester.
 *
 * Licensed under the Apache License, Version 2.0 (the "License");
 * you may not use this file except in compliance with the License.
 * You may obtain a copy of the License at
 *
 * http://www.apache.org/licenses/LICENSE-2.0
 *
 * Unless required by applicable law or agreed to in writing, software
 * distributed under the License is distributed on an "AS IS" BASIS,
 * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
 * See the License for the specific language governing permissions and
 * limitations under the License.
 *
 */

package uk.ac.manchester.tornado.examples.compute;

import java.awt.Component;
import java.awt.Dimension;
import java.awt.Graphics;
import java.awt.event.WindowAdapter;
import java.awt.event.WindowEvent;
import java.awt.image.BufferedImage;
import java.awt.image.WritableRaster;
import java.io.File;

import javax.imageio.ImageIO;
import javax.swing.JFrame;

import uk.ac.manchester.tornado.api.ImmutableTaskGraph;
import uk.ac.manchester.tornado.api.TaskGraph;
import uk.ac.manchester.tornado.api.TornadoExecutionPlan;
import uk.ac.manchester.tornado.api.annotations.Parallel;
import uk.ac.manchester.tornado.api.enums.DataTransferMode;
import uk.ac.manchester.tornado.api.types.arrays.ShortArray;

/**
 * <p>
 * How to run?
 * </p>
 * <code>
 * tornado --threadInfo -m tornado.examples/uk.ac.manchester.tornado.examples.compute.Mandelbrot
 * </code>
 */
public class Mandelbrot {
    // CHECKSTYLE:OFF

    public static final int SIZE = 1024;
    public static final boolean USE_TORNADO = true;

    public static void main(String[] args) {
        JFrame frame = new JFrame("Mandelbrot Example within Tornado");

        frame.addWindowListener(new WindowAdapter() {
            @Override
            public void windowClosing(WindowEvent event) {
                System.exit(0);
            }
        });

        frame.add(new MandelbrotImage());
        frame.pack();
        frame.setVisible(true);
    }

    @SuppressWarnings("serial")
    public static class MandelbrotImage extends Component {

        private BufferedImage image;

        public MandelbrotImage() {
        }

        private static ShortArray mandelbrotSequential(int size) {
            final int iterations = 10000;
            float space = 2.0f / size;

            ShortArray result = new ShortArray(size * size);

            for (int i = 0; i < size; i++) {
                int indexIDX = i;
                for (int j = 0; j < size; j++) {

                    int indexJDX = j;

                    float Zr = 0.0f;
                    float Zi = 0.0f;
                    float Cr = (1 * indexJDX * space - 1.5f);
                    float Ci = (1 * indexIDX * space - 1.0f);

                    float ZrN = 0;
                    float ZiN = 0;
                    int y;

                    for (y = 0; y < iterations && ZiN + ZrN <= 4.0f; y++) {
                        Zi = 2.0f * Zr * Zi + Ci;
                        Zr = 1 * ZrN - ZiN + Cr;
                        ZiN = Zi * Zi;
                        ZrN = Zr * Zr;
                    }
                    short r = (short) ((float)(y * 255 * 255) / iterations);
                    result.set(i * size + j, r);
                }
            }
            return result;
        }

        private static void mandelbrotTornado(int size, ShortArray output) {
            final int iterations = 10000;
            float space = 2.0f / size;

            for (@Parallel int i = 0; i < size; i++) {
                for (@Parallel int j = 0; j < size; j++) {
                    float Zr = 0.0f;
                    float Zi = 0.0f;
                    float Cr = (1 * j * space - 1.5f);
                    float Ci = (1 * i * space - 1.0f);
                    float ZrN = 0;
                    float ZiN = 0;
                    int y = 0;
                    for (int ii = 0; ii < iterations; ii++) {
                        if (ZiN + ZrN <= 4.0f) {
                            Zi = 2.0f * Zr * Zi + Ci;
                            Zr = 1 * ZrN - ZiN + Cr;
                            ZiN = Zi * Zi;
                            ZrN = Zr * Zr;
                            y++;
                        } else {
                            ii = iterations;
                        }
                    }
                    short r = (short) ((float)(y * 255 * 255) / iterations);
                    output.set(i * size + j, r);
                }
            }
        }

        private static BufferedImage writeFile(ShortArray output, int size) {
            BufferedImage img = null;
            try {
                img = new BufferedImage(size, size, BufferedImage.TYPE_INT_BGR);
                WritableRaster write = img.getRaster();
                File outputFile = new File("/tmp/mandelbrot.png");

                for (int i = 0; i < size; i++) {
                    for (int j = 0; j < size; j++) {
                        int colour = output.get((i * size + j));
                        write.setSample(i, j, 0, colour);
                    }
                }
                ImageIO.write(img, "PNG", outputFile);
            } catch (Exception e) {
                e.printStackTrace();
            }
            return img;
        }

        @Override
        public void paint(Graphics g) {
            if (!USE_TORNADO) {
                ShortArray mandelbrotSequential = mandelbrotSequential(SIZE);
                this.image = writeFile(mandelbrotSequential, SIZE);
            } else {
                ShortArray result = new ShortArray(SIZE * SIZE);
                TaskGraph taskGraph = new TaskGraph("s0") //
                        .task("t0", MandelbrotImage::mandelbrotTornado, SIZE, result) //
                        .transferToHost(DataTransferMode.EVERY_EXECUTION, result);

                ImmutableTaskGraph immutableTaskGraph = taskGraph.snapshot();
                TornadoExecutionPlan executor = new TornadoExecutionPlan(immutableTaskGraph);
                executor.execute();

                this.image = writeFile(result, SIZE);
            }
            // draw the image
            g.drawImage(this.image, 0, 0, null);
        }

        @Override
        public Dimension getPreferredSize() {
            if (image == null) {
                return new Dimension(SIZE, SIZE);
            } else {
                return new Dimension(image.getWidth(), image.getHeight());
            }
        }
    }
}
// CHECKSTYLE:ON

[otabuzzman]

[jjfumero]

Wow, nice! thank you for the patch! Feel free to push PRs. We welcome contributions.

Is the tutorial going to be in German? or also in English?

[otabuzzman]

In German. They once had an English edition, but it didn't sell and it was shut down. Am Di., 19. Dez. 2023 um 19:43 Uhr schrieb Juan Fumero < ***@***.***>:

…

Wow, nice! thank you for the patch! Feel free to push PRs. We welcome contributions. Is the tutorial going to be in German? or also in English? — Reply to this email directly, view it on GitHub <#302 (reply in thread)>, or unsubscribe <https://github.com/notifications/unsubscribe-auth/AD7PMXALIFPZUQHU4VVPLEDYKHN3LAVCNFSM6AAAAABA3TCNQ2VHI2DSMVQWIX3LMV43SRDJONRXK43TNFXW4Q3PNVWWK3TUHM3TSMBQHA3TS> . You are receiving this because you authored the thread.Message ID: ***@***.***>

[otabuzzman]

Regarding an English version of the tutorial: I could send you online translations of the articles as they are published and ask the editor to use them, for example on the TornadoVM website. I'm also thinking about a blog about parallelizing Java apps based on a project of my own. I did it a few years ago and used [PJ2]( https://www.cs.rit.edu/~ark/pj2.shtml) to parallelize the sequential version using CUDA. I also wrote about it at the time and also made an [infographic]( https://create.piktochart.com/output/25355418-java-parallelization-with-cuda) about it. My idea is to re-do the project this time using TornadoVM. If you want, feel free to share your ideas for a tutorial with me. I would be happy to take these into account so that you can use them as you wish. Am Di., 19. Dez. 2023 um 19:43 Uhr schrieb Juan Fumero < ***@***.***>:

…

Wow, nice! thank you for the patch! Feel free to push PRs. We welcome contributions. Is the tutorial going to be in German? or also in English? — Reply to this email directly, view it on GitHub <#302 (reply in thread)>, or unsubscribe <https://github.com/notifications/unsubscribe-auth/AD7PMXALIFPZUQHU4VVPLEDYKHN3LAVCNFSM6AAAAABA3TCNQ2VHI2DSMVQWIX3LMV43SRDJONRXK43TNFXW4Q3PNVWWK3TUHM3TSMBQHA3TS> . You are receiving this because you authored the thread.Message ID: ***@***.***>

[jjfumero]

Thanks for the info. All these projects look good. We welcome new examples and use-cases, and we are happy to help. Some of them for inspiration:

Ray-Tracer: we recently created a Ray-Tracer in pure Java (without external native GPU libraries) with TornadoVM:
https://github.com/beehive-lab/TornadoVM-Ray-Tracer/

We also have a library for running SLAM applications: https://github.com/beehive-lab/kfusion-tornadovm/
This application is very old and now we mainly use it for benchmarking.

[otabuzzman]

I created an installer script for Windows. It uses Windows tools instead of MSys2 with Linux tools. The changes to the script also affect the SPIR-V toolkit. The contribution policy states that PRs must target develop branches, but there is no such branch in the SPIR-V Toolkit repo. Will you add a develop branch to the repo or should the PR target the main branch? By the way, do you think it's even worth a PR (perhaps because TornadoVM is primarily intended for Linux)? Regards, Jürgen. Juan Fumero ***@***.***> schrieb am Mi. 20. Dez. 2023 um 10:12:

…

Thanks for the info. All these projects look good. We welcome new examples and use-cases, and we are happy to help. Some of them for inspiration: *Ray-Tracer*: we recently created a Ray-Tracer in pure Java (without external native GPU libraries) with TornadoVM: https://github.com/beehive-lab/TornadoVM-Ray-Tracer/ We also have a library for running *SLAM* applications: https://github.com/beehive-lab/kfusion-tornadovm/ This application is very old and now we mainly use it for benchmarking. — Reply to this email directly, view it on GitHub <#302 (reply in thread)>, or unsubscribe <https://github.com/notifications/unsubscribe-auth/AD7PMXDBZJZJZYMCUSTNIYTYKKTYXAVCNFSM6AAAAABA3TCNQ2VHI2DSMVQWIX3LMV43SRDJONRXK43TNFXW4Q3PNVWWK3TUHM3TSMBWGA3DM> . You are receiving this because you authored the thread.Message ID: ***@***.***>

[jjfumero]

Hi @otabuzzman . Do you mean that you improved the TornadoVM installer? or you improved the Mandelbrot example?

Either case, I think it should be fine by submitting a PR to the TornadoVM project. Regarding the SPIR-V Toolkit, you can open it in the master branch.

[otabuzzman]

Neither... :-) I stupidly used this old thread for my question, and will start a new one if necessary. It is an installation script intended exclusively for Windows. Anyway, I'll open the PR in the master branch of the SPIR-V toolkit. However, I want to do some tests first… Juan Fumero ***@***.***> schrieb am Sa. 16. März 2024 um 22:03:

…

Hi @otabuzzman <https://github.com/otabuzzman> . Do you mean that you improved the TornadoVM installer? or you improved the Mandelbrot example? Either case, I think it should be fine by submitting a PR to the TornadoVM project. Regarding the SPIR-V Toolkit, you can open it in the master branch. — Reply to this email directly, view it on GitHub <#302 (reply in thread)>, or unsubscribe <https://github.com/notifications/unsubscribe-auth/AD7PMXC7DUE7FUNLJKXFSBTYYSXSFAVCNFSM6AAAAABA3TCNQ2VHI2DSMVQWIX3LMV43SRDJONRXK43TNFXW4Q3PNVWWK3TUHM4DQMJUGM3DS> . You are receiving this because you were mentioned.Message ID: ***@***.***>