jWebGPU

Java bindings for WebGPU on desktop (JNI, FFM and TeaVM C), Android and the browser.

Get started

• Run the examples: jWebGPU-Examples includes desktop, Android and browser setup.
• Try it: browser demo or Android app.

Add to your project

Use these repositories and the same version for all jWebGPU dependencies, then choose the block for your platform:

repositories {
    mavenCentral()
    maven("https://central.sonatype.com/repository/maven-snapshots/")
}

val jWebGPUVersion = "-SNAPSHOT"

Shared core module

For code shared between platforms, add include(":core") to settings.gradle.kts. In core/build.gradle.kts, use the repositories and version above with:

plugins {
    `java-library`
}

dependencies {
    compileOnlyApi("com.github.xpenatan.jWebGPU:webgpu-core:$jWebGPUVersion")
}

In each platform module, add the shared project alongside its platform dependencies below:

dependencies {
    implementation(project(":core"))
}

compileOnlyApi exposes the shared API for compilation without packaging its stubs as the native runtime. Each platform supplies its implementation; jParser dependencies are transitive.

Desktop JNI

dependencies {
    implementation("com.github.xpenatan.jWebGPU:webgpu-desktop-jni:$jWebGPUVersion")
    runtimeOnly("com.github.xpenatan.jWebGPU:webgpu-desktop-jni-wgpu_windows_x64:$jWebGPUVersion")
}

Desktop FFM

Requires JDK 25.

dependencies {
    compileOnly("com.github.xpenatan.jWebGPU:webgpu-core:$jWebGPUVersion")
    implementation("com.github.xpenatan.jWebGPU:webgpu-desktop-ffm:$jWebGPUVersion")
    runtimeOnly("com.github.xpenatan.jWebGPU:webgpu-desktop-ffm-wgpu_windows_x64:$jWebGPUVersion")
}

For JNI/FFM, run with --enable-native-access=ALL-UNNAMED. Initialize through JWebGPULoader.init(...) before using WebGPU. Do not add webgpu-core to the runtime classpath; jParser dependencies are transitive.

Android

Add google() to the repositories and set minSdk = 29 or later.

dependencies {
    implementation("com.github.xpenatan.jWebGPU:webgpu-android-wgpu:$jWebGPUVersion")
}

The AAR includes the JNI runtime and native libraries. For Dawn, use webgpu-android-dawn instead; include only one backend AAR.

Browser WebGPU

Use Java 17+ and a TeaVM application.

dependencies {
    implementation("com.github.xpenatan.jWebGPU:webgpu-web:$jWebGPUVersion")
    runtimeOnly("com.github.xpenatan.jWebGPU:webgpu-web_wasm:$jWebGPUVersion")
}

Package the JavaScript/WASM resources with the TeaVM output and serve over localhost or HTTPS. The browser example includes the build and server configuration.

Desktop TeaVM C

Use a TeaVM C application and the target platform's native compiler.

dependencies {
    implementation("com.github.xpenatan.jWebGPU:webgpu-desktop-c-wgpu_windows_x64:$jWebGPUVersion")
}

This artifact brings the Java bindings and C runtime transitively.

For all desktop native artifacts, replace windows_x64 with linux_x64, mac_x64 or mac_arm64 to match the target. Replace wgpu with dawn to use Dawn; load it with JWebGPULoader.init(JWebGPUBackend.DAWN, listener). The browser chooses its own backend.

Complete desktop JNI/FFM, Android and browser applications are in jWebGPU-Examples.

Building from source

1. Install tools and generate bindings

Use JDK 25 with JAVA_HOME set. Install the tools for the target you want to build:

Target	Required tools
Windows x64	Visual Studio C++ Build Tools and Windows SDK; use an x64 Developer terminal
Linux x64	GCC/G++ and binutils (build-essential on Ubuntu/Debian)
macOS x64 / arm64	Xcode Command Line Tools (xcode-select --install)
Android	Android SDK 36 and NDK r27; set ANDROID_HOME and ANDROID_NDK_HOME to their installed directories
Browser	Emscripten SDK 6.0.3; activate it and load ./emsdk_env.ps1 in PowerShell, emsdk_env.bat in cmd, or source ./emsdk_env.sh on Linux/macOS so EMSDK is set

The Android NDK and Emscripten versions above match the build workflow. Build desktop natives on their corresponding OS; Android CI uses Linux, and web CI uses Linux with Emscripten.

git clone https://github.com/xpenatan/jWebGPU.git
cd jWebGPU

Run from this checkout. Commands below use the Windows wrapper; on Linux/macOS use ./gradlew instead of .\gradlew.bat and run chmod +x gradlew if needed.

.\gradlew.bat :webgpu:builder:jParser_generate

2. Download and build your target

Run the downloads before the native build; build tasks do not download dependencies automatically. The desktop and Android commands below select WGPU. For Dawn, use the same commands with _wgpu replaced by _dawn. Build either backend or repeat for both.

For every desktop target, download the shared GLFW headers first. Despite its name, this task supplies the headers for Windows, Linux and macOS:

.\gradlew.bat :webgpu:download:webgpu_download_glfw_windows

Each desktop block builds JNI, FFM and TeaVM C. Omit the tasks for bridges you do not need.

Windows x64

.\gradlew.bat :webgpu:download:webgpu_download_windows_x86_64_wgpu
.\gradlew.bat :webgpu:builder:jParser_build_windows64_jni_wgpu :webgpu:builder:jParser_build_windows64_ffm_wgpu :webgpu:builder:jParser_build_windows64_teavm_c_wgpu

Linux x64

./gradlew :webgpu:download:webgpu_download_linux_x86_64_wgpu
./gradlew :webgpu:builder:jParser_build_linux64_jni_wgpu :webgpu:builder:jParser_build_linux64_ffm_wgpu :webgpu:builder:jParser_build_linux64_teavm_c_wgpu

macOS Intel (x64)

./gradlew :webgpu:download:webgpu_download_macos_x86_64_wgpu
./gradlew :webgpu:builder:jParser_build_mac64_jni_wgpu :webgpu:builder:jParser_build_mac64_ffm_wgpu :webgpu:builder:jParser_build_mac64_teavm_c_wgpu

macOS Apple Silicon (arm64)

./gradlew :webgpu:download:webgpu_download_macos_aarch64_wgpu
./gradlew :webgpu:builder:jParser_build_macArm_jni_wgpu :webgpu:builder:jParser_build_macArm_ffm_wgpu :webgpu:builder:jParser_build_macArm_teavm_c_wgpu

Android JNI — download all four ABIs before building the native libraries:

.\gradlew.bat :webgpu:download:webgpu_download_android_i686_wgpu :webgpu:download:webgpu_download_android_x86_64_wgpu
.\gradlew.bat :webgpu:download:webgpu_download_android_armv7_wgpu :webgpu:download:webgpu_download_android_aarch64_wgpu
.\gradlew.bat :webgpu:builder:jParser_build_android_jni_wgpu

Browser WebGPU — uses Emdawnwebgpu, without a WGPU/Dawn variant switch:

.\gradlew.bat :webgpu:download:webgpu_download_emdawnwebgpu
.\gradlew.bat :webgpu:builder:jParser_build_web_wasm

3. Use the build

Native output is in webgpu/builder/build/c++/libs/.

To run the local build, clone jWebGPU-Examples beside this checkout, set useLocalJWebGPU=true and jWebGPUPath=../jWebGPU in its gradle.properties, then follow its platform run commands. includeBuild compiles the Java modules and packages the required artifacts automatically; no separate JAR tasks or publishing are needed. Rebuild the bindings and native bridge after source changes.

For LibGDX integration, see gdx-webgpu.

Apache License 2.0 · Sponsor