GPU deployment using Vulkan

IREE can accelerate model execution on GPUs via Vulkan, a low-overhead graphics and compute API. Vulkan is cross-platform: it is available on many operating systems, including Android, Linux, and Windows. Vulkan is also cross-vendor: it is supported by most GPU vendors, including AMD, ARM, Intel, NVIDIA, and Qualcomm.

Support matrix

As IREE and the compiler ecosystem it operates within matures, more target specific optimizations will be implemented. At this stage, expect reasonable performance across all GPUs and for improvements to be made over time for specific vendors and architectures.

GPU Vendor	Category	Performance	Focus Architecture
ARM Mali GPU	Mobile	Good	Valhall+
Qualcomm Adreno GPU	Mobile	Reasonable	640+
AMD GPU	Desktop/server	Good	RDNA+
NVIDIA GPU	Desktop/server	Good	Turing+

Prerequisites

In order to use Vulkan to drive the GPU, you need to have a functional Vulkan environment. IREE requires Vulkan 1.1 on Android and 1.2 elsewhere. It can be verified by the following steps:

Android Linux Windows

Android mandates Vulkan 1.1 support since Android 10. You just need to make sure the device's Android version is 10 or higher.

Run the following command in a shell:

vulkaninfo | grep apiVersion

If vulkaninfo does not exist, you will need to install the latest Vulkan SDK. Installing via LunarG's package repository is recommended, as it places Vulkan libraries and tools under system paths so it's easy to discover.

If the listed version is lower than Vulkan 1.2, you will need to update the driver for your GPU.

Run the following command in a shell:

vulkaninfo | grep apiVersion

If vulkaninfo does not exist, you will need to install the latest Vulkan SDK.

If the listed version is lower than Vulkan 1.2, you will need to update the driver for your GPU.

Get the IREE compiler

Vulkan expects the program running on GPU to be expressed by the SPIR-V binary exchange format, which the model must be compiled into.

Download the compiler from a release

Python packages are regularly published to PyPI. See the Python Bindings page for more details. The core iree-compiler package includes the SPIR-V compiler:

Stable releases Nightly releases

Stable release packages are published to PyPI.

python -m pip install iree-compiler

Nightly releases are published on GitHub releases.

python -m pip install \
  --find-links https://iree.dev/pip-release-links.html \
  --upgrade iree-compiler

Tip

iree-compile is installed to your python module installation path. If you pip install with the user mode, it is under ${HOME}/.local/bin, or %APPDATA%Python on Windows. You may want to include the path in your system's PATH environment variable:

export PATH=${HOME}/.local/bin:${PATH}

Build the compiler from source

Please make sure you have followed the Getting started page to build IREE for your host platform and the Android cross-compilation page if you are cross compiling for Android. The SPIR-V compiler backend is compiled in by default on all platforms.

Ensure that the IREE_TARGET_BACKEND_VULKAN_SPIRV CMake option is ON when configuring for the host.

Tip

iree-compile will be built under the iree-build/tools/ directory. You may want to include this path in your system's PATH environment variable.

Get the IREE runtime

Next you will need to get an IREE runtime that supports the Vulkan HAL driver.

You can check for Vulkan support by looking for a matching driver and device:

$ iree-run-module --list_drivers

        cuda: CUDA (dynamic)
  local-sync: Local execution using a lightweight inline synchronous queue
  local-task: Local execution using the IREE multithreading task system
      vulkan: Vulkan 1.x (dynamic)

$ iree-run-module --list_devices

  cuda://GPU-00000000-1111-2222-3333-444444444444
  local-sync://
  local-task://
  vulkan://00000000-1111-2222-3333-444444444444

Build the runtime from source

Please make sure you have followed the Getting started page to build IREE for Linux/Windows and the Android cross-compilation page for Android. The Vulkan HAL driver is compiled in by default on non-Apple platforms.

Ensure that the IREE_HAL_DRIVER_VULKAN CMake option is ON when configuring for the target.

Compile and run a program

With the SPIR-V compiler and Vulkan runtime, we can now compile programs and run them on GPUs.

Compile a program

The IREE compiler transforms a model into its final deployable format in many sequential steps. A model authored with Python in an ML framework should use the corresponding framework's import tool to convert into a format (i.e., MLIR) expected by the IREE compiler first.

Using MobileNet v2 as an example, you can download the SavedModel with trained weights from TensorFlow Hub and convert it using IREE's TensorFlow importer. Then run the following command to compile with the vulkan-spirv target:

iree-compile \
    --iree-hal-target-backends=vulkan-spirv \
    --iree-vulkan-target-triple=<...> \
    mobilenet_iree_input.mlir -o mobilenet_vulkan.vmfb

Note

Currently a target triple of the form <vendor/arch>-<product>-<os> is needed to compile towards a specific GPU architecture.

We don't support the full spectrum here(1); the following table summarizes the currently recognized ones.

If no triple is specified, then a safe but more limited default will be used.

This is more of a mechanism to help us develop IREE itself--in the long term we want to perform multiple targetting to generate to multiple architectures if no target triple is given.

1. It's also impossible to capture all details of a Vulkan implementation with a target triple, given the allowed variances on extensions, properties, limits, etc. So the target triple is just an approximation for usage.

GPU Vendor	Target Triple
ARM Mali GPU	e.g. valhall-unknown-{android30|android31}
Qualcomm Adreno GPU	e.g. adreno-unknown-{android30|android31}
AMD GPU	e.g. {rdna1|rdna2|rdna3}-unknown-unknown
NVIDIA GPU	e.g. {turing|ampere}-unknown-unknown
SwiftShader CPU	cpu-swiftshader-unknown

Run a compiled program

In the build directory, run the following command:

tools/iree-run-module \
    --device=vulkan \
    --module=mobilenet_vulkan.vmfb \
    --function=predict \
    --input="1x224x224x3xf32=0"

The above assumes the exported function in the model is named as predict and it expects one 224x224 RGB image. We are feeding in an image with all 0 values here for brevity, see iree-run-module --help for the format to specify concrete values.