arm-software / vulkan-sdk

Github repository for the Vulkan SDK

Vulkan SDK for Android

---

This project is not supported. For the latest tutorials and samples please go to:

• Vulkan-Samples

---

Introduction

The Vulkan Software Development Kit is a collection of resources to help you build Vulkan applications for a platform with a Mali GPU and an ARM processor. You can use it for creating new applications, training, and exploration of implementation possibilities.

Setting up development environment

Minimum NDK requirements

Minimum android-ndk-r12 is required. The NDK can be downloaded from Android Studio 2.2 or later. It is recommended to use the NDK provided in Android Studio.

OS requirements

The Vulkan SDK has been tested to build on Linux (Android Studio) and Windows (Android Studio). Partial support for running on Linux desktop is also included.

Android requirements

Not all Android devices support Vulkan. To make sure your Android device has the Vulkan API supported download the Hardware CapsViewer for Vulkan app to verify your device's status.

License

The software is provided under an MIT license. Contributions to this project are accepted under the same license.

Trademarks

Vulkan is a registered trademark of the Khronos Group Inc.

Building

Check out submodules

This repo uses GLM and STB as submodules, before building, make sure you pull those in.

git submodule init
git submodule update

Build and run samples from Android Studio

• Open Android Studio 3.4 or newer
• Open an existing Android Studio project
• Import a sample project
• You might be prompted to update or install the Gradle wrapper. Do so if asked.
• You might be prompted to download and/or update Android SDK tools if Android Studio has not downloaded these before.
• Under Tools -> Android -> SDK manager, install cmake, lldb and NDK components if these are not installed already.
• In the top menu, run Build -> Make Project.
• In the bottom of the screen, the Messages tab should display a build log which shows that libnative.so has been built and that build was successful.
• Run the app on the device by pressing the Play button on the top toolbar.
• To debug the code, Run -> Debug app. In the project view, you will find app/cpp/native/hellotriangle.cpp or similar. You can set breakpoints and step through the code.

Build samples for desktop Linux

mkdir build
cd build
cmake ..
make -j8

will build samples with a PNG backend. Running the binary on desktop should dump out PNG images. This is useful when developing samples and for creating screenshots.

X11 or Wayland backends can be used instead on Linux by passing in extra parameters to cmake:

cmake .. -DPLATFORM=wayland   # or xcb for X11

Documentation

For online tutorials, documentation and explanation of the samples, see Vulkan SDK documentation.

To build the same documentation for yourself for offline use, build Doxygen documentation with ./build_documentation.sh. This requires Doxygen to be installed on your machine.

Adding new samples

The build system for samples is designed to be as general as possible. To create a new sample based on hellotriangle:

cd samples
$EDITOR CMakeLists.txt             # add_subdirectory(newsample)
mkdir newsample
cp -r hellotriangle/{CMakeLists.txt,app,build.gradle,settings.gradle} newsample/
$EDITOR CMakeLists.txt             # Edit add_sample
$EDITOR app/AndroidManifest.xml    # Edit manifest:package
$EDITOR app/res/values/strings.xml # Edit resources:string

Source files go in newsample/, GLSL source files go in newsample/shaders and general assets (if needed) go in newsample/assets.

Samples must implement the VulkanApplication interface as well as implementing MaliSDK::create_application().

#include "framework/application.hpp"
#include "framework/context.hpp"
#include "framework/common.hpp"
#include "framework/assets.hpp"
#include "platform/platform.hpp"

class MyVulkanApplication : public VulkanApplication
{
    // ...
};

VulkanApplication* MaliSDK::create_application()
{
    return new MyVulkanApplication();
}

vulkan-sdk's Issues

Header image for README.md

I hava a problem about barrier scope. Can you help me?

/////////// A set of commands
vkCmd..(…)
…
vkCmd.. (…)
////////////

///////Barrier1
vkCmdPipelineBarrier (…)
///////

///////// B set of commands
vkCmd…(…)
…
vkQueueSubmit(…)
vkQueuePresentKHR(..)
//////////

/////////Barrier2
vkCmdPipelineBarrier (…)
////////

//////C set of commands
vkCmd…(…)
…
vkQueueSubmit(…)
vkQueuePresentKHR(..)
//////

I use one queue to submit commands.

I want to know:

Do the "Barrier1" act on "B set of commands" and "Barrier2" and "C set of commands"?

Thanks you!

why one queue ?

thanks

Tell people to verify capabilities of phone

I personally have run into trying to use the Vulkan SDK with my Galaxy S6 with the Mali-T760 GPU to find that it was not Vulkan supported.

I think giving people a link to the Hardware CapsViewer for Vulkan app will save many people hours of headaches from finding out they can't even run the SDK on their android device. All these Vulkan pages neglect to remind people that not all phones work with it.

projects fail to compile with android studio 3.0

running android studio 3.0. on win 10 (latest update).
projects fail to build. lots cmake and Gradle errors.

I miss a makefile for maked your github

Hello,

I miss a makefile for 'make' your GitHub :

Thank you in advance to bring to myself a makefile,

Regards.

Azaretdodo.

Why do write-after-read and read-after-read hazards only require execution dependencies to synchronize?

Memory dependencies are used to solve data hazards, e.g. to ensure that write operations are visible to subsequent read operations (read-after-write hazard), as well as write-after-write hazards. Write-after-read and read-after-read hazards only require execution dependencies to synchronize.

Why?
Can you help me? Thanks

Snap Package Support (Feature Request)

I was thinking the install and setup could be improved for Ubuntu and other Linux distros by having a snap package version. If it's not a priority would you mind if I made the snap as a contribution?

Crafting your first snap

Build and test snap

Getting Started snap template

Incorrect external dependency in hellotriangle sample

This code:
https://github.com/ARM-software/vulkan-sdk/blob/master/samples/hellotriangle/hellotriangle.cpp#L211-L226

Has misleading comments and incorrect dstAccessMask. As per offline discussion with @Tobski, we don't need a write-after-read barrier here - the semaphore establishes that - but we need a write-after-write dependency to make sure that the writes to color attachment don't get reordered with the (implied) write operations from the image layout transition. This requires dstAccessMask to be COLOR_ATTACHMENT_OPTIMAL.

We haven't discussed this so I'm less sure about this but it looks like this also needs an external dependency at the end of the pass (so 0 -> EXTERNAL) to establish a read-after-write barrier to make sure that (implied) read operations from the image layout transition wait until the color attachment writes complete.

One day validation layers will check for things like this.

About "dependency.srcAccessMask = 0;" question

I hava a question about "srcAccessMask = 0". Is it say "can't read and write"?

weston 10.0 deprecated wl-shell and will remove in later version

wl-shell deprecated caused the vulkan sdk app can not run on wayland platform.

Could you have any plan to convert wl-shell to xdg-shell?
https://gitlab.freedesktop.org/wayland/weston/-/merge_requests/559

"This is bad because the presentation engine can still be reading from the image. ", why?

The image layout transition as part of a subpass will happen without any waiting.
Why "This is bad because the presentation engine can still be reading from the image. "?

I have manualy installed glm and stb and keep stb asked by 'cmake ..'

Hello,

I have manualy installed glm and stb and keep stb asked by 'cmake ..' :

Why I am kept stb asked ?

Thank you in advance to help myself go more far,

Regards.

Azaretdodo.

when I make, some mistake happend

root@linaro-alip:~/jack_vulkan/vulkan-sdk/build# make
[ 3%] Built target vulkan-stub
[ 8%] Built target platform-png
[ 11%] Built target platform-asset-manager
[ 14%] Built target platform
[ 25%] Built target framework
[ 30%] Built target vulkan-sdk
[ 33%] Built target platform-wsi
[ 35%] Generating assets/shaders/triangle.frag.spv
/bin/sh: GLSLC-NOTFOUND: command not found
samples/hellotriangle/CMakeFiles/hellotriangle.dir/build.make:67: recipe for target 'samples/hellotriangle/assets/shaders/triangle.frag.spv' failed
make[2]: *** [samples/hellotriangle/assets/shaders/triangle.frag.spv] Error 127
CMakeFiles/Makefile2:470: recipe for target 'samples/hellotriangle/CMakeFiles/hellotriangle.dir/all' failed
make[1]: *** [samples/hellotriangle/CMakeFiles/hellotriangle.dir/all] Error 2
Makefile:94: recipe for target 'all' failed
make: *** [all] Error 2

this error is always exist, my platform is ARM 3399, on Ubuntu
I need some help

Release ARM Mali userspace driver on github

Thanks for this release, but it's not only this what everyone in the opensource community is looking for

About The thread pool.......

I think the queue that is consist all the interesting jobs only for one thread don't need the "mutex".

Android Studio compilation error: \vulkan-sdk\platform\android\android_assets.cpp:21:

Hi! i tried to compile multisampling sample using Android studio 3.4.2.
I have error during compilation native code using cmake.
I Attached Build.log

Can I build it in aarch64 ?

Now I have a rk3399 and linux runs on it. I just want to use mali t860 to test vulkan. Can I make it?

About the num of multisampledRenderTargets in the sample that is named "Multisampling in Vulkan".

Why only have one multisampledRenderTarget that is used to rendering？ Why not it is the number of mulitsampledRenderTargets that is euqal to the number of swapchians to avoids all hazards of "Asynchronous GPU"?

If I hava uniform buffer that will be updated by dynamic, I hope not need 3x uniform. How do?

If I hava uniform buffer that will be updated by dynamic, I hope not need 3x uniform. How do?

I think I can use some barriers to avoid this problem.

the issue of imageMemoryBarrier that is in multithreading.cpp

In multithreading.cpp. The the method of "createTextureFromAsset" .

imageMemoryBarrier(cmd, image, 0, VK_ACCESS_TRANSFER_WRITE_BIT, VK_PIPELINE_STAGE_TOP_OF_PIPE_BIT,
VK_PIPELINE_STAGE_TOP_OF_PIPE_BIT, VK_IMAGE_LAYOUT_UNDEFINED,
VK_IMAGE_LAYOUT_TRANSFER_DST_OPTIMAL);

the srcstage is VK_PIPELINE_STAGE_TOP_OF_PIPE_BIT,
the dststage is VK_PIPELINE_STAGE_TOP_OF_PIPE_BIT
it seem like this is a bug.

is it a bug in multipass.cpp?

// These cubes do not intersect with the camera.
static const vec4 lightPositions[] = {
	vec4(-20.0f, 20.0f, -30.0f, 25.0f), vec4(-20.0f, -20.0f, -30.0f, 20.0f), vec4(+20.0f, +20.0f, -15.0f, 12.0f),
	vec4(+20.0f, -20.0f, -15.0f, 12.0f),
};

I use codes to test, the code is:

bool checkPostInCamera(const mat4& vpMat, const vec4& pos) { std::vector vertices = { // Front face vec4(-1.0f, -1.0f, +1.0f, 1.0f), vec4(+1.0f, -1.0f, +1.0f, 1.0f), vec4(-1.0f, +1.0f, +1.0f, 1.0f), vec4(+1.0f, +1.0f, +1.0f, 1.0f),

            // Back face
            vec4(+1.0f, -1.0f, -1.0f, 1.0f),
            vec4(-1.0f, -1.0f, -1.0f, 1.0f),
            vec4(+1.0f, +1.0f, -1.0f, 1.0f),
            vec4(-1.0f, +1.0f, -1.0f, 1.0f)
    };

    for(int i =0; i < vertices.size(); ++i)
    {

        vec4 v = vec4(vertices[i].xyz() * pos.w, 1.0) + vec4(pos.xyz(), 0.0f);


        vec4 ndcPos = vpMat  * v;
        float xPos = ndcPos.x;
        float yPos = ndcPos.y;
        float zPos =  ndcPos.z;
        float wPos =  ndcPos.w;
        ndcPos = ndcPos / ndcPos.w;

        xPos = ndcPos.x;
        yPos = ndcPos.y;
        zPos =  ndcPos.z;
        wPos =  ndcPos.w;

        if (fabs(ndcPos.x) < 1.0f && fabs(ndcPos.y) < 1.0f  && fabs(ndcPos.z) < 1.0f)
        {
            return true;
        }
    }

    return false;
}

and the use case is:
auto result = checkPostInCamera(mvp[1], lightPositions[0]);

It is intersect with the camera!

anything isn't declare

Hello,

Anythink isn't declare in your make command line :

Thank you in advance to repair your error,

Regards.

Azaretdodo.

The promble of the uniform buffer in in the sample that is named "multipass"

The initialization of the uniform buffer is :
A：
uboAlignment = std::max(size_t(pContext->getPlatform().getGpuProperties().limits.minUniformBufferOffsetAlignment),
sizeof(mat4));
uniformBuffer = createBuffer(nullptr, backbuffers.size() * uboAlignment, VK_BUFFER_USAGE_UNIFORM_BUFFER_BIT);
VK_CHECK(vkMapMemory(device, uniformBuffer.memory, 0, VK_WHOLE_SIZE, 0, reinterpret_cast<void **>(&uboData)));

The VkDescriptorBufferInfo is :

B:
VkDescriptorBufferInfo uboInfo = { uniformBuffer.buffer, 0, sizeof(mat4) }.

The above codes is in the sample that is named "multipass".

About the B. Why not " VkDescriptorBufferInfo uboInfo = { uniformBuffer.buffer, 0, backbuffers.size() * uboAlignment }." or " VkDescriptorBufferInfo uboInfo = { uniformBuffer.buffer, 0, VK_WHOLE_SIZE }."

VkDescriptorBufferInfo uboInfo = { uniformBuffer.buffer, 0, sizeof(mat4) }