ash | ||
examples | ||
vk_loader | ||
vk_loader2 | ||
.gitignore | ||
Cargo.toml1 | ||
LICENSE-APACHE | ||
LICENSE-MIT | ||
README.md |
#Ash A low level Vulkan wrapper for Rust
Why Ash?
Functions return a type VkResult<T> = Result<T, vk::Result>
instead of an error code. No mutable references for the output are required.
pub fn create_swapchain_khr(&self,
create_info: &vk::SwapchainCreateInfoKHR)
-> VkResult<vk::SwapchainKHR>;
let swapchain = device.create_swapchain_khr(&swapchain_create_info).unwrap();
Always returns a Vec<T>
for functions that output multiple values.
pub fn get_swapchain_images_khr(&self,
swapchain: vk::SwapchainKHR)
-> VkResult<Vec<vk::Image>>;
let present_images = device.get_swapchain_images_khr(swapchain).unwrap();
Ash always uses slices in functions.
// C
void vkCmdPipelineBarrier(
VkCommandBuffer commandBuffer,
VkPipelineStageFlags srcStageMask,
VkPipelineStageFlags dstStageMask,
VkDependencyFlags dependencyFlags,
uint32_t memoryBarrierCount,
const VkMemoryBarrier* pMemoryBarriers,
uint32_t bufferMemoryBarrierCount,
const VkBufferMemoryBarrier* pBufferMemoryBarriers,
uint32_t imageMemoryBarrierCount,
const VkImageMemoryBarrier* pImageMemoryBarriers);
// Rust
pub fn cmd_pipeline_barrier(&self,
command_buffer: vk::CommandBuffer,
src_stage_mask: vk::PipelineStageFlags,
dst_stage_mask: vk::PipelineStageFlags,
dependency_flags: vk::DependencyFlags,
memory_barriers: &[vk::MemoryBarrier],
buffer_memory_barriers: &[vk::BufferMemoryBarrier],
image_memory_barriers: &[vk::ImageMemoryBarrier]);
device.cmd_pipeline_barrier(setup_command_buffer,
vk::PIPELINE_STAGE_TOP_OF_PIPE_BIT,
vk::PIPELINE_STAGE_TOP_OF_PIPE_BIT,
vk::DependencyFlags::empty(),
&[],
&[],
&[layout_transition_barrier]);
Ash still uses raw Vulkan structs. The only difference is type safety. Everything that can be an enum is an enum like vk::StructureType
, flags are implemented similar to the Bitflags
crate. Ash also follows the Rust style guide. The reason that Ash uses raw Vulkan structs is to be extensible, just like the Vulkan spec.
let pool_create_info = vk::CommandPoolCreateInfo {
s_type: vk::StructureType::CommandPoolCreateInfo,
p_next: ptr::null(),
flags: vk::COMMAND_POOL_CREATE_RESET_COMMAND_BUFFER_BIT,
queue_family_index: queue_family_index,
};
let pool = device.create_command_pool(&pool_create_info).unwrap();
Ash also takes care of loading the function pointers. Function pointers are split into 3 categories. Entry, Instance and Device. The reason for not loading it into a global is that in Vulkan you can have multiple devices and each device must load its own function pointers.
// Looks for the vulkan lib in your path, alternatively you can supply the path explicitly.
let entry = Entry::load_vulkan().unwrap();
let instance: Instance = entry.create_instance(&create_info).expect("Instance creation error");
let device: Device = instance.create_device(pdevice, &device_create_info)
.unwrap();
You don't have to pass an Instance or Device handle anymore, this is done implicitly for you.
// C
VkResult vkCreateCommandPool(
VkDevice device,
const VkCommandPoolCreateInfo* pCreateInfo,
const VkAllocationCallbacks* pAllocator,
VkCommandPool* pCommandPool);
// Rust
pub fn create_command_pool(&self,
create_info: &vk::CommandPoolCreateInfo)
-> VkResult<vk::CommandPool>;
let pool = device.create_command_pool(&pool_create_info).unwrap();
Complete
In progress
- Wrapping the complete spec
- Optional extension loading
Not started
- Custom allocators