Add an example demonstrating the basics of occlusion queries (#1548)

* Add an example demonstrating the basics of occlusion queries

* Make some functions const
This commit is contained in:
Rua 2021-04-10 18:07:15 +02:00 committed by GitHub
parent fc42de0c48
commit bb51359d32
No known key found for this signature in database
GPG Key ID: 4AEE18F83AFDEB23
3 changed files with 508 additions and 5 deletions

View File

@ -31,6 +31,7 @@
- `get_results` has been added to `QueriesRange`, to copy query results to the CPU.
- The following functions have been added to both `SyncCommandBufferBuilder` and `AutoCommandBufferBuilder`: `begin_query` (still unsafe), `end_query` (safe), `write_timestamp` (still unsafe), `copy_query_pool_results` (safe), `reset_command_pool` (still unsafe).
- Better documentation of everything in the `query` module.
- An example demonstrating occlusion queries.
- The deprecated `cause` trait function on Vulkano error types is replaced with `source`.
- Vulkano-shaders: Fixed and refined the generation of the `readonly` descriptor attribute. It should now correctly mark uniforms and sampled images as read-only, but storage buffers and images only if explicitly marked as `readonly` in the shader.
- Fixed bug in descriptor array layers check when the image is a cubemap.

View File

@ -0,0 +1,502 @@
// Copyright (c) 2016 The vulkano developers
// Licensed under the Apache License, Version 2.0
// <LICENSE-APACHE or
// https://www.apache.org/licenses/LICENSE-2.0> or the MIT
// license <LICENSE-MIT or https://opensource.org/licenses/MIT>,
// at your option. All files in the project carrying such
// notice may not be copied, modified, or distributed except
// according to those terms.
// This is a modification of the triangle example, that demonstrates the basics of occlusion queries.
// Occlusion queries allow you to query whether, and sometimes how many, pixels pass the depth test
// in a range of draw calls.
use std::sync::Arc;
use vulkano::buffer::{BufferAccess, BufferUsage, CpuAccessibleBuffer};
use vulkano::command_buffer::{AutoCommandBufferBuilder, DynamicState, SubpassContents};
use vulkano::device::{Device, DeviceExtensions, DeviceOwned};
use vulkano::format::Format;
use vulkano::image::{view::ImageView, AttachmentImage, ImageUsage, SwapchainImage};
use vulkano::instance::{Instance, PhysicalDevice};
use vulkano::pipeline::viewport::Viewport;
use vulkano::pipeline::GraphicsPipeline;
use vulkano::query::{QueryControlFlags, QueryPool, QueryResultFlags, QueryType};
use vulkano::render_pass::{Framebuffer, FramebufferAbstract, RenderPass, Subpass};
use vulkano::swapchain;
use vulkano::swapchain::{
AcquireError, ColorSpace, FullscreenExclusive, PresentMode, SurfaceTransform, Swapchain,
SwapchainCreationError,
};
use vulkano::sync;
use vulkano::sync::{FlushError, GpuFuture};
use vulkano_win::VkSurfaceBuild;
use winit::event::{Event, WindowEvent};
use winit::event_loop::{ControlFlow, EventLoop};
use winit::window::{Window, WindowBuilder};
fn main() {
let required_extensions = vulkano_win::required_extensions();
let instance = Instance::new(None, &required_extensions, None).unwrap();
let physical = PhysicalDevice::enumerate(&instance).next().unwrap();
println!(
"Using device: {} (type: {:?})",
physical.name(),
physical.ty()
);
let event_loop = EventLoop::new();
let surface = WindowBuilder::new()
.build_vk_surface(&event_loop, instance.clone())
.unwrap();
let queue_family = physical
.queue_families()
.find(|&q| q.supports_graphics() && surface.is_supported(q).unwrap_or(false))
.unwrap();
let device_ext = DeviceExtensions {
khr_swapchain: true,
..DeviceExtensions::none()
};
let (device, mut queues) = Device::new(
physical,
physical.supported_features(),
&device_ext,
[(queue_family, 0.5)].iter().cloned(),
)
.unwrap();
let queue = queues.next().unwrap();
let (mut swapchain, images) = {
let caps = surface.capabilities(physical).unwrap();
let alpha = caps.supported_composite_alpha.iter().next().unwrap();
let format = caps.supported_formats[0].0;
let dimensions: [u32; 2] = surface.window().inner_size().into();
Swapchain::new(
device.clone(),
surface.clone(),
caps.min_image_count,
format,
dimensions,
1,
ImageUsage::color_attachment(),
&queue,
SurfaceTransform::Identity,
alpha,
PresentMode::Fifo,
FullscreenExclusive::Default,
true,
ColorSpace::SrgbNonLinear,
)
.unwrap()
};
let vertex_buffer = {
#[derive(Default, Debug, Clone)]
struct Vertex {
position: [f32; 3],
color: [f32; 3],
}
vulkano::impl_vertex!(Vertex, position, color);
CpuAccessibleBuffer::from_iter(
device.clone(),
BufferUsage::all(),
false,
[
// The first triangle (red) is the same one as in the triangle example.
Vertex {
position: [-0.5, -0.25, 0.5],
color: [1.0, 0.0, 0.0],
},
Vertex {
position: [0.0, 0.5, 0.5],
color: [1.0, 0.0, 0.0],
},
Vertex {
position: [0.25, -0.1, 0.5],
color: [1.0, 0.0, 0.0],
},
// The second triangle (cyan) is the same shape and position as the first,
// but smaller, and moved behind a bit.
// It should be completely occluded by the first triangle.
// (You can lower its z value to put it in front)
Vertex {
position: [-0.25, -0.125, 0.6],
color: [0.0, 1.0, 1.0],
},
Vertex {
position: [0.0, 0.25, 0.6],
color: [0.0, 1.0, 1.0],
},
Vertex {
position: [0.125, -0.05, 0.6],
color: [0.0, 1.0, 1.0],
},
// The third triangle (green) is the same shape and size as the first,
// but moved to the left and behind the second.
// It is partially occluded by the first two.
Vertex {
position: [-0.25, -0.25, 0.7],
color: [0.0, 1.0, 0.0],
},
Vertex {
position: [0.25, 0.5, 0.7],
color: [0.0, 1.0, 0.0],
},
Vertex {
position: [0.5, -0.1, 0.7],
color: [0.0, 1.0, 0.0],
},
]
.iter()
.cloned(),
)
.unwrap()
};
// Create three buffer slices, one for each triangle.
let buffer_slice = vertex_buffer.into_buffer_slice();
let triangle1 = buffer_slice.clone().slice(0..3).unwrap();
let triangle2 = buffer_slice.clone().slice(3..6).unwrap();
let triangle3 = buffer_slice.clone().slice(6..9).unwrap();
// Create a query pool for occlusion queries, with 3 slots.
let query_pool = Arc::new(QueryPool::new(device.clone(), QueryType::Occlusion, 3).unwrap());
// Create a buffer on the CPU to hold the results of the three queries.
// Query results are always represented as either `u32` or `u64`.
// For occlusion queries, you always need one element per query. You can ask for the number of
// elements needed at runtime by calling `QueryType::result_size`.
// If you retrieve query results with `with_availability` enabled, then this array needs to
// be 6 elements long instead of 3.
let mut query_results = [0u32; 3];
mod vs {
vulkano_shaders::shader! {
ty: "vertex",
src: "
#version 450
layout(location = 0) in vec3 position;
layout(location = 1) in vec3 color;
layout(location = 0) out vec3 v_color;
void main() {
v_color = color;
gl_Position = vec4(position, 1.0);
}
"
}
}
mod fs {
vulkano_shaders::shader! {
ty: "fragment",
src: "
#version 450
layout(location = 0) in vec3 v_color;
layout(location = 0) out vec4 f_color;
void main() {
f_color = vec4(v_color, 1.0);
}
"
}
}
let vs = vs::Shader::load(device.clone()).unwrap();
let fs = fs::Shader::load(device.clone()).unwrap();
let render_pass = Arc::new(
vulkano::single_pass_renderpass!(
device.clone(),
attachments: {
color: {
load: Clear,
store: Store,
format: swapchain.format(),
samples: 1,
},
depth: {
load: Clear,
store: DontCare,
format: Format::D16Unorm,
samples: 1,
}
},
pass: {
color: [color],
depth_stencil: {depth}
}
)
.unwrap(),
);
let pipeline = Arc::new(
GraphicsPipeline::start()
.vertex_input_single_buffer()
.vertex_shader(vs.main_entry_point(), ())
.triangle_list()
.viewports_dynamic_scissors_irrelevant(1)
.fragment_shader(fs.main_entry_point(), ())
.render_pass(Subpass::from(render_pass.clone(), 0).unwrap())
// Enable depth testing, which is needed for occlusion queries to make sense at all.
// If you disable depth testing, every pixel is considered to pass the depth test, so
// every query will return a nonzero result.
.depth_stencil_simple_depth()
.build(device.clone())
.unwrap(),
);
let mut dynamic_state = DynamicState {
line_width: None,
viewports: None,
scissors: None,
compare_mask: None,
write_mask: None,
reference: None,
};
let mut framebuffers =
window_size_dependent_setup(&images, render_pass.clone(), &mut dynamic_state);
let mut recreate_swapchain = false;
let mut previous_frame_end = Some(sync::now(device.clone()).boxed());
event_loop.run(move |event, _, control_flow| match event {
Event::WindowEvent {
event: WindowEvent::CloseRequested,
..
} => {
*control_flow = ControlFlow::Exit;
}
Event::WindowEvent {
event: WindowEvent::Resized(_),
..
} => {
recreate_swapchain = true;
}
Event::RedrawEventsCleared => {
previous_frame_end.as_mut().unwrap().cleanup_finished();
if recreate_swapchain {
let dimensions: [u32; 2] = surface.window().inner_size().into();
let (new_swapchain, new_images) =
match swapchain.recreate_with_dimensions(dimensions) {
Ok(r) => r,
Err(SwapchainCreationError::UnsupportedDimensions) => return,
Err(e) => panic!("Failed to recreate swapchain: {:?}", e),
};
swapchain = new_swapchain;
framebuffers = window_size_dependent_setup(
&new_images,
render_pass.clone(),
&mut dynamic_state,
);
recreate_swapchain = false;
}
let (image_num, suboptimal, acquire_future) =
match swapchain::acquire_next_image(swapchain.clone(), None) {
Ok(r) => r,
Err(AcquireError::OutOfDate) => {
recreate_swapchain = true;
return;
}
Err(e) => panic!("Failed to acquire next image: {:?}", e),
};
if suboptimal {
recreate_swapchain = true;
}
let clear_values = vec![[0.0, 0.0, 1.0, 1.0].into(), 1.0.into()];
let mut builder =
AutoCommandBufferBuilder::primary_one_time_submit(device.clone(), queue.family())
.unwrap();
// Beginning or resetting a query is unsafe for now.
unsafe {
builder
// A query must be reset before each use, including the first use.
// This must be done outside a render pass.
.reset_query_pool(query_pool.clone(), 0..3)
.unwrap()
.begin_render_pass(
framebuffers[image_num].clone(),
SubpassContents::Inline,
clear_values,
)
.unwrap()
// Begin query 0, then draw the red triangle.
// Enabling the `precise` bit would give exact numeric results. This needs
// the `occlusion_query_precise` feature to be enabled on the device.
.begin_query(query_pool.clone(), 0, QueryControlFlags { precise: false })
.unwrap()
.draw(
pipeline.clone(),
&dynamic_state,
triangle1.clone(),
(),
(),
vec![],
)
.unwrap()
// End query 0.
.end_query(query_pool.clone(), 0)
.unwrap()
// Begin query 1 for the cyan triangle.
.begin_query(query_pool.clone(), 1, QueryControlFlags { precise: false })
.unwrap()
.draw(
pipeline.clone(),
&dynamic_state,
triangle2.clone(),
(),
(),
vec![],
)
.unwrap()
.end_query(query_pool.clone(), 1)
.unwrap()
// Finally, query 2 for the green triangle.
.begin_query(query_pool.clone(), 2, QueryControlFlags { precise: false })
.unwrap()
.draw(
pipeline.clone(),
&dynamic_state,
triangle3.clone(),
(),
(),
vec![],
)
.unwrap()
.end_query(query_pool.clone(), 2)
.unwrap()
.end_render_pass()
.unwrap();
}
let command_buffer = builder.build().unwrap();
let future = previous_frame_end
.take()
.unwrap()
.join(acquire_future)
.then_execute(queue.clone(), command_buffer)
.unwrap()
.then_swapchain_present(queue.clone(), swapchain.clone(), image_num)
.then_signal_fence_and_flush();
match future {
Ok(future) => {
previous_frame_end = Some(future.boxed());
}
Err(FlushError::OutOfDate) => {
recreate_swapchain = true;
previous_frame_end = Some(sync::now(device.clone()).boxed());
}
Err(e) => {
println!("Failed to flush future: {:?}", e);
previous_frame_end = Some(sync::now(device.clone()).boxed());
}
}
// Retrieve the query results.
// This copies the results to a variable on the CPU. You can also use the
// `copy_query_pool_results` function on a command buffer to write results to a
// Vulkano buffer. This could then be used to influence draw operations further down
// the line, either in the same frame or a future frame.
query_pool
.queries_range(0..3)
.unwrap()
.get_results(
&mut query_results,
QueryResultFlags {
// Block the function call until the results are available.
// Note: if not all the queries have actually been executed, then this
// will wait forever for something that never happens!
wait: true,
// Blocking and waiting will never give partial results.
partial: false,
// Blocking and waiting will ensure the results are always available after
// the function returns.
//
// If you disable waiting, then this can be used to include the
// availability of each query's results. You need one extra element per
// query in your `query_results` buffer for this. This element will
// be filled with a zero/nonzero value indicating availability.
with_availability: false,
},
)
.unwrap();
// If the `precise` bit was not enabled, then you're only guaranteed to get a boolean
// result here: zero if all pixels were occluded, nonzero if only some were occluded.
// Enabling `precise` will give the exact number of pixels.
// Query 0 (red triangle) will always succeed, because the depth buffer starts empty
// and will never occlude anything.
assert_ne!(query_results[0], 0);
// Query 1 (cyan triangle) will fail, because it's drawn completely behind the first.
assert_eq!(query_results[1], 0);
// Query 2 (green triangle) will succeed, because it's only partially occluded.
assert_ne!(query_results[2], 0);
}
_ => (),
});
}
fn window_size_dependent_setup(
images: &[Arc<SwapchainImage<Window>>],
render_pass: Arc<RenderPass>,
dynamic_state: &mut DynamicState,
) -> Vec<Arc<dyn FramebufferAbstract + Send + Sync>> {
let dimensions = images[0].dimensions();
let viewport = Viewport {
origin: [0.0, 0.0],
dimensions: [dimensions[0] as f32, dimensions[1] as f32],
depth_range: 0.0..1.0,
};
dynamic_state.viewports = Some(vec![viewport]);
let depth_attachment = ImageView::new(
AttachmentImage::with_usage(
render_pass.device().clone(),
dimensions,
Format::D16Unorm,
ImageUsage {
depth_stencil_attachment: true,
transient_attachment: true,
..ImageUsage::none()
},
)
.unwrap(),
)
.unwrap();
images
.iter()
.map(|image| {
let view = ImageView::new(image.clone()).unwrap();
Arc::new(
Framebuffer::start(render_pass.clone())
.add(view)
.unwrap()
.add(depth_attachment.clone())
.unwrap()
.build()
.unwrap(),
) as Arc<dyn FramebufferAbstract + Send + Sync>
})
.collect::<Vec<_>>()
}

View File

@ -310,7 +310,7 @@ impl<'a> QueriesRange<'a> {
debug_assert!(buffer_start % std::mem::size_of::<T>() == 0);
let count = self.range.end - self.range.start;
let per_query_len = self.pool.ty.data_size() + flags.with_availability as usize;
let per_query_len = self.pool.ty.result_size() + flags.with_availability as usize;
let required_len = per_query_len * count as usize;
if buffer_len < required_len {
@ -440,7 +440,7 @@ impl QueryType {
/// If the results are retrieved with [`QueryResultFlags::with_availability`] enabled, then
/// an additional element is required per query.
#[inline]
pub fn data_size(&self) -> usize {
pub const fn result_size(&self) -> usize {
match self {
Self::Occlusion | Self::Timestamp => 1,
Self::PipelineStatistics(flags) => flags.count(),
@ -526,7 +526,7 @@ impl QueryPipelineStatisticFlags {
/// Returns the number of flags that are set to `true`.
#[inline]
pub fn count(&self) -> usize {
pub const fn count(&self) -> usize {
let &Self {
input_assembly_vertices,
input_assembly_primitives,
@ -555,7 +555,7 @@ impl QueryPipelineStatisticFlags {
/// Returns `true` if any flags referring to compute operations are set to `true`.
#[inline]
pub fn is_compute(&self) -> bool {
pub const fn is_compute(&self) -> bool {
let &Self {
compute_shader_invocations,
..
@ -565,7 +565,7 @@ impl QueryPipelineStatisticFlags {
/// Returns `true` if any flags referring to graphics operations are set to `true`.
#[inline]
pub fn is_graphics(&self) -> bool {
pub const fn is_graphics(&self) -> bool {
let &Self {
input_assembly_vertices,
input_assembly_primitives,