Remove all the uniforms #346
There are no files selected for viewing
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
| Original file line number | Diff line number | Diff line change |
|---|---|---|
|
|
@@ -2,15 +2,24 @@ | |
| * License, v. 2.0. If a copy of the MPL was not distributed with this | ||
| * file, You can obtain one at http://mozilla.org/MPL/2.0/. */ | ||
|
|
||
| use arrayvec::ArrayVec; | ||
| use hal; | ||
| use hal::device::Device as BackendDevice; | ||
| use rendy_descriptor::{DescriptorAllocator, DescriptorSet}; | ||
| use rendy_memory::{Block, Heaps, Kind, MappedRange, MemoryBlock, MemoryUsage, MemoryUsageValue, Write}; | ||
|
|
||
| use std::cell::Cell; | ||
| use std::sync::Arc; | ||
| use std::mem; | ||
|
|
||
| use super::{MAX_FRAME_COUNT}; | ||
| use super::descriptor::{DescriptorData, DescriptorGroup}; | ||
|
|
||
| pub const DOWNLOAD_BUFFER_SIZE: usize = 10 << 20; // 10MB | ||
| // Maximum number of bound projections per frame. | ||
| // Projections are bound on a per render target basis, so those should fit in the 96 limit. | ||
| // We need this limit to define a fixed size for our uniform buffers. | ||
| pub const PROJECTION_PER_FRAME: usize = 96; | ||
|
There was a problem hiding this comment. where are we checking if we hit this limit? There was a problem hiding this comment. Added an assert for now https://github.com/szeged/webrender/pull/346/files#diff-5dbe2f57bb9273f9a363bcb71ec6834eR765 |
||
|
|
||
| #[derive(MallocSizeOf)] | ||
| pub struct BufferMemorySlice { | ||
|
|
@@ -325,6 +334,7 @@ impl<B: hal::Backend> Buffer<B> { | |
| let size = (data.len() * self.stride) as u64; | ||
| let range = | ||
| offset..((offset + size + non_coherent_atom_size_mask) & !non_coherent_atom_size_mask); | ||
|
|
||
| unsafe { | ||
| let mut mapped = self | ||
| .memory_block | ||
|
|
@@ -333,7 +343,7 @@ impl<B: hal::Backend> Buffer<B> { | |
| mapped | ||
| .write(device, 0..size) | ||
| .expect("Writer creation failed") | ||
| .write(data); | ||
| } | ||
| self.memory_block.unmap(device); | ||
| size as usize | ||
|
|
@@ -703,56 +713,142 @@ impl<B: hal::Backend> VertexBufferHandler<B> { | |
| } | ||
| } | ||
|
|
||
| struct DynamicBufferBundle<B: hal::Backend> { | ||
| buffer: Buffer<B>, | ||
| buffer_offset: u32, | ||
| descriptor_set: DescriptorSet<B>, | ||
| } | ||
|
|
||
| impl<B: hal::Backend> DynamicBufferBundle<B> { | ||
| unsafe fn new( | ||
| buffer_usage: hal::buffer::Usage, | ||
| data_stride: usize, | ||
| min_uniform_buffer_offset_alignment_mask: usize, | ||
| device: &B::Device, | ||
| heaps: &mut Heaps<B>, | ||
| descriptor_set: DescriptorSet<B> | ||
| ) -> Self { | ||
| debug_assert!(data_stride.is_power_of_two()); | ||
| let data_stride = ((data_stride - 1) | min_uniform_buffer_offset_alignment_mask) + 1; | ||
| let buffer = Buffer::new( | ||
| device, | ||
| heaps, | ||
| MemoryUsageValue::Dynamic, | ||
| buffer_usage, | ||
| min_uniform_buffer_offset_alignment_mask, | ||
| PROJECTION_PER_FRAME, | ||
| data_stride, | ||
| ); | ||
| device.write_descriptor_sets(Some(hal::pso::DescriptorSetWrite { | ||
| set: descriptor_set.raw(), | ||
| binding: 0, | ||
| array_offset: 0, | ||
| descriptors: Some(hal::pso::Descriptor::Buffer( | ||
| &buffer.buffer, | ||
| Some(0)..None, | ||
| )), | ||
| })); | ||
| DynamicBufferBundle { | ||
| buffer, | ||
| buffer_offset: 0, | ||
| descriptor_set, | ||
| } | ||
|
|
||
| } | ||
|
|
||
| fn add<T: Copy>( | ||
| &mut self, | ||
| device: &B::Device, | ||
| data: &[T], | ||
| non_coherent_atom_size_mask: u64, | ||
| ) { | ||
| debug_assert!(self.buffer_offset < PROJECTION_PER_FRAME as _); | ||
| let _ = self.buffer.update(device, data, self.buffer_offset as usize, non_coherent_atom_size_mask); | ||
| self.buffer_offset += 1; | ||
|
|
||
| } | ||
|
|
||
| fn last_update_start_offset(&self) -> u32 { | ||
| // buffer_offset points to the end of the last update, but we need the starting offset | ||
| (self.buffer_offset - 1) * self.buffer.stride as u32 | ||
| } | ||
|
|
||
| fn reset(&mut self) { | ||
| self.buffer_offset = 0; | ||
| } | ||
|
|
||
| unsafe fn deinit(self, device: &B::Device, heaps: &mut Heaps<B>, allocator: &mut DescriptorAllocator<B>) { | ||
| self.buffer.deinit(device, heaps); | ||
| allocator.free(std::iter::once(self.descriptor_set)); | ||
| } | ||
| } | ||
|
|
||
| pub(super) struct UniformBufferHandler<B: hal::Backend> { | ||
| buffer_bundles: ArrayVec<[DynamicBufferBundle<B>; MAX_FRAME_COUNT]>, | ||
| min_uniform_buffer_offset_alignment_mask: usize, | ||
| } | ||
|
|
||
| impl<B: hal::Backend> UniformBufferHandler<B> { | ||
| pub(super) fn new( | ||
| data_stride: usize, | ||
| min_uniform_buffer_offset_alignment_mask: usize, | ||
| frame_count: usize, | ||
| device: &B::Device, | ||
| heaps: &mut Heaps<B>, | ||
| descriptor_allocator: &mut DescriptorAllocator<B>, | ||
| group_data: &DescriptorData<B>, | ||
| group: &DescriptorGroup, | ||
| set_index: usize, | ||
| ) -> Self { | ||
| debug_assert!(frame_count <= MAX_FRAME_COUNT); | ||
| let mut descriptors = Vec::new(); | ||
| let mut buffer_bundles = ArrayVec::new(); | ||
| unsafe { | ||
| descriptor_allocator.allocate( | ||
| device, | ||
| group_data.descriptor_layout(group, set_index), | ||
| group_data.ranges(group, set_index), | ||
| frame_count as u32, | ||
| &mut descriptors, | ||
| ).expect("Allocate descriptor sets failed"); | ||
| for _ in 0..frame_count { | ||
| buffer_bundles.push(DynamicBufferBundle::new( | ||
| hal::buffer::Usage::UNIFORM, | ||
| data_stride, | ||
| min_uniform_buffer_offset_alignment_mask, | ||
| device, | ||
| heaps, | ||
| descriptors.pop().unwrap(), | ||
| )); | ||
| } | ||
| } | ||
|
|
||
| UniformBufferHandler { | ||
| buffer_bundles, | ||
| min_uniform_buffer_offset_alignment_mask, | ||
| } | ||
| } | ||
|
|
||
| pub(super) fn add<T: Copy>( | ||
| &mut self, | ||
| device: &B::Device, | ||
| data: &[T], | ||
| next_id: usize, | ||
| ) { | ||
| self.buffer_bundles[next_id].add(device, data, self.min_uniform_buffer_offset_alignment_mask as u64); | ||
| } | ||
|
|
||
| pub(super) fn buffer_info(&self, next_id: usize) -> (&B::DescriptorSet, u32) { | ||
| let ref buffer = self.buffer_bundles[next_id]; | ||
| (buffer.descriptor_set.raw(), buffer.last_update_start_offset()) | ||
| } | ||
|
|
||
| pub(super) fn reset(&mut self, next_id: usize) { | ||
| self.buffer_bundles[next_id].reset() | ||
| } | ||
|
|
||
| pub(super) unsafe fn deinit(self, device: &B::Device, heaps: &mut Heaps<B>, allocator: &mut DescriptorAllocator<B>) { | ||
| for bundle in self.buffer_bundles { | ||
| bundle.deinit(device, heaps, allocator); | ||
| } | ||
| } | ||
| } | ||
Oops, something went wrong.
Add this suggestion to a batch that can be applied as a single commit.
This suggestion is invalid because no changes were made to the code.
Suggestions cannot be applied while the pull request is closed.
Suggestions cannot be applied while viewing a subset of changes.
Only one suggestion per line can be applied in a batch.
Add this suggestion to a batch that can be applied as a single commit.
Applying suggestions on deleted lines is not supported.
You must change the existing code in this line in order to create a valid suggestion.
Outdated suggestions cannot be applied.
This suggestion has been applied or marked resolved.
Suggestions cannot be applied from pending reviews.
Suggestions cannot be applied on multi-line comments.
Suggestions cannot be applied while the pull request is queued to merge.
Suggestion cannot be applied right now. Please check back later.
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
why are we still having the transform here?
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
I think we have a misunderstanding here, looks like I didn't payed attention what you mentioned in #345
Unfortunately we don't have descriptor sets on a per target basis, we have the following groups: global (for each rendering loop), per WR pass and per draw. (The per-target variant would fit somewhere between second and third one.) That's the reason why I kept the uniform block around. If you think that getting rid of the uniform block and introducing a new descriptor binding with a new set index which points to a buffer could benefit us, I can make those changes as well.
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
I think we can safely just adopt the "per pass" descriptor set and increase its frequency just a bit to bind whenever the projection happens. This is still relatively rare, i.e. dozens per frame, not thousands.
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
Yes that's sounds totally reasonable. Also I will add a more descriptive name to it.