Skip to content
New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Optimization: avoid a memory allocation when padding RGB u8 to RGBA #2345

Merged
merged 1 commit into from
Jun 9, 2023
Merged

Optimization: avoid a memory allocation when padding RGB u8 to RGBA #2345

Changes from all commits
Commits
Show all changes
1 commit
Select commit
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
10 changes: 6 additions & 4 deletions crates/re_viewer_context/src/gpu_bridge/tensor_to_gpu.rs
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -72,9 +72,10 @@ fn color_tensor_to_gpu(
// Normalize sRGB(A) textures to 0-1 range, and let the GPU premultiply alpha.
// Why? Because premul must happen _before_ sRGB decode, so we can't
// use a "Srgb-aware" texture like `Rgba8UnormSrgb` for RGBA.
(3, TensorData::U8(buf)) => {
(pad_and_cast(buf.as_slice(), 255), TextureFormat::Rgba8Unorm)
}
(3, TensorData::U8(buf)) => (
pad_rgb_to_rgba(buf.as_slice(), u8::MAX).into(),
TextureFormat::Rgba8Unorm,
),
(4, TensorData::U8(buf)) => {
(cast_slice_to_cow(buf.as_slice()), TextureFormat::Rgba8Unorm)
}
Expand Down Expand Up @@ -318,7 +319,7 @@ fn general_texture_creation_desc_from_tensor<'a>(
// TODO(emilk): tell the shader to ignore the alpha channel instead!
match &tensor.data {
TensorData::U8(buf) => (
pad_and_cast(buf.as_slice(), u8::MAX),
pad_rgb_to_rgba(buf.as_slice(), u8::MAX).into(),
TextureFormat::Rgba8Uint,
),
TensorData::U16(buf) => (pad_and_cast(buf, u16::MAX), TextureFormat::Rgba16Uint),
Expand Down Expand Up @@ -428,6 +429,7 @@ fn narrow_f64_to_f32s(slice: &[f64]) -> Cow<'static, [u8]> {

fn pad_and_cast<T: Copy + Pod>(data: &[T], pad: T) -> Cow<'static, [u8]> {
re_tracing::profile_function!();
// TODO(emilk): optimize by combining the two steps into one; avoiding one allocation and memcpy
let padded: Vec<T> = pad_rgb_to_rgba(data, pad);
let bytes: Vec<u8> = pod_collect_to_vec(&padded);
bytes.into()
Expand Down