Pixel format conversions in pure Rust with SIMD optimizations on x86_64 and aarch64. The performance goal is to approximately reach the memory bandwidth limit and thus optimal performance when the input and output are not already in CPU cache.
Limitations and future work:
- Supports exactly one conversion: UYVY to I420. More will be added as needed.
- Expects to process full horizontal lines. This is likely to change to allow working on cropped regions.
- Does not support output to a frame with padding, as required by some APIs/devices.
- The ARM NEON code is less optimized than the AVX2 code today.
You may find the notes in docs/simd.md
helpful if you are new
to SIMD and thinking of contributing.
The main alternative to pixelfmt
is Chrome's C++
libyuv library.
Rust bindings are available via the yuv-sys
or libyuv
crates.
Some reasons to prefer pixelfmt
:
pixelfmt
is pure Rust and thus may be easier and quicker to build, particularly when cross-compiling.pixelfmt
is less total code.pixelfmt
'suyvy_to_i420
implementation benchmarks as slightly faster thanlibyuv
's. This appears to come down to design choices:libyuv
has separate routines to extract the Y and U+V planes, wherepixelfmt
gains some performance by doing it all with a single read of the UYVY data.pixelfmt
also has a hand-optimized conversion function wherelibyuv
uses a more generalized autovectorization approach.pixelfmt
uses safe Rust when possible, wherelibyuv
is entirely written in (unsafe) C++. That said,pixelfmt
still has a fair bit ofunsafe
logic for vendor SIMD intrinsics.
Some reasons to prefer libyuv
:
libyuv
is much more widely used.libyuv
is much more comprehensive and flexible.
SPDX-License-Identifier: MIT OR Apache-2.0
See LICENSE-MIT.txt or LICENSE-APACHE, respectively.