Skip to content

Commit

Permalink
Require nasm for AV1 decoding, unless on arm64 macOS (#7742)
Browse files Browse the repository at this point in the history
### What
* Closes #7669
* Closes #7671
* Disable AV1 decoding unless the `nasm` feature is turned on
* Turn on the `nasm` feature when building pre-built binaries
* Print enabled video features with `rerun --version` to make testing of
the above easy

### Checklist
* [x] I have read and agree to [Contributor
Guide](https://github.com/rerun-io/rerun/blob/main/CONTRIBUTING.md) and
the [Code of
Conduct](https://github.com/rerun-io/rerun/blob/main/CODE_OF_CONDUCT.md)
* [x] I've included a screenshot or gif (if applicable)
* [x] I have tested the web demo (if applicable):
* Using examples from latest `main` build:
[rerun.io/viewer](https://rerun.io/viewer/pr/7742?manifest_url=https://app.rerun.io/version/main/examples_manifest.json)
* Using full set of examples from `nightly` build:
[rerun.io/viewer](https://rerun.io/viewer/pr/7742?manifest_url=https://app.rerun.io/version/nightly/examples_manifest.json)
* [x] The PR title and labels are set such as to maximize their
usefulness for the next release's CHANGELOG
* [x] If applicable, add a new check to the [release
checklist](https://github.com/rerun-io/rerun/blob/main/tests/python/release_checklist)!
* [x] If have noted any breaking changes to the log API in
`CHANGELOG.md` and the migration guide

- [PR Build Summary](https://build.rerun.io/pr/7742)
- [Recent benchmark results](https://build.rerun.io/graphs/crates.html)
- [Wasm size tracking](https://build.rerun.io/graphs/sizes.html)

To run all checks from `main`, comment on the PR with `@rerun-bot
full-check`.
  • Loading branch information
emilk authored Oct 15, 2024
1 parent 57b15b0 commit 6fb673b
Show file tree
Hide file tree
Showing 11 changed files with 144 additions and 110 deletions.
4 changes: 2 additions & 2 deletions .github/workflows/contrib_rerun_py.yml
Original file line number Diff line number Diff line change
Expand Up @@ -61,11 +61,11 @@ jobs:
# this stops `re_web_viewer_server/build.rs` from running
RERUN_IS_PUBLISHING: true
run: |
cargo build \
pixi run cargo build \
--locked \
-p rerun-cli \
--no-default-features \
--features native_viewer,web_viewer \
--features release \
--release \
--target x86_64-unknown-linux-gnu
Expand Down
4 changes: 2 additions & 2 deletions .github/workflows/reusable_build_and_upload_rerun_cli.yml
Original file line number Diff line number Diff line change
Expand Up @@ -176,11 +176,11 @@ jobs:
# this stops `re_web_viewer_server/build.rs` from running
RERUN_IS_PUBLISHING: true
run: |
cargo build \
pixi run cargo build \
--locked \
-p rerun-cli \
--no-default-features \
--features native_viewer,web_viewer \
--features release \
--release \
--target ${{ needs.set-config.outputs.TARGET }}
Expand Down
3 changes: 1 addition & 2 deletions crates/store/re_video/Cargo.toml
Original file line number Diff line number Diff line change
Expand Up @@ -30,9 +30,8 @@ av1 = ["dep:dav1d"]

## Enable faster native video decoding with assembly.
## You need to install [nasm](https://nasm.us/) to compile with this feature.
# TODO(#7671): this feature flag currently does nothing on Linux.
nasm = [
# The default feature set of our dav1d fork has asm enabled (except on Linux, see above)
# The default feature set of our dav1d fork has asm enabled
"dav1d?/default",
]

Expand Down
24 changes: 16 additions & 8 deletions crates/store/re_video/src/decode/av1.rs
Original file line number Diff line number Diff line change
Expand Up @@ -37,14 +37,22 @@ impl SyncDav1dDecoder {
pub fn new(debug_name: String) -> Result<Self> {
re_tracing::profile_function!();

// TODO(#7671): enable this warning again on Linux when the `nasm` feature actually does something
#[allow(clippy::overly_complex_bool_expr)]
if !cfg!(target_os = "linux") && !cfg!(feature = "nasm") {
re_log::warn_once!(
"NOTE: native AV1 video decoder is running extra slowly. \
Speed it up by compiling Rerun with the `nasm` feature enabled. \
You'll need to also install nasm: https://nasm.us/"
);
if !cfg!(feature = "nasm") {
// The `nasm` feature makes AV1 decoding much faster.
// On Linux the difference is huge (~25x).
// On Windows, the difference was also pretty big (unsure how big).
// On an M3 Mac the difference is smalelr (2-3x),
// and ever without `nasm` emilk can play an 8k video at 2x speed.

if cfg!(target_os = "macos") && cfg!(target_arch = "aarch64") {
re_log::warn_once!(
"The native AV1 video decoder is unnecessarily slow. \
Speed it up by compiling Rerun with the `nasm` feature enabled."
);
} else {
// Better to return an error than to be perceived as being slow
return Err(Error::Dav1dWithoutNasm);
}
}

// See https://videolan.videolan.me/dav1d/structDav1dSettings.html for settings docs
Expand Down
5 changes: 5 additions & 0 deletions crates/store/re_video/src/decode/mod.rs
Original file line number Diff line number Diff line change
Expand Up @@ -100,6 +100,11 @@ pub enum Error {
#[cfg(not(target_arch = "wasm32"))]
#[error("dav1d: {0}")]
Dav1d(#[from] dav1d::Error),

#[cfg(feature = "av1")]
#[cfg(not(target_arch = "wasm32"))]
#[error("To enabled native AV1 decoding, compile Rerun with the `nasm` feature enabled.")]
Dav1dWithoutNasm,
}

pub type Result<T = (), E = Error> = std::result::Result<T, E>;
Expand Down
107 changes: 16 additions & 91 deletions crates/store/re_video/src/lib.rs
Original file line number Diff line number Diff line change
@@ -1,102 +1,27 @@
//! Video decoding library.
mod time;

pub mod decode;
pub mod demux;

pub use decode::{Chunk, Frame, PixelFormat};
pub use demux::{Config, Sample, VideoData, VideoLoadError};
pub use re_mp4::{TrackId, TrackKind};

/// A value in time units.
#[derive(Debug, Default, Clone, Copy, PartialEq, Eq, PartialOrd, Ord, Hash)]
pub struct Time(i64);

impl Time {
pub const ZERO: Self = Self(0);
pub const MAX: Self = Self(i64::MAX);

/// Create a new value in _time units_.
///
/// ⚠️ Don't use this for regular timestamps in seconds/milliseconds/etc.,
/// use the proper constructors for those instead!
/// This only exists for cases where you already have a value expressed in time units,
/// such as those received from the `WebCodecs` APIs.
#[inline]
pub fn new(v: i64) -> Self {
Self(v)
}

#[inline]
pub fn from_secs(v: f64, timescale: Timescale) -> Self {
Self((v * timescale.0 as f64).round() as i64)
}

#[inline]
pub fn from_millis(v: f64, timescale: Timescale) -> Self {
Self::from_secs(v / 1e3, timescale)
}

#[inline]
pub fn from_micros(v: f64, timescale: Timescale) -> Self {
Self::from_secs(v / 1e6, timescale)
}

#[inline]
pub fn from_nanos(v: i64, timescale: Timescale) -> Self {
Self::from_secs(v as f64 / 1e9, timescale)
}

/// Convert to a duration
#[inline]
pub fn duration(self, timescale: Timescale) -> std::time::Duration {
std::time::Duration::from_nanos(self.into_nanos(timescale) as _)
}

#[inline]
pub fn into_secs(self, timescale: Timescale) -> f64 {
self.0 as f64 / timescale.0 as f64
}
pub use self::{
decode::{Chunk, Frame, PixelFormat},
demux::{Config, Sample, VideoData, VideoLoadError},
time::{Time, Timescale},
};

#[inline]
pub fn into_millis(self, timescale: Timescale) -> f64 {
self.into_secs(timescale) * 1e3
/// Which features was this crate compiled with?
pub fn features() -> Vec<&'static str> {
// TODO(emilk): is there a helper crate for this?
let mut features = vec![];
if cfg!(feature = "av1") {
features.push("av1");
}

#[inline]
pub fn into_micros(self, timescale: Timescale) -> f64 {
self.into_secs(timescale) * 1e6
}

#[inline]
pub fn into_nanos(self, timescale: Timescale) -> i64 {
(self.into_secs(timescale) * 1e9).round() as i64
}
}

impl std::ops::Add for Time {
type Output = Self;

#[inline]
fn add(self, rhs: Self) -> Self::Output {
Self(self.0.saturating_add(rhs.0))
}
}

impl std::ops::Sub for Time {
type Output = Self;

#[inline]
fn sub(self, rhs: Self) -> Self::Output {
Self(self.0.saturating_sub(rhs.0))
}
}

/// The number of time units per second.
#[derive(Debug, Clone, Copy, PartialEq, Eq, PartialOrd, Ord, Hash)]
pub struct Timescale(u64);

impl Timescale {
pub(crate) fn new(v: u64) -> Self {
Self(v)
if cfg!(feature = "nasm") {
features.push("nasm");
}
features
}
93 changes: 93 additions & 0 deletions crates/store/re_video/src/time.rs
Original file line number Diff line number Diff line change
@@ -0,0 +1,93 @@
/// The number of time units per second.
#[derive(Debug, Clone, Copy, PartialEq, Eq, PartialOrd, Ord, Hash)]
pub struct Timescale(u64);

impl Timescale {
pub(crate) fn new(v: u64) -> Self {
Self(v)
}
}

/// A value in time units.
#[derive(Debug, Default, Clone, Copy, PartialEq, Eq, PartialOrd, Ord, Hash)]
pub struct Time(pub i64);

impl Time {
pub const ZERO: Self = Self(0);
pub const MAX: Self = Self(i64::MAX);

/// Create a new value in _time units_.
///
/// ⚠️ Don't use this for regular timestamps in seconds/milliseconds/etc.,
/// use the proper constructors for those instead!
/// This only exists for cases where you already have a value expressed in time units,
/// such as those received from the `WebCodecs` APIs.
#[inline]
pub fn new(v: i64) -> Self {
Self(v)
}

#[inline]
pub fn from_secs(v: f64, timescale: Timescale) -> Self {
Self((v * timescale.0 as f64).round() as i64)
}

#[inline]
pub fn from_millis(v: f64, timescale: Timescale) -> Self {
Self::from_secs(v / 1e3, timescale)
}

#[inline]
pub fn from_micros(v: f64, timescale: Timescale) -> Self {
Self::from_secs(v / 1e6, timescale)
}

#[inline]
pub fn from_nanos(v: i64, timescale: Timescale) -> Self {
Self::from_secs(v as f64 / 1e9, timescale)
}

/// Convert to a duration
#[inline]
pub fn duration(self, timescale: Timescale) -> std::time::Duration {
std::time::Duration::from_nanos(self.into_nanos(timescale) as _)
}

#[inline]
pub fn into_secs(self, timescale: Timescale) -> f64 {
self.0 as f64 / timescale.0 as f64
}

#[inline]
pub fn into_millis(self, timescale: Timescale) -> f64 {
self.into_secs(timescale) * 1e3
}

#[inline]
pub fn into_micros(self, timescale: Timescale) -> f64 {
self.into_secs(timescale) * 1e6
}

#[inline]
pub fn into_nanos(self, timescale: Timescale) -> i64 {
(self.into_secs(timescale) * 1e9).round() as i64
}
}

impl std::ops::Add for Time {
type Output = Self;

#[inline]
fn add(self, rhs: Self) -> Self::Output {
Self(self.0.saturating_add(rhs.0))
}
}

impl std::ops::Sub for Time {
type Output = Self;

#[inline]
fn sub(self, rhs: Self) -> Self::Output {
Self(self.0.saturating_sub(rhs.0))
}
}
11 changes: 8 additions & 3 deletions crates/top/rerun-cli/Cargo.toml
Original file line number Diff line number Diff line change
Expand Up @@ -37,13 +37,18 @@ path = "src/bin/rerun.rs"
doc = false

[features]
# The default is what the user gets when they call `cargo install rerun-cli --locked`,
# so wer have all the bells and wistles here
## The default is what the user gets when they call `cargo install rerun-cli --locked`,
## so we have all the bells and wistles here, except those that may require extra tools
## (like "nasm").
## That is: `cargo install rerun-cli --locked` should work for _everyone_.
default = ["native_viewer", "web_viewer"]

## The features we enable when we build the pre-built binaries during our releases.
## This may enable features that require extra build tools that not everyone heas.
release = ["default", "nasm"]

## Enable faster native video decoding with assembly.
## You need to install [nasm](https://nasm.us/) to compile with this feature.
# TODO(#7671): this feature flag currently does nothing on linux.
nasm = ["rerun/nasm"]

## Support spawning a native viewer.
Expand Down
1 change: 0 additions & 1 deletion crates/top/rerun/Cargo.toml
Original file line number Diff line number Diff line change
Expand Up @@ -73,7 +73,6 @@ log = ["dep:env_logger", "dep:log"]

## Enable faster native video decoding with assembly.
## You need to install [nasm](https://nasm.us/) to compile with this feature.
# TODO(#7671): this feature flag currently does nothing on linux.
nasm = ["re_video/nasm"]

## Support spawning a native viewer.
Expand Down
1 change: 1 addition & 0 deletions crates/top/rerun/src/commands/entrypoint.rs
Original file line number Diff line number Diff line change
Expand Up @@ -553,6 +553,7 @@ where

if args.version {
println!("{build_info}");
println!("Video features: {}", re_video::features().join(" "));
return Ok(0);
}

Expand Down
1 change: 0 additions & 1 deletion rerun_py/Cargo.toml
Original file line number Diff line number Diff line change
Expand Up @@ -30,7 +30,6 @@ extension-module = ["pyo3/extension-module"]

## Enable faster native video decoding with assembly.
## You need to install [nasm](https://nasm.us/) to compile with this feature.
# TODO(#7671): this feature flag currently does nothing on linux.
nasm = ["re_video/nasm"]

## Support serving a web viewer over HTTP with `serve()`.
Expand Down

0 comments on commit 6fb673b

Please sign in to comment.