diff --git a/Cargo.lock b/Cargo.lock index b590c228..e7a5debc 100644 --- a/Cargo.lock +++ b/Cargo.lock @@ -181,7 +181,7 @@ dependencies = [ [[package]] name = "image" -version = "0.23.2" +version = "0.23.3" source = "registry+https://github.com/rust-lang/crates.io-index" dependencies = [ "bytemuck 1.2.0 (registry+https://github.com/rust-lang/crates.io-index)", @@ -295,7 +295,7 @@ dependencies = [ [[package]] name = "num_cpus" -version = "1.12.0" +version = "1.13.0" source = "registry+https://github.com/rust-lang/crates.io-index" dependencies = [ "hermit-abi 0.1.8 (registry+https://github.com/rust-lang/crates.io-index)", @@ -311,12 +311,11 @@ dependencies = [ "clap 2.33.0 (registry+https://github.com/rust-lang/crates.io-index)", "cloudflare-zlib 0.2.5 (registry+https://github.com/rust-lang/crates.io-index)", "crc 1.8.1 (registry+https://github.com/rust-lang/crates.io-index)", - "image 0.23.2 (registry+https://github.com/rust-lang/crates.io-index)", + "image 0.23.3 (registry+https://github.com/rust-lang/crates.io-index)", "indexmap 1.3.2 (registry+https://github.com/rust-lang/crates.io-index)", "itertools 0.9.0 (registry+https://github.com/rust-lang/crates.io-index)", "libdeflater 0.2.0 (registry+https://github.com/rust-lang/crates.io-index)", "miniz_oxide 0.3.6 (registry+https://github.com/rust-lang/crates.io-index)", - "num_cpus 1.12.0 (registry+https://github.com/rust-lang/crates.io-index)", "rayon 1.3.0 (registry+https://github.com/rust-lang/crates.io-index)", "rgb 0.8.16 (registry+https://github.com/rust-lang/crates.io-index)", "wild 2.0.2 (registry+https://github.com/rust-lang/crates.io-index)", @@ -353,7 +352,7 @@ dependencies = [ "crossbeam-queue 0.2.1 (registry+https://github.com/rust-lang/crates.io-index)", "crossbeam-utils 0.7.2 (registry+https://github.com/rust-lang/crates.io-index)", "lazy_static 1.4.0 (registry+https://github.com/rust-lang/crates.io-index)", - "num_cpus 1.12.0 (registry+https://github.com/rust-lang/crates.io-index)", + "num_cpus 1.13.0 (registry+https://github.com/rust-lang/crates.io-index)", ] [[package]] @@ -478,7 +477,7 @@ dependencies = [ "checksum either 1.5.3 (registry+https://github.com/rust-lang/crates.io-index)" = "bb1f6b1ce1c140482ea30ddd3335fc0024ac7ee112895426e0a629a6c20adfe3" "checksum glob 0.3.0 (registry+https://github.com/rust-lang/crates.io-index)" = "9b919933a397b79c37e33b77bb2aa3dc8eb6e165ad809e58ff75bc7db2e34574" "checksum hermit-abi 0.1.8 (registry+https://github.com/rust-lang/crates.io-index)" = "1010591b26bbfe835e9faeabeb11866061cc7dcebffd56ad7d0942d0e61aefd8" -"checksum image 0.23.2 (registry+https://github.com/rust-lang/crates.io-index)" = "9062b90712d25bc6bb165d110aa59c6b47c849246e341e7b86a98daff9d49f60" +"checksum image 0.23.3 (registry+https://github.com/rust-lang/crates.io-index)" = "bfc5483f8d5afd3653b38a196c52294dcb239c3e1a5bade1990353ea13bcf387" "checksum indexmap 1.3.2 (registry+https://github.com/rust-lang/crates.io-index)" = "076f042c5b7b98f31d205f1249267e12a6518c1481e9dae9764af19b707d2292" "checksum inflate 0.4.5 (registry+https://github.com/rust-lang/crates.io-index)" = "1cdb29978cc5797bd8dcc8e5bf7de604891df2a8dc576973d71a281e916db2ff" "checksum itertools 0.9.0 (registry+https://github.com/rust-lang/crates.io-index)" = "284f18f85651fe11e8a991b2adb42cb078325c996ed026d994719efcfca1d54b" @@ -492,7 +491,7 @@ dependencies = [ "checksum num-iter 0.1.40 (registry+https://github.com/rust-lang/crates.io-index)" = "dfb0800a0291891dd9f4fe7bd9c19384f98f7fbe0cd0f39a2c6b88b9868bbc00" "checksum num-rational 0.2.3 (registry+https://github.com/rust-lang/crates.io-index)" = "da4dc79f9e6c81bef96148c8f6b8e72ad4541caa4a24373e900a36da07de03a3" "checksum num-traits 0.2.11 (registry+https://github.com/rust-lang/crates.io-index)" = "c62be47e61d1842b9170f0fdeec8eba98e60e90e5446449a0545e5152acd7096" -"checksum num_cpus 1.12.0 (registry+https://github.com/rust-lang/crates.io-index)" = "46203554f085ff89c235cd12f7075f3233af9b11ed7c9e16dfe2560d03313ce6" +"checksum num_cpus 1.13.0 (registry+https://github.com/rust-lang/crates.io-index)" = "05499f3756671c15885fee9034446956fff3f243d6077b91e5767df161f766b3" "checksum png 0.16.1 (registry+https://github.com/rust-lang/crates.io-index)" = "46060468187c21c00ffa2a920690b29997d7fd543f5a4d400461e4a7d4fccde8" "checksum rayon 1.3.0 (registry+https://github.com/rust-lang/crates.io-index)" = "db6ce3297f9c85e16621bb8cca38a06779ffc31bb8184e1be4bed2be4678a098" "checksum rayon-core 1.7.0 (registry+https://github.com/rust-lang/crates.io-index)" = "08a89b46efaf957e52b18062fb2f4660f8b8a4dde1807ca002690868ef2c85a9" diff --git a/Cargo.toml b/Cargo.toml index 55e409d9..bcbfa0eb 100644 --- a/Cargo.toml +++ b/Cargo.toml @@ -32,7 +32,6 @@ bit-vec = "^0.6.0" byteorder = "^1.0.0" crc = "^1.2.0" itertools = "^0.9.0" -num_cpus = "^1.0.0" zopfli = "^0.4.0" miniz_oxide = "0.3" rgb = "0.8.11" diff --git a/src/evaluate.rs b/src/evaluate.rs index f60a6e2f..254a0928 100644 --- a/src/evaluate.rs +++ b/src/evaluate.rs @@ -12,13 +12,15 @@ use crate::png::STD_WINDOW; #[cfg(not(feature = "parallel"))] use crate::rayon; use crate::Deadline; -#[cfg(feature = "parallel")] -use rayon; use rayon::prelude::*; +#[cfg(not(feature = "parallel"))] +use std::cell::RefCell; use std::sync::atomic::AtomicUsize; use std::sync::atomic::Ordering::SeqCst; +#[cfg(feature = "parallel")] use std::sync::mpsc::*; use std::sync::Arc; +#[cfg(feature = "parallel")] use std::thread; struct Candidate { @@ -46,30 +48,46 @@ pub(crate) struct Evaluator { nth: AtomicUsize, best_candidate_size: Arc, /// images are sent to the thread for evaluation + #[cfg(feature = "parallel")] eval_send: SyncSender, // the thread helps evaluate images asynchronously + #[cfg(feature = "parallel")] eval_thread: thread::JoinHandle>, + // in non-parallel mode, images are evaluated synchronously + #[cfg(not(feature = "parallel"))] + eval_best_candidate: RefCell>, } impl Evaluator { pub fn new(deadline: Arc) -> Self { + #[cfg(feature = "parallel")] let (tx, rx) = sync_channel(4); Self { deadline, best_candidate_size: Arc::new(AtomicMin::new(None)), nth: AtomicUsize::new(0), + #[cfg(feature = "parallel")] eval_send: tx, + #[cfg(feature = "parallel")] eval_thread: thread::spawn(move || rx.into_iter().min_by_key(Candidate::cmp_key)), + #[cfg(not(feature = "parallel"))] + eval_best_candidate: RefCell::new(None), } } /// Wait for all evaluations to finish and return smallest reduction /// Or `None` if all reductions were worse than baseline. + #[cfg(feature = "parallel")] fn get_best_candidate(self) -> Option { drop(self.eval_send); // disconnect the sender, breaking the loop in the thread self.eval_thread.join().expect("eval thread") } + #[cfg(not(feature = "parallel"))] + fn get_best_candidate(self) -> Option { + self.eval_best_candidate.into_inner() + } + pub fn get_result(self) -> Option { self.get_best_candidate().map(|candidate| candidate.image) } @@ -91,6 +109,7 @@ impl Evaluator { let best_candidate_size = self.best_candidate_size.clone(); // sends it off asynchronously for compression, // but results will be collected via the message queue + #[cfg(feature = "parallel")] let eval_send = self.eval_send.clone(); rayon::spawn(move || { let filters_iter = STD_FILTERS.par_iter().with_max_len(1); @@ -126,7 +145,18 @@ impl Evaluator { nth, }; - eval_send.send(new).expect("send"); + #[cfg(feature = "parallel")] + { + eval_send.send(new).expect("send"); + } + + #[cfg(not(feature = "parallel"))] + { + match &mut *self.eval_best_candidate.borrow_mut() { + Some(prev) if prev.cmp_key() < new.cmp_key() => {} + best => *best = Some(new), + } + } } }); }); diff --git a/src/lib.rs b/src/lib.rs index e4af09be..9b0d1446 100644 --- a/src/lib.rs +++ b/src/lib.rs @@ -13,7 +13,6 @@ #![warn(clippy::range_plus_one)] #![allow(clippy::cognitive_complexity)] -use num_cpus; #[cfg(feature = "parallel")] extern crate rayon; #[cfg(not(feature = "parallel"))] @@ -206,10 +205,6 @@ pub struct Options { /// /// Default: `false` pub use_heuristics: bool, - /// Number of threads to use - /// - /// Default: number of CPU cores - pub threads: usize, /// Maximum amount of time to spend on optimizations. /// Further potential optimizations are skipped if the timeout is exceeded. @@ -313,7 +308,6 @@ impl Default for Options { strip: Headers::None, deflate: Deflaters::Zlib, use_heuristics: false, - threads: num_cpus::get(), timeout: None, } } @@ -321,12 +315,6 @@ impl Default for Options { /// Perform optimization on the input file using the options provided pub fn optimize(input: &InFile, output: &OutFile, opts: &Options) -> PngResult<()> { - // Initialize the thread pool with correct number of threads - #[cfg(feature = "parallel")] - let _ = rayon::ThreadPoolBuilder::new() - .num_threads(opts.threads) - .build_global(); - // Read in the file and try to decode as PNG. if opts.verbosity.is_some() { eprintln!("Processing: {}", input); @@ -421,12 +409,6 @@ pub fn optimize(input: &InFile, output: &OutFile, opts: &Options) -> PngResult<( /// Perform optimization on the input file using the options provided, where the file is already /// loaded in-memory pub fn optimize_from_memory(data: &[u8], opts: &Options) -> PngResult> { - // Initialize the thread pool with correct number of threads - #[cfg(feature = "parallel")] - let _ = rayon::ThreadPoolBuilder::new() - .num_threads(opts.threads) - .build_global(); - // Read in the file and try to decode as PNG. if opts.verbosity.is_some() { eprintln!("Processing from memory"); diff --git a/src/main.rs b/src/main.rs index f1f394ec..517123d8 100644 --- a/src/main.rs +++ b/src/main.rs @@ -486,7 +486,12 @@ fn parse_opts_into_struct( } if let Some(x) = matches.value_of("threads") { - opts.threads = x.parse::().unwrap(); + let threads = x.parse::().unwrap(); + + rayon::ThreadPoolBuilder::new() + .num_threads(threads) + .build_global() + .map_err(|err| err.to_string())?; } Ok((out_file, out_dir, opts))