1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 98 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147 148 149 150 151 152 153 154 155 156 157 158 159 160 161 162 163 164 165 166 167 168 169 170 171 172 173 174 175 176 177 178 179 180 181 182 183 184 185 186 187 188 189 190 191 192 193 194 195 196 197 198 199 200 201 202 203 204 205 206 207 208 209 210 211 212 213 214 215 216 217 218 219 220 221 222 223 224 225 226 227 228 229 230 231 232 233 234 235 236 237 238 239 240 241 242 243 244 245 246 247 248 249 250 251 252 253 254 255 256 257 258 259 260 261 262 263 264 265 266 267 268 269 270 271 272 273 274 275 276 277 278 279 280 281 282 283 284 285 286 287 288 289 290 291 292 293 294 295 296 297 298 299 300 301 302 303 304 305 306 307 308 309 310 311 312 313 314 315 316 317 318 319 320 321 322 323 324 325 326 327 328 329 330 331 332 333 334 335 336 337 338 339 340 341 342 343 344 345 346 347 348 349 350 351 352 353 354 355 356 357 358 359 360 361 362 363 364 365 366 367 368 369 370 371 372 373 374 375 376 377 378 379 380 381 382 383 384 385 386 387 388 389 390 391 392 393 394 395 396 397 398 399 400 401 402 403 404 405 406 407 408 409 410 411 412 413 414 415 416 417 418 419 420 421 422 423 424 425 426 427 428 429 430 431 432 433 434 435 436 437 438 439 440 441 442 443 444 445 446 447 448 449 450 451 452 453 454 455 456 457 458 459 460 461 462 463 464 465 466 467 468 469 470 471 472 473 474 475 476 477 478 479 480 481 482 483 484 485 486 487 488 489 490 491 492 493 494 495 496 497 498 499 500 501 502 503 504 505 506 507 508 509 510 511 512 513 514 515 516 517 518 519 520 521 522 523 524 525 526 527 528 529 530 531 532 533 534 535 536 537 538 539 540 541 542 543 544 545 546 547 548 549 550 551 552 553 554 555 556 557 558 559 560 561 562 563 564 565 566 567 568 569 570 571 572 573 574 575 576 577 578 579 580 581 582 583 584 585 586 587 588 589 590 591 592 593 594 595 596 597 598 599 600 601 602 603 604 605 606 607 608 609 610 611 612 613 614 615 616 617 618 619 620 621 622 623 624 625 626 627 628 629 630 631 632 633 634 635 636 637 638 639 640 641 642 643 644 645 646 647 648 649 650 651 652 653 654 655 656 657 658 659 660 661 662 663 664 665 666 667 668 669 670 671 672 673 674 675 676 677 678 679 680 681 682 683 684 685 686 687 688 689 690 691 692 693 694 695 696 697 698 699 700 701 702 703 704 705 706 707 708 709 710 711 712 713 714 715 716 717 718 719 720 721 722 723 724 725 726 727 728 729 730 731 732 733 734 735 736 737 738 739 740 741 742 743 744 745 746 747 748 749 750 751 752 753 754 755 756 757 758 759 760 761 762 763 764 765 766 767 768 769 770 771 772 773 774 775 776 777 778 779 780 781 782 783 784 785 786 787 788 789 790 791 792 793 794 795 796 797 798 799 800 801 802 803 804 805 806 807 808 809 810 811 812 813 814 815 816 817 818 819 820 821 822 823 824 825 826 827 828 829 830 831 832 833 834 835 836 837 838 839 840 841 842 843 844 845 846 847 848 849 850 851 852 853 854 855 856
//! Lockless Work Stealing Job System. //! //! # Features //! - Lockless work stealing queue //! - 0 runtime allocations //! - Jobs are executed from closures //! - Chaining and grouping of jobs //! - Parallel for_each abstraction //! //! # General Notes //! This crate provides job system with 0 allocation overhead at runtime. What this means in //! practice is, that once the instance is created, there are no more memory allocations required //! to execute jobs. //! //! Jobs are allocated from a Job Pool with a fixed capacity. Each thread gets their own pool and //! queue. The system uses a fixed storage space where it stores the closure for each job. The //! default storage is set to 64 bytes. This can be extended to 128 by enabling the feature //! _"job_storage_128"_. //! //! When a Job finishes execution there is a possibility to start other jobs. See //! [JobScope::chain()] for more details. //! //! Additionally, it's recommended that, in order to avoid running out of jobs during execution, you //! regularly ensure that you all your previous jobs have finished executing. //! //! Finally, this crate runs on rust stable and has been tested with rust 1.50.0 with the 2018 edition. //! //! # Safety //! Due to the implementation performing type erasure to store the closures in the pre-allocated //! space, I have not been able to figure out a way to ensure the compiler is aware of this so it //! can enforce lifetime and ownership checks. Until that is addressed [JobScope::create], //! [JobScope::create_with_parent] and [JobScope::run] are marked as unsafe and is up to caller //! to guarantee that they can enforce their safty requirements. //! //! # Panics //! //! Due to the queues having a fixed size, if we start filling all the queues up to full capacity there //! is a small chance that a worker thread may not be able to queue chained jobs. When this happens //! the system will panic as there is no way to recover from this. //! The queues come with debug asserts which will detect this situation. //! //! # Examples //! ## Safe //! These example provide safe variants that uphold the safety requirements. //! //! ### Parallel //! See [for_each()][JobScope::for_each()] and [for_each_with_result()][JobScope::for_each_with_result] //! for more details. //! ``` //! let mut job_sys = jobsys::JobSystem::new(4, 512).unwrap(); //! let job_scope = jobsys::JobScope::new_from_system(&job_sys); //! let mut array = [0_u32; 100]; //! job_scope.for_each(&mut array, |slice: &mut [u32], start, _end| { //! for i in 0..slice.len() { //! slice[i] = (start + i) as u32; //! } //! }).expect("Failed to start jobs"); //! ``` //! //! ### Single Work Item //! See [JobInstance][JobInstance] for more details. //! ``` //! let job_sys = jobsys::JobSystem::new(2, 128).expect("Failed to init job system"); //! let job_scope = jobsys::JobScope::new_from_system(&job_sys); //! let job_instance = jobsys::JobInstance::create(&job_scope, || { //! println!("Hello from Job Instance"); //! }).unwrap(); //! job_instance.wait_with(|| println!("Waiting on Job to Finish")).expect("Failed to wait on job"); //! ``` //! //! ## Unsafe //! The following example are unsafe due to certain requirements that need to be upheld by the caller //! but can be used as building block for safer alternatives. //! ### Start and Wait //! ``` //! let job_sys = jobsys::JobSystem::new(4, 512).unwrap(); //! let job_scope = jobsys::JobScope::new_from_system(&job_sys); //! let mut handle = unsafe{job_scope.create(|| { println!("Hello World!");}).unwrap()}; //! unsafe{job_scope.run(&mut handle).expect("Failed to run job");} //! job_scope.wait(&mut handle); //! ``` //! ### Grouping //! Ensure one job does not complete until other jobs have finished as well. //! ``` //! let job_sys= jobsys::JobSystem::new(4, 512).unwrap(); //! let job_scope = jobsys::JobScope::new_from_system(&job_sys); //! let mut parent = job_scope.create_noop().unwrap(); // Create a job that does nothing //! let mut child = unsafe{job_scope.create_with_parent(&mut parent, || { println!("Hello World!");}).unwrap()}; //! unsafe{job_scope.run(&child).expect("Failed to start child");} //! unsafe{job_scope.run(&parent).expect("Failed to start parent");} //! job_scope.wait(&parent); // Parent will only finish when both it and its child have finished //! ``` //! ### Chaining //! Launch new jobs as soon as one job completes. //! ``` //! let job_sys = jobsys::JobSystem::new(4, 512).unwrap(); //! let job_scope = jobsys::JobScope::new_from_system(&job_sys); //! let mut first = unsafe{job_scope.create(|| { println!("Hello World Chained!");})}.unwrap(); //! let mut second = unsafe{job_scope.create(|| { println!("Hello World Chained!");})}.unwrap(); //! job_scope.chain(&mut first, &second).expect("Failed to chain job, maximum chain count exceeded"); //! unsafe{job_scope.run(&first).expect("Failed to start job");} //! job_scope.wait(&second); // Second will only be executed after first completes //! ``` //! ### Creating jobs on job threads //! ``` //! let job_sys = jobsys::JobSystem::new(4, 512).unwrap(); //! let job_scope = jobsys::JobScope::new_from_system(&job_sys); //! let mut handle = unsafe{job_scope.create(|| { //! let thread_job_scope = jobsys::JobScope::new_from_thread().unwrap(); //! let thread_job_handle = thread_job_scope.create(|| { //! println!("Created on job thread"); //! }).unwrap(); //! thread_job_scope.run(&thread_job_handle).expect("Failed to run job"); //! }).unwrap()}; //! unsafe{job_scope.run(&mut handle).expect("Failed to run job");} //! job_scope.wait(&mut handle); //! ``` mod job; mod queues; mod thread; use crate::thread::JobThread; use crate::thread::ThreadData; use std::marker::PhantomData; use std::sync::Arc; type ThreadDataList = Vec<ThreadData>; /// Work Stealing JobSystem. pub struct JobSystem { thread_data: std::sync::Arc<ThreadDataList>, threads: Vec<JobThread>, } #[derive(Debug)] pub enum Error { /// Failed to create/start a job Thread ThreadCreate, /// The submitted closure exceeds the storage capacity StorageSizeExceeded, /// The thread queue is full and can't be submitted to QueueFull, /// The number of chained jobs has exceed the limit ChainCountExceeded, /// Job handle could not be resolved HandleResolve, /// We have detected multiple instances of a job system created on the same thread. This is not /// supported. MultipleInstances, /// Parent and child handles are the same ParentEqualsChild, /// Thread Local Job Pool has not been initialized InvalidThread, } impl JobSystem { /// Create a new instance of a JobSystem. To create jobs, please create a new instance of a /// [JobScope]. /// /// * `thread_count` - Number of worker threads. /// * `job_capacity` - Maximum number of jobs the system can allocate. If this is not a power of /// 2 it will be rounded up to the next value which is a power of 2. pub fn new(thread_count: usize, job_capacity: usize) -> Result<Self, Error> { let final_job_capacity = job_capacity.next_power_of_two(); if !thread::tls_set_main_thread_index() { return Err(Error::MultipleInstances); } let actual_thread_count = thread_count.max(1) + 1; let mut data_vec: ThreadDataList = Vec::with_capacity(actual_thread_count); data_vec.resize_with(actual_thread_count, || ThreadData::new(final_job_capacity)); let thread_data = Arc::new(data_vec); let mut threads = Vec::with_capacity(thread_count); for index in 1..actual_thread_count { let mut thread = JobThread::new(thread_data.clone(), index); if thread.start(final_job_capacity).is_err() { return Err(Error::ThreadCreate); } threads.push(thread); } thread::tls_setup_job_pool(final_job_capacity); thread::tls_set_thread_data(thread_data.clone()); Ok(Self { thread_data: thread_data.clone(), threads, }) } fn shutdown(&mut self) { for thread in &mut self.threads { thread.finish().unwrap() } self.threads.clear(); self.thread_data = Arc::new(vec![]); thread::tls_reset_main_thread_index(); thread::tls_reset_job_pool(); thread::tls_reset_thread_data(); } } impl Drop for JobSystem { fn drop(&mut self) { self.shutdown(); } } #[derive(Debug, Hash, Eq, PartialEq, Default)] /// Handle which represents an allocated job. pub struct ScopedJobHandle<'scope, T> { h: usize, // The phantom data here is required to prevent rust from happily accepting mutable borrows // to the same value. E.g.: Without PhantomData<T> the code below will compile without errors. // let js = JobSystem::new()?; // let mut v:u32 = 0; // for _ in 100 { // let x = &mut v; // let handle = js.create(|| { // x += 10; // })?; // } // Note: This only works correctly if the handles are stored somewhere for the duration // of the task. p: PhantomData<T>, s: PhantomData<&'scope mut ()>, } /// Manages the creation and scheduling of jobs. This type can either be created from a JobSystem /// instance or retrieved from a worker thread. /// If you tried to use this with a thread that is not being tracked by the job system, all /// the functions will fail with `Error::InvalidThread`. #[derive(Debug)] pub struct JobScope { thread_data: std::sync::Arc<ThreadDataList>, } impl JobScope { /// Create a new instance from a job system. pub fn new_from_system(job_system: &JobSystem) -> Self { Self { thread_data: job_system.thread_data.clone(), } } /// Create a new instance on a thread managed by a job system. /// If this calling thread is not managed by the job system, this will fail pub fn new_from_thread() -> Result<Self, Error> { if let Some(thread_data) = thread::tls_get_thread_data() { return Ok(Self { thread_data }); } Err(Error::InvalidThread) } /// Allocates a new Job that does nothing. /// /// This can be useful in cases where you need a parent job for grouping control. pub fn create_noop(&self) -> Result<ScopedJobHandle<fn()>, Error> { unsafe { self.create(|| {}) } } /// Allocate a new Job /// /// This function will check whether the closure fits in the job storage space and return an /// error if that's the case. /// # Safety /// This function is unsafe since it's not possible to accurately determine whether the provided /// closure's capture will be captured multiple times or be alive until this closure has /// finished executing. /// This method is safe to call provided that you guarantee: /// * The closure captures are valid during the execution of the job /// * The closure doesn't capture a writeable reference more than once pub unsafe fn create<T>(&self, job: T) -> Result<ScopedJobHandle<T>, Error> where T: Sized + FnOnce() + Send, { if !job::can_store_type::<T>() { return Err(Error::StorageSizeExceeded); } let tls_job_pool = thread::tls_get_job_pool(); tls_job_pool.with(|tls| { if let Some(pool) = tls.borrow_mut().as_mut() { let (job_instance, handle) = pool.allocate(); job_instance.store(job); return Ok(ScopedJobHandle { h: handle, p: PhantomData, s: PhantomData, }); } Err(Error::InvalidThread) }) } /// Allocate a new Job as a child of another job. /// /// The newly allocated will be created as a child of a parent job. What this means in practice /// is that the new job (child) can run parallel with the parent job and the parent job will /// not reach completion status until all of it's children have finished. /// # Safety /// This function is unsafe since it's not possible to accurately determine whether the provided /// closure's capture will be captured multiple times or be alive until this closure has /// finished executing. /// This method is safe to call provided that you guarantee: /// * The closure captures are valid during the execution of the job /// * The closure doesn't capture a writeable reference more than once pub unsafe fn create_with_parent<'a, T, Y>( &'a self, parent: &ScopedJobHandle<'a, Y>, job: T, ) -> Result<ScopedJobHandle<'a, T>, Error> where T: Sized + FnOnce() + Send, { if !job::can_store_type::<T>() { return Err(Error::StorageSizeExceeded); } let tls_job_pool = thread::tls_get_job_pool(); tls_job_pool.with(|tls| { if let Some(pool) = tls.borrow_mut().as_mut() { if let Some(parent_job) = pool.get_mut_ptr(parent.h) { let (job_instance, handle) = pool.allocate(); job_instance.store(job); job_instance.set_parent_job(parent_job); return Ok(ScopedJobHandle { h: handle, p: PhantomData, s: PhantomData, }); } return Err(Error::HandleResolve); } Err(Error::InvalidThread) }) } /// Register the child job as a follow up to the parent job. /// /// Every Job has the ability to chain follow up jobs on completion. This ensures the child /// job only runs when the parent is finished. /// /// There is a limited capacity for the number of jobs you can chain with one job. This function /// will return error if we are unable to register the child job. pub fn chain<T, Y>( &self, parent: &mut ScopedJobHandle<'_, T>, child: &ScopedJobHandle<'_, Y>, ) -> Result<(), Error> where T: Sized + FnOnce() + Send, Y: Sized + FnOnce() + Send, { if parent.h == child.h { return Err(Error::ParentEqualsChild); } let tls_job_pool = thread::tls_get_job_pool(); tls_job_pool.with(|tls| { if let Some(job_pool) = tls.borrow_mut().as_mut() { let child_job = if let Some(child_job) = job_pool.get_mut_ptr(child.h) { child_job } else { return Err(Error::HandleResolve); }; if let Some(parent_job) = job_pool.get_mut(parent.h) { return match parent_job.chain_job(child_job) { Err(_) => Err(Error::ChainCountExceeded), _ => Ok(()), }; } return Err(Error::HandleResolve); } Err(Error::InvalidThread) }) } /// Execute a job. /// # Safety /// This is currently marked unsafe due to the reasons described in /// [JobScope::create()] and [JobScope::create_with_parent()]. pub unsafe fn run<T>(&self, handle: &ScopedJobHandle<T>) -> Result<(), Error> where T: Sized + FnOnce() + Send, { let tls_job_pool = thread::tls_get_job_pool(); tls_job_pool.with(|tls| { if let Some(job_pool) = tls.borrow_mut().as_mut() { if let Some(job) = job_pool.get_mut(handle.h) { return match thread::start_job(&self.thread_data, job) { true => Ok(()), false => Err(Error::QueueFull), }; } return Err(Error::HandleResolve); } Err(Error::InvalidThread) }) } /// Block and wait until the job has finished. /// /// While we wait, the system will try to execute other jobs. pub fn wait<T>(&self, handle: &ScopedJobHandle<T>) -> Result<(), Error> where T: Sized + FnOnce() + Send, { // can't loop directly with TLS Pool here otherwise we trigger the RefCell panic // since we would keep the queue borrowed and it is possible that run // will run a job on the thread that is currently waiting. loop { match self.is_finished(handle) { Ok(finished) => { if !finished { if let Some(other_job) = thread::get_job(&self.thread_data) { thread::run_job(&self.thread_data, other_job); } } else { return Ok(()); } } Err(error) => return Err(error), }; } } /// Check if a job has finished execution. Does not block. pub fn is_finished<T>(&self, handle: &ScopedJobHandle<T>) -> Result<bool, Error> where T: Sized + FnOnce() + Send, { let tls_job_pool = thread::tls_get_job_pool(); tls_job_pool.with(|tls| { if let Some(job_pool) = tls.borrow().as_ref() { if let Some(job) = job_pool.get(handle.h) { return Ok(job.is_finished()); } return Err(Error::HandleResolve); } Err(Error::InvalidThread) }) } /// Given a slice of data and read-only closure, divide the slice into unique sub-slices which /// are distributed to the worker threads. /// /// The closure will receive the following parameters in order: /// * unique sub-slice over which the the current thread is operating on /// * start index of the full slice /// * end index of the the full slice /// Note use [for_each_with_result()][JobScope::for_each_with_result()] if you wish to /// produce output. pub fn for_each<'env, T, Y>(&self, slice: &'env mut [Y], cb: T) -> Result<(), Error> where T: Fn(&mut [Y], usize, usize) + 'env + Send + Sync, Y: Send, { let parent_job = self.create_noop()?; const DEFAULT_GROUP_SIZE: usize = 64; let divisor = self.thread_data.len() - 1; let group_size = (slice.len() / divisor).max(DEFAULT_GROUP_SIZE); let mut offset = 0_usize; for work_slice in slice.chunks_mut(group_size) { let callback = &cb; let offset_copy = offset; offset += work_slice.len(); let child_job = unsafe { // This is safe since we guarantee that in this function that we wait until // all children have finished executing and the closure's captures are // non-overlapping self.create_with_parent(&parent_job, move || { callback(work_slice, offset_copy, offset_copy + work_slice.len()); })? }; unsafe { // See safety comment above self.run(&child_job)? }; } unsafe { // See safety comment above self.run(&parent_job)? }; self.wait(&parent_job)?; Ok(()) } /// Same as [for_each()][JobScope::for_each()] but allows a result type to be returned for every individual group. pub fn for_each_with_result<'env, T, Y, Z>( &self, slice: &'env mut [Y], cb: T, ) -> Result<Vec<Z>, Error> where T: Fn(&[Y], usize, usize) -> Z + 'env + Send + Sync, Z: Sized + Default + Send, Y: Send, { let mut parent_job = self.create_noop().unwrap(); const DEFAULT_GROUP_SIZE: usize = 64; let divisor = self.thread_data.len() - 1; let group_size = (slice.len() / divisor).max(DEFAULT_GROUP_SIZE); let group_count = (slice.len() as f64 / group_size as f64).ceil() as usize; let mut offset = 0_usize; let mut group_index = 0_usize; // While we could use Vec![z::default;group_count] here, this would not work for types that // are not cloneable let mut result_vec = Vec::<Z>::default(); result_vec.resize_with(group_count, || Z::default()); let mut result_iter = result_vec.iter_mut(); for work_slice in slice.chunks_mut(group_size) { debug_assert!(group_index < group_count); let result_ref = result_iter .next() .expect("Result Vector calculation is incorrect"); let offset_copy = offset; offset += work_slice.len(); let callback = &cb; let child_job = unsafe { // This is safe since we guarantee that in this function that we wait until // all children have finished executing and the closure's captures are // non-overlapping self.create_with_parent(&mut parent_job, move || { *result_ref = callback(work_slice, offset_copy, offset_copy + work_slice.len()); }) }?; unsafe { // See safety comment above self.run(&child_job)?; } group_index += 1; } unsafe { // See safety comment above self.run(&parent_job)?; } self.wait(&parent_job)?; Ok(result_vec) } } /// JobInstance provides a safe interface to schedule a Job and ensures waits until the job /// has finished execution before it goes out of scope. /// /// **NOTE** It is recommended you call one of the [wait][JobInstance::wait] or [wait_with][JobInstance::wait_with] /// functions before the type goes out of scope and gets dropped. If we detect and error during the /// `drop()` call, we will issue a panic. #[derive(Debug)] pub struct JobInstance<'scope, Job> where Job: Sized + Send + FnOnce(), { scope: &'scope JobScope, handle: ScopedJobHandle<'scope, Job>, } impl<'scope, Job> JobInstance<'scope, Job> where Job: Sized + Send + FnOnce(), { /// Attempt to create a new instance of job. The job handle will be allocated and scheduled /// to run immediately. pub fn create( job_scope: &'scope JobScope, job: Job, ) -> Result<JobInstance<'scope, Job>, Error> { match unsafe { job_scope.create(job) } { Ok(handle) => match unsafe { job_scope.run(&handle) } { Ok(()) => { return Ok(JobInstance { scope: job_scope, handle, }) } Err(e) => return Err(e), }, Err(e) => return Err(e), }; } /// Blocking wait until the job has finished. pub fn wait(&self) -> Result<(), Error> { self.scope.wait(&self.handle) } /// Blocking wait until the job has finished, but before performing the blocking wait on the /// job handle, execute the provide closure once. pub fn wait_with<FN>(&self, cb: FN) -> Result<(), Error> where FN: FnOnce(), { (cb)(); self.wait() } /// Check whether the job has finished, does not block. pub fn is_finished(&self) -> Result<bool, Error> { self.scope.is_finished(&self.handle) } } impl<'scope, Job: FnOnce() + Sized + Send> Drop for JobInstance<'scope, Job> { /// If we run into an error, this function will panic. It is recommend that you call /// one of the [wait][JobInstance::wait] or [wait_with][JobInstance::wait_with] functions on /// the [JobInstance][JobInstance] type instead. fn drop(&mut self) { if let Err(_) = self.scope.wait(&self.handle) { panic!("Failed to wait on job handle"); } } } #[cfg(test)] mod tests { use crate::{JobInstance, JobScope, JobSystem}; use std::cell::RefCell; const THREAD_COUNT: usize = 4; const JOB_CAPACITY: usize = 1024; #[test] fn start_stop() { let r = JobSystem::new(THREAD_COUNT, JOB_CAPACITY); assert!(r.is_ok()); } #[test] fn launch_jobs_check_overflow() { let job_sys = JobSystem::new(THREAD_COUNT, 8).expect("Failed to init job system"); let job_scope = JobScope::new_from_system(&job_sys); let mut _counter = 0_usize; for _ in 0..64_u32 { const JOB_COUNT: usize = 7; let mut jobs = Vec::<_>::with_capacity(JOB_COUNT); for _ in 0..JOB_COUNT { let handle = job_scope.create_noop().unwrap(); unsafe { assert!(job_scope.run(&handle).is_ok()); } jobs.push(handle); } for job in jobs { job_scope.wait(&job).expect("Wait Failed"); } _counter += JOB_COUNT; } } #[test] #[cfg_attr(miri, ignore)] fn launch_jobs_from_job_threads() { // This functions fails on miri since with an error related to multiple write access to the // job pools job array. As long as the system doesn't overflow, this is safe since the // write will be independent and the array will not change size during execution. This // most likely related to the the mutable borrow from the tls key. And this only // happens when thw thread that is actively waiting on a job to finish schedules a job // during JobScope::wait() on the same thread. let job_sys = JobSystem::new(THREAD_COUNT, 128).expect("Failed to init job system"); let job_scope = JobScope::new_from_system(&job_sys); const JOB_COUNT: u32 = 32; let mut jobs = Vec::<_>::with_capacity(JOB_COUNT as usize); for _ in 0..JOB_COUNT { let handle = unsafe { job_scope .create(|| { let thread_job_scope = JobScope::new_from_thread().unwrap(); let thread_job_handle = thread_job_scope.create_noop().unwrap(); thread_job_scope .run(&thread_job_handle) .expect("Failed to run job"); }) .unwrap() }; unsafe { assert!(job_scope.run(&handle).is_ok()) }; jobs.push(handle); } for job in jobs { job_scope.wait(&job).expect("Wait Failed"); } } #[test] fn launch_jobs_with_ref() { let job_sys = JobSystem::new(THREAD_COUNT, JOB_CAPACITY).expect("Failed to init job system"); let job_scope = JobScope::new_from_system(&job_sys); const JOB_COUNT: usize = 100; let mut jobs = Vec::<_>::with_capacity(JOB_COUNT); let val = std::sync::Arc::new(std::sync::atomic::AtomicU32::new(0)); for _ in 0..JOB_COUNT { let val_copy = val.clone(); let handle = unsafe { job_scope .create(move || { val_copy.fetch_add(10, std::sync::atomic::Ordering::Release); }) .unwrap() }; unsafe { assert!(job_scope.run(&handle).is_ok()); } jobs.push(handle); } for job in jobs { job_scope.wait(&job).expect("Wait failed"); } assert_eq!( val.load(std::sync::atomic::Ordering::Acquire), 10 * JOB_COUNT as u32 ); } #[test] fn launch_jobs_chained() { let job_sys = JobSystem::new(THREAD_COUNT, JOB_CAPACITY).expect("Failed to init job system"); const JOB_COUNT: usize = 20; let job_scope = JobScope::new_from_system(&job_sys); let mut jobs = Vec::<_>::with_capacity(JOB_COUNT); for i in 0..JOB_COUNT { let handle = unsafe { job_scope .create(move || { println!("Chained {:?}: Job {:02}", std::thread::current().id(), i); }) .unwrap() }; jobs.push(RefCell::new(handle)); if i > 0 { let cur_handle = &jobs[i]; let prev_handle = &jobs[i - 1]; job_scope .chain(&mut prev_handle.borrow_mut(), &cur_handle.borrow()) .expect("Failed to chain"); } } unsafe { assert!(job_scope.run(&jobs.first().unwrap().borrow_mut()).is_ok()); } job_scope .wait(&jobs.last().unwrap().borrow_mut()) .expect("Wait failed"); } #[test] fn parallel_for() { let job_sys = JobSystem::new(THREAD_COUNT, JOB_CAPACITY).expect("Failed to init job system"); let job_scope = JobScope::new_from_system(&job_sys); let mut array = [0_u32; 100]; let r = job_scope.for_each(&mut array, |slice: &mut [u32], start, _end| { for i in 0..slice.len() { slice[i] = (start + i) as u32; } }); assert!(r.is_ok()); for i in 0..array.len() { assert_eq!(array[i] as usize, i); } } #[test] fn launch_with_parent() { let job_sys = JobSystem::new(THREAD_COUNT, JOB_CAPACITY).expect("Failed to init job system"); let job_scope = JobScope::new_from_system(&job_sys); const JOB_COUNT: usize = 20; let mut parent = job_scope.create_noop().unwrap(); let mut jobs = Vec::<_>::with_capacity(JOB_COUNT); for _i in 1..JOB_COUNT { let handle = unsafe { job_scope .create_with_parent(&mut parent, move || { /* println!( "Hello from thread {:?}: Job {:02}", std::thread::current().id(), i );*/ }) .unwrap() }; jobs.push(handle); } unsafe { assert!(job_scope.run(&parent).is_ok()); } for job in &jobs { unsafe { assert!(job_scope.run(job).is_ok()); } } job_scope.wait(&parent).expect("Wait failed"); } #[test] fn parallel_for_with_result() { let job_sys = JobSystem::new(THREAD_COUNT, JOB_CAPACITY).expect("Failed to init job system"); let job_scope = JobScope::new_from_system(&job_sys); let mut array = [0_u32; 200]; let r = job_scope.for_each_with_result(&mut array, |_slice: &[u32], start, end| -> u32 { (end - start) as u32 }); assert!(r.is_ok()); let result: u32 = r.unwrap().iter().sum(); assert_eq!(result, 200_u32); } /* This test is actually unsafe as it captures multiple references to the same variable. Atm I haven't figure out a way to make this safe, but this is here for reference. #[test] fn test_ref_creation() { //let mut jobs = Vec::<_>::with_capacity(200); { let job_sys = JobSystem::new(2, 128).unwrap(); let js = JobScope::new_from_system(&job_sys); let mut v: u32 = 0; for _ in 0..100 { let x = &mut v; let handle = unsafe { js .create(move || { *x += 10; }) .unwrap() }; js.run(&handle).unwrap(); //jobs.push(handle); } } }*/ #[test] fn job_instance_wait_drop() { let job_sys = JobSystem::new(THREAD_COUNT, JOB_CAPACITY).expect("Failed to init job system"); let job_scope = JobScope::new_from_system(&job_sys); let _job_instance = JobInstance::create(&job_scope, || { println!("Hello from Job Instance (wait drop)"); }); } #[test] fn job_instance_wait_with() { let job_sys = JobSystem::new(THREAD_COUNT, JOB_CAPACITY).expect("Failed to init job system"); let job_scope = JobScope::new_from_system(&job_sys); let job_instance = JobInstance::create(&job_scope, || { println!("Hello from Job Instance"); }) .unwrap(); job_instance .wait_with(|| println!("Waiting on Job to Finish")) .expect("Failed to wait on job"); } }