Release Notes » Release 3.5.0 (2023/01/05)

Taskflow 3.5.0 is the 6th release in the 3.x line! This release includes several new changes, such as pipeline parallelism, improved work-stealing performance, profiling, documentation, examples, and unit tests.

Download

Taskflow 3.5.0 can be downloaded from here.

System Requirements

To use Taskflow v3.5.0, you need a compiler that supports C++17:

GNU C++ Compiler at least v8.4 with -std=c++17
Clang C++ Compiler at least v6.0 with -std=c++17
Microsoft Visual Studio at least v19.27 with /std:c++17
AppleClang Xcode Version at least v12.0 with -std=c++17
Nvidia CUDA Toolkit and Compiler (nvcc) at least v11.1 with -std=c++17
Intel C++ Compiler at least v19.0.1 with -std=c++17
Intel DPC++ Clang Compiler at least v13.0.0 with -std=c++17 and SYCL20

Taskflow works on Linux, Windows, and Mac OS X.

Release Summary

This release introduces a new data-parallel pipeline programming model, solves the busy-waiting problem in our work-stealing scheduler, and adds a new text-based feature for profiler report.

New Features

Taskflow Core

Added tf::WorkerInterface to allow changing properties of workers upon their creations
Added tf::Executor::loop_until to allow looping a worker with a custom stop predicate
Added tf::DataPipeline to implement data-parallel algorithms
- See Data-parallel Pipeline
Extended tf::TaskQueue to include priority (tf::TaskPriority)
- See Prioritized Tasking
Extended tf::Executor to include tf::WorkerInterface
Improved parallel algorithms (e.g., tf::Taskflow::for_each) with tail optimization
Resolved the busy-waiting problem in our work-stealing algorithm (#400)

cudaFlow

This release has no update on tf::cudaFlow.

Utilities

Added tf::unroll to unroll loops using template techniques
Added tf::CachelineAligned to create a cacheline-aligned object
Replaced std::aligned_union (deprecated in C++23) with a custom byte type (#445)

Taskflow Profiler (TFProf)

Added a new feature to generate a profile summary report
- See Display Profile Summary

Bug Fixes

Fixed the compilation error in taking move-only types for tf::Taskflow::transform_reduce
Fixed the compilation error in the graph pipeline benchmark
Fixed the compilation error in unknown OS (replaced with TF_OS_UNKNOWN)

If you encounter any potential bugs, please submit an issue at issue tracker.

Breaking Changes

This release has no breaking changes.

Deprecated and Removed Items

This release has no deprecated and removed items.

Documentation

Revised Executor
- Added Execute a Taskflow from an Internal Worker
Added Prioritized Tasking
Added Data-parallel Pipeline

Miscellaneous Items

We have published Taskflow in the following venues:

Tsung-Wei Huang and Leslie Hwang, "Task-Parallel Programming with Constrained Parallelism," IEEE High-Performance Extreme Computing Conference (HPEC), MA, 2022
Tsung-Wei Huang, "Enhancing the Performance Portability of Heterogeneous Circuit Analysis Programs," IEEE High-Performance Extreme Computing Conference (HPEC), MA, 2022
Dian-Lun Lin, Haoxing Ren, Yanqing Zhang, and Tsung-Wei Huang, "From RTL to CUDA: A GPU Acceleration Flow for RTL Simulation with Batch Stimulus," ACM International Conference on Parallel Processing (ICPP), Bordeaux, France, 2022

Please do not hesitate to contact Dr. Tsung-Wei Huang if you intend to collaborate with us on using Taskflow in your scientific computing projects.