Bordeaux Threads makes writing portable multi-threaded apps simple.
Parallelism for Common Lisp
A promise implementation for Common Lisp.
A simple semaphore class for bordeaux-threads inspired by SBCL's semaphore.
Thread-safe message-passing channels, in the style of
the occam programming language.
Communicating Sequential Process support for Common Lisp
Asynchronous operations for Common Lisp.
A futures implementation for Common Lisp. Plugs in nicely to cl-async.
Cl-cuda is a library to use NVIDIA CUDA in Common Lisp programs.
A lightweight threading / cooperative multitasking library.
PCall is a concurrency library that implements simple 'result-oriented' parallelism on top of Bordeaux-Threads. See also this article for background information.
Elegant High Performance Computing
STMX is an actively maintained, high-performance concurrency library providing Transactional Memory for Common Lisp.Home page and downloads: http://github.com/cosmos72/stmx Main features
Extremely intuitive to use and to write correct, thread-safe concurrent code
Brings database-style transactions to Common Lisp by introducing transactional memory
High performance implementation, benchmarked to exceed 7 millions transactions per CPU core per second on commodity PC hardware
Support for hardware memory transactions (requires 64-bit SBCL and CPUs with Intel TSX instructions - currently Core i5 4570, Core i7 4670, Core i7 4770 and some others). They increase STMX performance up to almost 40 millions transactions per CPU core per second
Removes the need for traditional locks, mutexes and conditions - writing correct concurrent code with them is well known to be hard
Transactional code is intrinsically deadlock-free: if two transactions conflict one of them will be re-executed
Automatic commit and rollback: if a transaction completes normally it will be committed, if it signals an error it will be rolled back
Transactions are composable: they can be executed in a larger transaction, either in sequence (all-or-nothing) or as alternatives (try them until one succeeds)
Offers freedom of choice between blocking and non-blocking transactional functions: given either behaviour, it is trivial to transform it into the other.
Features transactional versions of popular data structures: hash tables, red-black trees, stack, fifo
Includes transactional data structure for multicast publish/subscribe
Creating new transactional data structures is easy
Extensive test suite
Tested on SBCL, ABCL, CCL, CMUCL and ECL.
Very simple to install with Quicklisp
A quick-start guide and installation instructions are provided in README.md fileLicense: LLGPL What STMX is NOT
In order not to confuse programmers - less experienced ones in particular - and to avoid rising unrealistic hopes, the author himself stated the following about STMX:
it is NOT a quick hack to automagically transform existing, slow, single-threaded programs into fast, concurrent ones. No matter how much transactions can help, writing concurrent code still requires careful design and implementation - and testing. And refactoring takes time too.
it is NOT for optimization-focused programmers trying to squeeze the last cycle from their Common Lisp programs. STMX records an in-memory transaction log containing all reads and writes from/to transactional memory, then later (during commit) validates the transaction log against the latest data present in transactional memory and finally copies the transaction log onto the transactional memory while holding locks. STMX is quite optimized, but this machinery comes at an obvious performance cost with respect to hand-made, highly optimized locking code (but a good reality check is to ask yourself how many people have the skill and patience to write such code without bugs).
it is NOT supposed to be used for all data structures in a Common Lisp program. STMX is intended only for the data accessed concurrently by multiple threads while being modified by at least one thread. And even in that case, transactional memory is not always needed: it depends on the kinds of modifications.
it is NOT a serialization or persistence framework. Rather, messing with metaclasses and playing (allowed) tricks with slots contents as STMX does, quite likely does not mix well with serialization or persistence libraries such as CL-STORE, because they typically need full control on the slots of the objects to be serialized and de-serialized.
it is NOT a million dollar library from some deep-pocket company. At the moment, it is the work of a single person.
thread-pool is library that allows asynchronous function execution from a static thread pool.The usage of this library allows programmers to control the number of threads used in a process and avoid the thread creation overhead when new asynchronous computation are needed.Therefore thread-pool may increase overall performance in a multi-thread program.
Licence: BSDDependencies: Bordeaux-Threads, arnesiHome page: https://github.com/kiuma/thread-pool.
Topics: system programming thread concurrency
Trivial timer library
A collection of basic utilities for syntactic
extension and basic data structure hadling,
developed over the years of CL history by efforts of different individuals,
and gathered under the unbrella of a hierarchy of packages
which can be used on-demand.