The implementation of the Cilk-5 multithreaded language

Abstract

The fth release of the multithreaded language Cilk uses a provably good \work-stealing" scheduling algorithm similar to the rst system, but the language has been completely redesigned and the runtime system completely reengineered. The e ciency of the new implementation was aided by a clear strategy that arose from a theoretical analysis of the scheduling algorithm: concentrate on minimizing overheads that contribute to the work, even at the expense of overheads that contribute to the critical path. Although it may seem counterintuitive to move overheads onto the critical path, this \work-rst" principle has led to a portable Cilk-5 implementation in which the typical cost of spawning a parallel thread is only between 2 and 6 times the cost of a C function call on a variety of contemporary machines. Many Cilk programs run on one processor with virtually no degradation compared to equivalent C programs. This paper describes how the work-rst principle was exploited in the design of Cilk-5's compiler and its runtime system. In particular, we present Cilk-5's novel \two-clone" compilation strategy and its Dijkstra-like mutual-exclusion protocol for implementing the ready deque in the work-stealing scheduler.

Lace: Non-blocking Split Deque for Work-Stealing

“…Frameworks like Cilk [3,9] and Wool [7,8] allow writing parallel programs in a style similar to sequential programs [1].…”

“…The first provably efficient work-stealing scheduler for fully strict computations was presented in 1994 [4] and its implementation in Cilk in 1996 [3]. One improvement of the original Cilk algorithm is the THE protocol in Cilk-5 [9], which eliminates acquiring the lock in push and in most executions of pop, but every steal still requires locking.…”

“…Memory fences are expensive. For example, the THE protocol spends half of its execution time in the memory fence [9].…”

“…Tasks are often stored as pointers that are removed from the deque when the task is stolen [9,2,10,5]. To virtually eliminate the overhead of task creation for tasks that are never stolen, Faxén proposed a direct task stack, storing tasks instead of pointers in the work queue, implemented in Wool [7,8].…”

Exploring Traditional and Emerging Parallel Programming Models Using a Proxy Application

“…In the rest of the paper, we refer to the collective set of parallel programming models, languages, runtimes, and specific implementations as "programming models" for ease of reference. Programming models can range from purely task based models like Intel's TBB [4], Cilk [5], or PLASMA [6], to globally synchronous approaches, such as BSP [7]. Furthermore, they can be based on a common memory paradigm like Global Arrays [8] or HPF [9] or be data-driven, such as CHARM++ [10].…”