Compiler support of the workqueuing execution model for Intel SMP architectures

Programs with irregular patterns of dynamic data structures and/or those with complicated control structures such as recursion are hard to parallelize efficiently. The workqueuing model allows the user to exploit irregular parallelism, which is beyond the scope of what is supported by OpenMP *. We have integrated this model into the Intel @ C++ high-performance compiler's OpenMP parallelizer. In this paper, we give an overview of the workqueuing model and describe in detail both the generated multithreaded code as well as the multithreaded run-time library routines supporting it. We also present preliminary performance results of a set of benchmarks and applications measured on Intel SMP architectures.