Parallel Studio 2011: Now We Know What Happened to Ct, Cilk++, and RapidMind
Intel has released Parallel Studio 2011, a suite of tools intended to simplify and speed up the development of parallel programs running on multicore architectures. Based on the Parallel Studio package released last year, Parallel Studio 2011 adds expanded threading libraries and a threading advisor to its C++ compiler, error-checking and profiling tools.
The enhancements that Parallel 2011 brings to the table focus on two key areas, the addition of:
- Parallel Advisor, a design tool intended to simplify development by identifying parts of a program that can benefit from parallelization, giving architects the guidance they need to transition from serial to parallel code. Once those opportunities have been identified, Intel Parallel Advisor provides step-by-step recommendations for the best way to implement parallelism and helps developers analyze potential implementation scenarios.
- Parallel Building Blocks (PBB), a set of tools that let developers exploit parallelism with a new set of task and data parallel models that are portable, scalable, reliable, and future proof. These models easily integrate into existing applications, preserving software and hardware investments.
The Parallel Building Blocks component is built on Threading Building Blocks 3.0, Intel's existing C++ template library. What's most interesting, however, is that PBB also includes Cilk Plus, a technology Intel acquired with the acquisition of Cilk Arts. Cilk Plus is a C++ language extension consisting of three keywords that simplify the implementation and addition of task and data parallelism. Cilk Plus provides support for task, data, and vector parallelism.
Equally interesting is the introduction of Array Building Blocks, an API backed by a runtime library that provides a generalized data parallel programming solution that frees developers from dependencies on particular low-level parallelism mechanisms or hardware architectures. ArBB stems from the combination of RapidMind technology which Intel acquired a year ago, and Ct, which was developed by Intel Labs. ArBB is a C++ template library for data parallelism that automatically scales to future Intel platforms. It's use of cores and threads are determined by the runtime compiler.
From a high-level view, Parallel Studio addresses all phases of the development lifecycle, enabling developers to efficiently:
- Design by finding threading opportunities using the Parallel Advisor threading guide
- Build and debug C/C++ applications using the compilers and libraries in Parallel Composer, which now also includes Parallel Building Blocks.
- Verify to improve code reliability and quality by finding crash-causing bugs using the Parallel Inspector memory and threading error checker.
- Tune application performance with the Parallel Amplifier threading and performance profiler.
In addition, Parallel Studio now further assists C/C++ development by offering compatibility with and enhancement of the new parallelism capabilities introduced in Microsoft Visual Studio 2010. (Intel Parallel Studio also continues to offer full parallel programming support for versions 2005 and 2008.)