IBM has made available version 2 of its Multicore Software Development Kit for Java, a freely available all-in-one toolkit for the development of parallel programs on multicore platforms. The SDK offers various useful tools -- including those for testing, debugging, and analyzing applications targeting multicore systems -- to help developers and testers to find concurrency errors and locate performance bottleneck in parallel/multi-threaded programs.

The tools are designed for analyzing various concurrency-related problems in the applications which arise due to different types of memory and synchronization errors. Also, the SDK contains tools to do lock analysis and synchronization coverage analysis. Writing scalable and high-performance applications requires detailed analysis and profiling tools to help point out problem areas in the code. The MSDK provides tools to do profiling of all layers of the application execution stack.

The tools in the Multicore SDK include both static and dynamic analysis tools. The following toolsets are included in MSDK 2.0:

• Synchronization Coverage Analysis. A testing tool to figure out whether the different synchronization primitives used in the application code is doing something useful, or whether it is redundant. It is a dynamic tool which uses on-the-fly bytecode rewriting techniques to instrument the application.

• Orange Box Analysis. Useful for debugging program crashes. It provides the last few (configurable) read/write values of variables by each thread. This is a dynamic analysis tool which uses on-the-fly byte code rewriting techniques to instrument the application.

• Race Detection Tool. A debugging tool to detect data races in the application code. This is a dynamic analysis tool which uses on-the-fly byte code rewriting techniques to instrument the application.

• Deadlock Detection Tool. A debugging tool to find deadlocks (and potential deadlocks) in the application code. This is a dynamic analysis tool that uses on-the-fly bytecode rewriting techniques to instrument the application.

• Lock Status Report. Useful for debugging concurrent applications. It provides details about which thread holds, and which lock waits for a lock, and at what point in the code. This is a dynamic tool which uses on-the-fly bytecode instrumentation.

• MHP Analysis. A static analysis tool. It does a may-happen-in-parallel analysis and provides the statements that can execute in parallel. This analysis can be used in conjunction with the Race detector tool to reduce overheads and improve performance.

• Performance Analysis Tool. A detailed performance analysis tool to figure out performance bottlenecks. It instruments the complete application execution stack. Monitoring of the hardware performance counters, various system calls, operating system thread switches, various Java virtual machine functions and user application can be performed. Detailed profiling data can be visualized during application execution and can also be stored to do post-mortem analysis.

All tools are part of an Eclipse environment. The Performance Analysis tool is platform dependent and runs on x86/Linux. The other tools work with IBM Java 5 or higher.