Breakthroughs are sought for architecture as well as devices. The success of the Cell Broadband Engine™, developed for game consoles, and BlueGene, the world's fastest supercomputer, demonstrated that radically new computer architectures are needed to cope with the increasing variation of computer workloads. Using the experience and skills gained from the development of the world's fastest Java JIT compiler, TRL is advancing research and development for system software for these new architectures.
Competency fields
Java Synchronization
Because of the built-in support for multi-threaded programming, Java programs perform large numbers of lock operations to synchronize threads. Since most of these locks cannot be eliminated by JIT compi-lers, accelerating the lock performance itself is very important when trying to execute Java-based applications more quickly. Based on deep analyses of lock behavior in Java, we are developing new algo-rithms for Java synchronization, some of which are already includedin IBM's production Java virtual machines.
JVM on Hypervisors
Hypervisor technology is becoming increasingly popular. It virtualizes real hardware, thereby allowing multiple operating systems in a singlemachine. Jointly with the T.J. Watson Research Center and the Austin Research Laboratory, we are building a Java virtual machine (JVM) which runs directly on a hypervisor. This JVM on Hypervisor makes it possible to dynamically construct a scalable Java execution environment on a large cluster of machines. In addition, since the general-purpose OS layer could be removed, it allows for highly tuned resource manage-ment for the Java execution environment.
Advanced JIT Compiler Optimization
When Java was announced in 1995, TRL started to develop JIT compilers, which drastically improved the performance by dynamically translating bytecode to native code. We are currently focusing on two projects. One is idiom recognition, which replaces a specific pattern with faster code. The other is more effective register allocation to reduce memory accesses. These techniques contribute significantly to the performance of IBM JIT compilers, especially on the IBM System z. Papers on these developments have been presented at prestigious international conferences.
Java JIT compiler
Computer Science
Commercial Scale-Out (CSO)
Using a large cluster of commodity machines, such as PCs or blades, we aim to provide both scalable performance and fault tolerance when running commercial applications in parallel. Specifically, we are developing a new Java parallel programming model and an execution envi-ronment based on the parallel skeleton concept, with automatic failover upon machine failure, including data replication and checkpoints. With this technology, users are able to construct scalable applications without worrying about the details of parallelization or fault tolerance. We are working to apply this technology to batch programs, such as public utility billing and credit reimbursement, supporting both files and databases as primary data sources with transaction capabilities.
High Performance Computing
Optimization technologies for applications and system software are crucial for supercomputers, such as the world's fastest supercomputer, BlueGene/L (according to the November 2007 list of the Top 500 supercomputers). Our research topics include application-specific optimization technologies, especially for Quantum Chromodynamics, Fast Fourier Transforms, and Computational Fluid Dynamics. We are also working on automatic program optimization technologies for the US DARPA HPCS project. These activities involve collaboration with the T.J. Watson Research Center.
