Titanium: a high-performance Java dialect

Abstract

Titanium is a language and system for high‐performance parallel scientific computing. Titanium uses Java as its base, thereby leveraging the advantages of that language and allowing us to focus attention on parallel computing issues. The main additions to Java are immutable classes, multidimensional arrays, an explicitly parallel SPMD model of computation with a global address space, and zone‐based memory management. We discuss these features and our design approach, and report progress on the development of Titanium, including our current driving application: a three‐dimensional adaptive mesh refinement parallel Poisson solver. © 1998 John Wiley & Sons, Ltd.

MPJava: High-Performance Message Passing in Java Using Java.nio

“…Titanium [22] is a dialect of Java that provides new features useful for highperformance computation in Java, such as immutable classes, multidimensional arrays, and zone-based memory management. Titanium's backend produces C code with MPI calls.…”

Evaluating support for global address space languages on the Cray X1

“…Designed as parallel extensions for popular sequential programming languages, GAS languages such as UPC [15], Titanium [29,16], and Co-Array Fortran [22] provide better programmability through the support of a user-level global address space, leading to more flexible remote accesses through language-level one-sided communication. GAS languages thus offer a more convenient and productive programming style than explicit message passing (e.g., MPI [21]), and good performance can still be achieved because programmers retain explicit control of data placement and load balancing.…”

Hierarchical Pointer Analysis for Distributed Programs

“…The Titanium programming language [28] is a high performance dialect of Java designed for distributed machines. It is a single program, multiple data (SPMD) language, so all threads execute the same code image.…”

“…These chips are building blocks in larger shared and distributed memory parallel systems, resulting in machines that are increasingly hierarchical and use a combination of cache-coherent shared memory, partitioned memory with (remote) direct memory access (DMA or RDMA), and message passing. The partitioned global address space (PGAS) model is a natural fit for programming these machines, and languages that use it include Unified Parallel C (UPC) [7,26], Co-Array Fortran (CAF) [25], Titanium [28,12] (based on Java [10]), Chapel [8], X10 [24], and Fortress [1]. In all of these languages, pointers to shared state is permitted, and a fundamental question is whether a given pointer can be proven to access data in only a limited part of the machine hierarchy.…”