Thanos Stratikopoulos

The emergence of heterogeneous hardware resources has posed a significant obstacle for the Java programming language to keep up with the “Write Once Run Anywhere” paradigm. The reality is that several parts of the Java Virtual Machine must be modified to make Java portable for execution on modern heterogeneous hardware, such as GPUs and FPGAs. To tackle that challenge, TornadoVM is implemented in the University of Manchester as an open-source software technology that enables OpenJDK and other JDK distributions to offload parts of Java applications onto heterogeneous hardware for parallel execution.

This talk will present the latest features of TornadoVM, and it will showcase examples of how the TornadoVM API can be utilized to abstract heterogeneous hardware and increase the performance of Java applications. The goal is to explain the basic terms of TornadoVM including all new API extensions, in order for developers to take advantage of heterogeneous hardware with minimal effort. Finally, this presentation will discuss how TornadoVM is used in research and industry for the acceleration of Java applications.

Hardware acceleration has become prevalent in most application domains as a means to increase performance, while also achieving high energy efficiency. However, the programming models for heterogeneous hardware accelerators inherently support C/C++, thereby hindering the exploitation of heterogeneous resources from managed languages, such as Java. In the University of Manchester, we have been developing TornadoVM; an open-source software technology that can be used as a plugin to OpenJDK and other JVM distributions to enable hardware acceleration in a programmer friendly manner.

In this talk, we will present a practical view of TornadoVM and focus on two parts: (i) analyze what can impact the performance of applications on heterogeneous co-processors, and (ii) how Java developers can utilize TornadoVM to increase the performance of their applications.

https://github.com/beehive-lab/TornadoVM

The proliferation of heterogeneous hardware in recent years has driven us to consider that every system we program, most likely includes a mix of computing elements; each of which with different hardware characteristics enabling programmers to improve performance while reducing energy consumption. These new heterogeneous devices include multi-core CPUs, GPUs and FPGAs. This trend has been accompanied by changes in software development norms that do not necessarily favor programmers. A prime example is the two most popular heterogeneous programming languages, CUDA and OpenCL, which expose several low-level features to the API making them difficult to use by non-expert users.

Instead of using low-level programming languages, developers in industry and academia tend to use higher-level, object-oriented programming languages, typically executed on managed runtime environments, such as Java, R, and Javascript. Although many programmers might expect that such programming languages would have already been adapted for transparent execution on heterogeneous hardware, the reality is that their support is either very limited or absent.

In this talk, we present TornadoVM (https://github.com/beehive-lab/TornadoVM), a heterogeneous programming framework for Java programs. TornadoVM co-exists with standard JVMs (e.g., OpenJDK) that implement the JVMCI. TornadoVM consists of three components: 1) a simple API for composing pipelines of existing Java methods, 2) an optimizing JIT compiler that extends the Graal compiler with hardware-aware optimizations that generate OpenCL C code, and 3) a runtime system that executes TornadoVM specific bytecodes, performs memory management, and schedules the code for execution on GPUs, multicore CPUs, and FPGAs. Essentially, TornadoVM is a “VM-in-a-VM” that can adapt execution completely dynamically and transparently to the user, always finding the highest-performing combination of hardware accelerators through dynamic reconfiguration.