Designing a whole-program static analysis requires trade-offs between precision and scalability. While a context-insensitive points-to analysis is often considered a good compromise, it still has non-linear complexity that leads to scalability problems when analyzing large applications. On the other hand, rapid type analysis scales well but lacks precision. We use saturation in a context-insensitive type-based points-to analysis to make it as scalable as a rapid type analysis, while preserving most of the precision of the points-to analysis. With saturation, the points-to analysis only propagates small points-to sets for variables. If a variable can have more values than a certain threshold, the variable and all its usages are considered saturated and no longer analyzed. Our implementation in the points-to analysis of GraalVM Native Image, a closed-world approach to build standalone binaries for Java applications, shows that saturation allows GraalVM Native Image to analyze large Java applications with hundreds of thousands of methods in less than two minutes.

Inlining is one of the most important compiler optimizations. It reduces call overheads and widens the scope of other optimizations. But, inlining is somewhat of a black art of an optimizing compiler, and was characterized as a computationally intractable problem. Intricate heuristics, tuned during countless hours of compiler engineering, are often at the core of an inliner implementation. And despite decades of research, well established inlining heuristics are still missing. In this paper, we describe a novel inlining algorithm for JIT compilers that incrementally explores a program's call graph, and alternates between inlining and optimizations. We devise three novel heuristics that guide our inliner: adaptive decision thresholds, callsite clustering, and deep inlining trials. We implement the algorithm inside Graal, a dynamic JIT compiler for the HotSpot JVM. We evaluate our algorithm on a set of industry-standard benchmarks, including Java DaCapo, Scalabench, Spark-Perf, STMBench7 and other benchmarks, and we conclude that it significantly improves performance, surpassing state-of-the-art inlining approaches with speedups ranging from 5% up to 3×.

Established benchmark suites for the Java Virtual Machine (JVM), such as DaCapo, ScalaBench, and SPECjvm2008, lack workloads that take advantage of the parallel programming abstractions and concurrency primitives offered by the JVM and the Java Class Library. However, such workloads are fundamental for understanding the way in which modern applications and data-processing frameworks use the JVM's concurrency features, and for validating new just-in-time (JIT) compiler optimizations that enable more efficient execution of such workloads. We present Renaissance, a new benchmark suite composed of modern, real-world, concurrent, and object-oriented workloads that exercise various concurrency primitives of the JVM. We show that the use of concurrency primitives in these workloads reveals optimization opportunities that were not visible with the existing workloads. We use Renaissance to compare performance of two state-of-the-art, production-quality JIT compilers (HotSpot C2 and Graal), and show that the performance differences are more significant than on existing suites such as DaCapo and SPECjvm2008. We also use Renaissance to expose four new compiler optimizations, and we analyze the behavior of several existing ones. We use Renaissance to compare performance of two state-of-the-art, production-quality JIT compilers (HotSpot C2 and Graal), and show that the performance differences are more significant than on existing suites such as DaCapo and SPECjvm2008. We also use Renaissance to expose four new compiler optimizations, and we analyze the behavior of several existing ones.

Arbitrary program extension at run time in language-based VMs, e.g., Java's dynamic class loading, comes at a startup cost: high memory footprint and slow warmup. Cloud computing amplifies the startup overhead. Microservices and serverless cloud functions lead to small, self-contained applications that are started often. Slow startup and high memory footprint directly affect the cloud hosting costs, and slow startup can also break service-level agreements. Many applications are limited to a prescribed set of pre-tested classes, i.e., use a closed-world assumption at deployment time. For such Java applications, GraalVM Native Image offers fast startup and stable performance. GraalVM Native Image uses a novel iterative application of points-to analysis and heap snapshotting, followed by ahead-of-time compilation with an optimizing compiler. Initialization code can run at build time, i.e., executables can be tailored to a particular application configuration. Execution at run time starts with a pre-populated heap, leveraging copy-on-write memory sharing. We show that this approach improves the startup performance by up to two orders of magnitude compared to the Java HotSpot VM, while preserving peak performance. This allows Java applications to have a better startup performance than Go applications and the V8 JavaScript VM.

Compilers perform a variety of advanced optimizations to improve the quality of the generated machine code. However, optimizations that depend on the data flow of a program are often limited by control-flow merges. Code duplication can solve this problem by hoisting, i.e. duplicating, instructions from merge blocks to their predecessors. However, finding optimization opportunities enabled by duplication is a non-trivial task that requires compile-time intensive analysis. This imposes a challenge on modern (just-in-time) compilers: Duplicating instructions tentatively at every control flow merge is not feasible because excessive duplication leads to uncontrolled code growth and compile time increases. Therefore, compilers need to find out whether a duplication is beneficial enough to be performed.

In large-scale software applications, programmers often combine different programming languages because this allows them to use the most suitable language for a given problem, to gradually migrate existing projects from one language to another, or to reuse existing source code. However, different programming languages have fundamentally different implementations, which are hard to combine. The composition of language implementations often results in complex interfaces between languages, insufficient flexibility, or poor performance. We propose TruffleVM, a virtual machine (VM) that can execute different programming languages and is able to compose them in a seamless way. TruffleVM supports dynamically-typed languages (e.g., JavaScript and Ruby) as well as statically typed low-level languages (e.g., C). It consists of individual language implementations, which translate source code to an intermediate representation that is executed by a shared VM. TruffleVM composes these different language implementations via generic access . Generic access is a language-agnostic mechanism that language implementations use to access foreign data or call foreign functions. It features language-agnostic messages that the TruffleVM resolves to efficient foreign-language-specific operations at runtime. Generic access supports multiple languages, enables an efficient multi-language development, and ensures high performance. We evaluate generic access with two case studies. The first one explains the transparent composition of JavaScript, Ruby, and C. The second one shows an implementation of the C extensions application programming interface (API) for Ruby. We show that generic access guarantees good runtime performance. It avoids conversion or marshalling of foreign objects at the language boundary and allows the dynamic compiler to perform its optimizations across language boundaries.

Most high-performance dynamic language virtual machines duplicate language semantics in the interpreter, compiler, and runtime system. This violates the principle to not repeat yourself. In contrast, we define languages solely by writing an interpreter. The interpreter performs specializations, e.g., augments the interpreted program with type information and profiling information. Compiled code is derived automatically using partial evaluation while incorporating these specializations. This makes partial evaluation practical in the context of dynamic languages: It reduces the size of the compiled code while still compiling all parts of an operation that are relevant for a particular program. When a speculation fails, execution transfers back to the interpreter, the program re-specializes in the interpreter, and later partial evaluation again transforms the new state of the interpreter to compiled code. We evaluate our approach by comparing our implementations of JavaScript, Ruby, and R with best-in-class specialized production implementations. Our general-purpose compilation system is competitive with production systems even when they have been heavily optimized for the one language they support. For our set of benchmarks, our speedup relative to the V8 JavaScript VM is 0.83x, relative to JRuby is 3.8x, and relative to GNU R is 5x.

Measuring the time spent on small individual fractions of program code is a common technique for analysing performance behavior and detecting performance bottlenecks. The benefits of the approach include a detailed individual attribution of performance and understandable feedback loops when experimenting with different code versions. There are however severe pitfalls when following this approach that can lead to vastly misleading results. Modern dynamic compilers use complex optimisation techniques that take a large part of the program into account. There can be therefore unexpected side-effects when combining different code snippets or even when running a presumably unrelated part of the code. This talk will present performance paradoxes with examples from the domain of dynamic compilation of Java programs. Furthermore, it will discuss an alternative approach to modelling code performance characteristics that takes the challenges of complex optimising compilers into account.

Programmers combine different programming languages because it allows them to use the most suitable language for a given problem, to gradually migrate existing projects from one language to another, or to reuse existing source code. However, existing cross-language mechanisms suffer from complex interfaces, insufficient flexibility, or poor performance. We present the TruffleVM, a multi-language runtime that allows composing different language implementations in a seamless way. It reduces the amount of required boiler-plate code to a minimum by allowing programmers to access foreign functions or objects by using the notation of the host language. We compose language implementations that translate source code to an intermediate representation (IR), which is executed on top of a shared runtime system. Language implementations use language-independent messages that the runtime resolves at their first execution by transforming them to efficient foreign-language-specific operations. The TruffleVM avoids conversion or marshaling of foreign objects at the language boundary and allows the dynamic compiler to perform its optimizations across language boundaries, which guarantees high performance. This paper presents an implementation of our ideas based on the Truffle system and its guest language implementations JavaScript, Ruby, and C.

When building a compiler for a high-level language, certain intrinsic features of the language must be expressed in terms of the resulting low-level operations. Complex features are often expressed by explicitly weaving together bits of low-level IR, a process that is tedious, error prone, difficult to read, difficult to reason about, and machine dependent. In the Graal compiler for Java, we take a different approach: we use snippets of Java code to express semantics in a high-level, architecture-independent way. Two important restrictions make snippets feasible in practice: they are compiler specific, and they are explicitly prepared and specialized. Snippets make Graal simpler and more portable while still capable of generating machine code that can compete with other compilers of the Java HotSpot VM.

Self-optimizing AST interpreters dynamically adapt to the provided input for faster execution. This adaptation includes initial tests of the input, changes to AST nodes, and insertion of guards that ensure assumptions still hold. Such specialization and speculation is essential for the performance of dynamic programming languages such as JavaScript. In traditional procedural and objectoriented programming languages it can be tedious to write selfoptimizing AST interpreters, as those languages fail to provide constructs that would specifically support that. This paper introduces a declarative domain-specific language (DSL) that greatly simplifies writing self-optimizing AST interpreters. The DSL supports specialization of operations based on types of the input and other properties. It can then use these specializations directly or chain them to represent the operation with the minimum amount of code possible. The DSL significantly reduces the complexity of expressing specializations for those interpreters. We use it in our high-performance implementation of JavaScript, where 274 language operations have an average of about 4 and a maximum of 190 specializations. In addition, the DSL is used in implementations of Ruby, Python, R, and Smalltalk.

An abstract syntax tree (AST) interpreter is a simple and natural way to implement a programming language. However, it is also considered the slowest approach because of the high overhead of virtual method dispatch. Language implementers therefore define bytecodes to speed up interpretation, at the cost of introducing inflexible and hard to maintain bytecode formats. We present a novel approach to implementing AST interpreters in which the AST is modified during interpretation to incorporate type feedback. This tree rewriting is a general and powerful mechanism to optimize many constructs common in dynamic programming languages. Our system is implemented in Java and uses the static typing and primitive data types of Java elegantly to avoid the cost of boxed representations of primitive values in dynamic programming languages.

Dynamic updates to running programs improve development productivity and reduce downtime of long-running applications. This feature is however severely limited in current virtual machines for object-oriented languages. In particular, changes to classes often apply only to methods invoked after a class change, but not to active methods on the call stack of threads. Additionally, adding and removing methods as well as fields is often not supported.

While dynamic code evolution in object-oriented systems is an important feature supported by dynamic languages, there is currently only limited support for dynamic code evolution in high-performance, state-of-the-art runtime systems for statically typed languages, such as the Java Virtual Machine. In this tool demonstration, we present the Dynamic Code Evolution VM, which is based on a recent version of Oracle's state-of-the-art Java HotSpot(TM) VM and allows unlimited changes to loaded classes at runtime. Based on the Dynamic Code Evolution VM, we developed an enhanced version of the Mantisse GUI builder (which is part of the NetBeans IDE) that allows adding GUI components without restarting the application under development. Furthermore, we redesigned the dynamic AOP framework HotWave to take advantage of the enhanced dynamic code evolution capabilities. The new version, HotWave2, now supports most AspectJ constructs, including around() advice and static cross-cutting. We will demonstrate both the enhanced Mantisse GUI builder as well as HotWave2, weaving several aspects for dynamic analysis in sizable applications at runtime.

Intense research on virtual machines has highlighted the need for flexible software architectures that allow quick evaluation of new design and implementation techniques. The interface between the compiler and runtime system is a principal factor in the flexibility of both components and is critical to enabling rapid pursuit of new optimizations and features. Although many virtual machines have demonstrated modularity for many components, significant dependencies often remain between the compiler and the runtime system components such as the object model and memory management system. This paper addresses this challenge with a carefully designed strict compiler-runtime interface and the XIR language. Instead of the compiler backend lowering object operations to machine operations using hard-wired runtime-specific logic, XIR allows the runtime system to implement this logic, simultaneously simplifying and separating the backend from runtime-system details. In this paper we describe the design and implementation of this compiler-runtime interface and the XIR language in the C1X dynamic compiler, a port of the HotSpotTM Client compiler. Our results show a significant reduction in backend complexity with XIR and an overall reduction in the compiler-runtime interface complexity while still generating comparable quality code with only minor impact on compilation time.