Tracing vs. partial evaluation: comparing meta-compilation approaches for self-optimizing interpreters

Abstract

Tracing and partial evaluation have been proposed as metacompilation techniques for interpreters to make just-in-time compilation language-independent. They promise that programs executing on simple interpreters can reach performance of the same order of magnitude as if they would be executed on state-of-the-art virtual machines with highly optimizing just-in-time compilers built for a specific language. Tracing and partial evaluation approach this metacompilation from two ends of a spectrum, resulting in different sets of tradeoffs.This study investigates both approaches in the context of self-optimizing interpreters, a technique for building fast abstract-syntax-tree interpreters. Based on RPython for tracing and Truffle for partial evaluation, we assess the two approaches by comparing the impact of various optimizations on the performance of an interpreter for SOM, an objectoriented dynamically-typed language. The goal is to determine whether either approach yields clear performance or engineering benefits. We find that tracing and partial evaluation both reach roughly the same level of performance. SOM based on meta-tracing is on average 3x slower than Java, while SOM based on partial evaluation is on average 2.3x slower than Java. With respect to the engineering, tracing has however significant benefits, because it requires language implementers to apply fewer optimizations to reach the same level of performance.

Building efficient and highly run-time adaptable virtual machines

“…We may lead to conclusions that may not generalize to other programs. To mitigate this threat we selected a set of benchmarks included in the literature of the field [12,14]. We also evaluated the MOP in use cases appearing in previous works (read-only [2], tracing [15]).…”

“…TruffleSOM [12,14] is a SOM implementation using Truffle. When used in combination with Graal it is just 2.7x slower than Java for a set of classic VM benchmarks (cf.…”

A flexible framework for studying trace-based just-in-time compilation

“…This enables the interpreter's implementers to employ the benefits of trace-based compilation without having to create their own dedicated tracing compiler. Furthermore, this makes it possible for developers to rapidly create interpreters with an acceptable performance level [16].…”

“…The Truffle [23] and RPython [4] frameworks address this problem by reducing the language-specific engineering that is required by applying partial evaluation and meta-tracing to relatively simple interpreters. It has recently been shown [16] that the technique of metatracing is capable of lifting the performance of a meta-traced interpreter to the same order of magnitude of a dedicated just-in-time compiler, while requiring less engineering effort from the developers of this interpreter.…”

Applying Optimizations for Dynamically-typed Languages to Java

“…All benchmarks reach their peak performance within the irst 5 iterations on HotSpot C2. For a more detailed discussion about Trule warm-up performance, we would like to refer the reader to [19].…”