Citation Context

... compiler is an error-prone task that is made even harder by the dynamic features of a language. A recent approach to getting better performance for dynamic languages is that of tracing JIT compilers =-=[16, 7]-=-. Writing a tracing JIT compiler is relatively simple. It can be added to an existing interpreter for a language, the interpreter takes over some of the functionality of the compiler and the machine c...

Citation Context

.... The savings are primarily the work associated with looking up the destination given the class of the receiver. The virtual guard is an example of a trace exit that guards a speculative optimization =-=[12]-=-. Inlining Traces are agnostic towards method invocation and return, treating them like any other multiple-destination virtual branch instructions. However, when a return corresponds to an invoke in t...

Citation Context

...ance barrier. Our approach is to make justin-time (JIT) compilation effective in a dynamically typed language runtime environment. We accomplish this by using a novel tracebased compilation technique =-=[13]-=-, which is effective in eliminating many of the problems that arise when building traditional JIT compilers for dynamically typed languages. Our trace-based compiler selects and compiles only frequent...

Citation Context

...is important to note that because we are dealing with paths, the propagation/check only has to be invariant on iterations of that particular loop path, and might be variant on other paths in the loop =-=[12]-=-. Similarly, loop-dead propagations are propagations that are only live on the exit of the loop path. During the monitoring run, the path handler invokes loop-invariant handlers only on the first iter...

Citation Context

...based compilation. They use heuristics and profiles to identify and eliminate rarely executed sections of code, but rely on expensive runtime code instrumentation for trace identification. Gal et al. =-=[13, 15]-=- proposed an approach to building dynamic compilers in which no CFG is ever constructed, and no source code level compilation units such as methods are used. Instead, they use runtime profiling to det...

Citation Context

...ction Algorithms, Performance Trace-based Compilation, Trace Selection, Repetition Trace-based compilation has been explored in binary translation systems [1, 7, 8], lightweight JIT for embedded JVMs =-=[12, 20]-=-, and more recently in compilers for dynamically typed languages such as JavaScript [4, 13], Lua [17], and Python [6]. In a tracebased compiler, a trace, typically defined as a single-entry multipleex...

Citation Context

...ly executed hot traces and translates them into optimized machine-code trace fragments which can be cached for subsequent dispatch. Traces are used as the regions of analysis in many lightweight JVMs =-=[6, 12]-=-, e.g., those designed for embedded use where simple interpretation is too slow, yet the inclusion of a full-featured JIT compiler is not viable. Traces are also used as the regions of analysis in man...

Citation Context

...ace runtime management. The optimizer is extremely lightweight and performs basic redundancy elimination optimizations in a forward and a backward pass as well as basic register allocation. HotpathVM =-=[10]-=-, YETI [11] and Maxpath [12] are traceJITs that target Java. HotpathVM emphasizes its efficiency in domains where resources are quite constrained. It introduces a tree-SSA representation for fast anal...