Abstract | ||
---|---|---|
One of the nice properties of a tracing just-in-time compiler (JIT) is that many of its optimizations are simple, requiring one forward pass only. This is not true for loop-invariant code motion which is a very important optimization for code with tight kernels. Especially for dynamic languages that typically perform quite a lot of loop invariant type checking, boxed value unwrapping and virtual method lookups. In this paper we explain a scheme pioneered within the context of the LuaJIT project for making basic optimizations loop-aware by using a simple pre-processing step on the trace without changing the optimizations themselves. We have implemented the scheme in RPython's tracing JIT compiler. PyPy's Python JIT executing simple numerical kernels can become up to two times faster, bringing the performance into the ballpark of static language compilers. |
Year | DOI | Venue |
---|---|---|
2012 | 10.1145/2384577.2384586 | DLS |
Keywords | Field | DocType |
boxed value unwrapping,python jit,static language compiler,luajit project,basic optimizations loop-aware,simple numerical kernel,loop-aware optimizations,simple pre-processing step,jit compiler,just-in-time compiler,loop-invariant code motion,performance,loop invariant code motion,tracing jit,natural sciences,optimization,languages | Programming language,Loop-invariant code motion,Type checking,Computer science,Parallel computing,Compiler,Loop invariant,Virtual function,Tracing just-in-time compilation,Python (programming language),Tracing | Conference |
Volume | Issue | ISSN |
48 | 2 | 0362-1340 |
Citations | PageRank | References |
3 | 0.39 | 12 |
Authors | ||
3 |
Name | Order | Citations | PageRank |
---|---|---|---|
Håkan Ardö | 1 | 60 | 8.29 |
Carl Friedrich Bolz | 2 | 292 | 18.69 |
Maciej FijaBkowski | 3 | 31 | 1.65 |