Check out the new USENIX Web site. next up previous
Next: Floating-Point precision toggle Up: Optimizations for the FastVM Previous: Register Allocation

Method inlining

This section shows the impact of inlining small methods. Before register allocation takes place we inline short methods before register allocation takes place. Figure 11 shows that inlining of short methods generally has only a marginal influence on performance for most benchmarks. Javac is the only exceptions with a speedup of 1.15.

Figure: Benchmark speedup for inlining of short methods. The performance number represents the ratio between the measured runtime with method inlining enabled and the measured runtime with method inlining disabled.
\includegraphics[height=2.5in]{inline.eps}