One of the nice programming article for your weekend.
This Monday, I met with Marc Feeley, my PhD advisor. I showed him graphs comparing the performance of Higgs to that of the TraceMonkey compiler. Reviewers at conferences we’ve previously submitted papers to have been very skeptical, and pressed us multiple times to produce some kind of comparison of basic block versioning against tracing. We believe that faring well against TraceMonkey could increase our odds of getting our work published.
So far, we’re beating TraceMonkey on about half of our benchmark set. There are, however, cases where we’re not doing well at all. The above graph compares speedups produced by Higgs relative to V8 baseline and TraceMonkey. The scale is logarithmic and higher bars are better. If you look at the last bars at the far right, you’ll notice that they’re actually negative. That’s because our performance on that benchmark is about 100 times slower than that of TraceMonkey…
View original post 950 more words