Re: Streamtokenizer
"Bryce McKinlay" <[email protected]>
| Newsgroups | gmane.comp.gcc.java.devel |
|---|---|
| Message-ID | <[email protected]> |
On Sat, May 10, 2008 at 3:36 AM, Tom Tromey <[email protected]> wrote: >>>>>> "OneGuy" == OneGuy <[email protected]> writes: > > OneGuy> Memory usage was even worse (55 MB with Classpath vs 35 MB with Sun's). > > It looks like our nextToken will often create a new StringBuffer. > Perhaps it would be better to make one and cache it. > > FWIW I'd accept a clean patch to further optimize StreamTokenizer. > Most likely the only reason it performs poorly is that it never popped > up as a performance limiter before. Double.parseDouble() could also be used, avoiding creating a Double each time. I haven't really been following this stuff, but what is the situation with merging in stuff from OpenJDK? It looks like they have a better implementation here which could very easily be dropped in. Bryce