On Tue, Jul 22, 2003 at 09:14:02PM -0400, Andrew C. Oliver wrote: > 2. Metrics proving that its faster/better (it should be theoretically)
FYI, just did some quick performance tests. They involved simply parsing a 50M xls (using POIFS only, no HSSF) and then dumping it out to /dev/null. On average POIFS took 1.79 seconds and used 60MB. "POIFS2" backed by a RandomAccessFile took 0.82 seconds and used 4MB. I'm sure the speed up from not having to allocate the memory, but it is nice to know that the memory improvements will not come at the expense of performance. Interestingly, there's hardly any difference between a buffered RandomAccessFile and mmapping the entire thing. The implementation of the latter is certainly cleaner (it handles things like endianness for you), and maybe it makes a bigger difference on some platforms, but right now it's not that compelling. Chris --------------------------------------------------------------------- To unsubscribe, e-mail: [EMAIL PROTECTED] For additional commands, e-mail: [EMAIL PROTECTED]
