Re: [xml] Performance of tokenization?

[Bjorn Reese <breese mail1 stofanet dk>]

On Tue, 2004-09-07 at 18:38, Daniel Veillard wrote:

>   Completely linear with raw CPU cycle speed, so depends on your
> CPU. With cheap but recent processors (like an athlon 2600) 
> parsing to generate empty SAX feedback is around 25 MBytes/s. It
> also depends a lot on the content, this is for relatively regular
> and flat ascii data. It's just an order of magnitude there is just
> too many parameters modifying this.

Well, I am thinking more about how large a percentage the scanning and
tokenization takes compared to the entire parsing time (e.g. for
building a DOM.)

To be more precise, I am trying to estimate how much faster the parsing
of binary XML (which has already been tokenized) is compared to textual
XML.