Re: [xml] libxml2 and SSE4.2 optimizations

[Cory Nelson <phrosty gmail com>]

Based purely on my own experience, these provide mixed benefit to a
fully validating parser. It depends on your input and how willing you
are to specialize.

For XML 1.0, valid code points are explicit and there are a ton of
them. It's not something easily fit into this SSE model. You're still
going to end up checking each code point individually and a simple LUT
already does this really quickly. You may be able to speed this up
somewhat by special-casing when you're decoding UTF-8 and all bytes
represent a single code point.

On Fri, Jun 13, 2014 at 10:04 AM, Chris Healy <cphealy gmail com> wrote:
> I am working on a project that uses libxml2 extensively that also uses a
> newer Intel chipset.  I'm looking to improve it's performance and have read
> that the newer Intel chipsets have new instructions (SSE4.2) that can help
> accelerate XML parsing:
>
> https://software.intel.com/en-us/articles/xml-parsing-accelerator-with-intel-streaming-simd-extensions-4-intel-sse4/
>
>
> Are the methods discussed in this whitepaper applicable to libxml2?  Are
> there any technical reasons why this could not / should not be done?
>
> Regards,
>
> Chris
>
> _______________________________________________
> xml mailing list, project page  http://xmlsoft.org/
> xml gnome org
> https://mail.gnome.org/mailman/listinfo/xml



-- 
Cory Nelson
http://int64.org