lucenenet-dev mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From Elad Margalit <eladm...@gmail.com>
Subject Re: Lucene.net vs Lucene with IKVM
Date Thu, 05 Feb 2015 14:33:45 GMT
Totally Agree with Syn-Hershko,

Currently, I'm using the ikvm ported version, i can tell for sure, there is
at least 20% penalty in performance.

when you're using the ivkm version, the JIT avoid many optimizations,
in the future, we'll be able to do much more optimizations - async/await,
SIMD instructions when RyuJIT is out in few month, bcl collections,
stackalloc for small inline arrays and many other stuff.
we'll also be able to monitor the critical performance paths in tools such
as ants profiler or dotTrace, and make lucene even more faster.

i have no doubt after the porting will be done, it will be about 15-20%
faster than the native java version.

its a main goal for all of us to contribute to this port

Thank you all guys

Cheers,
E


On Thu, Feb 5, 2015 at 2:47 PM, Itamar Syn-Hershko <itamar@code972.com>
wrote:

> It robs us of many optimization opportunities on the CLR. Java Lucene
> compiled using IKVM will not perform as well as a native port. Think
> async/await, BCL data structures, different GC considerations (LOH for
> example), etc. There's also the issue of supporting PCL, Mono, Mobile and
> Azure natively. Just to name a few.
>
> I've had a chat with the lead developer of IKVM and promised to run an IKVM
> version head to head with the native port once we are done.
>
> --
>
> Itamar Syn-Hershko
> http://code972.com | @synhershko <https://twitter.com/synhershko>
> Freelance Developer & Consultant
> Lucene.NET committer and PMC member
>
> On Thu, Feb 5, 2015 at 2:43 PM, Omri Suissa <omri.suissa@diffdoof.com>
> wrote:
>
> > Hi,
> >
> > It looks like it takes a lot of time to manually port Lucene to .net
> (still
> > in version 3), why not using ikvm to port (at least as a base line and
> > improve from there)?
> >
> >
> > What are the disadvantages of ikvm in this case?
> >
> >
> > Thanks,
> >
> > Omri
> >
>

Mime
  • Unnamed multipart/alternative (inline, None, 0 bytes)
View raw message