lucene-dev mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From "Michael McCandless (JIRA)" <>
Subject [jira] Commented: (LUCENE-1410) PFOR implementation
Date Mon, 06 Oct 2008 19:31:44 GMT


Michael McCandless commented on LUCENE-1410:

bq. The only concern I have there is that when using a byte[] directly from the file getting
the int values may result in non 4 byte aligned fetches. Can current hardware do this well?

Good question, though if that's the case we could presumably work
around it by ensuring the header of the file is 0 mod 4?

Or... is this because a full block (header + bits + exception data)
may not be 0 mod 4?  (Though we too could pad if necessary)

I think the API we want to reach (eventually) is an IntBlockInput in
Directory where you call read(int[]) and it returns the next 128 (say)
ints in an array and moves itself to the end of that block (start of
the next block).

It's tempting to move to a full C implementation directly now. Should we do that?
A similar move was made in the past by letting gcc deal with vInts, but meanwhile the jvms
caught up there.

Maybe we should explore this eventually, but I think for now we should
first try to get the Java version online?

I'd really love to get a prototype integration working so that we can
then do an end-to-end performance test (ie, actual searches).  I'm still
wondering how much speedups at this layer will actually affect overall
search time.

> PFOR implementation
> -------------------
>                 Key: LUCENE-1410
>                 URL:
>             Project: Lucene - Java
>          Issue Type: New Feature
>          Components: Other
>            Reporter: Paul Elschot
>            Priority: Minor
>         Attachments: autogen.tgz, LUCENE-1410b.patch,,,
>   Original Estimate: 21840h
>  Remaining Estimate: 21840h
> Implementation of Patched Frame of Reference.

This message is automatically generated by JIRA.
You can reply to this email to add a comment to the issue online.

To unsubscribe, e-mail:
For additional commands, e-mail:

View raw message