hc-dev mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From "Tony Poppleton (JIRA)" <j...@apache.org>
Subject [jira] Commented: (HTTPCORE-177) Optimize AbstractSessionInputBuffer#read(byte[], int, int)
Date Sat, 16 Jan 2010 00:30:54 GMT

    [ https://issues.apache.org/jira/browse/HTTPCORE-177?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=12801029#action_12801029
] 

Tony Poppleton commented on HTTPCORE-177:
-----------------------------------------

One possible implementation would be for any read method to always fill up the buffer completely
from 0 to the buffer size.  The buffer would then be processed appropriately by each method,
until the buffer position hits the end of the buffer, at which point a whole new buffer would
be read in entirely replacing the previous buffer content.

This would remove the compacting of the buffer in fillBuffer().  If I understand the code
correctly, then if you were to call the single byte read method repeatedly then it would continually
shift the whole buffer back by 1, and then read in a single byte from the stream to fill the
gap?  I may have misunderstood though, as that seems very inefficient.

Anyway, would this implementation work?  I have not studied the class is enough detail about
exactly how each method processes the buffer to know yet...

> Optimize AbstractSessionInputBuffer#read(byte[], int, int)
> ----------------------------------------------------------
>
>                 Key: HTTPCORE-177
>                 URL: https://issues.apache.org/jira/browse/HTTPCORE-177
>             Project: HttpComponents HttpCore
>          Issue Type: Improvement
>          Components: HttpCore
>    Affects Versions: 4.0-beta3
>            Reporter: Oleg Kalnichevski
>             Fix For: 4.1-beta1
>
>
> At the moment AbstractSessionInputBuffer#read(byte[], int, int) always copies input to
the internal data buffer prior to copying it to the destination buffer. The performance of
this method may potentially be improved by eliminating intermediate data copy operation and
reading input directly to the destination buffer, especially when dealing with large chunks.
> Oleg

-- 
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: dev-unsubscribe@hc.apache.org
For additional commands, e-mail: dev-help@hc.apache.org


Mime
View raw message