Performance patch for AIX system api support of sun.nio.ch.FileChannelImpl.transferTo0()

[Steve Poole]

On Wed, 2011-10-19 at 12:46 +0800, Charles Lee wrote:
> On 10/18/2011 01:48 PM, Deven wrote: 
> > Hi nio-devs,
> >     
> >     sun.nio.ch.FileChannelImpl.transferTo0() native method will take
> > advantage of Linux/Solaris system snedfile() apis to improve the
> > performance of
> > FileChannel.transferTo(long position,long count,WritableByteChannel target). 
> > 
> >     On AIX there is a similar system api called send_file() which
> > will do the similar things Linux/Solaris sendfile() will do, however
> > current transferTo0() dose not use it to optimize AIX performance.
> > 
> >      This patch add AIX send_file() into the transferTo0() native
> > method, so AIX could also take advantage of AIX system api support
> > for transferTo operation and will improve the performance of
> > FileChannelImpl.transferTo() .
> > 
> >      The attached TransferToClient.java and TransferToServer.java is
> > the performance test which can test the FileChannelImpl.transferTo()
> > performance.
> > 
> >      If there is anybody could review this patch and give your
> > suggestion? 
> > 
> > Best regards!
> > 
> 
> Hi Deven,
> 
> Can you help to make a webrev for other's review?
> 
> Just with a quick glance, your patch has some format issues. Please
> keep to follow the code convention from
> http://www.oracle.com/technetwork/java/codeconvtoc-136057.html
> 
> Hi Alan,
> 
> Are there any examples of the performance test cases?

The testcases do not have a copyright header - and to be honest they are
not very good testcases.   They have commented out lines, no comments on
what the testcase is doing and (finally) they need to be formed  to fit
into the jtreg format. I'm not clear what that format would be in this
case though :-)


> -- 
> Yours Charles